Motion detection and multi-feature extraction-based flame intelligent-recognition method

Abstract

The invention discloses a motion detection and multi-feature extraction-based flame intelligent-recognition method, which relates to the field of computer vision. According to the method, firstly, animproved mixed Gaussian model is utilized for motion region extraction, then OHTA-based color segmentation is combined for color segmentation for obtaining a suspected flame region; then flame static-features and dynamic-features are extracted; and finally, a support vector machine-based algorithm is used for flame recognition. According to the method, the suspected flame region with a high accuracy degree is extracted, interference of flame-similar objects is largely eliminated, and the various flame distinctive-features are fused. The algorithm has robustness, anti-interference performance and adaptability, and is high in a recognition rate on flame at the same time.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to an intelligent flame recognition method based on motion detection and multi-feature extraction. Background technique [0002] With the rapid development of national economy and society, fire accidents of various large space buildings are increasing day by day. Traditional fire detectors can be divided into smoke detectors, temperature detectors and light detectors. However, the traditional smoke detector itself has the defect that the threshold value is set too high, so that the fire can only respond and issue an alarm when the fire reaches a certain scale. Therefore, in scenes with large spaces, such as empty stadiums, forests, large factory warehouses and other complex environments, the application of conventional smoke and temperature detectors is limited. With the continuous advancement of science, preventing fires has become particularly important, and people have begun to f...

AI Technical Summary

Problems solved by technology

Among them, the flame recognition method that only uses static features such as color is susceptible to interference from scenes similar to flame colors, thus affecting the robustness of the system

Paulo et al. first extracted the flame color area according to the RGB color Gaussian model. After extracting the area change rate, surface and boundary roughness, slope and other features of the flame color area, they trained with a Bayesian classifier, but the selection of candidate areas by this algorithm Not accurate enough

Images