Part Recognition and Pose Estimation Based on Convolutional Neural Network

Automated sorting of parts is the basis and premise of an automated production system. In order to accurately and quickly solve the problem of part type recognition and pose estimation, this paper proposes a method of part recognition and pose estimation based on convolutional neural network. By adapting and optimizing the application of the existing mature network model, the training can be used for the type recognition of mechanical parts and the posture estimation of the network model, which provides an accurate information foundation for industrial automation sorting and assembly. At the same time, a virtual data set generation method based on OSG is proposed. A large number of training data sets are generated in the virtual scene to ensure the accuracy of model training.