| Polarimetric Synthetic Aperture Radar(PolSAR),as an active remote sensing sensor,can work all day and all day,and has certain penetration ability to some ground objects,so it has been widely used in the fields of ground object classification,urban planning,geological exploration,ocean monitoring,target detection and so on.This subject has been established based on the fact that aircraft target detection,as one of these important application fields is of great significance to military and national defense,aiming to achieve aircraft target detection in PolSAR images through polarimetric scattering information.This work takes the aircraft target on the ground in the PolSAR image as the research object,and uses the region segmentation method and the polarimetric scattering characteristics of the aircraft target to realize the detection of aircraft target detection.The specific research contents are as follows:(1)In order to solve the problems of the low resolution of PolSAR image,the irregular geometric shape of aircraft target,this thesis presents an aircraft target detection method based on the combination of maximum entropy region segmentation and multi-feature judgment based on the polarimetric scattering characteristics of aircraft target.Firstly,the regions such as runway and apron are extracted by using the maximum entropy region segmentation method,and the suspected aircraft targets are extracted by using the dissimilation scattering power based on the regions.Then,combining Yamaguchi decomposition and coherence matrix,the detection feature is constructed.Finally,the thresholds of the suspected aircraft target are determined by the detection feature and background power(the average power around target).Through these two thresholds,the final detection result is obtained.(2)In order to solve the problems that the threshold needs to be set manually in the detection stage of the above algorithm,and the adaptability is not high,a second algorithm is proposed: aircraft target detection method combining region segmentation with Support Vector Machine(SVM)classification.Firstly,the PauliRGB image is directly segmented by the method of combining two-dimensional maximum entropy and two-dimensional minimum cross entropy.Then,the runway,apron and other areas are extracted by using other information such as power.Under the condition of low resolution,the covariance matrix of PolSAR data generally obeys Wishart distribution with constant texture coefficient,so this algorithm uses constant false alarm rate(CFAR)to extract suspected targets.Finally,the SVM classifier trained online is used to judge the suspected aircraft target and get the real aircraft targets.Both algorithms firstly segment the image to extract the runway and apron area,then extract the suspected aircraft target in this area,and finally make multi-feature judgment to obtain detection results.The first algorithm does not need offline training,and directly performs threshold judgment on suspected targets.The second algorithm is complex,which needs to make a training sample set and train the SVM classifier,but it does not need to set the threshold manually in the final feature decision.The experimental results show that the two algorithms can effectively detect aircraft targets through the measured data collected by American UAVSAR and AIRSAR systems. |
