基于神经网络和深度相机的工件检测与定位系统  

Workpiece detection and localization system based on neural network and depth camera

作　　者：蒋俊华[1] 蒋叶华 张镭[1] JIANG Junhua;JIANG Yehua;ZHANG Lei(School of Physics and Electronics,Henan University,Kaifeng 475000,China)

机构地区：[1]河南大学物理与电子学院,河南开封475000

出　　处：《传感器与微系统》2020年第8期82-85,共4页Transducer and Microsystem Technologies

基　　金：河南省高等学校青年骨干教师培养计划项目(2019GGJS032);河南大学本科教学改革研究与实践项目(HDXJJG2018-114)。

摘　　要：针对工业生产线上环境复杂、干扰多的问题,提出了基于深度学习技术和双目IR结构光深度相机相结合的方法,实现了基于SSD-Mobilenet网络和Intel Realsense D435i相机的工件检测和定位系统,可以对目标工件进行实时类别检测和三维坐标获取。通过一系列实验表明:该系统具有抗干扰能力强、检测帧率快、准确率高、定位精度高等优点,可以快速准确引导机械手工作,具有很大的应用意义和价值。Aiming at the problem that environment is complicated and interference of the workpiece on industrial production line,a method for workpiece’s detection and positioning based on deep learning technology and binocular IR structured light depth camera is proposed,which can realize real-time category detection and 3 D coordinate acquisition of target workpiece. Through a series of experiments,it shows that the system has the advantages of strong anti-interference ability,fast detection frame rate,high accuracy and high positioning precision,and can guide manipulator to work quickly and accurately. It has great application significance and value.

关 键 词：深度学习 目标检测 深度相机 三维测量 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]