摘要
针对传统工业中机械手的人机交互方式不够直观的问题,利用卷积神经网络(CNN)设计一种基于手势动作的机器视觉型控制系统。采用OpenCV构建手势数据集,以CNN中的AlexNet结构为基础,改进和优化为一个更适合手势识别的13层CNN模型;通过串口通信技术将上位机的手势识别结果传给下位机,利用STM32单片机实现对机械手的相应控制。实验结果表明,该方式在测试集上的手势识别准确率平均为98%,能直观且便捷地控制机械手作业。
Aiming at the problem that the man-machine interaction mode of the manipulator in the traditional industry is not direct,we use a convolutional neural network(CNN)to design a machine vision control system based on gesture.OpenCV was used to construct a gesture data set.Based on the AlexNet structure of CNN,a 13 layer CNN model which was more suitable for gesture recognition was improved and optimized.Through serial communication technology,the gesture recognition results of the upper computer were transmitted to the lower computer.Finally,the corresponding control of the manipulator was realized by STM32.The experimental results show that the accuracy of gesture recognition on the test set is 98%on average,which can directly and conveniently control the manipulator operation.
作者
张行健
张建新
Zhang Xingjian;Zhang Jianxin(Faculty of Mechnical Engineering and Automation,Zhejiang Sci-Tech University,Hangzhou 310018,Zhejiang,China)
出处
《计算机应用与软件》
北大核心
2020年第10期220-224,共5页
Computer Applications and Software
基金
浙江省高等教育“十三五”第一批教学改革研究项目(jg20180090)
浙江理工大学教育教学改革研究项目(jgyl201704)。
