Two-stage behavior identification method and system based on key frame sequence and behavior information

Abstract

The invention discloses a two-stage behavior identification method and system based on a key frame sequence and behavior information, and the method comprises the steps: screening out similar adjacent frames through calculating the similarity between sparse representation results of video frames, and obtaining a key frame sequence; calculating the similarity between the categories by using the behavior information of the behavior category label, and dividing all behavior categories into a plurality of large categories; establishing and training a two-stage behavior identification method model, wherein the first-stage training enables the network to have a rough classification capability, and the second-stage training enables the network to have a fine classification capability; and finally, identifying the video by using the trained model. According to the method, the key frame sequence of the video is acquired as input data, so that the key frame sequence contains more information; in addition, the network training process and the recognition process are divided into two stages by utilizing the information of the behavior category label, so that the learning process of the network is easier, and the recognition accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a two-stage behavior recognition method and system based on key frame sequences and behavior information. Background technique [0002] With the improvement of computer computing power and the development of streaming media, there are more and more video data, and people are no longer satisfied with the computer's ability to process image data. People hope that computers can also process video data in the same way as image data, and analyze the information contained in video data, so video analysis has become an important problem that needs to be solved urgently in the field of artificial intelligence. Behavior recognition is one of the contents of video analysis. It is also called action recognition. Its purpose is to analyze human behavior from a video containing complete actions and identify the types of actions made by people in the video. Different from ...

AI Technical Summary

Problems solved by technology

However, because of the 3D convolution used in this type of method, the amount of calculation and parameters are relatively large. To reduce the amount of calculation, it can only reduce the length of the input video frame sequence, and most of these methods use random interception of fixed-length video frames. Sequence method to obtain the input video frame sequence, which is easy to make the input video frame sequence contain too much redundant information

In addition, the current behavior recognition methods based on 3D convolutional networks are end-to-end training of the network model, directly allowing the network to learn the correspondence between data and real labels, which may increase the difficulty of network learning

Images