It is a challenge to timely detecte microleakage of natural gas which is stored in underground repository or pipeline. The response and other remote sensing characteristics of stressed vegetation were used to indirectly detect the microleakage point of natural gas via controlled experiments in the field. In detail, the canopy reflectance of stressed area and control area of soybean and grassland were measured respectively. Singular values were removed and spectrum was smoothed, then spectra of the canopy reflectance was analyzed using the method of continuous wavelet based on first-derivative, which showed that wavelet energy coefficients of stressed and control canopy reflectance at 685 and 715 nm were good features to separate the stressed and control groups. (DW685－DW715)/(DW685+DW715) (DW) was designed in this paper using 685 and 715 nm and compared to PRI, NPCI, NDVI, and D725/D702, which showed that better performance, universality and robustness were possessed by DW in identifying the stressed grassland and soybean. The results showed that it is feasible to indirectly detect natural gas microleakage points through hyperspectral technology, which can provide technical support and theoretical basis for future engineering applications.