Establishing the regression model of smoke-flow opacity for adaptive intelligent smoke identification using remote hyperspectral sensing

Severe air pollution has far-reaching effects on human health, which is why in 2016 the United States implemented a standard test method for determining plume opacity in the atmosphere of an outdoor environment (ASTM 7520-16). This method uses a digital camera to photograph the smoke stream and a digital image to calculate the opacity of the smoke stream. This method has a strong reliance on the subjective judgment of the inspector. In view of the shortcomings of traditional inaccurate quantification and only quantification of specific air pollutants, this paper proposes to combine the image recognition technology and deep learning analysis of remote hyperspectral cameras to adaptively determine the background of smoke, sunlight and atmospheric environmental conditions, so as to improve the accuracy of intelligent smoke concentration. An attempt was also made to establish a regression model of the opacity of smoke flow of various colors to increase the feasibility of intelligent smoke identification in the future.