A video data typical dynamic target detection method and system

Abstract

The invention discloses a video data typical dynamic target detection method and system. The method comprises the following steps of S1, receiving a video real-time stream; S2, preprocessing the videoreal-time stream; S3, carrying out target analysis and identification on the preprocessed video real-time stream; and S4, sending out feedback according to the identification result in the step S3. Acamera sensor registered for collecting video real-time streams is further included in front of the S1. The step S1 further comprises sending a corresponding video real-time stream to the user according to a user instruction, the step S2 specifically comprises decoding the video real-time stream and preliminarily filtering image redundant information. According to the method and the system, a distributed deep neural network computing technology platform framework is adopted, the image convolution target segmentation model is constructed, real-time and high-accuracy recognition and segmentation of typical object contents such as people and vehicles in the current environment are achieved, and the problems that a traditional image is poor in recognition capability, high in false alarm rateand poor in computing real-time performance can be well solved.

Description

technical field [0001] The invention relates to visual processing, in particular to a method and system for detecting a typical dynamic target of video data. Background technique [0002] Surveillance camera is the main means of security monitoring system. With the development of digital image processing technology and pattern recognition technology, the application range of surveillance cameras has been further expanded. For example, monitoring cameras in coal mines are more widely used in the fields of automatic coal mining, personnel and equipment monitoring and positioning. Surveillance cameras are generally fixedly installed on the pan-tilt, and are controlled by the pan-tilt control system to change the angle, so as to realize image acquisition in different areas; for example, by setting up surveillance cameras at key traffic intersections, real-time monitoring of the flow of people and vehicles can be achieved. monitoring, so how to identify and detect key dynamic t...

AI Technical Summary

Problems solved by technology

Surveillance cameras are generally fixedly installed on the pan-tilt, and are controlled by the pan-tilt control system to change the angle, so as to realize image acquisition in different areas; for example, by setting up surveillance cameras at key traffic intersections, real-time monitoring of the flow of people and vehicles can be achieved. Therefore, how to identify and detect key dynamic targets on the real-time stream of video data, and how to overcome the problems of weak recognition ability, high false alarm rate, and poor real-time performance of traditional image recognition for random and complex environments such as video have become an urgent need in this field. Key Issues to Solve

Images