Video flame detection method based on fusion of multiple classifiers

Abstract

The invention relates to a video flame detection method based on fusion of multiple classifiers. The method comprises: firstly, using a color space base on YCbCr to primarily segment a candidate flameregion; then extracting a whole moving feature, a texture feature and a stroboscopic feature of flame from the candidate flame region; respectively inputting the three flame features into four classifiers of a support vector machine, a naive Bayesian, a decision-making tree and a stochastic forest for analysis; finally, providing a detection identification method for fusion of multiple classifiers based on a confidence coefficient function (D-S evidence theory), especially when the same feature is input to different classifiers, output results having conflict. The invention provides a rejection criterion to solve the conflict. The invention can effectively improve reliability and applicability of the video flame detection technology, and has characteristics of high accuracy and low falsealarm rate.

Description

technical field [0001] The invention belongs to the field of flame detection, in particular to a video flame detection method based on multi-classifier fusion. Background technique [0002] With the continuous development of video surveillance technology and the increasing density of monitoring points, video-based flame detection has become one of the topics with great theoretical research value and practical application value, and it is also a hot issue in the field of fire detection; traditional non- Most of the contact detectors can only be used for the detection of small indoor spaces. They have the disadvantages of small detection range, great influence of the environment and single fire criterion. The traditional detection method solves the problem of limited detection range and propagation delay through video monitoring of flames. However, the current flame detection still has the problems of low algorithm detection rate, high misrecognition rate, and scene change alg...

AI Technical Summary

Problems solved by technology

[0002] With the continuous development of video surveillance technology and the increasing density of monitoring points, video-based flame detection has become one of the topics with great theoretical research value and practical application value, and it is also a hot issue in the field of fire detection; traditional non- Most of the contact detectors can only be used for the detection of small indoor spaces. They have the disadvantages of small detection range, great influence of the environment and single fire criterion. The traditional detection method solves the problem of limited detection range and propagation delay through video monitoring of flames. However, the current flame detection still has the problems of low algorithm detection rate, high misrecognition rate, and scene change algorithm failure. To solve these problems, a flame detection method that integrates dynamic and static features of flames and integrates multiple classifiers is proposed. method, which effectively improves the accuracy of flame detection and reduces the misrecognition rate, and has better detection results for flames in different scenes

Images