基于深度学习与光流算法的工业机器人异常姿态检测方法  

Abnormal Attitude Detection for Industrial Robots Based on Deep Learning and Optical Flow Algorithm

作　　者：白渊铭 魏艳龙 Bai Yuanming;Wei Yanlong(Network and Information Center,Shanxi Energy Institute,Jinzhong Shanxi 030600,China;School of Computer Science and Technology,Taiyuan Normal University,Jinzhong Shanxi 030600,China)

机构地区：[1]山西能源学院网络与信息中心,山西晋中030600 [2]太原师范学院计算机科学与技术学院,山西晋中030600

出　　处：《山西电子技术》2025年第2期35-37,48,共4页Shanxi Electronic Technology

基　　金：山西省自然科学基金项目(20210302124541);山西省自然科学基金项目(202103021224337);山西省青年科学基金项目(202203021212185)。

摘　　要：针对工业机器人作业时可能出现的姿态异常故障行为,提出了一种Mask R-CNN算法与光流算法相结合的检测方法。通过摄像头以非接触的方式采集机械运动数据,采用Mask R-CNN算法,自适应地找到机械关节点,并建立机器人的机械骨骼框架,再应用光流算法对机械运动进行姿态估计。实验结果表明,本文提出的方法可精确捕捉机器人的机械臂异常行为,算法模型泛化能力强,姿态异常诊断的准确率可达90%以上,该方法有效地解决了工业机器人运行姿态异常检测这一问题。A detection method combining the Mask R-CNN algorithm and the optical flow algorithm is proposed for the faulty behaviors such as abnormal attitude that may occur during the operation of industrial robots.The mechanical motion data is captured in a non-contact way by a camera,and the Mask R-CNN algorithm is used to adaptively find the mechanical joint points and establish the robot’s mechanical skeleton framework,and then the optical flow algorithm is applied to estimate the attitude of the mechanical motion.Experimental results show that the method proposed in this paper can accurately capture the abnormal behavior of the robot’s mechanical arm,the algorithm model generalization ability is strong,and the accuracy of the posture abnormality diagnosis can reach more than 90%,and the method effectively solves the problem of detecting abnormalities in the operating posture of industrial robots.

关 键 词：工业机器人 Mask R-CNN 光流估计 机械臂 异常检测 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]