The Detection Of Airplane Based On Multi-feature Fusion In PolSAR Image

PolSAR is an active imaging radar,and it operates at a lower frequency than Sband microwave band,so it can avoid the effects of clouds,rain,dust,fog,etc.Because it is active working,so it doesn't need to rely on the sun and images day and night.Unlike single-polarized SAR,the PolSAR data is stored in the form of a scattering matrix,which is often associated with the scattering matrix of different models,so that the polarization information of the feature can be extracted more abundantly.With its advantages,PolSAR has been widely used in various fields,such as feature classification,terrain mapping and mapping,impact evaluation,and natural disaster prevention.The thesis focuses on the PolSAR image data,and mainly studied the application of PolSAR in the classification of objects and target detection.Aiming at the unsupervised classification of PolSAR images,this thesis mainly analyzed the problem of insufficient classification accuracy in the unsupervised classification method based on H/ ? decomposition.The main reason for this problem is that the structural information inside the target is not fully exploited.Therefore,this thesis proposed an unsupervised PolSAR image classification algorithm which combines H/ ? decomposition and scattering characteristics.The algorithm first uses the plane to roughly divide the image into eight categories,and then uses the scattering similarity feature to analyze the internal structural information of the target according to the internal scattering information of the target,then the results of the rough classification are further divided.Finally,the Wishart classifier is used to iteratively optimize the initial class,and the optimized classification result is obtained.The algorithm has obvious improvement in classification details.Aiming at the research content of aircraft target detection in PolSAR image,this thesis analyzed the structural characteristics of the aircraft target and its scattering characteristics in the PolSAR image,and combined the a priori knowledge of the parking area of the aircraft target as the runway or the apron.The aircraft target detection algorithm of PolSAR image combined with regional filtering and multi-level feature decision is proposed.The algorithm firstly uses the threshold segmentation method to extract the strong scattering power target of interest,and then uses the region screening method to separate the suspected aircraft from the strong scattering target.The target is finally based on the characteristics of the aircraft target(the characteristics of the aircraft target in the PolSAR image are: the scattering power of the aircraft target relative to the area in which it is located is large,the area in which the aircraft target is parked is homogeneous,and the polarization of the aircraft target is crossed.The entropy is relatively large.The false target is eliminated to obtain the final test result.In order to test the effectiveness of the proposed method,experiments were performed with the PolSAR data acquired by AIRSAR and UVASAR system from NASA Laboratories in the United States.The experimental results verify the effectiveness of the classification and aircraft target detection methods proposed in this thesis.