Research And Application Of Point Cloud Objective Classification And Segmentation Technology Based On Deep Learning

A point cloud is one of the common representations of 3D shapes.The classification and segmentation of point clouds is a popular research topic in the field of deep learning.There are many deep learning methods in this stage,but there are still problems and challenges.On the one hand,most of the methods can not effectively learn the local features of the point cloud,which can not further improve the accuracy and robustness of classification and segmentation;on the other hand,the high complexity of the neural network leads to need a great amount of memory and requires a lot of training time.In this paper,a new method is proposed to solve the above problems:Firstly,A robust deep neural network,RMFP-DNN,for multi-feature point cloud classification and segmentation is proposed.RMFP-DNN extracts the local features between points through the improved self-attention module,uses the Multi Layer Perceptron(MLP)to learn the global feature of points,and finally improves the accuracy and robustness by combining the local features and the global features through feature fusion.RMFP-DNN is trained and tested with Model Net40 data for object classification and the Shape Net data for part segmentation.The results show that RMFP-DNN has better accuracy and robustness than Point Net,Point Net++ and DGCNN.Secondly,A lightweight point cloud classification deep neural network ADLP-DNN with adaptive Down Sampling is proposed.By adaptive Down Sampling,the feature point which is important to the classification results are retained,which can reduce the complexity of computational after feature extraction.On the basis of RMFP-DNN,adaptive Down Sampling is added to reduce dimension after feature extraction,and multi GPU parallel computing mode is adopted to reduce training time.ADLP-DNN is trained and tested with modelnet40 data.The results show that ADLP-DNN is more efficient than Point Net,Point Net++ and RMFP-DNN.Finally,Based on the above research,designs a point cloud classification and segmentation system based on deep learning,which is implemented by using Python and PYQT5.Using Graphical User Interface(GUI)system to show the classification and segmentation of point cloud,which can reveal and analyze the classification and segmentation of point cloud more clearly in a clear way.