A baseline slope index to detect natural gas microleakage-stressed vegetation considering shadow removal in hyperspectral imagery

Natural gas storage leakages cause significant energy waste, pollution, and even challenge human safety. Despite difficulties in direct detection by sensors, underground natural gas microleakages can be identified through stressed vegetation based on hyperspectral remote sensing. However, previous models often require substantial computing power due to complex network, while existing index-based models cause inaccurate identifications. Moreover, shadow effects, which can obscure spectral changes and cause misdetection, were not fully considered in previous models. Therefore, this study proposed a baseline slope index model considering shadow removal (BLSI-SR) to detect natural gas microleakage-stressed vegetation, which consists of two steps: (1) shadow removal. The endmembers were extracted from the sunlit regions in an unsupervised way, and input into the triple shadow multilinear mixing (triple-SMLM) model for vegetation canopy shadow removal. (2) Stressed range extraction. A spectral index called baseline slope index (BLSI), derived from the baseline slope between near-infrared and green bands, was used to extract the stressed range and locate the gas leakage. The results indicated that compared to other methods, the BLSI-SR achieves 100 % Precision and Recall with the best mean absolute location error of 6.52 cm. This study has the potential for enhancing leakage monitoring in underground natural gas storage.