جستجوی مقالات مرتبط با کلیدواژه « image texture » در نشریات گروه « عمران »

Close-range photogrammetry aims to produce accurate 3D geometric models of objects using images taken from the subject. Nowadays, the creation of realistic 3D models and their visualisation is a common practice that is becoming more popular every day. On the other hand, choosing the right modelling software for photogrammetry has always been a challenge and a topic of discussion among experts and researchers. Therefore, it is essential to examine and evaluate the models produced by different software tools. Due to the widespread use of Agisoft software among engineers and researchers in this field, this study aimed to perform image processing and modelling using two versions of this software, namely Photoscan and Metashape. In previous research, the criterion for optimising the image mesh has been based on improving the accuracy of the modelling. In order to assess and evaluate the 3D models produced by the two versions of the software, we defined different scenarios for the design of the image mesh. We compared the 3D models generated for each scenario with a mathematical reference model. We also examined the complete modelling in the software under different conditions using two different textures, as the texture of the image directly affects the quality of the point cloud. It is important to analyse the role of the image texture together with the geometry of the image mesh. Therefore, we evaluated the image texture as a radiometric index and investigated how these two factors affect the quality of the point cloud. As a result, we determined the optimal number of images with appropriate texture required to produce an accurate and high-quality 3D model.

close-range photogrammetry, we capture a series of images of an object using a specific image network. These images are then used with the Structure from Motion (SfM) method to generate point clouds and 3D models. The concept behind SfM is inspired by how our eyes perceive objects. This approach offers a quick, automated, and cost-effective way to obtain 3D data. It involves creating 3D coordinate models by processing a sequence of overlapping images of the object. Finally, the resulting 3D models are compared with a reference point cloud using the Cloud Compare point cloud processing software.

 The results of using images with simple texture show that in Photoscan software, increasing the number of images not only leads to noise in the point cloud, but also reduces the similarity of the generated model to the cube. According to the results, the best 3D model with a high similarity to the cube is associated with the fourth scenario (45 images) with an error of 0.01 millimetres. In the case of the Metashape software, the best model is associated with the third scenario (90 images) with an error of 0.05 millimetres. On the other hand, in cases where images with complex textures were used, the best point cloud is related to the fourth scenario (45 images) with an error of 0.02 millimetres in Photoscan software and to the third scenario (90 images) with an error of 0.04 millimetres in Metashape software. In general, the use of objects with complex textures leads to a better match and therefore to denser point clouds due to the presence of complex and non-uniform gradients in the images.

The results show that the optimal number of images and the presence of a complex image texture have a significant impact on the improvement of the quality of the 3D point cloud of the object. Despite the increased processing time, the quality of the 3D model does not increase significantly with a large number of images; it only leads to denser point clouds due to increased noise in the point cloud.

Evaluation of pavement performance plays a major role in pavement management systems for determination of optimum strategy in repair and maintenance of the road. One of the most prominent assets in evaluation of the pavement is identification and survey of pavement surface distresses. In the past two decades, extensive studies have been carried out in order to develop automatic methods for pavement distress evaluation. Most of these methods are based on computer vision and image processing techniques. Of the most important components of machine vision systems is the feature extraction process. Textural features present more detailed information about the image regions characteristics compared to other features such as color and geometrical (shape) properties. In the present study, after acquisition of six different groups of asphalt pavement distress images under controlled condition, in order to analyze and describe their texture, second order statistics based on grey level co-occurrence matrix has been employed. In order to generate the images co-occurrence matrices, four distinct directions and three different distance (offset) parameters have been utilized. Based on the results of the classification of distress images acquired by Mahalanobis minimum distance classifier, it can be concluded that statistical indices extracted from grey level co-occurrence matrix having distance parameter equal to one, have superior discrimination performance in camparison to other selected distance values. The classification accuracy rates of asphalt pavement distress images based on grey level co-occurrence matrix with one, two and three distance parameters values are 80%, 75% and 60%, respectively.

Inspection of the pavement distresses is one of the most prominent phases of pavement management process in regard with determining optimum pavement maintenance strategies. Over the past few decades, a considerable number of efforts have been carried out on developing automatic methods for objectively distress detection all of which rely on machine vision and image processing techniques. One of the most important assets comprising machine vision systems is the feature extraction process. In the past few years, multi-resolutional analysis approaches, namely wavelet transforms has provided a great tool for fast and accurate image texture representation. In the present study, after acquisition of six different types of asphalt pavement distresses under controlled condition, in order to identify and categorize them, four 2-D multi-resolution transforms including Haar discrete wavelet, Daubechies3 discrete wavelet, Coiflet1 discrete wavelet and dual-tree complex wavelet were utilized. After decomposition of the distress images by applying the aforementioned transforms, first-order statistical indices based on histogram and second-order statistics based on gray level co-occurrence matrix were employed, in order to describe the wavelet frequency sub-bands texture. The distress classification results based on minimum Mahalanobis distance classifier indicate that extracting second-order statistics from the sub-bands of the dual-tree complex wavelet and Haar discrete wavelet transforms, yielding classification accuracy of 99% and 95% respectively, outperform other feature extraction algorithms in distress recognition. Furthermore, statistical indices acquired from gray level co-occurrence matrix with average classification rate of 87%, obtained superior performance in distress images discrimination compared to histogram statistics.

Evaluation of pavement performance is one of the most prominent assets in choosing the beneficial strategy for pavement management operations. In the past two decades, a considerable number of investigations have been carried out on developing automatic methods for distress rocognition all of which rely on the machine vision and image processing techniques. In the past few years multi-resolutional analysis methods, namely wavelet transform has provided a great tool for fast and accurate auto-detection of distresses. In the present study, a method has been proposed utilizing the wavelet transform method which can analyze the texture surface of pavement considering the longitudinal, transverse and diagonal textural structures as the key elements. In this paper, after performing the discrete wavelet transform and decomposing the image into frequency sub-bands using 4 different wavelet families, properties of sub-bands texture has been acquired (based on grey level co-occurrence matrix) and compared to the results acquired based on image texture analysis in spatial domain. Finally, the minimal Mahalanobis distance method was applied in order to categorize the acquired images into seven classes including alligator cracking, without distress, longitudinal cracking, transverse cracking, bleeding, patching and raveling. Based on the results of validation and evaluation of the classifiction performance it was observed that the distress image classification using image texture analysis in the transformation domain leads to the more accurate results in comparison to spatial domain. The mean accuracy of distress image classification in transformation domain is 67% while the accuracy rate in classification of distress data based on extraction of texture features in spatial domain is 49/76%. In case of transformation domain, although Daubechies 2 filter has a better sensitivity rate in discrimination of bleeding distress, in general, the Haar filter outperformed other utilized wavelets in recognition and classification of asphalt pavement surface distresses with 95% accuracy.

Nowadays the generation of a three-dimensional (3D) model is mainly achieved using non-contact systems based on light waves, in particular using active or passive sensors. We can currently distinguish four alternative methods for object and scene modeling: (i) image-based rendering, which does not include the generation of a geometric 3D model but, it might be considered a good technique for the generation of virtual views; (ii) image-based modeling (e.g. photogrammetry), the widely used method for geometric surfaces of architectural objects, precise terrain and city modeling as well as Cultural Heritage documentation; (iii) range-based modeling (e.g. laser scanning), which is becoming a very common approach for the scientific community but also for non-expert users such as Cultural Heritage professionals; (iv) combination of image- and range-based modeling, as they both have advantages and disadvantages and their integration can allow the generation of complete and detailed 3D models efficiently and quickly. Moreover, 3D models of outdoors objects by means of terrestrial laser scanner (TLS) point clouds has been one of the most important and reliable technical methods. However, TLS point clouds have two main drawbacks. They dont have sufficient information about the object texture, and sometimes a complete perspective of the facades cannot be obtained due to irregular circumstances. Advances in both TLS hardware and photogrammetric solutions combined are creating increasingly more precise and rich 3D colored models. The combining of close-range photogrammetry and TLS data have been shown to be effective and accurate techniques for the 3-dimensional documentation of complex buildings and sites, which is considered, by many institutions and organizations. In this paper the 3D model generation of two historical building based on the combination of two different surveying techniques has been presented. The main goal of the work was to merge in a unique result data derived from close-range photogrammetry and from TLS in such a way that the final model could be seamlessly explored. Close range photogrammetry and Laser scan, complementary, and combine their data in most applications is increasingly common. Laser scanner while collecting 3D data, are able to produce images of texture mapping applications, but for some reason these images cannot be used. Texture images separately with good quality, at the suitable time and the appropriate conditions are provided using the principles of photogrammetry. This paper explores the projection matrix method is a comprehensive reconstruction of ancient buildings. This research is a case study on two series of real data from our ancient monuments, Abarkooh dome in Yazd province and stone bridge of Lorestan, which can be used to provide documentation verifying in national or international scale and three-dimensional modeling to be used for different purposes. In this study along with projection matrix method, other mathematical models were examined using multiple texture images for produce 3D models with true textures. The use of multiple texture images improved quality, geometry and accuracy of the results in texture mapping. The results indicated that the errors and distortions of RGB colors, is depends to the accuracy of the measurement in the space of the image and cloud points and the mathematical models. Also, check average of errors in the two transverse and longitudinal coordinates, indicating the projection matrix method was suitable in compared to other methods. In general, the projection matrix method is computationally more easily and geometrically more accurate than other models which is tested in this study. The values of errors in the projection matrix method are 0.15 and 0.33 (pixel) in the stone bridge of Lorestan and -4.24E-13 and 7.67E-14 in the Abarkooh dome in pixel units, respectively. In contrast, the errors of iterative direct linear transformation method are -2.38 and -5.79 E-10 for the first point clouds and 5.40E-5 and 3.64E-5 in pixel units for the second point clouds, respectively. Therefore, the main advantage of the projection matrix method is that it requires no initial values and it operated as matrix multiplication for producing three-dimensional models with realistic textures that is recommended as a stable method.

Most of common classification algorithms in remote sensing are based on spectral characteristics of pixels. These approaches ignore the spatial information of data, such as texture, in classification process. Simultaneous usage of texture and spectral information is a new trend in remote sensing image classification which has been considered in this study. We have evaluated the efficiency of gray-level co-occurrence matrix texture features (GLCM) extracted from panchromatic (PAN) image of ALI detector in improving the classification accuracy of Hyperion hyperspectral (HS) data in urban regions of Tehran. Classification is performed using a support vector machine (SVM) classifier with a Gaussian kernel. In our experiments, we have considered three different cases: a) classifying original Hyperion data, b) classifying Hyperion data pansharpened by color normalized transform (CNT), and finally, c) simultaneous use of GLCM texture features of panchromatic data and the spectral features of pansharpened data in classification process.In case b and c, for pansharpening HS data we have performed the following steps:1.Registering HS data with PAN data using ◦nineteen ground control points,◦polynomial warping of second order, and◦Nearest neighbor interpolation.1.Selecting a subset of HS bands which spectrally overlap with PAN image.1.Fusing the spatial information of PAN image into the HS subset bands, obtained in step 2, using CNT method.Moreover, as GLCM features, we have extracted 8 texture features from GLCM matrices: mean, variance, homogeneity, contrast, dissimilarity, entropy, angular second moment, and correlation. In order to assess the influence of the size of GLCM extraction window on the quality of texture features, we have considered various window sizes: 3×3, 5×5, 7×7, and 9×9.At the first phase of our experiments, we compared classification results obtained using the original HS data with the results obtained from the pansharpened HS subset – see Table 1. The results showed an increase of about 15% in the average classification accuracy when using the pansharpened data. At the second phase, we combined each of the texture features individually with the pansharpened HS subset. The results are given in Table 2. As the table suggests, regardless of the type of texture feature and the size of the GLCM extraction window, the combinations improve the overall classification accuracy (OA) of data. However, texture features show better quality when extracted from GLCM matrices obtained using 9×9 neighborhood windows. In addition, we observe that regardless of the size of GLCM window, dissimilarity feature delivers the best results.To summarize, by using the pansharpened HS subset instead of the original HS data, we achieved about 15% gain in the classification accuracy. Moreover, combining dissimilarity texture features –extracted from GLCM matrices obtained using 9×9 neighborhood windows– with the pansharpened HS subset improved classification results. In our experiments, we achieved about 5% increase in OA compared to that of using pansharpened HS subset alone.