Classification of small sample part models based on deep learning
To solve the problem of low automatic classification accuracy of mechanical parts with small sam-ples in equipment manufacturing industry,a classification detection method of mechanical parts based on ACGAN is proposed.Firstly,based on PCL point cloud technology,the two-dimensional view information of 10 kinds of representative parts such as valve seats under a fixed visual angle is obtained.Secondly,the ACGAN algorithm based on small sample data classification is selected,and 42 two-dimensional views of each part obtained by point cloud technology are used as the data set of the algorithm model to complete the design and training of the algorithm model.The experiment results show that the training set accuracy of the two-dimensional view obtained by PCL point cloud can reach 99.80％,and the test set accuracy can reach 99.67％.Therefore,this algorithm can realize the automatic classification of small sample mechanical parts with high accuracy.
万方数据
维普
