Study on Spectral Features of Ginkgo Biloba Leaves Based on Foliar Dust Content.

In this study, we collected 120 samples of Ginkgo biloba leaves along the main road network: two, three, four, fifth Ring Road in Beijing to study the relationship between the spectral reflectance and foliar dust. Through electronic balance and analytical spectral devices field spec pro (ASD) analysis, information of dust content and reflectance spectroscopy of the dust leaves and the clean leaves was obtained. For the sake of seeking variation characteristics of spectral region of Ginkgo biloba leaves, spectral reflectance data was analyzed. The spectral reflectance, which had a high correlation with foliar dust content, was screened out. After that, in order to explore the relationship between foliar dust content and spectral reflectance, using traditional regression method and partial least squares (PLS) regression method established the model respectively, the effects of different foliar dust content on plant spectral characteristics were elaborated. The experiment showed that there was a negative correlation between foliar dust content and leaf spectral reflectance characteristics on the leaf spectrum curves of the clean and dust leaves, in the range of 300-710 nm; the spectral reflectance of dust was greater than the clean leaf in the range of 710-850 nm. In the study of the combination relationship between foliar dust content and leaf spectral bands, we could find out that there was a relatively good relevance between the area of red edge and foliar dust content, and the same as normalized difference index. The result of prediction of foliar dust content of Ginkgo biloba leaves was accurate based on visible and nearinfrared wavelengths. In the research of inversion of leaf surface dust content, following the traditional linear regression model and principal component regression model, the precision of inversion of foliar dust content could be improved by partial least squares (PLS) in a certain extent. Therefore, we concluded that the effect of partial least squares inversion was the best among the 3 kinds of regression model.