Urban green vegetation stress conditions diagnosis based on hyperspectral database — A case study of Xuzhou

In order to provide scientific support for the management of the urban green vegetation, taking Xuzhou city as an example, this paper proposed a method to diagnose stress conditions of vegetation rapidly by using a green vegetation spectral database. Under laboratory conditions, 303 samples of leaf reflectivity spectra taken from 25 kinds of green vegetation in urban areas were acquired by the means of AvaSpec-2048×14-USB2 spectrometer integrating sphere reflectivity measurement. When collecting spectral data, the visual observation and other traditional detection methods were used to diagnose the stress conditions of each vegetation sample, which could be divided into four levels including normal, mild stress, moderate stress and severe stress. After these procedures, a spectral database which can diagnose stress conditions of green vegetation in Xuzhou city rapidly was established on the platform of the software Environment for Visualizing Images (ENVI). To verify the ability of the database to diagnose stress conditions of green vegetation, 113 unknown test samples were introduced into the database. Firstly, spectral data were preprocessed by the methods of smoothing in order to eliminate the influence of background information. Secondly, on the basis of analyzing spectral feature of green vegetation in different stress conditions, the methods of spectral matching analysis in database, including Spectral Feature Fitting, Spectral Angle Mapper and Comprehensive Matching were used for matching analysis to diagnose the stress levels of 113 unknown vegetation samples. And then the matching accuracy which based on the traditional detection methods was evaluated. The results showed that the feature band which is capable of diagnosing the stress conditions of green vegetation mainly focused on the areas of green peak, redabsorption band and red edge in the reflection spectra curves. The matching accuracy of Spectral Feature Fitting, Spectral Angle Mapper and Co- prehensive Matching reached 78.5%, 75.6% and 83.4%, respectively. The result indicates that it is feasible to diagnose the stress conditions of green vegetation using the method of spectral matching, and this method is expected to be a supplementary and alternative of traditional detection methods.