融合光谱-纹理-颜色信息识别地下天然气微泄漏点

doi:10.3964/j.issn.1000-0593(2020)10-3123-07

摘要
参考文献
相关文章 (15)

全文: PDF (8029 KB)
输出: BibTeX | EndNote (RIS)

摘要：地下天然气储气库一旦发生泄漏将会导致严重的安全事故与经济损失，而快速准确地识别天然气泄漏点能够有效避免以上情况发生。针对地下储气库微泄漏的情况，传统方法因其准确率低、成本高等缺点难以适用。利用高光谱技术进行野外模拟实验，融合光谱信息、纹理信息和颜色信息对天然气泄漏点进行识别。基于地下天然气储气库泄漏胁迫相应区域植被生长的特征，以冬小麦为研究对象，利用SOC710VP高光谱成像仪（光谱范围：400～1 000 nm）分别获取第11天、第24天、第32天、第40天以及第49天控制组与胁迫组地块的高光谱影像，对所获取的影像进行光谱平滑、反射率校正和裁剪处理之后：（1）首先运用方差分析（ANOVA）选取特征波段，分别为：510，520，570，625，645，680和690 nm；其次，运用灰度共生矩阵（GLCM）计算特征波段影像纹理特征，并计算RGB合成影像的前三阶颜色矩；（2）运用NDVI对影像进行阈值分割，将其分为植被部分和裸露土壤部分。对于植被部分，基于最小二乘支持向量机（LSSVM）构建融合光谱、纹理以及颜色特征的识别模型，并对胁迫下的小麦进行识别；（3）融合胁迫下的小麦和裸露土壤的识别结果，并对融合后的结果进行形态学和圆拟合处理。结果表明：模型综合光谱、纹理以及颜色特征，能够在胁迫发生的第24天较好地识别出胁迫区域；受天然气胁迫的冬小麦，在光谱、纹理以及颜色特征方面均与控制组小麦存在较大差异；受天然气胁迫的区域在时序上呈现先扩大后缩小的趋势；研究还发现距胁迫边缘0.25 m的圆环内，小麦胁迫组长势较控制组小麦区旺盛，出现“绿晕现象”。为验证模型的适用性，运用所构建的识别模型对实验区内的大豆、玉米和草地进行识别验证，均取得较好的效果。该研究结果可为工程应用中识别天然气泄漏提供理论支持。

关键词：高光谱；天然气微泄漏；胁迫特征；识别模型

Abstract：The leakage of underground natural gas storage will lead to serious safety accidents and economic losses, which can be effectively avoided by the rapid and accurate identification of natural gas leakage point. For the micro leakage of underground gas storage, traditional methods are difficult to be applied due to its low accuracy and high cost. In this study, hyperspectral technology is used to conduct field simulation experiments. Meanwhile, spectral, texture and color information was fused to identify natural gas leakage points. Based on the characteristics of vegetation growth in the corresponding areas under leakage stress of underground natural gas storage, the SOC710VP (spectral range: 400～1 000 nm) is used to obtain the hyperspectral images of winter wheat in the control and stress areas on the 11th, 24th, 32nd, 40th and 49th days accordingly. After preprocessing steps such as spectral smoothing, reflectivity correction and clipping, (1) ANOVA (Analysis of Variation) is first used to select the following characteristic bands: 510, 520, 570, 625, 645, 680 and 690 nm, respectively. Secondly, the gray level co-occurrence matrix (GLCM) is used to calculate the texture features of the characteristic band images and the first three order color moments of RGB synthetic images are calculated. (2) NDVI is used to segment the image, which is divided into vegetation part and bare soil part. Based on least squares support vector machine (LSSVM), a recognition model was constructed to recognize wheat under stress by fusing spectral, texture and color features. (3) The recognition results of wheat and bare soil under stress are fused and processed by morphology and circle fitting. The results showed that there are significant differences in the spectrum, texture and color characteristics between wheat under natural gas stress and wheat without stress. The areas under natural gas stress tend to expand first and then shrink. The recognition model constructed in this paper synthesizes spectral, texture and color features can better identify the stress area on the 24th day of stress occurrence. It is also found that wheat grew more vigorously and appeared “green halo” in the circle 0.25 m from the stress edge. In order to verify the applicability of the model, this paper used the established identification model to identify the soybean, corn and grassland in the experimental area, and achieves satisfying results. The research results can provide theoretical support for the identification of natural gas leakage in engineering applications.

Key words：Hyperspectral; Natural gas micro-leakage; Stress characteristics; Recognition model

收稿日期: 2019-09-07 修订日期: 2020-01-16

中图分类号:

P237

基金资助: 吴自勇，1994年生，中国矿业大学（北京）地球科学与测绘工程学院硕士研究生 e-mail: wlw11266@163.com

通讯作者: 蒋金豹 E-mail: jjb@cumtb.edu.cn

作者简介: 吴自勇，1994年生，中国矿业大学（北京）地球科学与测绘工程学院硕士研究生 e-mail: wlw11266@163.com

引用本文:

吴自勇，蒋金豹，郭鉴威，王鑫达，季杨. 融合光谱-纹理-颜色信息识别地下天然气微泄漏点[J]. 光谱学与光谱分析, 2020, 40(10): 3123-3129.
WU Zi-yong, JIANG Jin-bao, GUO Jian-wei, WANG Xin-da, JI Yang. Recognition of Underground Gas Micro-Leakage Points by Fusing Spectrum-Texture-Color Information. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2020, 40(10): 3123-3129.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2020)10-3123-07 或 https://www.gpxygpfx.com/CN/Y2020/V40/I10/3123

[1] National Development and Reform Commission（国家发展和改革委委员会）. The 13th Five-Year Plan for Natural Gas Development（《天然气发展“十三五规划”》），2016.
[2] ZHANG Gang-xiong，LI Bin，ZHENG De-wen, et al（张刚雄，李彬，郑得文, 等）. Natural Gas Industry（天然气工业），2017，37(1)：153.
[3] CHEN Xiao-yuan，TAN Yu-fei（陈晓源, 谭羽非）. Oil & Gas Storage and Transportation（油气储运），2011, 30(7)：513.
[4] Lassalle G, Credoz A, Hédacq, Rémy, et al. Environmental Science & Technology, 2018, 52(4): 1756.
[5] Sanches I D, Filho C R S, Magalhães L A, et al. ISPRS Journal of Photogrammetry and Remote Sensing, 2013, 78(4): 85.
[6] Smith K L, Steven M D, Colls J J. Remote Sensing of Environment, 2004, 92(2): 207.
[7] Noomen M F, Werff H M A V D, Meer F D V D. Ecological Informatics, 2012, 8(2): 55.
[8] Werff H M A, Bakker W H, van der Meer F D, et al. Computers & Geosciences, 2006, 32(9): 1334.
[9] Jiang J, Qiao X, He R. Journal of Food Engineering, 2016, 169: 284.
[10] Qiao X, Jiang J, Qi X, et al. Food Chemistry, 2017, 220: 393.
[11] GAO Yun-liang, MA Ling(高运良，马玲). Mathematical Statistics(数理统计). Beijing: China Coal Industry Publishing House(北京：煤炭工业出版社)，2002.
[12] ZHANG Shuai-tang, WANG Zi-yan, ZOU Xiu-guo, et al(张帅堂, 王紫烟, 邹修国, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报)，2017，33(22)：200.
[13] HAN Ding, WU Pei, ZHANG Qiang, et al(韩丁，武佩，张强，等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2016, 32(23): 168.
[14] SHEN Qian, ZHU Chang-ming, ZHANG Xin(沈谦, 朱长明, 张新). Bulletin of Surveying and Mapping(测绘通报), 2019，(7): 33.
[15] ZHAO Chun-hui, WANG Li-guo, QI Bin(赵春晖，王立国，齐滨). Hyperspectral Remote Sensing Image Processing Method and Application (高光谱遥感图像处理方法及应用). Beijing: Publishing House of Electronics Industry(北京：电子工业出版社)，2016. 187.