Title :
Hyperspectral characteristics analysis of Aconitum leucostomum Worosch in the grassland: Taking Ily as an example
Author :
Amin Wen ; Jianghua Zheng ; Xiulan Wu ; Junwei Xuan ; Chen Mu ; Wei Wang
Author_Institution :
Sch. of Resources & Environ., Xinjiang Univ., Urumqi, China
Abstract :
Poisonous grasses disrupt landscape patterns and compromise the functionality of ecosystem process. Native Aconitum leucostomum Worosch poses significant threats to native vegetation and animal husbandry in the Tuohulasu grassland in Ily, XinJiang, and quantifying hyperspectral characteristics is vital for monitoring its spread. In this paper, field reflectance spectra were collected from two plant species (Aconitum leucostomum Worosch and mixed grasses). The original curves were preprocessed by Savitzky-Golay smoothing filter, spectral derivative analysis and continuum removal logarithm. Then the hyperspectral characteristics of the vegetation were extracted and analyzed. The method of Mahalanobis Distance was used to identify the most hypersensitive bands, then using stepwise discriminant analysis of the spectrum to identify the best band for recognizing and distinguishing Aconitum leucostomum Worosch from mixed grass in the study area, and calculating the identification accuracy. The results show that: 1) We could concluded from the reflectance spectral curves after spectral characteristics analysis that although the reflectance spectrum of the two dominating grass types have similar spectral curves, there were effective bands for distinguishing them, Aconitum leucostomum Worosch and mixed grasses had significant differences between 750 and 1150nm, and spectral curves of Aconitum leucostomum Worosch living in different growth status had differences too; 2) Through the spectral analysis, we found that there were obvious "Twin Peaks” in the spectral curves of Aconitum leucostomum Worosch and the mixed grasses only had one peak. The red edge parameters appeared Aconitum leucostomum Worosch\´s > mixed grasses; 3) this paper obtained the best hypersensitive and recognized bands, which were 676~695nm and 704~736nm. In conclusion, this paper was designed to provide initial theoretical foundation for growth condition and physiological parameters of Aconitum - eucostomum Worosch, and played a vital and significant role in making theoretical groundwork for the distribution and monitoring of Aconitum leucostomum Worosch in future.
Keywords :
geophysical signal processing; hyperspectral imaging; spectral analysis; vegetation mapping; Aconitum leucostomum Worosch; China; Ily; Mahalanobis distance; Savitzky-Golay smoothing filter; Tuohulasu grassland; XinJiang; animal husbandry; ecosystem process functionality; field reflectance spectra; growth status; hypersensitive bands; hyperspectral characteristics analysis; hyperspectral characteristics quantification; identification accuracy; landscape patterns; native vegetation; plant growth condition; plant physiological parameters; poisonous grasses; red edge parameters; reflectance spectral curves; stepwise discriminant analysis; Absorption; Accuracy; Hyperspectral imaging; Monitoring; Reflectivity; Vegetation mapping; Hyperspectral; Poisonous grasses; Spectral characteristics; Xinjiang;
Conference_Titel :
Agro-geoinformatics (Agro-geoinformatics 2014), Third International Conference on
Conference_Location :
Beijing
DOI :
10.1109/Agro-Geoinformatics.2014.6910650