基于三成分分解的极化SAR图像海岸带目标无监督检测OA

Polarimetric Synthetic Aperture Radar(SAR)image coastal zone target detection is an important research direction in the field of coastal remote sensing applications.SAR images are affected by coherent speckle noise and scattering interference,making it difficult for traditional geometric feature detection methods to extract precise contour features;while deep learning methods rely on massive annotated data,and the same model is difficult to apply to data from different sensors and significantly different multi-type targets.To address this,an unsupervised detection method based on polarimetric scattering features is proposed.This method first uses a modified three-component decomposition method to extract the double-bounce scattering power and volume scattering power of ground objects;then,through generalized likelihood ratio threshold segmentation,the volume scattering power is used for sea-land segmentation to extract the coastal narrow-band region,and the double-bounce scattering power is used for coastal strong scattering region segmentation to extract the target region of interest;finally,by calculating the distance between the target and the coastline,the classification and differentiation of different coastal zone targets are achieved.Experiments were conducted on five single-look fully polarimetric images from Gaofen-3,and the results show that the proposed method achieves a detection rate higher than 80%and a false alarm rate lower than 10%,enabling unified detection of different types and sizes of targets.The coastal zone target detection method proposed in this paper relies on the unique polarimetric scattering characteristics of targets to achieve detection with higher accuracy.