Multi-factor joint abnormal pedestrian discrimination method based on generative adversarial network

Abstract

The invention discloses a multi-factor joint abnormal pedestrian discrimination method based on a generative adversarial network, and the method employs an improved Pedestrian-Synthesis-GAN network for the detection and tracking of pedestrians, and provides a basis for the discrimination of a movement track, the recording of passing time and the head exposure duration of the pedestrians; predicting a pedestrian trajectory by using a Social-GAN network, and calculating the similarity between the actual trajectory and the predicted trajectory to obtain an abnormal score; designing an SVM for judging pedestrian passing time, and obtaining a corresponding abnormal score. Because abnormal pedestrians can generate a freezing effect, a GAN network is used for recognizing and detecting a human face on the basis of pedestrian detection, and abnormal scores of face exposure duration and exposure times are calculated; and finally setting a dynamic weight, and carrying out multi-factor fusion judgment on five aspects of the motion trail, the stay time, the face exposure duration, the exposure frequency and the behavior posture of the pedestrian. The accuracy of detecting the abnormal pedestrian is effectively improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a multi-factor joint discrimination method for abnormal pedestrians based on a generative confrontation network. Background technique [0002] As the coverage of surveillance cameras in public areas gradually increases, people's demand for using surveillance videos to pre-identify abnormal persons in surveillance is also increasing. The intelligent detection technology for abnormal persons will monitor pedestrians more efficiently and find out abnormal situations in advance, so as to better deal with various emergencies. It can be seen that intelligent monitoring will play an important role in social public security. [0003] In the detection of abnormal pedestrians, there are some outstanding problems, specifically: firstly, there are few abnormal samples, and there are many types of abnormal situations, which are difficult to enumerate completely; secondly,...

AI Technical Summary

Problems solved by technology

However, since the mentality, purpose, and focus of abnormal pedestrians are very different from normal pedestrians, the differences they exhibit are also various. Moreover, some abnormal personnel have certain anti-reconnaissance capabilities and will deliberately cover up some of their behavioral characteristics. As a result, it is impossible to identify successfully. It can be seen that it is unreliable to determine abnormal pedestrians with a single factor

Images