Segmentation Method of Polarized SAR Image Based on Eigenvalue Metric Spectral Clustering

Abstract

The invention discloses a polarization SAR image segmentation method based on eigenvalue metric spectrum clustering. It mainly solves the problem that many parameters are difficult to adjust adaptively in the existing polarization SAR image segmentation process. The implementation process is: (1) decompose the eigenvalues ​​of the polarimetric SAR image to form a feature sample set x; (2) calculate the mean value corresponding to the three eigenvalues ​​of its 8 neighborhoods for each pixel, and construct the average feature sample set (3) Use the Mahalanobis distance to construct a similarity matrix for the feature sample set x and the average feature sample set respectively, and obtain a mixed similarity matrix w′ based on the two similarity matrices; (4) For the mixed similarity matrix w′, pass The spectral clustering algorithm obtains the cluster label C1; (5) Repeat steps (3)-(4), and use the MCLA algorithm to integrate the obtained class label set to obtain the final segmentation result. The invention has the advantages of strong adaptability, low complexity and more detailed and accurate segmentation results, and can be used for target detection and target recognition of polarimetric SAR images.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, relates to polarization synthetic aperture radar image segmentation, and can be used for image target detection, image target segmentation and recognition. Background technique [0002] With the development of radar technology, polarimetric SAR has become the development trend of SAR, and polarimetric SAR can obtain richer target information. The understanding and interpretation of polarimetric SAR images involves signal processing, pattern recognition and many other disciplines. As one of the basic problems of polarimetric SAR image processing, polarimetric SAR image segmentation lays the foundation for later target recognition of polarimetric SAR images. [0003] Existing polarization SAR image segmentation methods can be divided into supervised and unsupervised categories. [0004] Supervised methods include: Kong et al. proposed to segment polarimetric SAR images usi...

AI Technical Summary

Problems solved by technology

For this reason, Ersahin and Anfinsen et al. used spectral clustering to segment polarimetric SAR images, and defined the similarity matrix through the Wishart distance of the polarimetric coherence matrix. Although this method can automatically complete clustering and does not require threshold determination, However, due to the special distribution of the polarization coherence matrix, the construction method of the similarity matrix is ​​limited. At the same time, the Gaussian kernel parameters need to be manually set accurately according to experience, and the calculation cost is high.

Images