Subway group abnormal behavior detection method based on station two-dimensional crowd density analysis

Abstract

The invention discloses a subway group abnormal behavior detection method based on station two-dimensional crowd density analysis. The method comprises the following steps: firstly, preprocessing original video data; then, using a full convolution U-Net neural network to extract target features; tracking the target feature points by using a pyramid Lucas-Kanade optical flow method and converting the target feature points into a station two-dimensional space coordinate system; and finally, training and learning the extracted group motion feature information by using a support vector machine anda random forest to complete the detection of group abnormal behaviors. The detection method provided by the invention improves the abnormal detection speed and recognition rate, ensures the applicability in different scenes, and reduces the false detection rate and omission ratio of group abnormal behavior recognition.

Description

technical field [0001] The invention relates to a method for detecting abnormal behavior of subway groups, in particular to a method for detecting abnormal behavior of subway groups based on two-dimensional crowd density analysis at stations, which belongs to the intelligent monitoring technology of urban rail transit. Background technique [0002] Urban rail transit, with its advantages such as exclusive lines, large transport capacity, and stable running time, has become an important way to relieve the pressure of urban public transport and improve the attractiveness of cities, and it has become an indispensable mode of transportation for the general public in daily travel. Followed by it is not to be ignored security issues. The scene of the subway station itself is complex, and the characteristics of high passenger flow and high density have brought great difficulties to the safety supervision work, making the subway station an area with a high incidence of mass security...

AI Technical Summary

Problems solved by technology

[0008] (1) Preprocessing: Use relevant algorithms to perform preprocessing operations on each frame of image data captured by the surveillance camera, perform image enhancement and sharpening, reduce image noise, remove image shadows, set background and regions of interest (based on HSV color space model to reduce the impact of lighting changes), to solve the problems of image deformation and blur caused by the image environment or shooting azimuth angle; and introduce the grid method to divide the video image into grids of different sizes. The size is inversely proportional to the pixel area and the distance to the surveillance camera;

Images