Real-time point cloud model classification method based on lightweight network lightpointnet

Abstract

The invention discloses a real-time point cloud model classification method based on the lightweight network LightPointNet, comprising the steps of: S1, analyzing the structural characteristics of the deep convolutional neural network, and designing a lightweight real-time point cloud network LightPointNet according to application requirements, the lightweight The magnitude real-time point cloud network LightPointNet includes at least an input layer, a convolutional layer, a fully connected layer, and an output layer. The last layer of the convolutional layer contains a pooling layer, and each feature that is convolved in the last convolutional layer The maximum pooling is used in the channel to generate feature values; S2, input data, and convolve the data through three convolutional layers to obtain the final feature values; S3, input the obtained feature values ​​to the fully connected layer for Classification; S4, using the classification operation of S3 to obtain the final classification accuracy. The invention has the advantages of few network layers, few parameters, fast processing speed, good classification performance and the like.

Description

technical field [0001] The invention relates to the technical fields of computer graphics, computer vision and intelligent recognition, in particular to a real-time point cloud model classification method based on a lightweight network LightPointNet. Background technique [0002] With the wide application of 3D sensors such as laser scanners and RGBD cameras in robotics, unmanned driving, and 3D scene roaming, the data of point cloud models has increased dramatically, and the high-level semantic understanding of point cloud models has gradually attracted people's attention. However, in sharp contrast to the rapid development of various technologies in the image field, due to the disorder, sparsity, and limited information of the point cloud model, various researches based on the point cloud model are progressing slowly. The present invention aims at the high-level semantic understanding requirements of the point cloud model, studies the classification method for the point cl...

AI Technical Summary

Problems solved by technology

[0009] The purpose of the present invention is to overcome the problems of complex network structure, long training time, and large amount of required data in the existing deep learning network oriented to point cloud model recognition, and propose a real-time point cloud model based on the lightweight network LightPointNet Classification method. This method only sets three layers of convolutional layers and one layer of fully connected layers. Without affecting the classification accuracy, it reduces the number of network layers, simplifies the network structure, and increases the processing speed of the network. It has the advantages of fewer network layers, Less parameters, fast processing speed, good classification performance, etc.

Images