Behavior recognition and location method based on multi-task joint learning

Abstract

The invention discloses a behavior recognition and positioning method based on multi-task joint learning, which combines the convolutional neural network in deep learning with multi-task joint learning to replace the single-task convolutional neural network algorithm, so as to realize human behavior in video Targets for identification and behavioral targeting. The invention improves the object detection depth network in the faster rcnn, and combines the behavior recognition depth network on this basis, so that the combined network can achieve the ability of multi-task joint learning, so that the two tasks can promote each other and enhance the robustness of the recognition algorithm and accuracy; at the same time, the present invention combines the video data set and the picture data set, which enhances the information diversity of the training set; in addition, if it will consume a huge amount of energy to mark the position of the human body in the video data set, the present invention can pass the algorithm The autonomous learning omits the labeling work of the data set, which can greatly reduce the labeling workload.

Description

technical field [0001] The invention belongs to the technical fields of computer vision, machine learning and deep learning, and specifically relates to the design of a behavior recognition and positioning method based on multi-task joint learning. Background technique [0002] In the field of security, there is a great demand for human behavior detection and positioning, such as the detection of violent behavior. If the violent behavior that endangers society and others can be detected in real time and measures taken, it will be of great significance to social stability. If it is possible to locate the specific thugs who committed violent acts in the video, combined with the application of face recognition, it will be of great value to quickly solve the case. However, the current video surveillance system is mainly based on manpower and supplemented by computers. It mainly identifies the content of the surveillance video manually. The workload is huge, and as the surveillan...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to solve the above-mentioned problems in the prior art, and propose a behavior recognition and positioning method based on multi-task joint learning, which uses the combination of convolutional neural network in deep learning and multi-task joint learning to replace a single task Convolutional neural network algorithm to meet the needs of human behavior recognition and behavior positioning in video clips

Images