People counting method based on multi-scale mask perception feedback convolutional neural network

Abstract

The invention discloses a people counting method based on a multi-scale mask perception feedback convolutional neural network. The people counting method comprises the following steps: (1) searching and manufacturing a people counting database; (2) generating training and testing samples; (3) carrying out data preprocessing on samples of the training and testing set; (4) constructing a people counting deep network model; (5) sending the generated training sample into the constructed deep network model for training, and optimizing the parameters of the network through an Adam optimization method; and (6) testing the deep network model. According to the method, a people counting network is constructed by using a multi-task learning strategy, the result of mask estimation branching is fused into picture features by the network, and then a robust people counting model is obtained; according to the method, a simple fusion strategy and a multi-scale learning strategy based on hole convolution are used, so that an accurate and stable detection result is realized, and complex configuration and memory consumption in application are avoided.

Description

technical field [0001] The invention relates to the technical fields of image processing and pattern recognition, in particular to a method for counting people based on a multi-scale mask perception feedback convolutional neural network. Background technique [0002] With the advancement of urbanization and the needs of people for entertainment and work, scenes of gatherings of many people and groups have become common in many places, which has also brought about increasingly severe safety management problems. In recent years, people often hear reports of mass incidents and stampedes around the world. Therefore, how to count the number of people in public places is a crucial task. This task can help manage crowds and provide crowd density distribution data, which is convenient for comprehensively guiding the flow of crowds, making safety problems in public places controllable and preventable. Therefore, this task has important application value for public safety. [0003]...

AI Technical Summary

Problems solved by technology

Through the analysis of the pictures in the scene, the following difficult problems have to be solved in order to achieve robust detection: (1) In a scene with a relatively high density, there is a serious occlusion problem between heads; (2) Due to the viewing angle of the monitoring equipment The change of the crowd scale is very obvious; (3) Different scenes will have different lighting, etc.

Images