Detecting unknown gas clouds with multispectral imagery based on spectral angle matching

Abstract

The fields of safety production, environmental monitoring, public safety, and other areas all benefit greatly from the use of gas detection technologies. The infrared image can represent the spatial distribution of the gas cloud and the background, allowing for long-distance and non-contact detection during hazardous chemical accident rescue. One of the major challenges in gas detection based on infrared imaging is how to choose and gather the spectral information of the gas. It determines the properties of the complete imaging system, including its complexity, the kinds of gases that may be detected, and the sensitivity of the detection. In this paper, a gas detection system based on multispectral infrared imaging was designed, which used short pass and long pass filters to separate light. It was composed of imaging optical system, uncooled focal plane detector, filter wheel and data acquisition and processing system. The rotating filter wheel was used to separate the radiation of the object to obtaining images with different spectral information. Using image processing techniques like image subtraction and spectral angle mapping, the diffusion zone of a gas was estimated. The identified gas cloud was color-mapped in the infrared image. The infrared image had a resolution of 640 × 512, and the time from gas leakage to warning was less than two seconds. The working band of the system was 6.5-14.5 μm, and the real time detection of NH₃, SF₆, CH₄, SO₂ was realized.