Human body behavior event real-time analysis method facing multi-channel videos

The invention discloses a human body behavior event real-time analysis method facing multi-channel videos. A front-end camera pushes a video content which is collected through an RTMP protocol to a server end. A server receives the video content and converts an RTMP protocol streaming medium into an HLS protocol. The server starts an index file timing scanning thread, reads a M3U8 index file in the HLS protocol, acquires an assigned TS video file name in an index and adds the file name into a work queue. Work in the queue is read in a circulation mode. Each piece of work starts a corresponding sub-thread at a server end and the sub-threads concurrently analyze and identify the video content of each piece of work. An analyzed and identified result is written into an original video fragment which is finally displayed on a client, or information is submitted to other early warning systems. In the invention, video stream data of the front-end camera is processed in a multi-channel and concurrent mode, and multipl