基于双目视觉的AGV障碍物检测与避障

Obstacle detection and obstacle avoidance of AGV based on binocular vision

针对智能车间中物料位置和运动状态的不确定性,提出一种基于双目视觉的自动导引车(AGV)障碍物检测与避障方法。在障碍物检测过程中,首先采用基于深度检测的障碍物判定算法判断障碍物是否存在;然后利用帧差法提出一种障碍物信息的获取算法,分别得出静、动态障碍物的方位与速度信息。为进一步提高障碍物检测的可靠性,分别对不同状态的障碍物进行检测实验,结果表明,该算法误差较小、结果可靠。针对不同障碍物设计避障策略,对比PID和模糊PID的方法控制AGV运动,通过分析AGV运动过程中的角度和距离偏差验证了避障策略的有效性。最后,开发了一个简单高效的上位机系统,进一步验证了所提算法的可行性和有效性。

According to the uncertainty of materials＇ position and motion state in intelligent workshop, a method of obstacle detection and avoidance for Automatic Guided Vehicle （AGV） based on binocular vision was proposed. During obstacle detection process, the obstacle detection algorithm based on depth detection was used to judge the existence of obstacles. An algorithm for obtaining the information of obstacles was proposed by frame subtraction to get the information of obstacle position and velocity in the static and dynamic state. To improve the reliability of ob stacle detection, the static and dynamic obstacles were detected by experiments. The result showed that the algo rithm was reliable and had little error. Aiming at the different obstacles, the obstacle avoidance strategy was pro posed. PID and fuzzy P1D method were used to control AGV motion, and the effectiveness of obstacle avoidance strategy was proved by analyzing the angle and distance deviation of motion. A simple and efficient PC software was designed to verify the feasibility and effectiveness of the proposed algorithm.