Linear object segmentation method and segmentation system

Abstract

The application discloses a linear object segmentation method and segmentation system, wherein the linear object segmentation method comprises: through a Hough transform method, mapping an edge image into a polar coordinate system so as to searching for a main peak value and a secondary peak value in the collinear cumulative graph under the polar coordinate system; and according to the main peak value and the secondary peak value, determining whether the position relation between a first straight line and a second straight line meets a segmentation requirement so as to determine whether an initial segmentation threshold value can accurately and effectively divide linear objects in a sonar image, if the determination result is no and the initial segmentation threshold value is not equal to a cut-off segmentation threshold value, correcting the initial segmentation threshold value, and then returning to the step of generating a binary image according to the initial segmentation threshold value and sonar image until the position relation between a first straight line and a second straight line meets a segmentation requirement or the initial segmentation threshold value is equal to a cut-off segmentation threshold value so as to achieve the purpose of searching for an ideal initial segmentation threshold value to perform linear object segmentation on the sonar image.

Description

technical field

[0001] The present application relates to the technical field of image segmentation, and more specifically, to a linear object segmentation method and a segmentation system. Background technique

[0002] Sonar imaging is less affected by water turbidity and light intensity, so it is widely used in underwater target detection, environmental perception, underwater target tracking and other fields. Choosing an appropriate image segmentation technique can improve the probability of detection, tracking and recognition.

[0003] The methods for image segmentation in the prior art mainly include big law, local big law and attribute histogram method. These three methods are all based on the selection of segmentation threshold. The main idea is to obtain grayscale histogram, and then find the segmentation threshold corresponding to the valley between the two peaks from the grayscale histogram, and use the segmentation threshold as the ideal segmentation threshold to ...

Images