基于视觉的工业机器人异常动作检测方法研究

Vision-Based Fault Action Detection of Industrial Robot

工业机器人的突发故障引发的安全问题时有发生。传统的基于数据分析的故障诊断方法存在传感器数据易受干扰,机器人通讯协议不统一,监测系统嵌入在执行系统内部相互影响等问题。提出一种基于机器视觉的工业机器人故障动作检测方法。对工业机器人作业视频进行实时分析,采用图像分割技术分离工业机器人本体并采用图像哈希技术生成工业机器人姿态编码,结合序列模式分析技术检测工业机器人异常动作并进行预警。不依赖于工业机器人通讯协议,以非接触式的方式对工业机器人进行实时监控,具有易于部署和成本低的特点。基于自主构建的工业机器人仿真视频数据集进行了实验研究,结果表明提出的方法可准确识别工业机器人异常动作,精确率和召回率均为100%。

Security problems caused by sudden failures of industrial robots frequently occur.Traditional fault detection methods of industrial robots are mainly based on data analysis techniques which have many limitations:1)the data received from sensors may be interfered by the external environment;2)industrial robots do not have a unified communication protocol,which increases the cost of collecting sensor data;3)the embedded monitoring component may affect the running of industrial robots.This paper proposes a vision-based detection method of fault actions for industrial robots.The method analyzes the working video of industrial robots and detects fault actions caused by robot failures.Firstly,the image segmentation technique was used to divide industrial robots from the background.Then the posture of robots was encoded into hash codes.Finally,fault actions of robots were detected from the sequence of hash codes of robot postures.The method does not depend on the communication protocols of industrial robots,and can monitor industrial robots in a non-contact way,which is easy to deploy and has a low cost for use.Based on the self-constructed video data set for industrial robots,experiments were conducted.Experimental results show that the proposed method can accurately identify the fault actions of industrial robots.Both the accuracy rate and the recall rate are 100%.