Three-dimensional point cloud single-stage target detection method for decoupling classification and regression tasks

A 3D point cloud, target detection technology, applied in neural learning methods, instruments, biological neural network models, etc., can solve problems such as accuracy loss, and achieve the effect of improving accuracy

Active Publication Date: 2021-08-06
HARBIN ENG UNIV
View PDF9 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This will lead to some loss of accuracy, which also happens in 3D object detection tasks

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
  • Three-dimensional point cloud single-stage target detection method for decoupling classification and regression tasks
  • Three-dimensional point cloud single-stage target detection method for decoupling classification and regression tasks
  • Three-dimensional point cloud single-stage target detection method for decoupling classification and regression tasks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0023] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0024] The present invention improves the traditional single-stage detection region proposal network and can be used to improve the effect of general object detection problems. The main steps are: (1) voxelize the point cloud to make the data from unnecessary point cloud into an ordered grid structure, (2) use 3D sparse convolution to extract the features of the network to obtain high-order Feature map, (3) In the feature map, use a double-headed detector to aggregate features and predict the classification, regression box, and direction of the target. In order to solve the problem of feature entanglement between target detection sub-tasks, the present invention designs a double-headed detection network structure, which can extract the features concerned by classification and regression tasks respectively from high-dimensional features, and pre...

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 three-dimensional point cloud single-stage target detection method for decoupling classification and regression tasks, and the method comprises the steps: (1) carrying out the voxelization processing of a point cloud, enabling the data to become an ordered grid structure from an unnecessary point cloud, (2) carrying out the feature extraction of a network through three-dimensional sparse convolution, and obtaining a high-order feature map, and (3) in the feature map, using a double-head detector to aggregate the features and predict the classification, regression frame and direction of the target. In order to solve the problem of feature entanglement between target detection subtasks, a double-head detection network structure is designed, features concerned by classification and regression tasks can be extracted from high-dimensional features, and the subtasks are predicted. On the basis of decoupling, related information in the two tasks is combined by using a joint detection method to jointly predict a target category. According to the method, the accuracy of three-dimensional target detection is improved, and the method can be easily migrated to other methods.

Description

technical field [0001] The invention relates to a three-dimensional point cloud single-stage target detection method for decoupling classification and regression tasks, belonging to the field of computer vision three-dimensional point cloud processing. Background technique [0002] Object detection is one of the fundamental works in 2D and 3D space in the field of computer vision. Accurate object detection results are the prelude to tasks such as tracking, and are crucial in applications such as intelligent transportation and indoor smart home. According to the network structure, the target detection framework is divided into one-stage structure and two-stage structure. The one-stage structure can directly detect objects, while the two-stage structure adds a network that can aggregate the complete local features of the entire object, thus trading more time and computing resources for higher accuracy. [0003] Object detection usually consists of two subtasks: classificatio...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06K9/46G06N3/08
CPCG06N3/08G06V10/44G06V2201/07G06F18/241G06F18/253
Inventor 何芸倩夏桂华张智苏丽王立鹏
Owner HARBIN ENG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap