Semantic segmentation method of multi-scale multi-feature algorithm based on spherical neighborhood

Abstract

A semantic segmentation method of a multi-scale multi-feature algorithm based on a spherical neighborhood comprises the following steps: registering obtained point cloud data with a remote sensing image to generate point cloud data fused with RGB information; selecting a spherical neighborhood to obtain local neighborhood features of the point cloud data fused with the RGB information, and extracting multi-scale point cloud features by changing the radius of the spherical neighborhood; the extracted basic features, the five-dimensional neighborhood features of at least two scales and xyz coordinate information of the point cloud are combined and input into an improved model MSMF-PointNet based on PointNet for semantic segmentation, and a classification result is output. According to the method, the classification precision far better than PointNet can be obtained in outdoor scene point cloud data obtained through airborne LiDAR scanning, building facades, fences and the like are better classified due to the fact that linearity, perpendicularity and other characteristics are added, the classification result of roughness, total variance, trees and shrubs is better, flatness is added, and the classification precision is greatly improved. And the classification result of the roof and the waterproof ground is better.

Description

technical field [0001] The invention belongs to the technical field of remote sensing and photogrammetry, and in particular relates to a semantic segmentation method based on a spherical neighborhood-based multi-scale and multi-feature algorithm. Background technique [0002] Deep learning is an emerging technology that can automatically learn to extract advanced features of input data through deep network structure. It is the most influential and fastest-growing cutting-edge technology in current pattern recognition, computer vision and data analysis. Before being applied to 3D data, deep learning has become an effective force for various tasks in 2D computer vision and image processing. After winning the championship with a score of more than ten percent in the second place, the deep neural network structure dominated by CNN has made major breakthroughs in the fields of image classification, segmentation and recognition. However, due to the high-density, massive and unstr...

AI Technical Summary

Problems solved by technology

[0004] However, the above methods will cause the loss of data information during the preprocessing process. In response to this shortcoming, Qi et al., a scholar at Stanford University, published a groundbreaking research work in 2017, and proposed a deep learning model PointNet, which can be directly applied to On the 3D point cloud data, the accuracy of semantic segmentation is further improved

However, the PointNet network structure mainly extracts the global features between point clouds, ignoring the local feature extraction associated with points in the point cloud, and the insufficient extraction ability of local features often leads to insufficient segmentation accuracy and poor segmentation of object details. And other issues

Images