Abstract: According to the needs of investigation the illicit cultivation of opium poppy, the spectral reflectance of wheat, opium poppy, corn poppy and green bristlegrass in the wavelength range of 450～1000 nm was measured using a fiber optical spectrometer in the laboratory. Using the stepwise discriminant analysis method of SPSS statistical software, eight characteristic wavelength for distinguishing these four species of plants were obtained. The discriminant model is established with the selected characteristic wavelength point. the discriminant model takes poppy recognition error rate and the non-poppy recognition error rate as evaluation index. The variation of the recognition effect with the number of characteristic wavelength points and characteristic wavelength points was studied. The results showed that the wavelengths near the “Red-edge” have an important influence on the identification of poppy and non-poppy. Using two combinations of the four characteristic wavelengths of 684.0, 706.4, 725.2, 919.2 nm or 684.0, 694.3, 706.4, and 725.2 nm, the poppy recognition error rate and non-poppy recognition error rate can be reduced to 0%.

Key words: remote sensing, opium poppy identification, spectral analysis, stepwise discriminant analysis