The design of intelligent expert classifier for featured crop mapping combining spectral library

In the study, a module was designed which was used to estimate yield of crop (such as lychee, banana and sugarcane etc) using spectral library of featured crops of South China. An approach combining spectral library with expert system classification methods by spatial data mining techniques is present. Being one of spatial data mining techniques, inductive learning algorithm is used to discover knowledge of spectral library for the expert system and is designed specially for monitoring featured crops of South China. So the intelligent expert classifier can make use of spectral data, attribute data and spatial data of spectral library to extract information of South China´s featured crops. The study pre-defined some attributes of inductive learning algorithm which are in favor of featured crops mapping, for improving the efficiency of algorithm and ensuring the usability of rules. The estimation of lychee planting area of Shenzhen city in 2005 was presented as a case study. The following classification rules of lychee were acquired by running inductive learning algorithm: for example, 0.062< reflectance of TM2 <0.071, 0.47<NDVI<0.54, 20<DEM<90, special rules of lychee planting in Guangdong province, 40<area of lychee orchard (pixel number) <5000 and the result of classification was presented by figure 2. Compared with traditional unsupervised classification, it improves classification accuracy greatly, and the rate of accuracy reaches 93.5% with KAPPA coefficient of 0.85. The result indicates that intelligent expert classifier is a better classification method for extracting of lychee planting area and is able to meet the need in agriculture application for quick crop area monitoring in South China.