Publication: Optimized robot mapping and obstacle avoidance using stereo vision
Loading...
Date
2020-07-01
Authors
Teo, Jen Son
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Robot mapping is the construction of map for an unknown environment with no priori map by using computer vision as it needs the information of surrounding such as distance from obstacles and features of obstacles. It is often used in the areas that are dangerous toward human. Although obstacle avoidance can be done with typical range finding sensors, but those sensors only provide distance information which is not enough for robot mapping of the autonomous robot. In this project, stereo vision is being chosen and applied to AGV because it can perform range finding and extract surrounding features. Two USB cameras are being calibrated for stereo vision. A region in the rectified image pair is selected as ROI for optimization and the ROI of rectified image pair is being processed to obtain disparity map which helps to estimate distance from obstacles. Through thresholding the disparity map, obstacle detection with detection from 30cm to 100cm is achieved. Then, the depth of obstacles is estimated with error less than 2%. For obstacle avoidance, the mobile robot with the system implemented able to avoid obstacles within the test field. Although the deviation of robot’s path may cause inaccuracy, a 2D occupancy map is able to be constructed using the depth information obtained through stereo vision. The processing time of the system is optimized by the ROI implementation in which it reduces the time taken for image processing and the final processing time per frame after optimization is 0.2143s.