Integration of RADAR and Multi-Spectral Remotely sensed Images for Target Detection (Case Study: Detection of Individual Structures)

Identifying the feature as a specific target, whether static or moving, is one of the main challenges in image processing and can be used in fields such as monitoring, determining the position of strategic features, co-referencing images, producing 3D models and etc. Due to the fact that satellite images can cover a large area with less time and cost, therefore, target detection on the satellite images is more important and useful. In the research, the aim is using the ability of radar images to identify geometric structures and the ability of multispectral images to identify the nature of features, to reduce the problems in this field. Some of the problems include: the presence of futures with a weak geometric pattern, the geometric or spectral similarity of features with each other, the presence of multiple candidates and the occurrence of problems to achieve a single target and the fading of the target due to high noise. In the algorithm designed and implemented in the research, step by step by analyzing the information extracted from the images, the target identification process is improved and the error is reduced until finally, by integrating the results of radar and spectral image processing, the number of candidates reaches to an acceptable level. The implementation of the proposed algorithm showed that in the integration process, the selection of target and specification of study area have a significant impact on the results. Also, there are many challenges in the target detection process. Among these challenges is the selection of the best, most useful image data and features to identify the target and also determine the optimal thresholds. Also, in the process of target detection with radar images, it was found that both the geometric features and the height information extracted from the images have valuable contents that should be used together to achieve the desired result.