SAR image segmentation method based on feature vector integration spectral clustering

Abstract

The invention discloses an SAR image segmentation method based on feature vector integration spectral clustering and mainly provides a solution to the problems of low accuracy and poor scale parameter sensitivity appeared when an existing spectral clustering method is used in SAR image segmentation. The realization process is as follows: (1) carrying out three-layer wavelet transformation on an input SAR image and extracting wavelet features; (2) setting scale parameters sigma to be in the range from 0.1 to 1 and setting the number of integration features to m; (3) randomly selecting m scale parameters from the scale parameters sigma and using approximation to calculate m similarity matrixes W1, W2, ..., Wm; (4) calculating Laplasse matrixes L1, L2,..., Lm according to the m similarity matrixes W1, W2,..., Wm and respectively carrying out non-negative matrix factorization to obtain feature vector matrixes V1, V2,..., Vm; (5) integrating the feature vectors V1, V2,..., Vm to obtain a new feature vector matrix U; and (6) standardizing the feature vector matrix U to obtain Y, carrying out K_means clustering on the Y and outputting the final segmentation result of the SAR image. The SAR image segmentation method based on feature vector integration spectral clustering has the advantages of high-accuracy segmentation result and insensitive scale parameter and can be used in target detection and target segmentation and recognition of SAR images.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to SAR image segmentation, and can be used for SAR image target detection, target segmentation and recognition. Background technique [0002] Synthetic Aperture Radar (SAR) imaging technology actively emits and receives electromagnetic waves, and forms images according to the reflection and scattering characteristics of objects. Because it has all-weather, all-weather, high-resolution detection and reconnaissance capabilities that can effectively identify camouflage and penetrate cover objects, the interpretation of SAR images has attracted more and more attention and attention in the fields of national defense and civilian use. As one of the key links in SAR image interpretation, SAR image segmentation is becoming more and more important and urgent. [0003] Among many segmentation methods, the clustering-based segmentation method is to divide the similar regions in the SAR ima...

AI Technical Summary

Problems solved by technology

However, when the approximate fast spectral clustering algorithm is used for SAR image segmentation, it is very sensitive to scale parameters and does not consider the non-negative characteristics of image pixels themselves, resulting in inaccurate edges of SAR image segmentation results

Images