Smoke Detection Based On Video

Smoke detection based on video is a new technology which gets the video information obtained through machine vision and recognizes smoke by using intelligent algorithm. Compared with the traditional detection method based on sensor, smoke detection has a lot of advantages, such as wider range of application, simpler hardware and more information; Compared with the method based on video flame detection, smoke detection respond to fire earlier, so timely measures can be taken at tireless stage and disaster can be controlled more effectively. Further research on smoke detection algorithm has been made in this thesis through using the intelligent fire alarm system.This thesis firstly analyzes the influence of atmospheric scattering caused by smoke in image and gives the model of smoke image. It is demonstrated that the scene radiance suffers from attenuation caused by smoke and it makes the image semitransparent. On this basis, dark channel prior theory, image restoration, image matching and region growing be used to get suspicious smoke area which has a translucent property. Background modeling and motion region extraction are carried out in the suspected smoke area of the image, and edge characteristics of smoke and non smoke are studied through experiments of a large number of statistics. Finally, a fuzzy clustering method of edge points in order to reduce the false alarm of video smoke detection is presented in this thesis.The experimental results show that the proposed smoke detection method is stable and reliable, and the algorithm achieves the desired effect with less false alarm and missed alarm. The time complexity of the algorithm is low, it can satisfy the requirement of the real-time system, and it can better adapt to the detection of early fire smoke in different environments.