Land Cover Classification With Expert Systems In Sakaerat Environmental Research Station, Thailand

The purpose of this research is to specify land cover type of Sakaerat Environmental Research Station (SERS) by applied the two classification methods, maximum like likelihood and expert classification to the Landsat5TM image which acquired on January18,2011.The expert classification systems is a hierarchy of rule that user can define the variables for describing the hypotheses or classes. The variables selected for this operation comprise of digital number of the image, digital elevation of SERS, ferrous minerals, principle component analysis, and tasseled cap transformation. The systems assign the classes under these conditions.The image was classified into seven categories:1) agricultural area,2) water body,3) grass and abandon land,4) bamboo,5) deciduous forest,6) plantation, and7) evergreen and restoration forest. The results show that the overall accuracy and overall Kappa statistic of expert classification is greater than the overall accuracy and overall Kappa statistic of maximum likelihood classification which are88.10%,0.7595,82.38%and0.6425, respectively. Besides, the producer’s and user’s accuracy of each category of the expert classification is also greater than the producer’s and user’s accuracy of the maximum likelihood classification.Therefore, land cover classified by expert classification systems would be useful in improving classification accuracy with suitable variables. A higher accuracy of classification shows a higher of map accuracy, which is more reliable for further use and study.