Abstract:
The spectral reflectance curves of 47 kinds of common tree species in Hangzhou city were choosed as data source, used correlation analysis method divided channel interval and spectrum resampling formula to get new spectral curve. The new spectral curves were classified by selforganizing competitive neural network method. Then the rationality of channel division was tested by the classification results of original spectral curve and the new spectral curves. The results showed that, in the 23 total spectrum channels, 11 of them were more narrow, which could be better identify the spectral information of plants, and which were mostly in the red and nearinfrared bands. The new spectrum based on channel division retained differences between species with less data, and achieved the dimensionality reduction of date. It provided a kind of reference for Hyperspectral sensing data.