A fast behavior recognition method and system

Abstract

The invention relates to a fast behavior recognition method and system. In the training process of the method, a feature extraction algorithm with low computational complexity is adopted, and a variety of dimension reduction means are used to reduce the dimension and normalize the obtained feature matrix, reducing the The time required for processing; in the identification process of this method, the captured video is processed with the same algorithm as the training process, and then put into the trained model for detection and identification; the system consists of five parts: video acquisition module , offline learning module, behavior recognition module, multi-thread processing module and video display module; the system uses a sliding window mechanism to process the real-time captured video into several video segments, and processes each video segment during behavior recognition to ensure Real-time update of recognition results. At the same time, the feature extraction and feature encoding process with high computational complexity is processed by multi-threading to ensure the real-time performance of processing and recognition, which can realize fast and accurate recognition and visual display of behaviors appearing in the video, and is easy to operate and easy to expand.

Description

technical field [0001] The invention belongs to the technical fields of computer vision, machine learning and intelligent monitoring, and relates to a fast behavior recognition method and system. Background technique [0002] With the rapid popularization and development of video capture equipment and broadband networks, video has become the main carrier of information, and the amount of video data has shown explosive growth. Compared with the vigorous development of video equipment, the identification and analysis methods of video content in various video processing solutions are relatively lagging behind. Most of the solutions still use manual management, and the existing video surveillance systems usually only record Video images can only be used for post-event evidence collection, without giving full play to the real-time and initiative of video surveillance. [0003] At the same time, due to the huge amount of video information, managers cannot make real-time and accur...

AI Technical Summary

Problems solved by technology

[0009] In view of this, the purpose of the present invention is to provide a method that can quickly identify the behaviors that appear in the video and a behavior recognition system based on the method. The method and system can overcome the existing behavior recognition algorithms and system calculations. The time and space complexity is high, and the problem of identifying behaviors and actions appearing in the video cannot be quickly completed

Images