Robot vision image segmentation method based on multi-scale fractal dimension and neural network

Abstract

The invention discloses a robot vision image segmentation method based on multi-scale fractal dimension and a neural network, and belongs to the technical field of robot vision. The robot vision image segmentation method adopts a method combing image wavelet transformation and fractal dimension to perform image segmentation, uses a neural network method for distinguishing image areas where extracted characteristics belong and can achieve rapid image segmentation. The robot vision image segmentation method utilizes a character that fractal dimension of surface texture of different objects in nature is different, and can effectively segment three different areas of the sky, trees and roads. In addition, segmentation speed can be improved when image area segmentation is conducted on the basis of existing neural network training results, and the robot vision image segmentation method is suitable for a robot vision system in a movement process.

Description

Technical field [0001] The invention belongs to the field of robot vision, and particularly relates to an image segmentation method based on multi-scale fractal dimensions and neural networks. Background technique [0002] Image segmentation is widely used in various fields related to image processing, including remote sensing images, medical images, traffic images, and so on. Because image segmentation technology occupies an important position in the image processing process, it has been highly valued by researchers for many years, and thousands of segmentation algorithms have been proposed so far. Currently, the widely used segmentation algorithms include threshold-based segmentation methods, edge-based segmentation methods, region-based segmentation methods, and so on. Among them, for the threshold-based segmentation method, the key problem is to determine an optimal threshold. However, when the grayscale difference of the image is not obvious or the grayscale range of differ...

AI Technical Summary

Problems solved by technology

If the detection accuracy is improved, the false edges generated by noise will lead to unreasonable contours; if the noise immunity is improved, contour missed detection and position deviation will occur

The essence of the region-based segmentation method is to connect pixels with similar properties to form the final segmented region, which often results in over-segmentation of the image.

Images