Method for detecting and counting distribution of dense crowds in video

Abstract

The invention provides a method for detecting and counting dense crowd distribution in a video. Firstly, acquiring a large number of videos containing crowds with different densities to construct a data set; then constructing a deep neural network of multi-scale feature fusion and an attention mechanism, inputting the training set into the network, outputting prediction results of a corresponding crowd density map and an attention map, constructing a loss function model in combination with the real density map and the attention map for training, and generating an optimized network; obtaining a density map of a crowd video image through optimized multi-scale feature fusion and deep neural network prediction of an attention mechanism, furthering performing point clustering on the estimated density map by using a grid-based hierarchical density space clustering method to identify a group, and obtaining the number of people and position information of the group quickly. According to the invention, the problems of perspective distortion, scale change and background noise influence of the camera can be solved, and the counting precision and stability are improved; and meanwhile, the crowd is divided into groups, so that the distribution condition of the crowd can be visually displayed.

Description

technical field [0001] The invention relates to a crowd detection method in an intelligent video monitoring neighborhood, in particular to a method for detecting and counting dense crowd distribution in a video. Background technique [0002] With the increasing demand for security precautions in areas such as public safety and traffic scenes, intelligent video surveillance has gradually replaced traditional video surveillance. Crowd distribution detection and counting is a research hotspot in the field of intelligent video surveillance, which has important social significance and market application prospects. For example, in public places where crowds tend to gather, crowd information can be used to warn of safety issues such as stampedes, and crowd distribution information can help rationally allocate manpower and material resources, thereby reducing accident casualties and even avoiding accidents; for urban public transportation systems, it can According to the number of ...

AI Technical Summary

Problems solved by technology

[0004] There are two difficulties in achieving accurate crowd counting: (1) camera perspective distortion

Pedestrians with different distances from the shooting device have different sizes, and the scale of pedestrians in an image changes significantly; (2) The background in the scene is complex

Images