Multi-feature combined crowd grouping detection method in dense scene

Abstract

The invention relates to a multi-feature combined crowd grouping detection method in a dense scene. The method comprises the following steps: step 1, obtaining pedestrian coordinate information for animage of a video image sequence; step 2, obtaining an arrangement mode of combining every two pedestrians; step 3, performing feature extraction on the pedestrian coordinate information; step 4, performing correlation clustering solution on the data set with the known track and the grouping result to obtain a maximum weight matrix; and step 5, obtaining a clustering result on the test set througha support vector machine. Compared with the prior art, pedestrian coordinates are obtained by adopting continuous energy minimization, and then, pedestrian motion track space-time characteristics, motion direction characteristics, Granger causality, heat energy diagram characteristics and motion correlation characteristics are extracted, the data set is trained through correlation clustering to obtain the classification mapping of the extracted features to the clustering result. Finally, the classification mapping is carried out on the data set to realize grouping, and the method has the advantages of high recognition accuracy, wide application range and the like.

Description

technical field [0001] The invention relates to the technical fields of computer vision and intelligent transportation, in particular to a method for grouping and detecting crowds in dense scenes combined with multiple features. And other systems are widely used. Background technique [0002] With the urbanization of the world, crowd phenomena become more and more frequent, such as sports competitions, demonstrations, terrorist activities, etc., and the safety of public places has attracted more and more attention. Therefore, the detection technology of small and medium groups in dense crowds has broad application prospects in intelligent monitoring, virtual reality, public safety, etc., and has important research value in maintaining social security and intelligently planning urban traffic. [0003] The sports scenes of dense crowds can be divided into two types macroscopically: structured sports scenes and unstructured sports scenes. In a structured motion scene, the cro...

AI Technical Summary

Problems solved by technology

For example, the construction of a hydrodynamic model proposed by Moore et al. This type of algorithm is effective in extremely dense crowds, but this type of algorithm ignores the existence of a single individual purpose and goal

Images