A multi-scale supervision based three-dimensional scene point cloud segmentation method

CN116012587BActive Publication Date: 2026-02-27ZHONGBEI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202310067111.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-01-16
Publication Date
2026-02-27
Estimated Expiration
2043-01-16

AI Technical Summary

Technical Problem

Existing deep learning-based 3D point cloud segmentation models lack direct supervision in the hidden layers, resulting in low segmentation accuracy and poor performance, especially in object boundary segmentation.

Method used

A multi-scale supervision mechanism is introduced. Through the encoder-decoder structure, the hidden layer is supervised by the category information vector. The decoded features are predicted by combining multilayer perceptron and farthest point sampling, thereby improving the segmentation ability of the network.

Benefits of technology

It improves the segmentation accuracy and object edge segmentation effect of the network, enhances the network's ability to learn local features, and improves the segmentation effect of 3D point clouds.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116012587B_ABST
    Figure CN116012587B_ABST
Patent Text Reader

Abstract

The present application belongs to the field of computer vision, and particularly relates to a three-dimensional scene point cloud segmentation method based on multi-scale supervision. The present application introduces a multi-scale supervision mode, judges the quality of the hidden layer feature map by performing additional supervision learning on each layer of the decoder, thereby improving the overall segmentation precision of the network, prompting the features learned by the network hidden layer to be easily distinguishable and more robust, and further improving the segmentation effect of the network on object edges, and being widely applicable to large indoor point cloud semantic segmentation.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Semantic segmentation method for point cloud in outdoor large scene

    CN112560865A

  • Improved U-net semantic segmentation model construction method and method and system for tea tender shoot recognition and picking point positioning

    CN114863112A