论文部分内容阅读
Sugarcane is the first economic plant in Guangxi Autonomous Region.Guangxi produced more than 60% of the national total output.But according to the 2012 yearbook, the yield per unit area of Guangxi was only 77.8% of the number of Guangdong province, yet the fertilizer amount per unit area was 1.23 times of that of Guangdong province.In order to give a scientific guidance for the crop management to get a higher output/input ratio, a spectroscopy method was used to predict sugarcane growth status in this study.The visible-NIR spectral features of the leaves of a sugarcane variety (ROC 22) from an experimental field was studied in this paper, the spectral reflectance, leaf chlorophyll, total nitrogen (TN), total potassium (TK), total phosphorus (TP) were measured in the lab.The correlations between the spectral reflectance and leaf chlorophyll & TN content were also analyzed, and prediction models were built eventually.The experimental field which was located in the campus of Guangxi University, was divided into 14 plots and was managed with "3414" fertilizer plan which had 4 different fertilization levels for K, P and N respectively.A total of 4 experiments were conducted in later seedling stage, tillering stage and jointing stage.The last opened leaves were sampled in each experiment, and 3 samples were randomly collected from each plot.The spectral reflectance (from 220 to 1400 nm) of each leaf sample was measured by Shimadzu UV-VIS spectrophotometer UV-2600 in the lab.In the meantime, the leaf chlorophyll contents were analyzed by extraction method, and the TN, TP, TK contents were measured by Nesslerization method, Mo-Sb colorimetric method and Flame spectrophotometry method, respectively.Analysis of the spectral data reveals significant differences at 550 nm which was the peak of visible range.The derivatives of the spectral signatures showed that the red-edge positions shifted slightly, but the slopes at these red-edge positions were all close to 1.The correlation analysis between spectral reflectance and chlorophyll content showed that they had negative correlations with R greater than 0.8 in 527-578 nm and 701-731 nm.Further, the selection of sensitive wave bands was implemented to calculate a NDVI with higher sensitivity.In this selection procedure, each one of the spectral reflectance in the red range (620-770 nm) was selected to combine with each one in the near-infrared range (780-1100 nm) to calculate NDVI, and all the correlations between each NDVI combination and chlorophyll & TN of all the samples were then analyzed.The distribution maps of the NDVI values of all the combinations were generated in a two-dimension space consisted of red and NIR bands,and then the distribution map of correlations was generated as well to choose the sensitive bands.The NDVI combination with highest correlation was used to build prediction models for chlorophyll and TN content.