Research On SAR Image Change Detection Technology Based On Image Segmentation

Synthetic aperture radar (SAR) is an active microwave sensor with the characteristics of alltime, all-weather conditions and strong penetration. Therefore, the SAR images compare withother images in remote sensing classification has a greater advantage and more suitable for thehigh mobility of the investigation and detection. This paper summarizes the SAR image changedetection in the current research status, based on the existing methods false alarm rate that weretoo high, and uses the SAR image segmentation of many research results to make a deep study ofthe SAR image change detection technology which based on image segmentation. The mainwork accomplished and achievements are as follows:1、 It makes a detail introduction of the development present situation and the keytechnologies involved in change detection and segmentation technology of SAR image,including SAR image filtering, radiometric calibration, geometric correction, matching, andsummarizes the advantages and disadvantages and accurate evaluation method of a variety ofchange detection methods.2、K-means clustering algorithm and the theory of segmentation based on MRF areintroduced into the paper, and the K-means segmentation algorithm is improved, then combinedthe improved K-means segmentation method with the MRF segmentation method to form anadaptive K-means MRF segmentation method. Through experiments and analysis of thealgorithm are verified the effectiveness of the proposed algorithm.3、A change detection ideas of joint segmentation is proposed and the Kullback LeiblerDivergence is introduced as a measure of the change detection. The experiment uses adaptiveK-means MRF segmentation method to make a segmentation of the SAR image. TheExperimental results show that this change detection based on segmentation can effectivelyreduce the false alarm rate.4、Another change detection method based on image segmentation is proposed, and theMahalanobis distance is introduced as a measure of the change detection, then using amulti-scale segmentation method to make a segmentation of the SAR image. The experimentalresults show that this method is also adaptive to the change detection in SAR images.