Human body abnormal behavior identification method based on monitoring system

Abstract

The invention discloses a human body abnormal behavior identification method based on a monitoring system; the method comprises the following steps: 1, sample image input; 2, foreground extraction; 3, background updating; 4, foreground image denoising process; 5, motion target tracking and recording: using a Mean-Shift target tracking algorithm to track the foreground image after denoising, and recording the motion information of the motion target; 6, behavior feature extraction: using a background differencing method to extract a motion area of the motion target according to the motion target motion information recorded in step5, and selecting a foreground image with features according to the motion target motion behaviors; 7, building a standard behavior database; 8, result analysis and abnormity determinations; 9, abnormity alarming. The method can improve the foreground image clearness, and can improve the abnormity behavior determination accuracy.

Description

technical field [0001] The invention relates to a method for identifying abnormal behavior of a monitoring system, in particular to a method for identifying abnormal behavior of a human body based on the monitoring system. Background technique [0002] Most of the current industrial monitoring systems only detect or track moving objects in the scene, and there are relatively few further detection and analysis of abnormal events or abnormal behaviors in the scene. The existing anomaly detection methods based on the monitoring system mainly use the model-based method to monitor and judge. This method first needs to determine a certain criterion, and then extract the shape, movement and other information of the moving target from the image sequence. According to these obtained Feature information manually or using a semi-supervised method to define a model of normal behavior, usually using a graphical model to model the state represented by the sequence image features, those ob...

AI Technical Summary

Problems solved by technology

The existing anomaly detection methods based on the monitoring system mainly use the model-based method to monitor and judge. This method first needs to determine a certain criterion, and then extract the shape, movement and other information of the moving target from the image sequence. According to these obtained Feature information manually or using a semi-supervised method to define a model of normal behavior, usually using a graphical model to model the state represented by the sequence image features, those observations that do not match the normal behavior model are considered abnormal, through artificial Observation is not only easy to cause misjudgment, but also affects the speed of judgment and recognition, and also increases labor costs; when extracting information such as the shape and motion of moving objects from image sequences, the traditional Gaussian background model is directly done in the RGB color space. Unified processing, in addition, because the outline of the person is an irregular arc, the motion detection of the person is generally applied outdoors, and the outdoor light is often changed, so the foreground image will appear discontinuous empty scene, resulting in the foreground image The clarity is low, which affects subsequent identification of abnormal behaviors

Therefore, the existing technology has the problems of poor foreground image clarity and prone to misjudgment of abnormal behavior

Images