Wavebands Selection for Rice Information Extraction Based on Spectral Bands Inter-Correlation
Wang Fu-min,Huang Jing-feng,Xu Jun-feng,Wang Xiu-zhen
DOI: https://doi.org/10.3964/j.issn.1000-0593.2008.05.032
2008-01-01
Spectroscopy and spectral analysis
Abstract:The hyperspectral remote sensing data usually involve hundreds or even thousands of narrow bands, which may be crucial for providing additional information with significant improvements over broad bands in quantifying biophysical and biochemical variables of agricultural crop. However, the huge data generated by hyperspectral systems, and the problems this presents for storage and analysis, have far prevented the routine use of such data. The objective of the present research was to identify the spectral bands in the visible and near-infrared range that were suitable for the study of rice. The hyperspectral reflectance of canopy in different development stages was measured in experimental field using a 1 nm-wide spectroradiometer but was aggregated to 10 nm-wide bandwidths to match the first spaceborne hyperspectral sensor, Hyperion. The correlation coefficients(r) between all the combinations of spectral bands were computed, and then they were converted to R2 , which constituted R2 matrices. The matrices were plotted against wavebands. The criterion of band selection is that the lower the R2 value, the less the redundancy between two wavebands while the higher R2 indicates that there is redundant information between two wavebands. According to the criterion, the wavebands corresponding to the first 100 minimum R2 values were selected from all canopy spectra collected on different dates. And then these bands were analyzed. The results indicate that the visible and infrared (NIR and SWIR) themselves contain redundant information. The wavebands containing abundant information of rice are located in specific bands in the longer wavelength portion of the visible region, with secondary clusters in red edge region, in strongly reflective near-infrared region with relatively higher reflectance, in one particular section of short wave near-infrared (SWIR) (1 530 nm) and in the second maximum reflectance region of SWIR (2 215 nm). Compared with the selected bands with other vegetation, rice seems to have three spectral regions of 400-410 nm, 630-650 nm and 1 520-1 540 nm, which exclusively depict the characteristics of rice. Moreover, this research identified 17 spectral bands in the visible and near-infrared region, which were 405, 565, 585, 605, 620, 640, 660, 680, 695, 705, 720, 740, 865, 910, 1 085, 1 530 and 2 215 nm. These bands contain the majority of the rice information content. A reduction in band number without significant information loss is important because it makes it possible to achieve fine spatial resolution without sacrificing the ability to characterize rice status.