Accessible Region Segmentation for Vision-based Autonomous Mobile Robots
This paper proposes an region segmentation method to improve the accuracy of accessible region used for vision-based navigation autonomous mobile robots, under the ground environment with illegible borderlines, and the method is based on Adjacent Pixel-Value Differencing (APVD). First of all, a process so-called Inverse Perspective Mapping (IPAf) is taken to map images obtained by robots vision system from imaging coordinate system to real world coordinate system, APVD-based edge detection algorithm is proposed later, which is used to draw out the edge information of ground environment. Then, after fusing the edge features and pixels gray information, borderline criterion function is established to partition the accessible region from current image, using method of region-growing, in which a growing seed is selected dynamically. According to field tests carried out in an autonomous mobile robot, the edge detection based on APVD, turns out to be robust, which can provide accurate accessible region segmentation results for following processes of autonomous mobile robots under dynamic environment
machine vision image segmentation edge detection region growing autonomous robots
Dong Liang Zhongke Shi
School of Automation Northwestern Polytechnical University, NPU Xian, China
国际会议
重庆
英文
459-463
2011-01-21（万方平台首次上网日期，不代表论文的发表时间）