Extraction of Natural Gas Microleakage Stress Regions Based on Hyperspectral Images of Winter Wheat
Natural gas has gradually occupied an important position in the energy structure.As natural gas pipelines and gas storage are buried underground all year round,oxygen-free corrosion,natural disasters,looseness of injection wells and pipelines,and other factors will lead to gas leakage.So,it is necessary to determine the location of leakage points and make early judgments and warnings before large-scale leakage from underground natural gas storage.This paper collected four periods of hyperspectral image data of winter wheat.It integrated the spatial-temporal-spectral features of hyperspectral data to explore the relationship between the radius and duration of winter wheat stress under natural gas stress,thus indirectly detecting the microleakage point of natural gas.On the one hand,the index CWTmexh(CW Tmexh=CW2770/(1-CW487)× CW550),constructed by continuous wavelet transform of the canopy spectra after continuum removal,was used to classify pixels into non-stress and stress with threshold segmentation.On the other hand,PCA features of hyperspectral image data are extracted,and natural gas stress regions are identified with the SVM classifier.Finally,the results of both threshold segmentation and SVM classification are analyzed by mathematical morphology,and the stress area is fitted with a circular curve using the least square to explore the relationship between the stress radius of natural gas leakage and the stress days.The results show that:(1)The CWTmexh index can be applied to imaging hyperspectral data,showing good recognition performance;(2)SVM classifier can recognize winter wheat stress areas based on spectral difference characteristics with good classification accuracy(i.e.,the maximum classification accuracy of 99.25％and kappa coefficient is 0.97)and the recognition accuracy increases with the continuation of natural gas stress;(3)There is a strong linear correlation between the radius of the stressed area and the ventilation days of winter wheat.Results of this study showed that it is feasible to indirectly identify natural gas micro-leakage points through hyperspectral remote sensing by monitoring surface vegetation at the canopy and low altitude scales and can predict time-dependent changes associated with underground natural gas micro leakage stress.The results can provide a theoretical basis for monitoring the leakage points of underground natural gas storage by spaceborne hyperspectral remote sensing and provide technical support for future engineering applications.
Natural gas micro-leakageWinter wheatHyperspectral imageSpatial characteristics
万方数据
维普
