Abstract: Preprocessing was made on remote sensing image by Hyperion of Lin’an, Yuhang of Hangzhou and Anji of Zhejiang province. 5 species and 2 groups were selected, including Phyllostachys edulis, Ph. violascens ‘Prevernalis’, Ph. heteroclada, Cunninghamia lanceolata, Pinus massoniana, evergreen broad-leaved forest (Ilex chinensis, Quercus glauca, Photinia serrulata, Camellia sinensis) and deciduous broad-leaved one (Carya cathayensis, Castanea mollissima, Quercus fabri, Liquidambar formosana, Morus alba, Amygdalus persica), to be classified by envelope removal. Non-forest pixel luminance value was classified by dimensionality reduction and training. Positional accuracy evaluation was used on classification accuracy of original and treated image. The results showed that the total classification accuracy and total Kappa coefficient of envelope removal method was 90.5% and 0.86 respectively, while that of original image classification was 80.2% and 0.78 respectively.