Multi-diagnosis Space Models of As Stress in Rice Based on Hyperspectral Indices
基于高光谱指数的水稻砷污染胁迫多重判别模型

High arsenic content in rice can influence the chlorophyll,water content and structure in their leaves, reduce the rate of photosynthesis and change their spectral features. Multiple models for diagnosing As contamination in rice based on spectral parameters were studied. Sixty samples belonging to mature rice in three different areas were scanned by ASD field pro3 for optical data. Arsenic reference values were obtained by atomic absorption spectrometry. First, correlation analysis was used between 9 hyperspectral indices and As content in rice, and three indices(PSNDa,fWBI,SIPI)were extracted to diagnose As contamination in rice, which were respectively sensitive to chlorophyll, water content and structure of leaves, then took the three indices to form a diagnosis spectral indices space (PSNDa-fWBI,PSNDa-SIPI,fWBI-SIPI) of As stress in rice. Second, principal component analysis and independent component analysis were also applied in these 9 hyperspectral indices, and two principal components(F1,F2) and two independent components(ICA1,ICA2) were extracted. These four components(F1,F2, ICA1,ICA2) were all correlated with As content in rice, and composed another two diagnosis spaces (F1-F2, ICA1-ICA2) for predicting As contamination. And these spaces composed a multiple diagnosis space model which diagnosed As contamination in rice of test area from different level, and showed a good result.