Multi-level attention scale perception crowd counting method

Abstract

The invention provides a multi-level attention scale perception crowd counting method, and belongs to the application of deep learning in computer vision. The method comprises the following specific steps: S1, acquiring a data set; S2, constructing a multi-level attention scale perception neural network; S3, debugging, training and testing the multi-level attention scale perception neural network; and S4, acquiring a camera image, and inputting the camera image into the trained neural network for testing to obtain a predicted density map and a predicted number of people of the image. In this way, the method can be suitable for crowd number detection in a large-scale scene, and the accuracy of a detection result is effectively improved.

Description

technical field [0001] The invention relates to a multi-level attention scale perception crowd counting method. Background technique [0002] With the acceleration of the country's urbanization and the rapid development of the urban economy, the scene of crowd gatherings has increased, and the number of tourists has increased, but at the same time there are security risks. Therefore, by designing a crowd counting method, predicting the number of people, and giving early warning to highly crowded scenes, it can help relevant personnel to carry out pre-warning and post-event decision-making for emergencies. occur. [0003] At present, there are mainly two types of crowd counting: 1) methods based on traditional methods, such as support vector machines, decision trees, etc.; 2) methods based on deep learning, such as MCNN, CSRNet and other network neural network methods. The above crowd counting methods based on deep learning all have certain limitations. Method 1) uses the ...

AI Technical Summary

Problems solved by technology

Method 1) using traditional methods, high complexity and poor precision; method 2) using existing neural networks, low precision and other issues

Images