Video analysis based pedestrian volume counting method

Abstract

The invention provides a video analysis based pedestrian volume counting method. The videos comprise at least one group of pedestrian monitoring sample videos and one group of pedestrian monitoring detection videos shot by a camera from a top region. The pedestrian volume counting method comprises the steps of S10, acquiring human head images in each pedestrian monitoring sample video and each pedestrian monitoring detection video, and forming a positive sample training image set and a negative sample training image set; S20, respectively extracting an HOG (Histogram of Oriented Gradient); S30, performing offline training of an SVM (Support Vector Machine) head classifier on HOG features of training input images; S40, performing real-time head detection; S50, carrying out target tracking by adopting an optical flow method; S60, setting an ROI (Region of Interest); and S70, counting the number of pedestrians. According to the method provided by the invention, the pedestrian volume in a specific region can be counted, pictures at different moments can be checked at a mobile phone terminal, and a problem of mutual shielding of detected targets can be avoided compared with a method of human body detection; and the method is established on the basis of widely applied video monitoring equipment, low in cost and high in counting accuracy.

Description

technical field [0001] The invention relates to an image acquisition and analysis technology, in particular to a method for counting people flow based on video analysis. Background technique [0002] In the management and decision-making of shopping malls, airports, museums and other public places, the flow of people is an indispensable data. Taking shopping malls as an example, the flow of people is a very basic and important indicator, which is closely related to the sales of shopping malls. The flow of people can be A reliable reference for sales, service and logistics. In some areas, the flow of people, as a parameter to control the population density of the area, is also an important consideration for its safety. Therefore, through the statistics of the flow of people, it is possible to effectively monitor and organize the operation of public places in real time, and provide people with a safe environment and better services. [0003] At present, the methods for obtai...

AI Technical Summary

Problems solved by technology

Using manual statistics on pedestrian flow, although the accuracy can meet the requirements, it consumes manpower and financial resources, and is not systematic and comprehensive; although contact equipment such as revolving doors and rotary handles are separated from manual work, they give pedestrians a sense of compulsion , which brings inconvenience, and generally only one person can pass through at a time, which is not suitable for people flow statistics in the public domain; and the method of using sensors to detect human bodies for statistics is currently the most widely used infrared detection system, which is also mainly used for passageways , this type of system is non-contact, which has greatly improved compared with contact equipment

However, there are also problems that cannot be ignored in actual operation. For example, the use of a single channel restricts the convenience of a large flow of people, and when a wide channel is used, when multiple pedestrians pass through the infrared device continuously or at the same time, there will be problems. The accuracy of people counting will be greatly reduced due to occlusion. In addition, the temporary stay of a single pedestrian in the detection area, or the interference of the body and carried items will also have a great impact on the accuracy of counting.

Images