Narrowband Chan-Vese model underwater multi-object segmentation method for adaptive step initialization

Abstract

The invention discloses a narrowband Chan-Vese model underwater multi-object segmentation method for adaptive step initialization. The method comprises the following steps: 1, after smoothing denoising is carried out on a solar image, initial segmentation is completed according to a block mode k-means clustering algorithm to initially judge the position of an underground target; 2, a step area is determined, an adaptive step initialization zero level set function is determined; 3, a Chan-Vese model narrowband level set is used for solar image segmentation to complete two-class and three-class segmentation on an underwater multi-object area; and 4, Chan-Vese model-based objective and quantitative analysis is carried out on a segmentation result. A Chan-Vese model narrowband level set is adopted for solar image segmentation, local optimization can be completed, global search in the existing level set method can be avoided, noise influences in the segmentation result can be reduced to the minimal, and the segmentation precision and the speed can be further improved.

Description

technical field [0001] The invention belongs to the technical field of underwater sonar image processing, and in particular relates to an underwater multi-target segmentation method of a narrowband Chan-Vese model with adaptive ladder initialization, which can realize underwater multi-target segmentation of sonar images. Background technique [0002] The research on the underwater multi-target recognition technology of sonar images is of great significance and value in the civilian field. The underwater multi-target recognition technology of sonar images will be one of the main technologies to be studied in future shipbuilding and ocean engineering. But before underwater target recognition in sonar images, target segmentation and feature extraction must be performed on sonar images, and underwater target segmentation and feature extraction in sonar images are key steps in the process of underwater target recognition. Only the correct segmentation of underwater targets can ma...

AI Technical Summary

Problems solved by technology

However, because the value of the level set function is too small, it is affected by the smooth term of the energy function, resulting in wrong segmentation results

When the value is selected larger, because the change value of each iteration is constant, when the change value is less than the selected value, the sonar image segmentation will hardly change after several evolution iterations, and then it will be slower, sonar image The segmentation is accurate, but the number of iterations increases and the time spent increases

Images