The Research Of Autonomous Rescue Robot Based On Machine Vision

Machine vision has the advantages of non-contact measurement, working stably in a long time and wide spectral response range, which is a research hotspot in the field of robot application in recent years and also the foundation of robot to realize the autonomous operation.Based on the rules of autonomous ability challenge in RoboCup Rescue, this paper analyzes the requirements of autonomous rescue robot, which also designs and implements the function of target detection, target tracking and autonomous obstacle avoidance based on machine vision of autonomous rescue robot.Target detection:according to the rules of the competition, autonomous rescue robots need to proactively identify qrcodes, hazard labels and simulated victims in the process of searching a maze environment. To solve these visual tasks, this paper detects qrcode using edge detection and hough transform, and extracts qrcodes's information using ZXing library; this paper detects hazard labels through the SURF feature matching, which uses RANSAC algorithm to optimize the matching accuracy; this paper detects simulated victims according to the sensor fusion results of infrared thermal imager and ASUS Xtion Pro live.Target tracking:on the basis of camera calibration, this paper uses particle filter algorithm and TLD algorithm to implement target tracking and selects the appropriate algorithm for this paper on the basis of their experimental results. According to the established coordinate system of target tracking, the relative position of robot and target are calculated and target is tracked by controlling robot.Autonomous obstacle avoidance:In order to search victims in a maze environment, the robot must have the function of autonomous obstacle avoidance. On the one hand, this paper realize autonomous obstacle avoidance using VFH+ algorithm with environmental perception information based on laser range finder; On the other hand, this paper extract more obstacles information through the stereo vision information. Finally, autonomous obstacle avoidance module with multi information fusion is realized by using the information of the barrier after fusion.The software system of this paper is based on the ROS platform of Ubuntu system, and all the experiments is tested on the Turtlebot robot. Experimental results show that the system can detect the target quickly in a multi-obstacles environment and track target in real time, which is a foundation for autonomous rescue robot to explore the unknown environment.