A Graph-Based Raster Map Segmentation Method

Abstract

The invention discloses a graph-based grid map segmentation method. In view of the fact that the corridor segmentation in the existing segmentation map algorithm is inconsistent with the real map, the invention makes full use of the obstacle information in the map constructed in the real environment, from the grid Semantic information obtained from the grid map mainly includes two parts: free region clustering and graph-based region merging; free region clustering is specifically ISODATA combined with the ray-throwing algorithm, so the clustering algorithm can ensure that the clustering results will not The clustering algorithm can effectively eliminate the noise in the map; the graph-based region merging is specifically to effectively use the obstacle information in the map, build and optimize the graph according to the connectivity of the classes, and then merge the classes to obtain a complete corridors and rooms.

Description

technical field [0001] The invention belongs to the field of computer vision, in particular to a map segmentation technology. Background technique [0002] It is self-evident that humans rely on the visual system in the process of understanding the world. Similarly, machines regard computer vision as the basis for recognizing the world. As one of the most important artificial intelligence technologies, computer vision is widely used in robot visual navigation, face recognition, image recognition, driverless and other technologies. In the field of robotics, maps are one of the main tools to help robots describe their environment and precisely locate them. The most common type of map used in robotic autonomous localization and navigation techniques is the grid map. The free areas in the grid map represent the areas that the robot can pass through. Segmenting the free area of ​​the grid map to obtain the area with semantic information is helpful for the robot to establish to...

AI Technical Summary

Problems solved by technology

[0010] (2) The segmentation results of long and straight corridors by existing methods are inconsistent with the segmentation results of real maps marked manually

[0011] In the above method, there is a situation that the segmentation of the corridor in the segmentation map is inconsistent with the real map, and there is no solution in the prior art that can solve this technical problem well

Images