Abnormal behavior detection method based on time-space Laplacian Eigenmaps learning

Abstract

The invention discloses an abnormal behavior detection method based on time-space Laplacian Eigenmaps learning, belongs to the technical field of digital image processing, and relates to theoretical knowledge related to computer vision, mode identification, machine learning and data mining. An optical flow histogram is used to extract optical flow features from two adjacent frames of pictures, movement characteristic information in a monitoring scene is obtained, the movement characteristic information is clustered in a spectral clustering manner by using a video expression form of low-dimension space, the clustering amount and characteristic sets in different classifications are obtained, Hausdorff distance is applied to the characteristic sets to measure the similarity between the sets, a characteristic set which is different from those of other classifications obviously is searched, and thus, an abnormal behavior is detected. According to the invention, data in high-dimension space is re-expressed in a low-dimension space, the operational complexity is reduced, and abnormal behavior detection in a crowded scene is helped. The detection rate of abnormal behaviors is 73.52-78.45%, the omission rate 17.05-21.45%, and the false detection rate 4.5-6.1%.

Description

Technical field [0001] The invention belongs to the technical field of digital image processing, and relates to relevant theoretical knowledge such as computer vision, pattern recognition, machine learning, and data mining. Background technique [0002] Accidents caused by public safety issues have shown a rapid growth trend in recent years. Therefore, visual analysis of dense crowd scenes has become an active research field. Scene analysis is based on digital image processing, pattern recognition, and computer vision. The captured scene images or video sequences are analyzed to complete the identification process. Through some subsequent processing, functions such as rapid acquisition of abnormal behaviors and early warning of public safety incidents can be realized. Abnormal behaviors refer to group behaviors realized by sudden gatherings, group fights, and riots. Abnormal event acquisition can usually be divided into the following steps: feature extraction of video sequences,...

AI Technical Summary

Problems solved by technology

[0003] 1. The amount of motion information is large. As mentioned in the title, scenes with dense crowds contain a large amount of human motion information, such as jogging, walking, and jumping, and there are a large number of human body occlusions, so it is impossible to track and detect each individual

[0004] 2. The background is complex. In densely populated areas such as squares, urban central business districts, stations, etc., there are often busy scenes of people coming and going. Not only that, the neon lights flicker and change at night, and the advertisements played on the LCD screen Lighting changes, etc., will have a negative impact on the effect of abnormal behavior detection

The problem with this method: For the set normal behavior sequence, there is only a single conversion method, for example: set normal behaviors: sit, walk, run, and the traditional abnormal behavior detection only exists by sitting -> walking -> running If the state transition is from sitting -> running, it will be judged as an abnormal behavior at this time. At the same time, the traditional abnormal behavior detection also needs attention in terms of computational complexity, especially in dense crowd scenes. The way of detection and tracking often requires a lot of time overhead and computational complexity

Images