Unlock instant, AI-driven research and patent intelligence for your innovation.

Building roof plane segmentation method based on PointNet and RANSAC algorithms

A plane segmentation and building technology, applied in computing, image analysis, computer components, etc., can solve the problems of plane lack of semantic information, increase and add semantic annotation, etc., to reduce the overall calculation time and ensure the effect of segmentation accuracy

Pending Publication Date: 2022-01-28
SHANGHAI UNIVERSITY OF ELECTRIC POWER
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, only using the RANSAC algorithm for plane segmentation also has practical problems, that is, the plane after segmentation lacks semantic information, especially when the structure of the roof of the building is complex and various types of roofs need to be finely segmented, only using the RANSAC algorithm will increase a lot. Work on adding semantic annotations

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Building roof plane segmentation method based on PointNet and RANSAC algorithms
  • Building roof plane segmentation method based on PointNet and RANSAC algorithms
  • Building roof plane segmentation method based on PointNet and RANSAC algorithms

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] refer to Figure 1 ~ Figure 2 , for the first embodiment of the present invention, this embodiment provides a kind of building roof plane segmentation method based on PointNet and RANSAC algorithm, comprising:

[0033] S1: Label the original point cloud, and downsample the labeled original point cloud to obtain the predicted point cloud.

[0034] (1) Label the original point cloud

[0035] The point cloud data of the original point cloud obtained by the 3D laser scanner must first be marked before segmentation. Specifically, the original point cloud is marked as void and roof by using the open source point cloud labeling tool Semantic-Segmentation-Editor developed by Hitachi. And in the labeling effect, white is used to indicate the void category, and purple is used to indicate the roof category.

[0036] (2) Down mining

[0037] After the point cloud is marked, the number of point clouds obtained by laser scanning of general buildings is usually very large, which ca...

Embodiment 2

[0075] To verify and explain the technical effects adopted in this method, this embodiment conducts experiments on three test samples respectively, and uses the means of scientific demonstration to test the results to verify the real effect of this method.

[0076] The test samples are a flat sloping roof building, a building with an undulating sloping roof, and a large building point cloud. The point cloud images of the three test samples are for example image 3 , the statistics of the experimental results of the three test samples are shown in Table 1, and the actual segmentation effect is as follows Figure 4 shown.

[0077] Table 1: Test sample experimental results.

[0078]

[0079] It can be seen from the test results that the overall segmentation accuracy of this method reaches 88.2%, and the average segmentation accuracy of the actual building roof point cloud can reach 90%. In terms of machine vision learning efficiency, it is 50% higher than the PointNet model; ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a building roof plane segmentation method based on PointNet and RANSAC algorithms, and the method comprises the steps of marking an original point cloud, carrying out the collection reduction of the marked original point cloud, and obtaining a predicted point cloud; inputting the predicted point cloud into a pre-segmentation network, and performing roof pre-segmentation to obtain a pre-segmentation point cloud; aligning the original point cloud with the pre-segmentation point cloud to obtain a restored point cloud; segmenting a roof part with semantic information in the restored point cloud through an RANSAC algorithm to obtain a roof point cloud; combining the non-roof point cloud with the roof point cloud to obtain a building point cloud, and completing building roof plane segmentation. According to the invention, the overall point cloud of the building can be processed, artificial semantic annotation work is simplified from annotation of each plane of the roof to annotation of the roof part of the building, and the overall operation time is shortened to a great extent while the segmentation precision is ensured.

Description

technical field [0001] The invention relates to the technical field of roof plane segmentation, in particular to a building roof plane segmentation method based on PointNet and RANSAC algorithms. Background technique [0002] With the development of building science and technology, many new types of roof structures have appeared, and a scientific and modern management method and technology is needed. The management and design are more artistic, and the roof of a building with a more aesthetic feeling can be effectively managed and designed throughout the life cycle. People It has begun to introduce machine vision technology in artificial intelligence, and the application research on the intelligence of small-area license plates using network deep learning technology is relatively mature. , the performance of network deep learning is over-fitting and under-fitting, its network regularization and data enhancement capabilities are not strong, its generalization ability of netwo...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06T7/10G06V30/19
CPCG06T7/10G06T2207/10028G06F18/241
Inventor 陈辉张傲吴仁杰杨宁张传林崔承刚
Owner SHANGHAI UNIVERSITY OF ELECTRIC POWER