Gesture depth image continuous detection method using multi-head mask balance fusion unit

A depth image and detection method technology, applied in neural learning methods, character and pattern recognition, biological neural network models, etc., can solve the problem that the single-frame gesture detection method cannot introduce timing and space dependencies at the same time, so as to improve continuity And stability and detection accuracy, improve stability, improve the effect of continuity and stability

Active Publication Date: 2020-06-12
ZHEJIANG UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide a gesture depth image sequence that relies on the quasi-Gaussian spatial distribution of hand joints to apply a multi-head mask mechanism to the spatial position relationship for the existing single-frame gesture detection method. Detection method, the method greatly improves the stability of gesture detection by introducing the quasi-Gaussian spatial distribution of hand joints

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
  • Gesture depth image continuous detection method using multi-head mask balance fusion unit
  • Gesture depth image continuous detection method using multi-head mask balance fusion unit
  • Gesture depth image continuous detection method using multi-head mask balance fusion unit

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0043] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0044] refer to figure 1 , according to the embodiment of the invention and the implementation process of the complete method of the present invention are as follows:

[0045] S1. Capture the depth image sequence of the dynamic changes of the human hand through the depth camera where D t Represents the depth image, T represents the image sequence, imgH and imgW represent the length and width of the image respectively, and the depth information stored in each pixel unit in the image is recorded as {z=D t (u,v)|u∈[0,imgH), v∈[0,imgW)}

[0046] S2 projects each frame of the depth image sequence into a 3D space point cloud For the correspondence between the pixels of each frame of depth image and the spatial point cloud, refer to the following calculation formul...

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 gesture depth image continuous detection method using a multi-head mask balance fusion unit. The method includes: capturing a depth image sequence dynamically changed by a human hand through a depth camera; gridding each frame of projection three-dimensional space point cloud as an input grid of the frame to form a grid sequence; traversing each frame, and inputting the frames into a volume convolutional neural network single-frame detection model for processing to obtain hand joint pseudo-Gaussian spatial distribution of each frame; selecting hand joint pseudo-Gaussian spatial distribution of a previous frame, inputting the hand joint pseudo-Gaussian spatial distribution into a multi-head mask equalization fusion unit to obtain a mask of a current frame, and thenperforming mask superposition on the mask of the current frame and an input grid; inputting the multi-frame detection model of the volume convolutional neural network to obtain hand joint pseudo-Gaussian spatial distribution of the current frame; repeating the steps, and converting the hand joint pseudo-Gaussian spatial distribution of each frame into three-dimensional coordinates of the hand bone joint points. According to the method, the hand joint pseudo-Gaussian space distribution is introduced, so that the gesture detection stability is greatly improved.

Description

technical field [0001] The invention belongs to a gesture depth image recognition and detection method in the technical field of image automatic recognition, and more specifically relates to a gesture depth image continuous detection method based on a multi-head mask equalization fusion unit. Background technique [0002] Gesture detection is the cornerstone technology in many gesture human-computer interaction fields, such as gesture servo of robotic arm, virtual reality and augmented reality interaction. The single-frame gesture detection solution based on the monocular depth camera to capture the depth image of the human hand and the supervised learning of the deep convolutional neural network has become the mainstream method for the coordinate detection of each joint of the gesture. However, the existing single-frame gesture depth image detection and recognition methods, due to the high-dimensional complexity of gesture space expression and the problems of self-occlusion...

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): G06K9/00G06N3/04G06N3/08
CPCG06N3/08G06V40/28G06N3/048G06N3/045Y02T10/40
Inventor 何再兴郭方泰赵昕玥张树有谭建荣
Owner ZHEJIANG 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