Ray completion region graph and characteristic learning-based SAR (synthetic aperture radar) image segmentation method

Abstract

The invention discloses a ray completion region graph and characteristic learning-based SAR (synthetic aperture radar) image segmentation method. The problem that a large homogenous region such as a farmland cannot be accurately segmented by a conventional segmentation method is mainly solved. An implementation process comprises the following steps: 1, obtaining a sketch map of an input SAR image by using an initial sketch model; 2, completing sketch lines by virtue of a ray clustering method, and acquiring a region graph; 3, segmenting the SAR image into a concentrated region, a homogenous region and a structural region of a pixel space by utilizing the region graph; 4, performing characteristic learning and clustering on the concentrated region and the homogenous region by utilizing a bag-of-words model respectively, performing watershed segmentation and sketch line-guided superpixel combination on the structural region, and combining combined superpixels to the homogenous region to finally obtain an SAR image segmentation result by utilizing gray characteristics. Compared with the prior art, the method has the advantages that the segmentation result is high in region consistency, and boundaries and line targets are positioned more accurately.

Description

Technical field [0001] The invention belongs to the field of image processing technology, relates to a method for segmentation of SAR images, and can be used for target detection or recognition. Background technique [0002] SAR image segmentation is one of the basic and key technologies in SAR image processing and interpretation. The segmentation results have an important impact on the subsequent processing of images. Due to the unique imaging principle of the SAR system, the SAR image contains a large amount of coherent speckle noise, complex targets and mixed shadows, all of which increase the difficulty of SAR image segmentation. At the same time, due to the essential difference between SAR image and optical image, many mature optical image segmentation methods cannot be applied to SAR image segmentation. Current SAR image segmentation methods can be roughly divided into two categories: gray-level segmentation methods and texture-based segmentation methods. However, in actu...

AI Technical Summary

Problems solved by technology

However, this method has two shortcomings: 1. The sketch lines used in this method are disconnected, and the non-aggregated area formed is not complete; 2. The wireless segment area is not accurately divided

Images