Behavior recognition system and method based on skeleton data

Abstract

The invention relates to a behavior recognition system and method based on skeleton data, and the system comprises a data collection unit, a server, a data transmission unit, and a display and early warning unit, and the data collection unit is used for collecting real-time video data; the server is used for processing the real-time video data, outputting behavior identification data and outputting an early warning instruction; the data transmission unit is used for transmitting the real-time video data, the behavior identification data and the early warning instruction; and the display earlywarning unit is used for displaying the behavior identification data and the early warning instruction and giving an alarm according to the early warning instruction. Compared with the prior art, according to the invention, the time-space diagram convolution is combined with the long-short period memory network; according to the invention, the monitoring camera is combined to realize the identification of the human body behaviors in the designated area, the identification speed is fast, and the identification accuracy is high.

Description

technical field [0001] The invention relates to the technical field of behavior recognition in computer vision, in particular to a skeleton data-based behavior recognition system and method. Background technique [0002] Human body behavior recognition is a research hotspot in the field of computer vision research. Automatic recognition of human body gestures will provide a new human-computer interaction mode. In crowded places or places that need to focus on security monitoring, cameras usually take real-time pictures, and mainly identify and judge the behaviors in the video pictures manually. This kind of behavior recognition method is inefficient and cannot guarantee behavior. For the accuracy of recognition judgment, if automatic behavior recognition is carried out by computer, real-time recognition of certain designated behaviors can be replaced by manual methods, which can save a lot of manpower, thereby effectively improving recognition efficiency and recognition accu...

AI Technical Summary

Problems solved by technology

[0005] 1. Processing based on RGB images requires a large amount of calculation and slow speed. It cannot handle non-European structure data and cannot achieve the effect of real-time recognition;

[0006] 2. Sensitive to light changes, once occlusion or complex weather occurs, the accuracy of behavior recognition will be reduced;

[0007] 3. Sensitive to irrelevant information in the picture scene, such as changes in clothing and background

Images