Small-cluster crowd recognition method based on dynamic primitives

Abstract

The present invention provides a small-cluster crowd identification method based on dynamic graphic elements, comprising the following steps: S1) extracting trajectory; S2) scene modeling based on dynamic graphic elements; S3) crowd classification based on convolutional neural network. The advantages of the present invention are: by extracting the semantic information of the pedestrian's trajectory map, the subsequent neural network can capture better feature information to complete the classification, and the model is compared with the traditional machine learning method in terms of accuracy and recall rate. The effect is obviously improved.

Description

technical field [0001] The invention relates to the technical field of data analysis, in particular to a method for identifying small clusters of people based on dynamic graphic elements. Background technique [0002] With the development and popularization of technologies such as the mobile Internet, the Internet of Things, cameras, social networks, and urban perception networks, user behavior in public places has been effectively recorded, and this kind of spatiotemporal information includes users' travel trajectories, travel intentions, etc. feature. These massive spatio-temporal information can not only describe people's movement patterns, but also become an important part of public safety early warning. There are many types of small groups of people in society, such as thieves, beggars, unemployed people, etc. The behavior characteristics of these people are different from ordinary residents, so it is possible to detect them by analyzing spatiotemporal data. [0003] ...

AI Technical Summary

Problems solved by technology

This method has the following defects: First, factors such as the uncertainty of pedestrian activities and the differences in the population's residence have caused the disadvantages of poor applicability of the research method; second, the number of objects for trajectory screening is very small, resulting in the classification model The positive sample cases are small and the samples are extremely unbalanced; third, big data analysis techniques and statistical analysis methods only analyze from the data level, although some features can be extracted from the user's trajectory, such as the trajectory length of the crowd, trajectory mode, speed, etc., but these features are still not enough to represent the semantic features of pedestrian activities

Images