Abstract:In order to solve the problems of inaccurate grasping and low efficiency of localization and classification in the working environment of the traditional parallel robots,a detection technique of workpiece recognition based on machine vision and Faster-RCNN neural network was proposed.they were collected by the parallel robot experimental platform,they were preprocessed then added to the network data set.The Faster-RCNN convolutional neural network in depth learning with Python 3.7-torch 1.7 was built as the basic framework for training workpiece image data sets.Finally,the workpiece detection results obtained by the trained convolutional neural network were compared with the original test workpiece identification system.The results show that the improved recognition average accuracy is better than the original recognition system,the reaction time is shortened and different types of workpieces can be recognized.