Coarse-to-fine video target behavior identification method

Abstract

The invention discloses a coarse-to-fine video target behavior identification method, and the method comprises the steps: firstly obtaining human body key points by using an attitude estimation algorithm or annotation information in a video, and cutting and zooming different body parts of a human body; taking the deep neural network as a feature extraction network, and extracting feature vectors of different part areas; and iteratively training a classifier by utilizing the extracted feature vectors of different parts, and searching for the optimal coarse classification of the behavior; for the coarse classifier and each fine-grained classifier, selecting different parts to be cascaded with the global feature vector; carrying out individual training of the classifiers; and carrying out probability fusion on classification results of the coarse-grained classifier and the fine-grained classifier to obtain a final behavior recognition result. According to the method, a behavior recognition framework from coarse to fine is constructed, cascading is utilized to train classifiers in a targeted mode for feature expressions of different body parts of a person with different granularities,and therefore the probability of wrongly dividing similar behaviors is effectively reduced, and the overall behavior recognition accuracy is improved.

Description

technical field [0001] The invention relates to the field of behavior recognition in videos, in particular to a coarse-to-fine video target behavior recognition method. Background technique [0002] Different from image-based recognition and detection, video-based content and human behavior analysis is currently a difficult and challenging task for human visual understanding. Video human behavior recognition, as the basic research of video abnormal behavior detection, relational reasoning and deep understanding of content, has been widely concerned by researchers. [0003] At present, more mature behavior recognition schemes can be divided into two categories according to different application scenarios and information sources: (1) template matching based on background modeling. This method mainly divides the scene of the moving target in the video, that is, the input static image, and uses the frame difference method or background modeling method to segment the foreground ...

AI Technical Summary

Problems solved by technology

However, this type of method lacks information guidance such as semantics and attention mechanisms, and requires a large-scale video data set for training, which increases storage consumption and computation.

Images