Behavior analysis method and device based on human body key point detection

Abstract

In order to solve the problem of low accuracy of behavior analysis in the technical problems, the invention provides a behavior analysis method and device based on human body key point detection, andthe accuracy of behavior analysis is improved. The behavior analysis method based on human body key point detection comprises the steps of detecting pedestrians in a to-be-identified video; tracking pedestrians in the detected to-be-identified video; extracting a human body key point sequence of pedestrians in the tracked to-be-identified video; and obtaining pedestrian behaviors based on the human body key point sequence and a preset behavior classifier, and the behavior classifier being a classifier obtained by training based on the human body key point sequence and the pedestrian behaviors.The device comprises a detection module, a tracking module, an extraction module and a judgment module. The pedestrian behavior is obtained based on the human body key point sequence and the preset behavior classifier, and compared with the prior art, the method has higher efficiency and accuracy.

Description

technical field [0001] The present disclosure relates to a behavior analysis method, in particular to a behavior analysis method and device based on human body key point detection. Background technique [0002] Behavior analysis is one of the hottest fields in current computer vision research. The current methods for behavior analysis mainly include traditional feature extraction and prediction methods. Typical algorithms include using optical flow algorithm to extract features, and then using SVM for classification. Another This kind of method is to use the method of deep learning training to analyze the behavior. The typical method is: use the behavior analysis of the two-stream neural network, first calculate the dense optical flow for every two frames in the video sequence, and obtain the dense optical flow sequence, and then analyze the video image and The dense optical flow is trained by CNN respectively, and the two branches of the network judge the category of the ac...

AI Technical Summary

Problems solved by technology

[0002] Behavior analysis is one of the hottest fields in current computer vision research. The current methods for behavior analysis mainly include traditional feature extraction and prediction methods. Typical algorithms include using optical flow algorithm to extract features, and then using SVM for classification. Another This kind of method is to use the method of deep learning training to analyze the behavior. The typical method is: use the behavior analysis of the two-stream neural network, first calculate the dense optical flow for every two frames in the video sequence, and obtain the dense optical flow sequence, and then analyze the video image and The dense optical flow is trained by CNN respectively, and the two branches of the network judge the category of the action respectively. Finally, the classification results of the two networks are fused to obtain the final classification result. The disadvantage is that the traditional method is very sensitive to noise and perspective changes. , the use of neural network method is time-consuming, and the use of neural network method is easily affected by appearance factors, such as color tone, image noise, etc., which leads to low accuracy of traditional analysis methods

Images