Traffic image segmentation method based on improved multi-object harmony search algorithm

Abstract

The invention discloses a traffic image segmentation method based on an improved multi-object harmony search algorithm, and mainly solves problems of a conventional traffic image segmentation technology that the evaluation index is single, the anti-noise performance is poor, the segmentation accuracy is not high, and the number of image segmentation types is required by be determined manually. The implementation steps mainly comprise reading a gray scale image and counting a gray scale histogram; initializing a harmony memory bank; generating a group of new explanations, wherein the size the new explanations is equal to the size the harmony memory bank; updating the harmony memory bank; determining whether a termination condition is met or not; selecting the optimal cluster center from the harmony memory bank according to a PBM evaluation index; and classifying pixels of the image according to the optimal cluster center to obtain a segmentation result. Compared with the conventional traffic image segmentation technology, the method provided by the invention is diversified in evaluation index, high in segmentation accuracy and high in anti-noise performance, can automatically determine the number of segmentation types of the image, and can be applied to the traffic image segmentation.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a traffic image segmentation method based on an improved multi-object harmony search algorithm. Background technique [0002] The quality of image segmentation results directly affects the quality of subsequent image processing. Therefore, image segmentation is one of the key and difficult points in the field of image understanding. This difficult task has brought great challenges to researchers. The purpose of image segmentation is to divide an image into several non-overlapping areas with specific meanings, the same area has similar characteristics, and different areas have great differences. [0003] The existing image segmentation methods can be mainly divided into edge-based segmentation methods, region-based segmentation methods, and cluster-based segmentation methods. The essence of most clustering methods is the optimization problem of the objective function. D...

AI Technical Summary

Problems solved by technology

At present, most image segmentation methods based on multi-objective optimal clustering only consider two segmentation criteria, and each type of image has its own characteristics, so there are shortcomings such as small application range, unsatisfactory segmentation results, and low segmentation accuracy.

[0005] In the existing traffic image segmentation methods, whether it is the multi-threshold segmentation method or the clustering segmentation method, it is necessary to artificially set the number of segmentation targets, while the vision-based intelligent vehicle navigation technology in intelligent transportation requires the computer to automatically divide the road, Objects such as vehicles and obstacles are separated. Obviously, traditional image segmentation methods cannot meet this specific requirement.

Images