Spectral Features Analysis of Pinus massoniana with Pest of Dendrolimus punctatus Walker and Levels Detection
XU Zhang-hua1,2, LIU Jian1,2,3*, YU Kun-yong1, 2, GONG Cong-hong1, 2, XIE Wan-jun1, 2, TANG Meng-ya1, 2, LAI Ri-wen1, 2, LI Zeng-lu1, 3
1. Institute of Geomatics Application, Fujian Agriculture and Forestry University, Fuzhou 350002, China 2. College of Forestry, Fujian Agriculture and Forestry University, Fuzhou 350002, China 3. Sanming University, Sanming 365000, China
Abstract：Taking 51 field measured hyperspectral data with different pest levels in Yanping, Fujian Province as objects, the spectral reflectance and first derivative features of 4 levels of healthy, mild, moderate and severe insect pest were analyzed. On the basis of 7 detecting parameters construction, the pest level detecting models were built. The results showed that (1) the spectral reflectance of Pinus massoniana with pests were significantly lower than that of healthy state, and the higher the pest level, the lower the reflectance; (2) with the increase in pest level, the spectral reflectance curves’ “green peak” and “red valley” of Pinus massoniana gradually disappeared, and the red edge was leveled; (3) the pest led to spectral “green peak” red shift, red edge position blue shift, but the changes in “red valley” and near-infrared position were complicated; (4) CARI, RES, REA and REDVI were highly relevant to pest levels, and the correlations between REP, RERVI, RENDVI and pest level were weak; (5) the multiple linear regression model with the variables of the 7 detection parameters could effectively detect the pest levels of Dendrolimus punctatus Walker, with both the estimation rate and accuracy above 0.85.
