People counting method and device based on video images

Abstract

The invention provides a people counting method and device based on video images, and the method comprises the steps of obtaining the setting parameters of a camera device in a shooting scene; according to the set parameters, obtaining an imaging weight matrix of a human body during imaging in the shooting scene; obtaining a pixel matrix of an image mask of the shooting scene; obtaining the current weighted pixel value of the shooting scene according to the pixel matrix and the imaging weight matrix of the image mask; and according to the current weighted pixel value and a preset mapping relation between the weighted pixel value and the number of people, obtaining the counted number of people of the shooting scene, so that the applicability of the method can be effectively improved. In addition, the operation complexity caused by the feature detection can be avoided, the system resource requirement is reduced, and therefore the real-time performance of the system is improved.

Description

technical field [0001] The invention relates to the technical field of video image processing, in particular to a method and device for counting people based on video images. Background technique [0002] With the continuous development of video image processing technology, it is possible to conduct real-time statistics of the number of people in supermarkets, banks, subways and other places through video surveillance, so as to obtain statistical information such as passenger flow distribution and congestion level, and provide effective information for public area management and resource scheduling. reference materials. [0003] In the prior art, people counting is performed by extracting features of pedestrians from images captured by a camera device, such as head and shoulder shape features, face features, pedestrian histogram of oriented gradient (HOG) features, etc. feature. Specifically, feature extraction and pedestrian positioning can be realized by machine learning...

AI Technical Summary

Problems solved by technology

[0004] However, in different shooting scenarios, the features selected by machine learning methods are quite different, which may lead to the inability to recognize portraits. When there are key features, missed detection may occur

Therefore, when the shooting scene changes, it is necessary to re-extract samples and retrain the pedestrian image detector, so that the system can work normally in the new shooting scene, resulting in a large workload

In addition, the computational complexity of feature detection is large, and it is necessary to extract the region of interest (Region of Interest, referred to as ROI), run the feature extraction algorithm and the classification algorithm, and the demand for computing resources is large, so it is difficult to meet the real-time requirements

Images