Lightweight point cloud noise reduction method based on convolutional neural network and related equipment

The invention relates to a lightweight point cloud noise reduction method based on a convolutional neural network and a related device. The lightweight point cloud noise reduction method comprises the following steps of S1, performing shallow feature extraction; s2, carrying out deep feature extraction and feature fusion at the same time; and S3, carrying out Dropout operation on the obtained final features to prevent overfitting, reducing the number of channels of the features to the number of categories of the point clouds by using convolution to obtain a classification result of the noise points, and removing the noise points. According to the method, Lidar point cloud data obtained in extreme weather is used for carrying out a denoising task, a LilaBlock module and an FEAR module are used for carrying out the denoising task, and compared with a traditional WeatherNet denoising network, a Head part (FEAR) is fully fused with shallow layer features obtained after LilaBlock convolution operation, so that the