Bp neural network image segmentation method and device based on adaptive genetic algorithm

Abstract

The invention relates to a BP neural network image segmentation method and device based on an adaptive genetic algorithm. The method comprises the following steps: 1) analyzing the image to be segmented to generate a training sample of the neural network; 2) setting neural network parameters and population parameters, and performing chromosome Encoding; 3) Input training samples to train the network, use a new adaptive genetic algorithm to optimize its weight and threshold, and introduce adjustment coefficients into adaptive crossover and mutation operations; 4) Input the image to be segmented, and use the trained neural network to classify it , to achieve image segmentation. The device includes a training sample generation module, a neural network structure determination module, a network training module and an image segmentation module. The present invention introduces an adjustment coefficient related to the evolutionary algebra, which solves the problem of stagnation of individual evolution in the early stage of population evolution, and also avoids the local convergence problem caused by similar individual fitness, thereby obtaining a neural network that can maximize the representation of image features more quickly , to achieve more accurate image segmentation.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a BP neural network image segmentation method and device based on an adaptive genetic algorithm. Background technique [0002] Image segmentation is the technology and process of accurately extracting specific and interesting objects in the image. As an important part of image analysis and computer vision systems, it determines the quality of digital image analysis and the quality of visual information processing results. , is a key step in image recognition and analysis. At present, image segmentation methods mainly include segmentation methods based on deformation models, segmentation methods based on regions, and segmentation methods based on statistics. [0003] When an image-designed segmentation algorithm is applied to a specific image, the effect is often unsatisfactory. Since various segmentation algorithms have certain pertinence and applicability, when it is n...

AI Technical Summary

Problems solved by technology

[0007] The present invention provides a BP neural network image segmentation method and device based on an adaptive genetic algorithm, aiming to solve the problem that the existing image segmentation method is prone to individual evolutionary stagnation, local convergence and low segmentation accuracy of the neural network method in the early stage of population evolution , the problem of slow training speed

Images