Abnormal behavior detection method and system based on spatial diagram convolutional network

A convolutional network and detection method technology, applied in neural learning methods, biological neural network models, instruments, etc., can solve problems such as low frequency of abnormal behavior, different abnormal definitions, and increased burden on abnormal detection models, so as to achieve accurate detection, Computationally efficient effects

Active Publication Date: 2021-08-17
SHANDONG UNIV
View PDF5 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, there are still some shortcomings in the application of video anomaly detection: on the one hand, different scenarios have different definitions of anomalies. For example, crowds fleeing are considered normal events when playing games, but they may be identified as abnormal events in train stations and airports. events; on the other hand, the frequency of abnormal behavior is low, resulting in insufficient number of positive samples in the experimental process, so the task of anomaly detection cannot be regarded as a binary classification problem (normal, abnormal), so that the traditional supervised classification method cannot be used to complete the task
[0006] Although semi-supervised methods have achieved good results in abnormal event detection tasks, they still face two problems: first, it is impractical to build a perfect normal model under the premise of knowing all normal events in advance; Training data is quite time consuming
Since pixel-based features contain a large amount of redundant information in the video background, and there are usually irrelevant target persons in the background, using pixel-based features will inevitably introduce noise, thereby increasing the burden on the anomaly detection model to distinguish effective signals from noise

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
  • Abnormal behavior detection method and system based on spatial diagram convolutional network
  • Abnormal behavior detection method and system based on spatial diagram convolutional network
  • Abnormal behavior detection method and system based on spatial diagram convolutional network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0052] Embodiment 1 of the present invention provides an abnormal behavior detection system based on a spatial graph convolutional network, the system includes:

[0053] The extraction module is used to extract the skeleton feature space diagram of all individuals in the video frame to be detected;

[0054] The first calculation module is used to process the extracted skeleton feature space map by using the trained abnormal score model to obtain the abnormal score of each skeleton in the video frame;

[0055] The second calculation module is used to perform a maximum pooling operation on the abnormal scores of all skeletons in the video frame to obtain the abnormal scores of the video frame;

[0056] The classification module is used to identify and classify the abnormal behavior level of the video frame according to the abnormal score of the video frame.

[0057] In this embodiment 1, the above-mentioned system is used to implement an abnormal behavior detection method based...

Embodiment 2

[0080] Embodiment 2 provides a brand-new unsupervised method based on human skeleton features to detect abnormal events related to people in videos. include:

[0081] Step 1: Decompose the skeleton space graph, specifically:

[0082] Extract all human skeleton features in each frame of video image, and use a spatial map to represent the skeleton features;

[0083] A decomposition model is used to decompose the set of feature vectors corresponding to the set of graph nodes of the skeleton space graph into global feature components and local feature components.

[0084] Step 2: Based on the global and local feature components of each skeleton map, an unsupervised anomaly detection algorithm iForest is used to generate an initial normal skeleton set and an initial abnormal skeleton set.

[0085] Step 3: Based on the initial set with The anomaly scoring module ρ is iteratively trained using a self-training mechanism to obtain a better skeleton anomaly score.

[0086] Step 4...

Embodiment 3

[0133] In Example 3, an unsupervised graph convolution network abnormal behavior detection method is proposed. The overall method includes the following steps:

[0134] Step 1: Decompose the skeleton space graph, specifically:

[0135] Extract all human skeleton features in each frame of video image, and use a spatial map to represent the skeleton features;

[0136] A decomposition model is used to decompose the set of feature vectors corresponding to the set of graph nodes of the skeleton space graph into global feature components and local feature components.

[0137] Step 2: Based on the global and local feature components of each skeleton map, use the unsupervised anomaly detection algorithm iForest to generate an initial normal skeleton set N and an initial abnormal skeleton set A.

[0138] The global feature components and local feature components of the skeleton graph are respectively input into the iForest algorithm, and the abnormal scores of the global and local com...

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 an abnormal behavior detection method and system based on a space diagram convolutional network, and belongs to the technical field of machine vision processing. The method includes extracting skeleton feature spatial diagrams of all individuals in a video frame to be detected; processing the spatial diagrams by using a trained abnormal score model to obtain an abnormal score of each skeleton in the video frame; performing maximum pooling operation on the abnormal scores of all the skeletons in the video frame to obtain the abnormal score of the video frame; and performing abnormal behavior level identification classification on the frame of video according to the abnormal score of the video frame. According to the method, a space diagram is adopted to represent human skeleton features and is decomposed into global and local feature components, the global components comprise skeleton rigid motion information, the local components describe non-rigid deformation in skeleton joint points, the global features and the local features are combined, and a detection model is established without manually calibrated normal data; abnormal behavior detection under completely unsupervised setting is realized, the detection is accurate, and the calculation efficiency is high.

Description

technical field [0001] The invention relates to the technical field of machine vision processing, in particular to a method and system for detecting abnormal behaviors based on a spatial graph convolutional network. Background technique [0002] The intelligent video surveillance system plays an important role in ensuring social public safety. The system can automatically analyze and process the video data collected by surveillance cameras in a timely manner, thereby reducing the waste of human and material resources. Among them, as an important branch of intelligent surveillance, Automatic Video Anomaly Detection in Complex and Crowded Scenes (Automatic Video Anomaly Detection in Complex and Crowded Scenes) has gradually become one of the research hotspots, which is dedicated to quickly and accurately detecting unconventional behaviors such as crowd riots. to ensure public safety. [0003] However, there are still some shortcomings in the application of video anomaly detec...

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 & AuthorityApplications(China)
IPC IPC(8): G06K9/00G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/53G06V10/44G06N3/045G06F18/241
Inventor常发亮李南君刘春生
OwnerSHANDONG UNIV