حمید عبادی

With the development of remote sensing science, the use of hyperspectral images is becoming more widespread. Classification is one of the most popular topics in hyperspectral remote sensing. In the last two decades, a number of methods have been proposed to address the problem of hyperspectral data classification.In the present study, a structure based on learning capsule networks has been used to classify hyperspectral images, so that the network structure can have the most optimal generation of features by using a convolution layer and a capsule layer, and at the same time Avoid overfitting the on training data. The obtained results show the high quality of production features in the proposed structure. With the development of remote sensing science, the use of hyperspectral images is becoming more widespread. Classification is one of the most popular topics in hyperspectral remote sensing. In the last two decades, a number of methods have been proposed to address the problem of hyperspectral data classification.In the present study, a structure based on learning capsule networks has been used to classify hyperspectral images, so that the network structure can have the most optimal generation of features by using a convolution layer and a capsule layer, and at the same time Avoid overfitting the on training data. The obtained results show the high quality of production features in the proposed structure.In order to improve the classification accuracy, the feature extraction approach through the designed network and the classification by the Extreme Gradient Boosting was compared with the classification method by the global deep network. The proposed capsule approach consists of 3 basic layers: 1) Prime caps, which are capsules of size 8 and 32 with 9 × 9 filters and movement step 2, 2) Digitcaps with 10 16-dimensional capsules, and 3) fully connected layer. The results of examining two approaches for deep networking as well as combining capsule networks with XGBoost reinforcement tree algorithm were compared. Approaches such as SVM, RF-200, LSTM, GRU and GRU-Pretanh were considered to compare the proposed approach based on the configurations mentioned in their research.Up in addition to the study and quality measurement of production vector deep features by the proposed method in different classifiers, the ability of deep global networks in the application of classification should also be examined. The results of examining two approaches for deep network and also combining CapsNet with XGBoost show that by using the proposed combined method, images are classified with 99% accuracy on training data and 97.5% accuracy on test data.Up in addition to the study and quality measurement of production vector deep features by the proposed method in different classifiers, the ability of deep global networks in the application of classification should also be examined.The results of examining two approaches for deep network and also combining CapsNet with XGBoost show that by using the proposed combined method, images are classified with 99% accuracy on training data and 97.5% accuracy on test data.

The use of accurate and up-to-date photogrammetric products as a basis for urban planning is very important because cities are complex and dynamic physical and social systems that are constantly changing. Urban planning experts and scientists need a wide range of spatial maps and technologies to make decisions to advance its benefits. This requires providing fast and accurate mapping methods. One of the low-cost and accurate mapping method is UAV-based photogrammetry but the limitations of this method require special solutions for developing the system in urban areas. Currently, most UAV photogrammetry projects are carried out in urban areas, without considering changes in the elevation of the ground and the heights of various features, as well as occlusion points. All three of these cases have a great impact on the accuracy and quality of output products in areas with altitude features. The developed three-dimensional flight planning in this article, in addition to paying attention to the height of various features such as buildings and terrain, makes the scale as uniform as possible and prevents the UAV from hitting obstacles. The flight design also includes a new technique such as oblique or semi-oblique imaging to reduce occlusion points. The proposed method was implemented in an urban area. The results of the algorithm were evaluated based on the accuracy, and quality of point clouds and output products, the accuracy of the checkpoints and comparison with the expected standard errors, scale stability in images, density of point clouds, and reduction of occlusion points. The results show the high quality of the products and the planimetric accuracy of 3.6 cm and the altitude accuracy of 7 and 4 cm at the requested scale of 1.750. Also, the results show that oblique images have a significant effect in generating information in the connection between walls and the ground and the density of building facades in the area. The results show that the design and implementation of the UAV photogrammetric flight planning algorithm based on the topographic model and features is successful.

Although groundwater withdrawal is considered to be the main and most common cause of land subsidence, extraction of oil and gas, or minerals from beneath the earth's surface also results in land subsidence. Land subsidence over oil and gas fields leads to several natural and manufactured disasters such as critical and strategic infrastructure destruction. In addition, it imposes significant costs on the oil industry, which indicates the need to control and monitor land subsidence in oil fields. Due to irrecoverable damage caused by this phenomenon, it is necessary to regularly monitor areas subject to subduction in oil and gas fields. There are different ways to measure crustal deformation in Geomatics engineering based on geodetic and remote sensing technologies. Despite high accuracy, the use of geodetic techniques such as Global Navigation Satellite System (GNSS) to regularly monitor crustal deformation in a wide coverage is limited by some disadvantages like discontinuous data collection, the need for installation of the equipment on the ground or direct contact with the ground which is time- and cost-consuming. InSAR technology has been widely used in recent decades to monitor crustal deformations with high spatiotemporal resolution and lower cost and time. This study aims to present a new methodology based on the Persistent Scatterer-InSAR (PS-InSAR) method to investigate the correlation between land subsidence and petroleum extraction in the southwestern region of Iran with active oil and gas fields during two periods from 2017/04 to 2017/10 and from 2019/04 to 2019/10, respectively with the highest and lowest petroleum extraction. The results showed that subductions of 10 to 30 cm during the period of 2017/04 to 2017/10 with an annual rate of 20 to 50 cm have occurred in southwestern Iran around the active oil fields which might be due to the significant petroleum extraction since 2016.

By improvement of the spatial resolution of remote sensing images, more accurate information are provided from the image scene such as texture structures. However, extraction of land cover information from these datas has become a challenging process due to the high spectral diversity and the heterogeneity of surface materials. Visual interpretation is costly and time consuming and automatic interpretation of images does not necessarily lead to high accuracy. Achieving optimal interpretation accuracy requires the design of automatic algorithms that are capable of dealing with the complexity of the image scene. To overcome this problem, object-based image analysis (OBIA) that is sensitive to the image scene morphology, can be particularly effective in an urban area where the density of man-made structures is high. In object-based classification, pixels of a segment are analyzed in combination with each other. So the dimensions of the problem space are reduced, in compared to the pixel-based method, which leads to increasing the computational speed. Meanwhile, due to the different sizes of image segments, supervised object-based classification faces challenges in creating an optimal training set. In this research, AdaBoost algorithm was selected for the object-based classification, to overcome the problem of feature space imbalance, due to the small number of training samples in comparison with the high dimensions of the feature space (including spectral, spatial and geometric features), two strategies were proposed. In the first approach an active learning mechanism was integrated with AdaBoost to produce optimal training data set (OTD) and in another approach based on the feature-to-feature correlation (redundancy) and the feature-to-class correlation (relevance), the candidate feature subset (CFS) was generated to reduce the size of the feature space. To evaluate the proposed method, the developed algorithm was performed on the standard dataset of Vaihingen in Germany and the results were compared with the pixel-based classification. In order to evaluate the signification of the results, the McNemar statistical test was used. The experimental results showed that the proposed object-based approach improved the overall accuracy by 6% and the kappa coefficient by 7% compared to the pixel-based approach. Also, the computational speed of proposed object-based AdaBoost was significantly increased compared to the pixel-based approach. These results indicate the superiority of the proposed approach both in terms of accuracy and processing speed.

Reference evapotranspiration (ET0) is a major research area of both hydrology and water resources management, especially in irrigation agriculture. The most important and direct application of ET0 is in the field of irrigation. One of the conventional methods for estimating reference evapotranspiration using meteorological data is the Penman-Monteith-FAO equation. This equation due to satisfactory results has been used in a variety of climates around the world. However, the lack of necessary meteorological data makes it difficult to estimate spatially distributed ET0 using the FAO-PM method in the wider ungauged areas. Penman Monteith method requires air temperature, wind speed, relative humidity, solar radiation and etc.To overcome the existing limits of the FAO-PM model, various attempts aiming to estimate ET0 with limited observed data have been conducted. At present, remote sensing methods are the only way to obtain the various variables at the temporal and spatial scales needed to estimate evapotranspiration. In recent years, several algorithms have been proposed to estimate reference evapotranspiration using remote sensing data. Some of these models, which are based on the relationship of energy balance, are called surface energy balance methods. In addition to remote sensing, data analysis techniques based on machine learning (ML) are more frequently used in agricultural studies in recent years, especially in evapotranspiration. Therefore, analyses performed with ML algorithms, when coupled with remote sensing data, have the potential to predict the biophysical variables, mainly due to the adaptive capacity of the models to find patterns in nonlinear behavior variable, such as ET0. Machine learning methods are well known and have been widely used in other engineering sciences. The purpose of this study is to estimate the reference evapotranspiration using machine learning algorithms and remote sensing data and finally to analyze the algorithms used. In general, the final results of evapotranspiration estimation depend on factors such as the type of data and the method for estimating evapotranspiration.In this study, the standard method of estimating ET0 with meteorological data, Penman-Monteith FAO equation has been used. The NDVI vegetation index indicates the amount of vegetation on the ground and is sensitive to the early stages of phenology. But the enhanced vegetation index (EVI) minimizes atmospheric effects and differences in blue and red reflections. The SAVI index is used to calculate the vegetation of the land surface that has moderated the effect of soil on it. Three machine learning algorithms were introduced to train the ET0 models, including random forest (RF), gradient boosting regressor (GBR) and support vector regression (SVR). Random forest are one of the machine learning methods that performs classification and regression using Bootstrap and Bagging methods. In this research, three machine learning algorithms with different input data (vegetation indices and all bands of Landsat 7 and 8) are used and after comparing the results, the best model was selected. Performance Evaluation Indicators to compare and evaluate the performance of the studied models, the parameters of mean square error (RMSE), mean absolute error (MAE), coefficient of determination (R2) and correlation coefficient (CC) are used. Finally, according to the results of the two approaches used in this study, using the values of all Landsat bands, the reference evapotranspiration can be estimated with more accuracy.Accurate estimating of reference evapotranspiration is necessary to estimate irrigation needs and in general, to accurately manage water resources. Conventional methods of measuring evapotranspiration are reference using meteorological data. These measurements are point-based, so they are only suitable for very small scale areas. At present, remote sensing methods are the only non-terrestrial way to obtain the various variables at the temporal and spatial scales needed to estimate reference evapotranspiration. In order to reduce the dependence on climatic data and better resolution, machine learning methods are used to calculate the reference evapotranspiration. In this research, RF, GBR and SVR models have been used. In the present study, there are two approaches Used. In the first approach, the values of all bands of Landsat images are as model input, while in the second approach, vegetation indices are calculated with only a few bands of Landsat images and then used as model inputs. By examining, it can be seen that the information obtained from the Landsat image bands is related to the phenological behavior of the products, and it is also possible to contract very relevant information related to agricultural products that are examined temporarily and spatially. One of the factors influencing the accuracy of estimating reference evapotranspiration is the use of other Landsat bands in addition to the bands related to vegetation indices.

Remote Sensing (RS), as one of the most efficient mapping technologies, is employed in wide areas due to its speed, cost-effectiveness, monitoring over wide areas and using time series data. So far, several data and methods are used for this purpose. In general, RS active and passive sensors provide useful information in various applications such as building extraction, natural resource management, agricultural monitoring, etc. The extraction of accurate information about the location, density and distribution of buildings in the urban areas is one of the major challenges in the urban study which is used in various applications. In this framework, the monitoring of the urban parameters, such as urban green space, public health, and environmental justice, urban density and so on has been accomplished by radar and optical image processing, in the last three decades. So far, various methods, including Artificial Intelligence (AI), Deep Learning (DL), object-based methods, etc. have been proposed to extract information in the urban areas. However, an important issue is access to the powerful computer hardware to process the time-series images. In such a situation, the use of the Google Earth Engine (GEE) as a web-based RS platform and its ability to perform spatial and temporal aggregations on a set of satellite images has been considered by many researchers. In this research, a semi-automatic method was developed building extraction in Tabriz, northwest of Iran, based on the satellite images using the GEE cloud computing platform. Since accessible data is one of the most important challenges in the use of space RS, in this study, the free Sentinel-1 and sentinel-2 data, which belongs to the European Space Agency (ESA), has been utilized.
 

2-1- Study Area The study area is central part of the city of Tabriz East Azerbaijan Province, which is located in northwestern of Iran. 2-2- Data Various data sources have been used in this study, including Sentinel-1, Sentinel-2, and the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM). In addition, 400 training samples were created using High-Resolution Google Earth Imagery (GEI) in two classes: urban-residential (buildings) and non-residential areas (vegetation, soil, road, water and etc.).
2-3-

The goal of this research is to develop a method for identifying the buildings in an urban area. For this purpose, after importing images and pre-processing them in the GEE Platform, a map of the Primary Urban Areas (PUA) and High-Potential Buildings (HPB) was produced from Sentinel-1 images according to the sensitivity of the radar images to the target physical parameters. Then, in order to remove the annoying features and extract the Secondary Urban Areas (SUA), spectral indices such as Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Renormalized Difference Vegetation Index (RDVI), Normalized Difference Water Index (NDWI), Soil Extraction Index (SOEI), Normalized Difference Built-up Index (NDBI), and Build-up Extraction Index (BUEI) were extracted from Sentinel-2 images. Also, the high slope of the area and the mountainous areas was extracted from the SRTM DEM data and used as a mask in the final results. Afterwards, the unimodal histogram thresholding method was used in order to determine the threshold value for each index. Finally, by merging the map of HPB and the map of SUA, the final map was produced and evaluated by other methods. In this research, the proposed method used images from GEI with a very high spatial resolution to validate the generated map. As a result, sampling was carried out using a visual interpretation of GEI in two classes: residential areas (buildings) and non-residential areas. The samples were selected randomly and 400 points were collected for each residential and non-residential class. In the study area, a total of 800 test points were used to evaluate the results of the proposed method. To evaluate the accuracy of the results, the criteria of overall accuracy (OA), kappa coefficient (KC), user accuracy (UA) and producer accuracy (PA) were used.

According to the visual interpretation, all buildings in urban areas with a length and width greater than 10 meters (spatial resolution of the four major bands of Sentinel2) can be extracted using the proposed method in this study, and the results are acceptable in various features. According to the proposed method, annoying features such as vegetation and water body areas were removed from the building identification process with high accuracy, and the accuracy in the study area was improved. The results showed that the OA and KC were 90.11 % and 0.803, respectively. Based on the quantitative and qualitative comparisons, the proposed method had a very satisfying performance.

Due to the spectral diversity and the presence of various features in urban environments, preparing a map related to it in a large area is extremely difficult. In this regard, the current study presented a very fast semi-automatic method for preparing the urban area map and extracting buildings in Tabriz using Sentinel-1 and Sentinel-2 satellite images as a time series in the GEE platform. One of the most significant benefits of the proposed method is that the data and processing system used in our study is free. Thus, in addition to not having to download large amounts of data, the method presented in the current study has the ability to eliminate many of the limitations of traditional methods, such as classification methods and their requirement for large training samples. The proposed method did not extract the map of buildings using heavy and complex algorithms, which was an important consideration in the discussion of computational cost. Therefore, it can be concluded that the simultaneous use of Radar and optical RS data in the GEE Web-Based platform has a very high potential in distinguishing features and building mapping.

Iran is one of the countries exposed to many natural disasters, of which the flood phenomenon is one of the most serious. According to official statistics, more than 50% of casualties due to natural disasters in Iran are due to floods. Accurate and timely information is needed to plan any flood management measures. Currently, many researchers have examined the methods of obtaining water area information. Among these methods, land use classification and change analysis are the most important applications of remote sensing techniques. One of the most widely practical classifications is the separation of water class from other classes.Also, Landsat images are one of the most common data sources in this field. In this regard researches have been done in the field of identifying water areas, due to its high importance in flood crisis management with the advancement of satellite technology in remote sensing. In this study, an automated method has been developed to prepare flood change maps and use different capabilities of different water and humidity indicators to provide automated training data. This can reduce the challenge of operator interferences. Besides, the efficiency of the high level of automation in the extraction of flooded areas in accelerating and facilitating the management of this crisis is indisputable. Therefore, when manual training data is not available, applying a fully automated approach will be helpful. The integration of indicators can also be effective in improving the accuracy of existing water indicators. Floods are one of the most important natural disasters in the world that occurred due to different factors such as human encroachment on rivers and vegetation damages. The purpose of this study is to develop an automated method for preparing training data for supervised classification of images and using the capabilities of integrated indicators to identify flooded areas. Therefore, several approaches for flood in Golestan province, 2019, were implemented by Landsat-8 data and the results of each were examined. The Kmeans clustering algorithm, Otsu, Multi and Adaptive thresholds were used to generate automated training data based on different water indices (NDWI, AWEI and MNDWI; Then the ANN, SVM, ML, MD and BE classifications) was implemented to monitor flood changes. The results compared with the manual classification method indicated that the proposed approach, in addition to providing a high-level of automation in monitoring changes, also had high efficiency and accuracy. In another development approach, different water and moisture indices were combined with the aim of improving the production process of automated training data. Among the 85 tests performed, the combined approaches had the highest accuracy. Due to the nature of different water indicators, their selection and appropriate combination, in addition to reducing the noise in the water indicator image was also effective in increasing the ability to identify impure water areas. Finally, a comprehensive deductive analysis of automated methods and hybrid indicators was carried out to help to manage flood crisis, facilitate and accelerate the process. It can be said that with the advancement of technology and the emergence of various aerial / satellite images, many opportunities and possibilities have been created for effective and sustainable management. One of the best ways to extract information from digital images is to classify the image. One of the challenges in image classification is to reduce the interference of users using algorithms that have high level of automation. This issue has high importance due to the sensitivity and sudden occurrence of the flood crisis. Therefore, to promote this challenge, in this research, an automated and efficient approach for supervised classifications with emphasis on automatic production of training data to prepare a map of flood changes is presented. In total, 85 different studies were performed to obtain flood change maps, using hybrid and non-hybrid automated methods to prepare automated training data for supervised classifications. It is considered that the obtained results had acceptable accuracy. For example, in all approaches studied, ANN classification offered better accuracy, which could be seen from the diagrams provided. Also, due to the lack of significant difference in the accuracy of the maps, obtained from this classification with different indicators and thresholds, it can be said that this classification is also highly robust. The results indicate that ANN classification in combined approaches with Otsu and Adaptive thresholds has the highest accuracy and maximum robustness. In comparison with methods related to execution speed, ML and MD classifications also provided good performance. However, the difference in the accuracy of these classifications compared to the ANN classification is not significant in some cases and will be possible to replace due to higher execution speed. Finally, by analysis of the results, it can be stated that the automated approaches proposed in this research can be replaced by traditional classification methods.

Advances in Remote Sensing technologies provide various information regarding object detection problems. This information makes the interpretation of optical remote sensing images easier. Especial kinds of these interpretations relate to Object Detection approaches that most researches in this field are carried out using Neural Networks and Deep Learning techniques; Design of the network is an important process that affects detection accuracy. Recent researches in the deep learning field and convolutional neural networks show that deeper networks can achieve better accuracy. However, in previous researches, sometimes too deep networks are the reason for other problems such as increasing the number of trainable parameters, vanishing gradients, unused extracted features, etc. These problems decrease the accuracy of the network in recognition of objects. This issue has been mentioned in many types of researches in the field of convolutional networks, and they have tried to meet the challenge by examining different topologies or presenting new training methods. In this article, a model was developed and tried to keep extracted features and transfer them to the next layers. The proposed architecture is a combination of several blocks stacked in a row. The blocks receive their input from the previous block and perform the relevant calculations. Each block consists of several cells that have two layers of convolution. To efficiently use all the features of the training images, the filters used in the convolution layers have kernels with sizes of 1×1 and 3×3. The output of the 3×3 layer in the combining stage is integrated with the information of the previous layers. The architecture of each cell in the proposed network keeps all the extracted features from previous layers to be used in subsequent cells. With these connections between layers, the networks can be deeper with fewer effects of vanishing gradient. In addition to solving gradient problem, this architecture decreases the number of trainable parameters and duration of the training phase impressively. The result of this process is an increase in the ability of existing models to distinguish multi-class objectives. For this purpose, first, a collection of 320 training images is proposed and preprocessed. The proposed method is defined as feature extractor of Faster R-CNN model, and it is trained on image collection. To evaluate the proposed method, a part of Beijing International Airport and a part of Imam Khomeini International Airport were selected as the first and second case study areas. The F1-Measure criterion values for both regions are 97.9 and 93.7, respectively. While, ResNet architecture with 101 layers of convolution and 14.4 million more trainable parameters than the proposed architecture has achieved values of 96.7 and 93% for the mentioned criterion. Finally, the results of applying the proposed model were compared with different famous models of the existing network. The experimental results indicated the reliability and efficiency of the proposed method. To improve the proposed architecture in this paper, dilated convolution operators can be used to extract more prominent features. On the other hand, with the aim of development and generalization, the proposed method can be applied in two stages on high resolution remote sensing images; In the first step, the goal is to identify the location of the airport, and in the next step, the planes inside each airport will be identified by the proposed method.

Natural hazards have always had devastating effects on human life, of which the flood is one of the most serious. Therefore, providing rapid flood identification methods for crisis management is necessary. The purpose of this study is to provide a method with proper accuracy and speed in preparing flood maps. In this study, two time series of Sentinel-1 and Landsat-8 data were used to prepare a flood intensity map by integrating statistical calculation and index extraction. The proposed algorithm is that first the map of permanent water surfaces is automatically prepared by optical images over a period of 5 years. Then, to determine the flood intensity in different regions, statistical calculations are used on the time series of radar images, and finally, using the Normalized difference flood index, which can quickly identify the flood, the final flood map is obtained. The proposed approach has been implemented following the occurrence of the flood of 1398 in two regions of Golestan and Khuzestan, which have different geographical conditions. Assessments performed with the help of ground truth maps and confusion matrices, and in addition, McNemar test was used for more complete analysis. The implementation of the algorithm in the Google-earth-engine environment showed that this method, in addition to having high accuracy, allows the use of hundreds of images without the need for special hardware. The overall accuracy in a period of time in Golestan and Khuzestan was 91.84 and 97.36, which indicates the high generalizability of the algorithm in regions with different extent.

Flood is a natural hazard that causes many deaths each year, and due to the effects of climate change, the number of occurrences is increasing worldwide. Therefore, natural disaster damage assessment, such as floods, provides important information to support decision-making and policy development in the field of natural hazard management and climate change planning. In this regard, in recent years, various methods for classifying remote sensing images have been developed, which always face challenges in differentiating a variety of land uses. Another challenge in flood crisis management is the lack of access to satellite imagery with high temporal resolution while maintaining spatial resolution, which is more pronounced in the presence of cloud cover in the area and occurs during floods. The purpose of this study is to identify flooded areas in Khuzestan province following the flood of 1398, which is based on the integration of optical images of Sentinel 2 and MODIS to produce a time series with relatively good spatial and temporal resolution. In order to classify and prepare maps, a patch-based hierarchical convolutional neural network has been designed, which solves the challenge of extracting deep features due to the relatively weak structure of images with a resolution of more than 10 meters. In addition, the effect of different neighborhood dimensions on the extraction of deep features in all images has been investigated. Finally, the area of damage to urban land cover and various agricultural lands has been estimated consecutively during the flood period.

The data used in this research are two series of different satellite images including Sentinel-2 MSI Level-1C images with a spatial resolution of 10 meters and the product of MODIS daily surface reflectance (MOD09GA) with a spatial resolution of 500 meters. The general process of implementing this research can be summarized in 7 phases. In the first phase, the data is first pre-processed. Then, in the second phase, the image fusion algorithm is implemented to predict the daily surface reflectivity of the images, and if the error and accuracy of the predicted images are appropriate, the time series of the flood period is obtained. In the third phase, Ground truth maps are prepared by the researcher using image interpretation. In the fourth phase, training samples are prepared from these data to perform various classifications such as deep learning approaches and machine learning, and the proposed network is implemented in different input dimensions. It should be noted that the number of training and validation samples in deep learning networks has been very limited and less than half a percent of images to automate and reduce user dependence. In the fifth phase, to perform damage assessment in the agricultural and vegetation regions, the relevant maps are prepared with the best approach tested in the previous phase, and finally, in the sixth phase, accuracy assessments are performed by the confusion matrix and related criteria. In the last phase which is the seventh phase, the area of flood-affected land uses is estimated.

The present study is implemented to improve one of the most important issues in crisis management in the country, namely the assessment of damage caused by the sudden phenomenon of floods. Therefore, presenting a method with appropriate speed compared to existing methods and also increasing the accuracy of final maps due to its challenging has been one of the objectives of this research. First, in order to prepare a suitable time series of optical data with an appropriate spatial and temporal resolution, the ESTARFM fusion algorithm was used. According to the evaluations performed for the two integrated images, this algorithm has high efficiency and accuracy in areas with heterogeneous coverage. Due to the change in environmental conditions between the images, the maximum errors have occurred in water-sensitive bands, but all errors due to their small values in each band indicate the efficiency of the algorithm used. In addition, since the two images are predicted in time series, so the generalizability of the algorithm has been investigated and proven. Furthermore, regarding the classification algorithms for preparing the destruction map, the proposed neural network has a significant difference in accuracy compared to other approaches. In addition, in the study of the extracted classes, in the proposed approach, the built-up areas benefit from a very high identification compared to other algorithms and the appropriateness of other uses, especially the use of water areas, is maintained. According to the studies, the highest rate of flooding in the study area was in the third week of April and after that, the area has been experiencing a decreasing trend. Therefore, the damage was estimated on April 14 and 21. According to the assessments, flooding has decreased from April 14 to April 21 in built-up areas, rainfed and fallow lands, and has increased in wetland and Aquatic cultivation areas.

In this research, the ESTARFM image fusion algorithm, which is known to be suitable for combining images in heterogeneous regions, has been used for April 8 and April 14 images, and the evaluations have been done with the help of scatter plots and least-squares error. The results showed the efficiency of the method in integrating relatively high-resolution Sentinel 2 images and low-resolution MODIS images in the field of flood management. In the field of identifying flooded areas and further due to the poor structure of images with a resolution above 10 meters, the possibility of extracting optimal and deep features is difficult. In the present study, a patch-based convolutional neural network has been designed with a minimum of layers and hyper-parameters, which provides the possibility of training from scratch with the least amount of training samples and without overfitting for images with different environmental conditions. Also, in order to find the optimal state, the dimensions of different inputs in all images have been tested to make a comparison of the effect of different neighborhoods. Thus, patches of sizes 3 to 11 were tested, patches 5 and 7 in the pre-flood image, and patches 9 and 11 in the post-flood images were the best. The results were compared with approaches such as object and pixel-based SVM, LCNN, and DCNN neural networks with dimensions of 3 × 3 and 5 × 5 according to the reference research, and had a significant improvement in accuracy. Time evaluations were performed between all approaches and the lowest time was related to the proposed approach with patch dimensions of 3 × 3 and 5 × 5 and the highest time was related to DCNN network with dimensions of 5 × 5. However, due to the importance of time in crisis management and the need to prepare a high-speed map, the proposed approach has provided an appropriate response. If the time and accuracy are proportionally considered in implementing the research, the designed network in 9 × 9 input dimensions is recommended because in this case, both the accuracy and the time superiority are satisfied.

The training data selection is an important and operative step in the classification and change detection procedure from remote sensing images, which needs to be provided with high sensitivity. These samples are often determined by the human factor, which is a time-consuming process and prone to high error. Old maps can be a valuable source of information for selecting and preparing training samples. If these samples are accurately refined, they can speed up, facilitate and also increase the accuracy of the change detection process. The main innovation of the present paper is the diligence in the sampling process, which has been made imaginable by proposing a model based on the chi-squared statistical test and k-means clustering. This method, while using Chi-square statistical test, tries to select pure training samples, by selecting samples that are close to the centers of internal clusters in each class with multiple k-means clustering that takes into account the internal spectral diversity of classes. In this method, the spectral and the first and second-order of co-occurrence matrix are extracted and used in the support vector machine (SVM) classification process. Furthermore, the feature selection and SVM parameters have been optimized by the genetic algorithm to more improve the classification and change detection accuracy. For implementation, bitemporal satellite images at 2011 and 2015 and land use map of 2009 related to the Shiraz has been employed. Using the proposed method led to update the thematic maps of the study area with an overall accuracy of 98.72% and 94.57%, and a from-to change map. Experimental results showed that the refinement process of the training samples improves the results of the 2011 image classification (increasing the kappa coefficient from 65% to 87% and increasing the overall accuracy from 73% to 91%) as well as the 2015 image (increasing the overall accuracy from 69% to 86.32% and Kappa coefficient has been increased from 59% to 80.48%).

Monitoring the earth and its biosphere is an essential task in any scale to achieve a sustainable development. Therefore, forests, as an invaluable natural resource, have an important role to control the climate changes and the carbon cycle. For this reason, biomass and consequently forest height have been known as the key information for monitoring the forest and its underlying surface. Several studies, it has been shown that Synthetic Aperture RADAR (SAR) imaging systems can greatly help to this purpose. In this framework, a novel technique called Polarimetric SAR Interferometry (PolInSAR) is an appropriate and an available tool for forest height estimation, due to its sensitivity to location and vertical distribution of the forest structural components. Based on this, from a view point, the methods employed in this field can be divided into two categories: a) based on Random Volume over Ground (RVoG) inversion model, and b) based on model-based decomposition techniques of PolInSAR data. In this study, in order to improve the forest height estimation, a novel method based on the combination of two mentioned categories has been proposed. The performance and the efficiency of the proposed method were demonstrated by two datasets simulated from the PolSARProSim software in L and P bands. So that, in L band, it obtained 0.69 m, 1.84 m and 3.38 m improvements for height estimation in Digital Elevation Model (DEM) differencing, coherence amplitude and combined (phase and coherence) methods, respectively. As well, in P band, in it obtained 1.53 m and 2.74 m improvements especially in DEM differencing and coherence amplitude methods, respectively.

Road detection and Extraction is one of the most important issues in photogrammetry, remote sensing and machine vision. A great deal of research has been done in this area based on multispectral images, which are mostly relatively good results. In this paper, a novel automated and hierarchical object-based method for detecting and extracting of roads is proposed. This research is based on the Mean Shift Segmentation Method and the HSI Color Model for Road Detection. Initially, the multispectral images were segmented and then NDVI and NDWI spectral indices were created. In addition, the segmented images were transformed to HSI color space. Then, primary road surfaces were detected by Hue, NDVI, and NDWI spectral indices. In addition, the centerlines of roads were extracted using Voronoi diagram-based technique. After extracting of centerlines of primary roads, dangle errors were removed with emphasis on the topological rules and the lengths of dangles. In order to evaluate the proposed method, the Moonah multi-spectral Image provided by the ISPRS was used. According to the evaluation results, the parameters of completeness, accuracy and quality of the proposed method are, on average, estimated to be 98%, 84% and 84%. In addition, the results of the proposed method were compared with the results of five state of-the-art methods. The results demonstrate the high capability of the proposed method in detecting and extracting roads from satellite multispectral images in urban areas.

Ever-increasing growth and development of urbanization and rapid land-based changes have increased necessity of continuous checking of these changes for urban and environmental planning. Classification of remote sensing high resolution images can be the most effective way to achieve this goal. The classification of these images has always been challenged due to similarities between different classes and differences through one class. Dense classification, also known as semantic segmentation, is also one of the open issues in remote sensing domain. The existence of these kinds of challenges reminds the need for precise methods for classifying images. Deep learning, because of ability to extract deep and powerful features and compatible potential with images, has been known as a good choice in this domain. In this article, in order to cope with the challenges, a convolutional neural networks method based on deep learning is presented for classifying images. The reason for this choice is using deep and comprehensive features by the mentioned method. These features are captured in a supervised manner. In deep learning methods, on the other hand, there is an underlying need for training data and Because of restriction of data in remote sensing, it has been tried to ensure that the number of training samples used in the project is adequate. In this paper, the underlying goal is determination of CNN structure based on deep networks for effective classifying of aerial imagery with high spatial resolution. For this purpose, First, a deep network is designed to extract the deep and optimal features of the aerial image. Architecture and configuration of the deep network are defined in this step. Then, to evaluate the impact of different dimensions of neighborhoods on producing optimal deep features, feature extraction in image patches with different dimensions has been investigated. These patches have been used for train network. After training network with Patches in different sizes, Finally, in order to investigate the classification ability of the deep learning method, in a different approach, a support vector machine has been used for classification based on the deep features produced by the CNN. Comparison of the classification results shows almost same results in the deep learning method in comparison with the conventional support vector machine model, in the same conditions to using deep features. To evaluate the method, aerial data with a spatial resolution of one meter in Des moines area in USA and other data from Royan district in Mazandaran province have been used. Finally, the results of the evaluations show improvement in all three criteria including precision, recall and f1-score in the condition of using larger patches. Also, in general, using of deep learning methods as feature extractor and classifying these deep features by the support vector machine has a bit better evaluation results than feature extraction and classification by CNN.

Monitoring and assessment of the biosphere are two essential tasks at any scale. Based on this, forests play an important role in controlling the climate and the global carbon cycle. For this reason, biomass and consequently forest height are known as the key information for forest monitoring. In the recent decade, several studies have shown that the Synthetic Aperture RADAR (SAR) imaging systems in Compact Polarimetry (CP) mode can overcome the disadvantages of Full Polarimetric (FP) SAR imaging systems and provide a good performance in various remote sensing applications such as monitoring and managing the important natural resources like forests. In this regard, a novel technique named Polarimetric Interferometry SAR (PolInSAR) has been further considered as a powerful tool for forest height estimation. 

In this research, the performance of the Compact PolInSAR (C-PolInSAR) data in Dual Circular Polarization (DCP) mode has been investigated in order to retrieve the forest height. For this reason, the common methods which are used for forest height estimation including Digital Elevation Model (DEM) differential method, coherence amplitude inversion, and phase & coherence inversion methods were applied and implemented on these data. In all of the aforementioned methods, LL+RR and LR polarizations were considered as the selected channels for estimating the volumetric and ground coherences, respectively. Then, the estimated coherences were considered as the input parameters for all of the mentioned methods. 

To evaluate the performance and the efficiency of C-PolInSAR data in DCP mode, the results obtained from these data were compared with those obtained from Full PolInSAR (F-PolInSAR) data. The results obtained in this study in two datasets simulated from PolSARProSim software in both L and P bands showed that the C-PolInSAR data in DCP mode yielded a similar result compared to the F-PolInSAR data for forest height estimation (when the HH+VV polarization is adopted as the ground backscattering), because, in this case the LL+RR and the LR polarizations are equal to the HV and the HH+VV polarizations, respectively, particularly, the C-PolInSAR data in DCP mode yielded 0.78 m and 0.55 m improvements for forest height estimation in L and P bands, respectively. In addition, all of the employed methods provided better and closer results compared to the real forest height (i.e. 18 m) in L band compared to P band, because the electromagnetic (EM) waves have a more penetration into the canopy in L band compared to P band. Thus, the attenuation of these waves is low and consequently the height estimation is more accurate. Without considering the used bands, the DEM method provided the lowest precision compared to other methods, because the HV (or LL+RR) phase center can lie anywhere between half the tree height and top of the canopy. The exact location of this phase depends on two vegetation parameters which are the wave mean attenuation and the vertical canopy structure variations. In this case, the trees have very thin canopies, and consequently, the attenuation is small, but the phase center is high due to the structure. In other words, when the canopy extends over the entire forest height, then the phase center can be at half the true height for low density (low attenuation), through to the top of the canopy for dense vegetation (high attenuation). This ambiguity is inherent in single baseline methods, and in order to overcome this, model-based correction methods need to be employed. It was also observed that the coherence amplitude method is among the weak algorithms due to ignoring the phase and its sensitivity to the attenuation and structural variations but it can be used as a backup solution when other approaches fail. Finally, the phase and the coherence inversion method had better results than two aforementioned methods for the forest height estimation. In this method, selecting the factor ‘’ is very important and it should be selected in a way to be strong towards the attenuation changes. In this study,  0.4 was adopted to maintain the height error variations.

As the final result, the C-PolInSAR data can be an efficient strategy due to its performance, when the full polarimetric imaging systems are either limited or not available. Moreover, utilizing these data in long wavelengths (e.g. P band) is more appropriate due to the effect of the Faraday rotation on the transmitted polarization.

With the development of digital sensors, an increasing number of high-resolution images are available. Interpretation of these images is not possible manually, which necessitates seeking for practical, fast and automatic solutions to solve the environmental and location-based management problems. The land cover classification using high-resolution imagery is a difficult process because of the complexity of the landscapes, and the spatial and spectral resolution of images. The predecessor studies of land cover classification were done using statistical methods such as maximum likelihood classification. However, the newer studies apply the artificial intelligence techniques such as artificial neural networks and support vector machines as a substitute for classification applications. A major problem with using these models is that the user cannot easily understand the final rules. In this paper, a hybrid algorithm is proposed in order to obtain the needed data by the knowledge-based system from the input data set. The proposed algorithm is designed to get better training data and improvement of the learning system in semi-urban areas by classes covered by different material and colors.
Classification of the remote sensing images refers to separation of the similar spectral sets and division of the units with the same spectral behavior. In remote sensing, due to the large size of data, the processing procedure is costly. On the other hand, to achieve better results, it has been recommended to use various features in the training procedure, which will consequently incrementally increase the volume of processing. Accordingly, the use of object-based process can increase velocity and homogeneity of the final interpreted image by reducing the computational base units. In the field of feature generation, a hybrid feature including both region-based (by kernels) and object-based (by segments) strategy, has been employed in this study. In order to produce the training data, needless of determining that by the user, utilize the capacity of integration of the multi-source data by KBS based system. For this purpose, the ontology concept that applying by the knowledge-based rules was used. Then to improve the obtained training samples and compensate its defects in the expression of the target class properties, the correction step is done. In the other words, the automatic Knowledge-based method was performed to apply the ontological relationships in order to train and control the object-based support vector machine system.
In order to evaluate the proposed method, a set of test images from two different geographic regions were used for validation of the method. In each geographic region, it was attempted to select different test images (various scene features). On this basis, in the first group, three test images belong to a region in northern Iran and Bandar Anzali city, and the second group includes two images in Germany. The GSD (Ground Sampling Distance) of all the 5 test images is equal to 9 cm. Finally, the proposed method has achieved an average accuracy of 82/80% in all test images. The evaluation of the results showed that the proposed technique could be desirable as an automatic and semi-supervised method for interpreting high-resolution images of the semi-urban regions.

 

In the past few decades, urban environments have expanded much larger than before. One of the most important problems in most metropolises and even small cities is the management of the transportation system. An advanced monitoring system of urban vehicles allows for overcoming the traffic problems. With the development of unmanned aerial vehicles (UAVs), continuous and accurate monitoring of urban environments has been provided for the users. In this research, an efficient method is presented that detects the vehicle in the UAV images. The proposed method is effective in terms of computational speed and accuracy.  

   The foundation of the proposed method is based on the characteristics of the local features in the UAV images. The presented method consists of two main stages of training classification model and detecting vehicles. In the first part, local features are extracted and described by the SIFT algorithm. The SIFT algorithm is one of the most powerful algorithms for extracting and describing local features that are used in various photogrammetry and machine vision applications. This algorithm is robust to geometric and radiometric changes of the images. Due to the high dimensions of extracted features from all the training samples, the BOVW (Bag of Visual Word) model has been applied. This model is used to reduce the dimensions of the features and display the images. Simple and efficient computing is one of the significant features of the BOVW model. At this stage, after producing a library of features, the SVM classification model is trained. In the detection part of the algorithm, the images are entered into the algorithm and the local features are extracted in all images by the SIFT algorithm. The BOVW model is often used to display an image patch. In most researches, this model is implemented by applying a search window to the whole image. This type of methods has a higher confidence level in detection, but it is a very time consuming process and increases the volume of the computations. For this purpose, the approach of points clustering and their representation by the BOVW model is proposed. In this method, features that are within a certain range are considered as a cluster. Euclidean distance is used in image space for clustering. Then, the clusters produced by the BOVW model are displayed. Then, a feature vector is constructed for each cluster. The trained SVM is applied to each of the production vectors and each cluster is classified as a vehicle and non-vehicle. If the cluster is detected as a car, the position of the center of that cluster is marked on the image.  

   The proposed method was implemented on 8 images with a number of different car targets. Also, considering the use of the search window approach in many researches, our results were compared with the results obtained by other researchers. The results show that the calculation time of the proposed method is 82 seconds, while the search window method takes 2496 seconds to run. In order to verify the accuracy of the algorithm, two criteria were used. The first criterion is the “Producer's accuracy”, which represents the proportion of correct detections of the vehicle to the entire vehicles existing in the images. This criterion is 75.79% for the proposed method. The second criterion is the “User's accuracy”. This criterion is obtained by dividing the correctly detected samples into the sum of the correctly and incorrectly detected samples. The User's accuracy criterion has been reported to be 59.50%.  

The value of the Producer's accuracy criterion is greater for the search window method which has led to a more accurate detection of vehicles compared to our method.  This is due to the small moving steps of the search window in the entire image. However, the search window method has increased the amount of the time spent on the calculations. The User's accuracy criterion shows that the proposed method has less incorrect detections. The results indicate that our method has a higher degree of reliability. The average of these two criteria indicates the superiority of the proposed method in terms of the accuracy of the calculations. On the other hand, the proposed algorithm has a great advantage in terms of computational speed compared to the search window method.

Close-range photogrammetry has been used in many applications in recent decades in various fields such as industry, cultural heritage, medicine and civil engineering. As an important tool for displacement measurement and deformation monitoring, close-range photogrammetry has generally been employed in industrial plants, quality control and accidents. Although close-range photogrammetric application in displacement measurement of large-scale structures was not introduced as much as its other application, but successful utilizations in this field prove it to be a potentially effective procedure in this area. In order to get familiar with these applications, this paper reviews the research conducted on the use of close-range photogrammetry for measuring displacement in large-scale structures. There are several unique advantages of having close-range photogrammetry employed in this field including the unnecessary of being in direct contact to the structure during the observation, the rapid acquisition of observations, and the immediate access to the results by automating the procedures. Moreover, the instant recording of observations in moving features and possibility of creating an archive from these observations to be used in future processing if required are more examples of these beneficial characteristics. In addition, the high flexibility and adaptability of this method to the project conditions made it possible to achieve high accuracies while, each photogrammetric target practically acts as an instrumentation sensor at a lower cost. Accordingly, based on the reviewed literature, on average, this method was able to reduce up to 60% of the costs and time compared to the other conventional methods as an effective and efficient tool.

According to the cities expansion, updating urban maps for urban planning is important and its effectiveness is depend on the information extraction / change detection accuracy. Information extraction methods are divided into two groups, including Pixel-Based (PB) and Object-Based (OB). OB analysis has overcome the limitations of PB analysis (producing salt-pepper results and features with holes). In the information extraction by SVM classification in complex urban areas, using various features was suggested to improve accuracy result. Also, in SVM, it is necessary to determine the values of the model parameters. In most of the previous OB research, the two important steps were determined by trial and error or based on an expert knowledge. The necessity of selecting independent features and determining the optimal values of SVM parameters, with the aim of minimizing the maximum user interaction, have resulted in proposing a novel method with a relatively high automation level based on SVM simultaneously optimization with meta-heuristic algorithms for large scale updating maps in  high spatial resolution and elevation data. Semi-automatic selection of train/test samples also has increased the automation level of the updating process. Therefore, according to the effect of information extraction on the updating results, the proposed method is trying to improve this step results. The results of the proposed method had no salt-pepper results in comparison with PB analysis. Also, the time processing of the proposed method in optimization and classification steps had been decreased. Finally, the results of change detection map obtained from the proposed method led to 9% and 5% improvement in comparison with other methods in changed class.

Object detection is one of the fundamental issues in image interpretation process, especially from remote-sensing imagery. One of the most effective and efficient methods in this field is the use of deep learning algorithm for feature extraction and interpretation. An object is a collection of unique patterns that differ with own adjacent properties. This difference usually occurs in one or more features simultaneously, which can be indicated by the difference in shape, color, and gray values. In this regard, the use of deep learning as an efficient branch of machine learning can be useful in generating high-level concepts through learning in different layers. In this research, a database based on the environmental and geographical conditions from some Iranian airports was created. Additionally, an optimal learner model was developed with a convolutional neural network. For this purpose, in the raw data processing section, besides using the transfer learning method, some vectors were extracted to classify the objects and delivered to an SVM model. The output values were compared with the values obtained from the test image for each object, and they were analyzed in a repeatable process for structural matching. Precision of 98.21% and F1-Measure of 99.1% was achieved, for identification of the target objects

Detecting agricultural land borders from high-resolution imagery is a major challenge due to the complex earth environment, which has become a significant issue in researches. Lack of location data to determine the area and location of agricultural land, to control and plan about cultivation and food, as well as to estimate the statistical parameters of agriculture for optimal management in agriculture, etc. shows the necessity of extract the boundaries of these lands. For this purpose, active contour models are a powerful tools for detecting and tracking image objects. In order to develop these models, the combination of two nonparametric region based Chan-Vese active contour and edge based Geodesic Active Contour models was considered to simultaneously apply edge and region information in a hybrid math structure to extract the agricultural land boundary. This combination is able to separate just one type of object from background; Therefore to increase the efficiency of the hybrid model and its development for color images, the hybrid model was combined with the Multiphase Active Contour model. After evaluating the proposed model with the reference produced map, the resulting values for the evaluation criteria were 90.3 for the overall accuracy index, 86.1 for the Kappa coefficient, 85.5 for the product's accuracy index, 87.7 for the user's accuracy index and 81.9 for the quality index, which were obtained from four images with different characteristics. After comparing the values obtained for the desired indices, the proposed model showed an average improvement of 9% for the CAPA index, 6% for the overall accuracy index, 12% for the product's accuracy index, 14% for the user's accuracy index and 12% for the quality index, increasing the accuracy rather than the basic model.

Land cover classification in remote sensing imagery is one of the most widely used spatial information extraction methods, which can facilitate generating object imagery classes of the ground surface in order to automate and accelerate meeting the basic needs of management, organization, and exploitation of the environment. Due to the similar behavior of pixels, remote-sensing image classification using merely the spectral and textural information would lead to inefficiency in the classification. In fact, in classification process, objects are commonly identified using spectral properties of image pixels. If the spatial and conceptual properties are also considered, it causes to a better distinction between image classes and closes the machine process to human interpretation and adds to the system's performance. The present research is mainly focused on the use of interactive segmentation and interpretation processes with respect to the geometry of the image classes. The accuracy of the results have improved by introducing the knowledge-based rules to control and regulate the interactive process, taking into account the geometric properties of target classes. To evaluate the efficiency of the proposed method, the results were evaluated and compared with some of the other methods on IRS satellite images in an urban area. The results showed that geometric and conceptual features as a complementary information source, improve classification results in the urban area with heterogeneous spectral effects. Overall, the proposed hybrid technique improved overall accuracy and Kappa coefficient by %8 and 11.18, respectively.

Building detection from remote sensing images has significant importance in updating the maps, urban monitoring, and a wide range of other applications. The high spatial resolution images are an important data source for geospatial information extraction. These images provide extraordinary facilities for feature extraction like buildings and spatial analysis in urban areas. However this task suffers from some problems due to spectral complexity in the image scene. Since high resolution images contain a lot of details about the scene such as; non-homogeneity of the roof of the buildings, sloping and flatness, it can create various spectral properties among other issues. Also, due to the use of similar materials, some buildings can’t be completely separated from the streets and parking. To overcome these issues, the use of neighborhood and height information is essential. Accordingly, the major part of this research is the use of spatial features of adjacent pixels in a multispectral image and elevation data to increase the accuracy of building classification. In this regard, on the one hand, with the expansion of the feature space in a quasi-deep method, the goal was to train the classification algorithm in higher level and more comprehensive information. But all the features available to distinguish between buildings and non-buildings are not useful. On the other hand, due to the large amount of input data and the increased computing time and memory required, to reduce the processing cost, it is necessary to perform the feature selection operation. Despite many efforts that have been made over the past decades to develop the methods for automatic building detection from these images, high-performance methods are still unavailable because of the uncertainties like optimal feature selection. Therefore, in this study, with a view to improving the automatic detection of the building from remote sensing data, a new hybrid approach is proposed to select the optimal features of the large dataset in a reasonable time. The proposed method of this research firstly extracts high-level features for optimal building detection by using quasi-deep texture structures. Then, it selects the optimal features based on the integration of the developed AdaBoost algorithm (Confidence Based AdaBoost) with the optimized support vector machines by particle swarm optimization (CB-SVMpso), and performs the binary classification of the building and background. The experiments were performed on the standard data set of Vaihingen in Germany and then the results of the proposed method were compared with efficient methods of machine learning. Also, a comparison was made between the quasi-deep feature sets with the traditional method of GLCM textures. In experiments, in order to purify the final results, no pre-processing and post-processing steps have been interfered. The experimental results showed that on average, the highest overall accuracy and kappa coefficient obtained from the proposed method were 93.25% and 83.06%, respectively, and in comparison to conventional methods, the accuracy of kappa coefficient has increased by 7.27%, as well as the computational time reduction by half, indicating the reliability and efficiency of the proposed method in detecting the majority of buildings.

In recent decades, close-range photogrammetry has been successfully used in various fields including industry, cultural heritage, medical, health and civil engineering. This method as a tool for measuring objects in industrial applications, has many advantages such as capability of real-time measurement, ease of observation, the ability to achieve high accuracy and also capability of creating an archiving observations for future processing.

3D models of ancient sites are produced and utilized for different purposes such as research, restoration and renovation of valuable ancient objects, creation of virtual museums and documentation of ancient sites. Nowadays, Geomatics techniques, as the most efficient methods for geometrical measurements, analysis and interpretations concerning issues in cultural heritage, are applied to produce geometric and thematic information. Buildings are susceptible to change and damage through the passage of time due to natural agents and disasters such as rain, wind, earthquake, flood, or damages imposed by human beings. The characteristics of these changes in some buildings with ancient value bear special importance. The first step to create 3D models, provide the information about ancient monuments and record them with documents is having accurate maps of their present condition to be able to add other information like type of construction materials. Special techniques should be employed to provide maps with high accuracy, in addition to other characteristics such as spending the least expense and time for continuous map production. The process of changes are recognized by comparing maps from different time spans based on which due decisions can be made. To provide these maps many different techniques have been used since past such as traditional surveying (using the usual total stations), photogrammetry (especially close-range photogrammetry) and laser scanners. In comparison to other techniques, photogrammetry has unique characteristics in documentation of ancient sites. No need to contact with the feature, the possibility to obtain the information of texture and color and the compliance of these characteristics with the 3D output data, high flexibility of this method to access the desired accuracy in measurements and its potential of access to accuracy at micrometer level as well as capability of low expense observations and archiving images, are parameters that have given rise to the more usage of techniques of photogrammetry in the modelling of ancient sites. Yet, the usual techniques of photogrammetry sometimes have limitations, for example, in rare cases of inaccessible features. As a result, the requirement to obtain accurate information from features, especially in dangerous and remote areas, and also, the necessity to economize expense and time have led to the usage of UAV-based photogrammetry.
UAV-based photogrammetry is a combination of aerial photogrammetry and close-range photogrammetry in which there is a sensor that can be a metric or non-metric camera or any other data collection tool. The images are acquired from low height. Access to imaging stations with appropriate angle toward all parts of a feature and low height of flight, result images with high spatial resolution, which consequently, bring about more accurate and precise 3D information from earth. Different categorizations have been presented for UAVs based on different criteria and applications. To mention some of these criteria we can refer to the criterion of flexibility, fixed or rotating blades or wings in UAVs and their source of energy. Based on the categorizations of platforms regarding this research, which is ancient sites, it is obvious at first glance that UAVs with fixed wings, fixed or semi-flexible parachutes and wingless are practically of no use due to low flexibility in flying and imaging , and also limited space of flight. Therefore, low expense, high flexibility and appropriate time of flight have contributed to the suitability of quadrotors as the best option among all systems with rotating blades in this research.
Low expense for production, no need to airports and long runways and better maneuverability are some particular parameters and characteristics of the functionality of UAVs. There are factors that limit the function of UAVs, such as instability while flying due to light weight, limited source of supply, limitation to carry bigger and more accurate measuring tools and requiring longer time for imaging, processing and calculations. Fortunately, all these limitations can be modified to some extent by an appropriate network design. In spite of all aforementioned capabilities of UAV systems, no specific standards have been designed to utilize them. Therefore, it is obviously necessary to investigate the feasibility of the usage of these systems, and to design appropriate networks to locate them in proper points to obtain images for photogrammetry.
Thus, the need for high accuracy in UAV-based photogrammetry for documentation and restoration of ancient sites necessitates more concern for the network geometry to achieve the desired accuracy. This article presents appropriate method for optimum locations of UAV for imaging. The proposed method for the optimal locations of UAV is based on the ellipsoid fitted on object, principles and constraints of photogrammetry network design and finally by exploring hidden areas. The results from images taken from a cultural heritage site showed that the number of images was reduced almost 4 times by applying network design principles. Consequently, the speed of 3D modelling would be increased almost eight times by applying the proposed method.