Road Image Clustering Method and Road Recognition Method Based on Color Density Features

Abstract

The invention relates to the technical field of quadruped robots, in particular to a road image clustering method and a road recognition method based on color density features. The image clustering method takes color as the basic feature, defines and extracts the color density of the image as the basis for image clustering and segmentation, and then through multiple clusters, reduces or eliminates the road recognition due to illumination changes, shadows, and road color inconsistencies. influence, enabling clustering of road regions even in the presence of environmental changes. The road recognition method provided by the present invention is based on the road image clustering method based on color density features. By re-clustering the mis-segmented interference areas, the recognition of unstructured roads is completed, and it solves problems caused by irregular block shadows or The problem of misclassification caused by irregular color blocks that randomly appear on the ground that are inconsistent with the color of the road; the technical solution provided by the invention provides an effective solution for the quadruped robot to recognize the unstructured road environment in the wild, and has important value.

Description

technical field [0001] The invention relates to the technical field of quadruped robots, in particular to a road image clustering method and a road recognition method based on color density features. Background technique [0002] In a complex field environment, in order to realize the autonomous walking of quadruped robots on unstructured rough roads, it is necessary to accurately identify the roads to assist humans in completing tasks such as field material transportation, patrolling, and exploration. Although there are differences in overall color characteristics between the irregular road in the wild and the surrounding environment, the color consistency of the road is not ideal due to the influence of changes in lighting conditions, reflections of uncertain objects, and road weeds. In addition, the unstructured road is not an ideal straight line, and the irregularity of the road shape also brings difficulties to the accurate identification of the road. [0003] In unstr...

AI Technical Summary

Problems solved by technology

However, due to the influence of factors such as changing light, road shadows, and inconsistency of road surface color on roads in the wild environment, it is difficult for roads to have consistent color features. Therefore, the adaptability and accuracy of the above clustering methods for road recognition are not ideal.

[0004] If the road contour feature is directly detected for recognition, when the road boundary is obvious and has good continuity, the road boundary line can be extracted by Hough straight line transformation, multiple spline curve fitting, etc., so as to realize the recognition of the road. The method has a good effect on structured roads. However, the shape and boundary of unstructured roads in the wild are uncertain and irregular in most cases, and the straight line outlines of disturbances in the environment (such as tree trunks) are very easy to identify road boundaries. Causes noise and misclassification, so does not perform well

Images