Improving the quality of SAR radar image detection and identification using neural network to increase the security factor of military tactics

In military science, the use of telecommunication technologies is very important because of receiving enemy information from a distance, and the artificial aperture radar system is an imaging radar that has a high detection and separation ability. According to the nature of formation of artificial aperture radar images, the presence of speckle noise is the most important factor in destroying the quality of these images and making a decision error.

Therefore, it is very important to have a pre-processing stage in order to reveal and identify the data and reduce the spot noise. The main goal is to provide a powerful algorithm based on artificial intelligence to improve the detection of SAR radar ground targets with object detection algorithms that are obtained through airplanes or satellites in order to monitor ground targets.

Analysis: 

In the pre-processing stage, after reducing the effect of speckle noise on SAR radar images, the proposed model is investigated to improve the detection of SAR radar ground targets with the help of Lee filter.

by using YOLO and RCNN algorithms, the RCNN algorithm has a better performance in detecting denoised MSTAR images with an average accuracy of 99.84% compared to YOLO with an average accuracy of 90.424%, but in detecting noisy MSTAR images, the YOLO method With an average accuracy of 80.875%, compared to the RCNN method with an average accuracy of 61.49%, it has a better performance for identifying and improving images and increasing the reliability of the system.