Research On Gas Detection Technology Based On Passive Infrared Imaging

With the rapid development of industrialization,various gases are widely used in industry,power and other industries.However,once these gases leak,they will seriously pollute the atmospheric environment and endanger human health.However,gas leakage will seriously pollute atmospheric environment and endanger human health,so the detection of gas leakage is particularly important.Each gas has a specific wavelength of infrared absorption spectrum,which can be imaged using uncooled infrared focal plane detectors.However,signal-to-noise ratio of infrared images is often low and gas edges are blurry.Therefore,aiming at the shortcomings of existing passive infrared imaging gas detection technologies,the infrared image non-uniformity correction method and gas leakage area instance segmentation algorithm are proposed in this paper,which improves infrared images clarity and achieves precise segmentation of gas areas.The research status at home and abroad of passive infrared imaging gas detection technology is summarized.For the shortcomings of current passive infrared imaging gas detection technology,the specific research content carried out includes:1.The theory of gas passive infrared imaging is deeply studied.The gas layer radiation transfer model and simplified model in passive infrared imaging are analyzed in detail.Based on the gas equivalent blackbody temperature difference model,the effects of gas absorption band,gas concentration,gas temperature,and background temperature on gas passive infrared imaging were simulated and analyzed.2.The non-uniformity of the readout circuit response in the infrared focal plane array unit detector can result in fixed pattern noise with stripe,which seriously affects the quality of the infrared images.Considering the problems of existing non-uniformity correction,such as the loss of image detail and edge blurring,a multi-scale residual network with attention mechanism is proposed for non-uniformity correction.A multi-scale feature representation module is designed to decompose the original image into varying scales to obtain more image information.The product of the direction structure similarity parameter and the Gaussian weighted Mahalanobis distance is used as the similarity metric;a channel spatial attention mechanism based on similarity ensures the extraction of a more discriminative channel and spatial feature.By adding skip connections between input and output,a residual learning strategy is introduced to subtract estimated non-uniform noise from original image.3.In response to the current difficulty in accurately determining gas range using infrared image gas detection methods,a Mask RCNN instance segmentation method based on variable convolution and edge preservation is designed to detect and identify gas leakage areas.Firstly,deformable convolution is placed in the backbone network combining Res Net50 and feature pyramid to enhance the network’s feature extraction ability and better adapt to the deformation and easy diffusion characteristics of gases.Then,based on classification,regression and segmentation,an edge branch is designed to fuse edge features with mask features,utilizing target edge to improve the accuracy of mask localization.Finally,a pseudo-color transformation is performed based on grayscale of gas leakage area to obtain a colored gas image.The experimental results show that non-uniformity correction of gas infrared images could remove stripe noise while maintaining edge and detail,which is helpful for subsequent identification of gas leakage areas.Based on gas infrared image characteristics,gas instance segmentation algorithm can accurately detect gas leakage and further highlight gas leakage using pseudo color transformation.Finally,the effectiveness of the proposed algorithm is verified through experiments using sulfur hexafluoride gas.