Study on Intelligent Diagnosis Method for Corn Nitrogen Status Estimation Based on Hyperspectral Data

摘要

Nitrogen nutrition index (NNI),a good indicator of the degree of nitrogen nutrition deficit of crop,is of special significance to precision agriculture.We develop a new method of corn nitrogen status estimation based on the three-layer dynamic structure of fuzzy neural network,which firstly acquires the sensitive spectral parameters related to leaf N content as a part of input variables by analyzing the correlation between leaf N content and canopy spectral parameters studied by the predecessors and comparing the effects of kinds of spectral parameters estimation on leaf N content.All together,influence factors (CWSI,heavy metal content (HMC) and soil pH value) as input variables of other observation data also and then NNI as output variable.The ultimate structure of network was 36-45-9 after training.The results showed that it had a high degree of fitting between the trained data and the measured data and network fitting accuracy reached 95％.Comparing to the statistical analysis methods just with sensitive spectral parameters related to leaf N content,the results got from two methods have good linearity with the measured data,but the study found that the results by statistical analysis had a big deviation with the actual data in the later filling stage,it was clearly unreasonable.This paper gets consideration to remote sensing data and other observation data affecting crop N to overcome the deviation problems of the statistical analysis.