A method for counting people in surveillance video

Abstract

The invention relates to a method for counting people in a surveillance video, comprising: establishing a pedestrian sample library; for each frame of video image, using a mixed Gaussian background modeling method combined with morphological filtering to obtain a foreground image; The number of foreground pixels is used to obtain the foreground area and the normalized scene area; using the image foreground as a template, the Harris corner information and SURF feature point information are extracted, and the number of effective feature points per unit area is used to represent the relationship between the crowd in the scene degree of occlusion; take the normalized scene area S2, crowd occlusion factors D1, D2 as input vectors, and take the statistics of the number of people in the scene as output vectors to train the BP network to complete the construction of the regression model T1; extract the HOG features of the pedestrian sample library, and use The Adaboost cascade classifier T2 trains the corresponding pedestrian detector; constructs a combined classifier to realize adaptive calculation of weights when classifiers are fused.

Description

technical field [0001] The invention belongs to the field of intelligent video monitoring. Specifically, it is a real-time people counting system based on computer vision. Background technique [0002] In recent years, with the improvement of people's attention to security and the development of modern security technology, video surveillance systems have been widely used in all aspects of social life, from the security of banks and exhibition halls to the monitoring of squares and campuses. , From the working environment to the family environment, the video surveillance system plays an irreplaceable role in social public security and punishing crimes, safeguarding the prosperity and stability of the society, and promoting the development and construction of a harmonious society. [0003] However, traditional video surveillance systems have their own limitations. First of all, its functions are relatively simple. It only has simple monitoring video storage and playback func...

AI Technical Summary

Problems solved by technology

[0003] However, traditional video surveillance systems have their own limitations

First of all, its functions are relatively simple. It only has simple monitoring video storage and playback functions. Its main function is forensic analysis after the event. However, the real-time analysis ability of the monitored scene is relatively lacking, and it does not have the function of providing real-time early warning for abnormal events that occur.

Secondly, in order to achieve the purpose of real-time monitoring, the monitoring room of the unit or department that has installed the monitoring camera needs security personnel to monitor without interruption around the clock, which is a great waste of human and material resources.

At the same time, security personnel are prone to fatigue when working continuously for a long time, so the probability of missed judgments and misjudgments will greatly increase

It can be seen that if we simply rely on traditional manpower for monitoring, we cannot adapt to the current development trend.

Images