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

An Action Detection Method Based on 3D Convolutional Neural Network

A convolutional neural network and motion detection technology, applied in the field of computer vision recognition, can solve the problems of video data processing, complex loss function, complex network structure, etc., to achieve the effect of improving positioning accuracy, improving reliability, and simple network structure

Active Publication Date: 2022-05-20
NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This method needs to train two classifiers. Compared with a single classifier, the loss function is more complex and difficult to train.
[0007] To sum up, although there are many researc

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
  • An Action Detection Method Based on 3D Convolutional Neural Network
  • An Action Detection Method Based on 3D Convolutional Neural Network
  • An Action Detection Method Based on 3D Convolutional Neural Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0043] The invention will be described in further detail below in conjunction with the accompanying drawings.

[0044] figure 1 Introduced the process of the present invention, the specific process is embodied in the following steps,

[0045] Video segmentation, each video is divided into multiple video segments with an overlap threshold of 75% between adjacent segments, each segment is composed of 16 consecutive frames of RGB images, and the number of overlapping frames between adjacent segments is 12 frames, Among them, if the last segment is less than 16 frames, it is discarded.

[0046] After the video is segmented, a video can be expressed as a 5-dimensional tensor. If a video is divided into N segments, the video can be expressed as a 5-dimensional tensor (N, 16, H, W, 3), where, N indicates the number of segments the video is divided into, 16 indicates that each segment includes 16 consecutive frames of pictures, H and W respectively represent the length and width of ...

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 proposes an action detection method based on a 3D convolutional neural network, which belongs to the technical field of computer vision recognition. The method includes the following steps: First, divide the video into multiple overlapping segments, use a trained 3D convolutional neural network to extract high-dimensional spatio-temporal features of each segment, and use a multi-class softmax classifier to extract the The features of each segment are predicted and classified, and the classification results are further smoothed and filtered in the time dimension; secondly, the background threshold is set, and the background score of each segment is compared with the set threshold to obtain the set of action segments; finally , combining the action segment set and frame rate information to realize the positioning of the action in the time dimension, so as to obtain the action category corresponding to the video and the start time segment set of the action. The invention realizes the end-to-end action detection and improves the reliability of the detection result.

Description

technical field [0001] The invention relates to an action detection method based on a 3D convolutional neural network, belonging to the technical field of computer vision recognition. Background technique [0002] In recent years, video processing technology has been developed rapidly. Among them, behavior detection for video has also attracted the attention of a large number of researchers due to its wide application prospects in security and other fields. With the development of deep learning, especially the extensive application of convolutional neural networks in computer vision and surprising results in the fields of recognition and detection, video behavior detection based on convolutional neural networks has received a lot of research. [0003] Application No. CN201611168185.2 "A Motion Detection Model Based on Convolutional Neural Network" uses a two-way convolutional neural network to extract the characteristics of RGB (red, green and blue three-channel) graphs and...

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): G06V40/20G06V40/10G06V10/764G06K9/62
CPCG06V40/20G06V40/10G06F18/24147
Inventor 宋佳蓉杨忠胡国雄韩家明张天翼朱家远
Owner NANJING UNIV OF AERONAUTICS & ASTRONAUTICS