OBJECT-ORIENTED CLASSIFICATION FOR ECOLOGICALLY SOUND LAND BASED ON HIGH-RESOLUTION IMAGES

Fine and fast classification for ecologically sound land use types is the perquisite for ecosystem management and protection. However, few studies focused on the fine classification of ecologically sound land use especially using high-resolution remote sensing data. Therefore, this paper is to test the feasibility and applicability of classifying ecologically sound land using object-oriented classification method based on high-resolution remote sensing images, Pleiades-1 satellite images in Dafeng County, Jiangsu Province. Results showed that the overall classification accuracy was 87.43%, the users' accuracies and producers' accuracies of classification results of seven ecologically sound land types were higher than 80.00%. The classification result was in agreement with the field survey results. In addition, the experiment also indicated the segmentation result based on multi-threshold algorithm was closer to the real features in the shape, size, figure and texture comparing with that based on quadtree or chessboard algorithm. Different land use types can easily and efficiently extracted based on their difference in features using the assign classification method. Our research provides data support for evaluating ecologically sound land and it also provides a relatively rapid, accurate and practical way for finely classifying ecologically sound land.