Discrimination of Land Use Patterns in Remote Sensing Image Data using Minimum Distance Algorithm and Watershed Algorithm

This paper describes a new method for extracting the agricultural land use pattern around the Miyajimanuma inland wetland based on remote sensing imagery. A land parcel and particle swarm optimization (PSO) K-means-based minimum distance classification (MDC) (LP-PSOK-MDC) method was developed. This method includes three steps: 1) considering the diversity of crop planting and growth state, a training sample pre-classification-based MDC method was developed; 2) the land parcels information was extracted by using watershed transform algorithm; finally, 3) pixels in the same land parcel were re-classified. Results of the study suggest that using this method the classification result was easily up to 96 %, much better than results obtained by using traditional supervised classification methods such as MDC and unsupervised classification method.