This study concerns land use classification system using Landsat TM data, which consists of following two hierarchical classification processes. The first is land cover classification using TM band data and the second is reclassification of land cover data into land use data using ancillary data.
Several methods, including maximum likelihood classifier and linear discriminant function method, have been used for Landsat digital data classification. However these methods have some problems in the theory and assumption of classification technique as well as computing time for classification.
In this study land cover classification method based on probit model is built. Main feature of probit model is that it can consider the difference in variance and covariance of band distribution between land cover categories which are not considered by linear discriminant function method. Through the applicability test of probit model, it was verified that it could achieve classification in very short computing time as accurately as by maximum likelihood classifier.
The method of converting the land cover data into land use data is also discussed. Binary tree method using zoning data as ancillary data is applied in this case.
View full abstract