基于深度学习和背景差分的泄漏气体图像检测方法研究

随着工业的不断发展，化工产业逐渐走向聚集化发展，产业效率和竞争力获得巨大提升，同时也带来巨大安全风险。在化工生产过程中，涉及多种危险气体，这些气体普遍有毒有害易燃易爆，一旦发生泄漏不仅对环境造成污染，更会对社会安全和人类健康构成了严重威胁。红外光学成像检测技术在泄漏气体检测方面具有多方面优势，能够在远距离对大范围视场中的空间检测泄漏气体，对气体在空间中的运动实时动态展示。但是，传统算法检测灵敏度低，需要整幅图像作背景，计算机视觉中的目标检测和语义分割算法，在气流边缘的分割准确度不足。因此，对基于深度学习和背景差分结合的气体泄漏检测算法进行研究，充分利用两种算法的优势，对提高泄漏气体预警检测和事故现场处置具有重要的意义。

本文通过调研和分析国内外气体泄漏红外光学成像检测算法的研究进展，结合相关理论知识，以泄漏气体红外图像为研究对象，基于深度学习和背景差分算法，提出了一种泄漏气体检测方法。针对目前基于深度学习的气体泄漏检测的研究缺少泄漏气体红外图像数据集的问题，在实验和演习中，释放安全气体模拟气体泄漏，构建了一个泄漏气体红外图像数据集，用于深度学习算法的训练。对于红外图像对比度低、气流痕迹微弱的特点；提出一种基于引导滤波和可视性恢复的图像增强算法，有效提高红外图像中气流的可辨识度，提高数据集的标注精度，为目标检测算法的训练奠定基础。将气体泄漏检测分为两个阶段，首先，利用目标检测算法强大的特征提取能力，完成对泄漏气体所在的矩形区域的初步确定，然后，在这些矩形区域上，完成改进的自适应阈值的背景差分处理，实现对泄漏气体的识别和分割。在两阶段中，分别采用不同的处理算法，二者的结合充分利用各自算法的优势，阈值自适应背景差分算法不仅对目标检测算法结果的复核，而且通过负反馈对检测阈值作动态调整。

With the continuous evolution of industry, the chemical industry has increasingly gravitated towards a centralized developmental paradigm, significantly enhancing both efficiency and competitive prowess. However, this transformation, while beneficial in many aspects, concurrently introduces substantial safety risks. Within chemical production processes, diverse toxic, harmful, flammable, and explosive gases are implicated. In the event of a leak, these gases can precipitate explosion incidents, resulting in loss of life and property damage. Infrared optical imaging detection technology offers manifold advantages in detecting gas leaks, capable of surveying large areas and long distances and dynamically presenting gas motion in real-time spatial settings. Nevertheless, traditional algorithms exhibit low detection sensitivity and require the entire image as background. Conversely, object detection and semantic segmentation in computer vision, while leveraging robust feature extraction capabilities encounter challenges in accurately segmenting gas flow edges. Hence, investigating gas leak detection algorithms predicated on deep learning and background subtraction holds significance for fortifying the detection prowess of gas detection equipment by  comprehensively harnessing the strengths of both algorithms.

 
This paper reviews both domestic and international advancements in the field of gas leak detection via infrared optical imaging.  It discusses relevant theoretical frameworks and algorithms and proposes a methodology for detecting leaking gases based on deep learning and background subtraction algorithms, utilizing gas leak infrared images as the research focal point.To address the existing scarcity of infrared image datasets  for training deep learning models in gas leak detection, this study has constructed a new dataset by releasing safe gases to simulate gas leaks in experiments. This dataset is utilized for the training deep learning algorithms.
Given the low contrast and faint traces of gas flows in infrared images, an image enhancement algorithm rooted in guided filtering and visibility restoration is proposed to effectively augment gas flow discernibility in infrared imagery, enhance dataset annotation precision, and establish a foundation for training the object detection algorithm.
Gas leak detection encompasses two phases: initially harnessing object detection algorithms' robust feature extraction capability to preliminarily locate rectangular regions harboring leaking gases, subsequently employing refined adaptive threshold background subtraction processing on these regions to achieve recognition and segmentation of leaking gases. In both phases, distinct processing algorithms are employed, leveraging the amalgamation of these algorithms to exploit their individual strengths. The adaptive threshold background subtraction algorithm not only scrutinizes the outcomes of the object detection algorithm but also dynamically adjusts the detection threshold through negative feedback

[1] WEIJIAN Z, CHUNCHAO Y, ZHIXIONG Y, et al. Remote sense for environment pollution gases in wide infrared spectral range[J/OL]. Infrared and Laser Engineering, 2019, 48: 1104002.https://api.semanticscholar.org/CorpusID:212909120.
[2] COLLINS D C, LEE M L. Developments in ion mobility spectrometry–mass spectrometry[J/OL]. Analytical and Bioanalytical Chemistry, 2002, 372: 66-73. https://api.semanticscholar.org/CorpusID:29195651.
[3] DESHMUKH K, KOVÁřÍK T, PASHA S K K. State of the art recent progress in two dimensional MXenes based gas sensors and biosensors: A comprehensive review[J/OL]. Coordination Chemistry Reviews, 2020, 424: 213514. https://api.semanticscholar.org/CorpusID:225022147.
[4] 聂小溪. 基于宽波段红外光谱成像技术的多种气体检测识别方法研究[D]. 电子科技大学,2023.
[5] ZENG Y, MORRIS J M. Detection limits of optical gas imagers as a function of temperature differential and distance[J/OL]. Journal of the Air & Waste Management Association, 2018,69: 351 - 361. https://api.semanticscholar.org/CorpusID:53093101.
[6] FOX T A, BARCHYN T E, RISK D, et al. A review of close-range and screening technologies for mitigating fugitive methane emissions in upstream oil and gas[J/OL]. Environmental Research Letters, 2019, 14. https://api.semanticscholar.org/CorpusID:159280839.
[7] MERIBOUT M. Gas Leak-Detection and Measurement Systems: Prospects and Future Trends[J/OL]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-13. DOI: 10.1109/TIM.2021.3096596.
[8] KASTEK M, PIATKOWSKI T, LAGUEUX P, et al. Passive Optoelectronics Systems ForStandoff Gas Detection: Results Of Tests[J/OL]. Artificial Intelligence Review, 2015, 198:89-104. https://api.semanticscholar.org/CorpusID:62579093.
[9] GAGNON M A, JAHJAH K A, MARCOTTE F, et al. Time-resolved thermal infrared multispectral imaging of gases and minerals[C/OL]//ANDRESEN B F, FULOP G F, HANSONC M, et al. Infrared Technology and Applications XL: Vol. 9070. SPIE, 2014: 90700J.https://doi.org/10.1117/12.2050569.
[10] KASTEK M, SOSNOWSKI T, OR T, et al. Multispectral Gas Detection Method[J/OL]. Artificial Intelligence Review, 2009, 123: 227-236. https://api.semanticscholar.org/CorpusID:60764657.
[11] BERNASCOLLE P, ELICHABE A, FERVEL F, et al. Stand-off CWA imaging system: second sight MS[C/OL]//Defense + Commercial Sensing. 2012. https://api.semanticscholar.org/CorpusID:120191421.
[12] WANG J, TCHAPMI L P, RAVIKUMAR A P, et al. Machine vision for natural gas methane emissions detection using an infrared camera[J/OL]. Applied Energy, 2020, 257: 113998. https://www.sciencedirect.com/science/article/pii/S030626191931685X. DOI: https://doi.org/10.1016/j.apenergy.2019.113998.
[13] BIN J, RAHMAN C A, ROGERS S, et al. Tensor-Based Approach for Liquefied Natural Gas Leakage Detection From Surveillance Thermal Cameras: A Feasibility Study in Rural Areas[J/OL]. IEEE Transactions on Industrial Informatics, 2021, 17(12): 8122-8130. DOI: 10.1109/TII.2021.3064845.
[14] WANG J, JI J, RAVIKUMAR A P, et al. VideoGasNet: Deep learning for natural gas methane leak classification using an infrared camera[J/OL]. Energy, 2022, 238: 121516. https://www.sciencedirect.com/science/article/pii/S0360544221017643. DOI: https://doi.org/10.1016/j.energy.2021.121516.
[15] SPATAFORA M A N, ALLEGRA D, GIUDICE O, et al. Natural Gas Leakage Detection: a Deep Learning Framework on IR Video Data[C/OL]//2022 26th International Conference on Pattern Recognition (ICPR). 2022: 636-642. DOI: 10.1109/ICPR56361.2022.9956523.
[16] RAVIKUMAR A P, WANG J, BRANDT A R. Are Optical Gas Imaging Technologies Effective For Methane Leak Detection?[J/OL]. Environmental Science & Technology, 2017, 51(1): 718-724. https://doi.org/10.1021/acs.est.6b03906.
[17] BIN J, BAHRAMI Z, RAHMAN C A, et al. Foreground Fusion-Based Liquefied Natural Gas Leak Detection Framework From Surveillance Thermal Imaging[J/OL]. IEEE Transactions on Emerging Topics in Computational Intelligence, 2023, 7(4): 1151-1162. DOI: 10.1109/TETCI.2022.3214826.
[18] BIN J, ROGERS S, LIU Z. Vision Fourier transformer empowered multi-modal imaging system for ethane leakage detection[J/OL]. Information Fusion, 2024, 106: 102266. https://www.sciencedirect.com/science/article/pii/S1566253524000447. DOI: https://doi.org/10.1016/j.inffus.2024.102266.
[19] LI J, WANG L, ZHANG C S, et al. Gas cloud infrared image enhancement based on anisotropic diffusion[C/OL]//Defense + Commercial Sensing. 2011. https://api.semanticscholar.org/CorpusID:121865474.
[20] LUO X, MA J, CHEN D, et al. Archimedean spiral push-broom differential thermal imaging for gas leakage detection[J/OL]. Opt. Express, 2019, 27(6): 9099-9114. https://opg.optica.org/oe/abstract.cfm?URI=oe-27-6-9099. DOI: 10.1364/OE.27.009099.
[21] PENG X, QIN H, HU Z, et al. Gas plume detection in infrared image using mask R-CNN with attention mechanism[C/OL]//GREIVENKAMP J, TANIDA J, JIANG Y, et al. AOPC 2019: AI in Optics and Photonics: Vol. 11342. SPIE, 2019: 113420U. https://doi.org/10.1117/12.2548179.
[22] LU Q, LI Q, HU L, et al. An Effective Low-Contrast SF₆ Gas Leakage Detection Method for Infrared Imaging[J/OL]. IEEE Transactions on Instrumentation and Measurement, 2021, 70:1-9. DOI: 10.1109/TIM.2021.3073443.
[23] WANG X, MIAO Y, CHEN Y. Deep Temporal Network for VOC Gas Recognition in the Middle Infrared Spectrum[C/OL]//2022 4th International Academic Exchange Conference on Science and Technology Innovation (IAECST). 2022: 1143-1146. DOI: 10.1109/IAECST57965.2022.10062280.
[24] SHI J, XIE W, HUANG X, et al. Real-time natural gas release forecasting by using physics guided deep learning probability model[J/OL]. Journal of Cleaner Production, 2022, 368:133201. https://www.sciencedirect.com/science/article/pii/S0959652622027895. DOI: https://doi.org/10.1016/j.jclepro.2022.133201.
[25] HUANG E, CHEN L, LV T, et al. GLRNet: Gas Leak Recognition via Temporal Difference in Infrared Video[C]//FANG L, POVEY D, ZHAI G, et al. Artificial Intelligence. Cham:Springer Nature Switzerland, 2022: 515-520.
[26] ZUO J, HU X, XU L, et al. CH4 gas leakage detection method for low contrast infrared images[J/OL]. Infrared Physics & Technology, 2022, 127: 104473. https://www.sciencedirect.com/science/article/pii/S1350449522004546. DOI: https://doi.org/10.1016/j.infrared.2022.104473.
[27] MA D, LIU A, FAN Z, et al. Gas Leakage Recognition Based on Wide-Band Infrared Imaging with the Auxiliary Excitation Method and Machine Learning Model[J/OL]. ACS Chemical Health & Safety, 2022, 29(5): 455-466. https://doi.org/10.1021/acs.chas.2c00045.
[28] HU Y, XU L, XU H, et al. Three-dimensional reconstruction of a leaking gas cloud based on two scanning FTIR remote-sensing imaging systems[J/OL]. Opt. Express, 2022, 30(14): 25581-25596. https://opg.optica.org/oe/abstract.cfm?URI=oe-30-14-25581. DOI: 10.1364/OE.460640.
[29] LI K, DUAN S, PANG L, et al. Chemical Gas Telemetry System Based on Multispectral Infrared Imaging[J/OL]. Toxics, 2023, 11(1). https://www.mdpi.com/2305-6304/11/1/83. DOI: 10.3390/toxics11010083.
[30] WANG Y, HUANG L, CHENG Z, et al. Flow Faster RCNN : Deep Learning Approach for Infrared Gas Leak Detection in Complex Chemical Plant Surroundings[C/OL]//2023 42nd Chinese Control Conference (CCC). 2023: 7823-7830. DOI: 10.23919/CCC58697.2023.10241164.
[31] HONG S Z, HU Y, YU H W. A VOCs Gas Detection Algorithm Based On Infrared Thermal Imaging[C/OL]//2019 Chinese Control And Decision Conference (CCDC). 2019: 329-334.DOI: 10.1109/CCDC.2019.8833058.
[32] WU S, ZHONG X, QU Z, et al. Infrared Gas Detection and Concentration Inversion Based on Dual-Temperature Background Points[J/OL]. Photonics, 2023, 10(5). https://www.mdpi.com/2304-6732/10/5/490. DOI: 10.3390/photonics10050490.
[33] ZIFEN H, HUIZHU C, YINHUI Z. Spatial information adaptive regulation and feature alignment for infrared methane instance segmentation[J/OL]. Optics and Precision Engineering,2023, 31(20): 3034-3049. DOI: 10.37188/OPE.20233120.3034.
[34] YU H, WANG J, WANG Z, et al. A lightweight network based on local–global feature fusion for real-time industrial invisible gas detection with infrared thermography[J/OL]. Applied Soft Computing, 2024, 152: 111138. https://www.sciencedirect.com/science/article/pii/S1568494623011560. DOI: https://doi.org/10.1016/j.asoc.2023.111138.
[35] WANG Q, XING M, SUN Y, et al. Optical gas imaging for leak detection based on improved deeplabv3+ model[J/OL]. Optics and Lasers in Engineering, 2024, 175: 108058. https://www.sciencedirect.com/science/article/pii/S0143816624000381. DOI: https://doi.org/10.1016/j.optlaseng.2024.108058.
[36] YAO J, XIONG Z, LI S, et al. TSFF-Net: A novel lightweight network for video real-time detection of SF6 gas leaks[J/OL]. Expert Systems with Applications, 2024, 247: 123219. https://www.sciencedirect.com/science/article/pii/S0957417424000848. DOI: https://doi.org/10.1016/j.eswa.2024.123219.
[37] 李家琨, 金伟其, 王霞, 等. 气体泄漏红外成像检测技术发展综述[J]. 红外技术, 2014, 36:513-520.
[38] FLANIGAN D F. Limits of passive remote detection of hazardous vapors by computer simulation[C/OL]//Defense, Security, and Sensing. 1996. https://api.semanticscholar.org/CorpusID:120080682.
[39] 郭安祥, 叶日新, 董明, 等. 基于红外吸收光谱法检测 SF_6 分解气体的仿真与实验研究[J].绝缘材料, 2017, 50(12): 72-77.
[40] ABITAN H, BOHR H, BUCHHAVE P. Correction to the Beer-Lambert-Bouguer law for optical absorption[J/OL]. Appl. Opt., 2008, 47(29): 5354-5357. https://opg.optica.org/ao/abstract.cfm?URI=ao-47-29-5354. DOI: 10.1364/AO.47.005354.
[41] 沈英, 邵昆明, 吴靖, 等. 气体光学检测技术及其应用研究进展[J]. 光电工程, 2020, 47:3-18.
[42] LAI R, TANG YANG Y, JIAN WANG B, et al. A quantitative measure based infrared image enhancement algorithm using plateau histogram[J/OL]. Optics Communications, 2010,283(21): 4283-4288. https://www.sciencedirect.com/science/article/pii/S003040181000684X.DOI: https://doi.org/10.1016/j.optcom.2010.06.072.
[43] WAN M, GU G, QIAN W, et al. Infrared Image Enhancement Using Adaptive Histogram Partition and Brightness Correction[J/OL]. Remote Sensing, 2018, 10(5). https://www.mdpi.com/2072-4292/10/5/682. DOI: 10.3390/rs10050682.
[44] SILVA E A, PANETTA K, AGAIAN S S. Quantifying image similarity using measure of enhancement by entropy[C/OL]//SPIE Defense + Commercial Sensing. 2007. https://api.semanticscholar.org/CorpusID:891075.
[45] KIM S. Sea-Based Infrared Scene Interpretation by Background Type Classification and Coastal Region Detection for Small Target Detection[J/OL]. Sensors, 2015, 15(9): 24487-24513. https://www.mdpi.com/1424-8220/15/9/24487. DOI: 10.3390/s150924487.
[46] BRANCHITTA F, DIANI M, CORSINI G, et al. New technique for the visualization of highdynamic range infrared images[J/OL]. Optical Engineering, 2009, 48: 096401. https://api.semanticscholar.org/CorpusID:62543772.
[47] LIU N, ZHAO D. Detail enhancement for high-dynamic-range infrared images based on guided image filter[J/OL]. Infrared Physics & Technology, 2014, 67: 138-147. https://api.semanticscholar.org/CorpusID:122293802.
[48] PITAS I. Digital Image Processing Algorithms and Applications[M]. 1st ed. USA: John Wiley& Sons, Inc., 2000.
[49] LAI R, YANG Y, WANG B, et al. A quantitative measure based infrared image enhancement algorithm using plateau histogram[J/OL]. Optics Communications, 2010, 283: 4283-4288.https://api.semanticscholar.org/CorpusID:123043458.
[50] PIZER S M, AMBURN E P, AUSTIN J D, et al. Adaptive histogram equalization and its variations[J/OL]. Graphical Models graphical Models and Image Processing computer Vision,Graphics, and Image Processing, 1987, 39: 355-368. https://api.semanticscholar.org/CorpusID:62771950.
[51] ZUIDERVELD K J. Contrast Limited Adaptive Histogram Equalization[C/OL]//Graphics gems. 1994. https://api.semanticscholar.org/CorpusID:62707267.
[52] TAREL J P, HAUTIÈRE N. Fast visibility restoration from a single color or gray level image[J/OL]. 2009 IEEE 12th International Conference on Computer Vision, 2009: 2201-2208. https://api.semanticscholar.org/CorpusID:6966375.
[53] HE K, SUN J, TANG X. Single image haze removal using dark channel prior[C/OL]//2009IEEE Conference on Computer Vision and Pattern Recognition. 2009: 1956-1963. DOI: 10.1109/CVPR.2009.5206515.
[54] ZHAO Q, LUO D, WANG J, et al. An Image Enhancement Method for Gas Leak Detection Based on Infrared Imaging *[C/OL]//2023 IEEE International Conference on Real-time Computing and Robotics (RCAR). 2023: 245-250. DOI: 10.1109/RCAR58764.2023.10249481.
[55] HE K, SUN J, TANG X. Guided Image Filtering[J/OL]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 35: 1397-1409. https://api.semanticscholar.org/CorpusID:1264129.
[56] GE Z, LIU S, WANG F, et al. YOLOX: Exceeding YOLO Series in 2021: abs/2107.08430[A/OL]. 2021. https://api.semanticscholar.org/CorpusID:236088010.
[57] ZHAO Q, NIE X, LUO D, et al. An Effective Method for Gas-Leak Area Detection and Gas Identification with Mid-Infrared Image[J/OL]. Photonics, 2022, 9(12). https://www.mdpi.com/2304-6732/9/12/992. DOI: 10.3390/photonics9120992.