| Fire threatens the safety of life and property of human society and damages the earth’s environment.It is of great significance to detect fire in the early stage of fire and promptly report to the police.Traditional fire alarms cannot meet the needs of alarms under complex environmental conditions.Therefore,based on the research of flame detection algorithm based on infrared video image,this paper proposes a new flame detection algorithm based on specific wavelength.The flame is a mixture of gas and solid produced by the intense oxidation reaction during the burning of a substance,which can release visible light,infrared and ultraviolet light.Its emission spectrum is determined by the chemical composition and intermediate products of the burning substance.According to Wien’s displacement law,the intensity of flame emission spectrum increases with increasing wavelength within the photosensitive range of CMOS image sensor at 1 100nm.The main interference of flame detection comes from sunlight,and is affected by the water vapor in the atmosphere.The solar radiation intensity is in the peak and valley at 940nm.Therefore,this paper uses a 940nm wavelength narrowband infrared filter to obtain filtered infrared video images.There are many flame image processing methods.This article studies a variety of filtering,foreground extraction,threshold segmentation,and morphological processing algorithms for specific wavelength infrared images,and analyzes and compares the collected data sets,and proposes the use of bilateral filtering and background difference methods,P-quantile method and open operation process the infrared image to extract the suspected flame target.Then,contour tracking and feature extraction are performed on the suspected flame target,and the appropriate threshold parameters are obtained by analyzing the data set.Finally,a flame infrared video image of a specific wavelength was used to test and analyze the algorithm proposed in this paper in multiple dimensions.The experimental data show that the algorithm has a good detection effect,and the real-time performance,detection distance and accuracy rate meet the requirements of fire detection. |
