Study On The Relationship Between The Amount Of Tree Leaf Dust And Reflectance Spectrum And Image Features

With the intensified impact of human, activities on the environment, fine particle groups increase in size, complexity and damage. Among them, the tiny particles or dust are closely related to haze, having negative influence not only on human health but also on vegetation reflectance spectra and the parameters of image. This research mainly determinated the characteristics of leaf reflectance spectra and images in the condition of different foliar dust quantity for 5 kinds of trees. Then the relationship model between the foliar dust content and spectral reflectance and image parameters were established respectively, influence factors on the leaf reflectance spectra and image parameters were discussed on the basis of model, which mainly includes the determination of amount for each kind of leaf dust, obtaining reflectance spectral band, leaf image segmentation and extraction of image parameters. Finally, the degree to which the foliar dust amount affected plant leaf reflectance spectra and image were discussed from the change of spectral characteristic band, now much the image parameters change before and after the dust remove, the changes between the leaves of trees. The purpose of this paper is to provide technical reference for the prevention and treatment of haze weather.According to the above research contents, the research results and conclusions are as follows:(1) Collected 120 of Euonymus japonicus Thunb, obtain the leaf spectral reflectance by using the portable near infrared spectrometer, measured foliar dust by high accuracy analysis of 1/10000 balance, analysis of the differences of leaf reflectance characteristics factor of before and after dust remove, establish the relationship model between the foliar dust content and leaf spectral reflectance. The model can predict the foliar dust amount of Euonymus japonicus by the reflectance spectra.(2) According to the data changes of leaf images parameters for Euonymus japonicus with and without dust, we can find out that there are different degrees of change for leaf image parameters in the condition of before and after dust remove, the logarithmic model being established by the change data of image parameters and spectral data is better than linear model that established by the same parameters. The model being established by mage parameters of hue (H) is better than the rest of image parameters, in a certain range of the dust amount, there is a negative correlation between foliar dust amount and image parameter H value.(3) Collection 80 Populus simonii Carr, Sophora japonica, Euonymus japonicus, Ginkgo biloba. Magnolia denudata leaf samples respectively, determination the amount of foliar dust, acquisition the spectral reflectance data, established the number relationship between spectral reflectance and foliar dust amount using different regression analysis model; although the general trend of leaf spectral curve with before and after dust remove is almost the same for 5 kinds of leaves, some differences are exist; except the Populus simonii Carr, using red edge parameters inversion the amount of foliar dust have a relatively good results for the other leaves. Thus, red edge parameters can predict the Sophora japonica, Euonymus japonicus, Ginkgo biloba, Magnolia denudata foliar dust amount in a certain precision range, the fitting accuracy of Partial least squares regression model (PLS) is the highest, the lower is the linear regression model, through hyperspectral data using partial least squares regression model can predict the amount of foliar dust.(4) Exploring the relationship between the amount of foliar dust and their image parameters for different tree species, segmenting the image of 5 kinds of leaves, establishing and fitting model, finding out that using appropriate image segmentation method segmenting image segmentation, getting a relatively good segmentation effect on the whole; In several image parameters in the test, taking image parameter H as the independent variable have a higher coefficient of determination, the results of fitting the relationship between foliar dust amount and H for Euonymus japonicus Thunb using the logarithmic model is the best, which indicate that we can predict the amount of foliar dust by using leaf image parameter H value.The model of this paper can be used in the estimation of foliar dust amount for main greening tree species in Beijing city, which can provide a reference for the treatment of environmental pollution.