SAR (specific absorption rate) image segmentation method based on hierarchy visual semantics

Abstract

The invention discloses an SAR (specific absorption rate) image segmentation method based on hierarchy visual semantics, which mainly solves the problems of difficulty in maintaining of area uniformity and inaccurate positioning of boundaries and line targets in the prior art. The SAR image segmentation method comprises the following steps of 1, according to a sketch map of SAR images, extracting an area map of the SAR images, and building the hierarchy visual semantics; 2, according to the area map, segmenting the SAR images into focusing, structural and homogenous areas; 3, segmenting the focusing area; 4, segmenting the structural area; 5, designing a positioning line object of the visual semantic rule in the structural area; 6, segmenting the homogenous area; 7, combining the segmenting results of the focusing, structural and homogenous areas and the line object, and obtaining the final segmenting result. The method has the advantages that the uniformity of the areas in the SAR images is maintained, the positioning accuracy of the boundaries and the line objects is improved, the good segmenting effect of the SAR images is realized, and the method can be used for the subsequent target classification, identifying and tracking.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to a SAR image segmentation method, and can be used for subsequent target identification and tracking. Background technique [0002] Synthetic aperture radar (SAR) has imaging characteristics such as all-weather, all-time, penetrability, long-distance, and high resolution, and has been widely used in various fields of national defense construction, national economy, and social development. SAR image segmentation is a basic and key problem in SAR image interpretation. It extracts the object of interest from the image background and provides the basis for subsequent processing such as classification, tracking, and recognition. [0003] The existing SAR image segmentation methods are mainly divided into two categories: feature-based methods and model-based methods. [0004] The feature-based method mainly emphasizes the features of the image, such as gray-level co-occurrence matrix...

AI Technical Summary

Problems solved by technology

Although this method can improve regional consistency to a certain extent, it still cannot obtain semantically consistent regions.

The multinomial hidden model method considers the gray feature and texture feature of the image as the likelihood model at the same time, and uses the multinomial logistic regression model with aggregation characteristics as the prior model of the image, and then uses the classification expectation maximization method to obtain the segmentation result of the image. The pre-defined context window cannot meet the multi-scale requirements of SAR image content, and it is easy to cause inaccurate boundary positioning and loss of line targets during the segmentation process

Images