Graph convolution behavior recognition method and device based on bone joint points

A recognition method and joint point technology, applied in the field of medical image processing, can solve problems such as insufficient expression, and achieve the effect of improving accuracy and good results

Pending Publication Date: 2021-02-23
BEIJING INSTITUTE OF TECHNOLOGYGY
View PDF1 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, this method also has certain problems. Since the action often involves the cooperation of different joints, the topological map of the human skeleton is not sufficient for the expression of the action. At the same time, the human action has differences in the time domain and the air domain. often have their own focus

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
  • Graph convolution behavior recognition method and device based on bone joint points
  • Graph convolution behavior recognition method and device based on bone joint points
  • Graph convolution behavior recognition method and device based on bone joint points

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0024] Such as figure 1 As shown, this graph convolution behavior recognition method based on skeletal joint points includes the following steps:

[0025] (1) Extract human bones through the OpenPose method;

[0026] (2) Perform dynamic modeling based on the time series of bones, and construct a spatio-temporal topology map;

[0027] (3) The structure of the graph convolutional network is improved, and the residual block of the graph convolutional network is reduced to a time domain graph convolution residual unit and a spatial domain graph convolution residual unit for the difference of spatiotemporal characteristics of different actions, so as to facilitate The network better learns the spatio-temporal features of different actions, and feature extraction is performed through an improved graph convolutional network;

[0028] (4) Use the Softmax classifier to perform behavior classification on the output features, and obtain the corresponding behavior category labels.

[0...

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 provides a graph convolution behavior recognition method and device based on the skeleton joint points. The real-time skeleton extraction and recognition can be completed. The accuracy of human body behavior recognition is improved. A good score is obtained, and the scene requirement of indoor monitoring is met. The method comprises the following steps: (1) carrying out human skeleton extraction through an OpenPose method; (2) performing dynamic modeling based on the time sequence of the skeleton, and constructing a space-time topological graph; (3) improving the structure of thegraph convolution network, reducing the residual block of the graph convolution network into a time domain graph convolution residual unit and a space domain graph convolution residual unit for the difference of the spatial and temporal features of different actions, so that the network can better learn the spatial and temporal features of different actions, and performing feature extraction through the improved graph convolution network; and (4) performing behavior classification on the output features by using a Softmax classifier to obtain corresponding behavior category labels.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, in particular to a method for recognizing a graph convolution behavior based on bone joint points, and a device for recognizing graph convolution behavior based on bone joint points. Background technique [0002] Behavior recognition is one of the important components of video understanding. It is of great significance to indoor intelligent monitoring. Accurate identification of indoor human behavior through methods can ensure the timeliness of indoor monitoring accident response and help strengthen the monitoring of indoor personnel. manage. [0003] At present, the main research methods of behavior recognition include traditional methods and deep learning methods, among which the graph convolution method in deep learning is a current research hotspot. It converts the recognition of human behavior based on video sequences into the research of human skeleton topology. By...

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): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V40/20G06N3/045G06F18/24G06F18/214
Inventor 宋红杨健李敏刘青
Owner BEIJING INSTITUTE OF TECHNOLOGYGY
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