A Modification of Weighted k-Nearest Neighbor and Its Application to Land-Cover Classification with Remote-Sensing Image

摘要

In this paper, we propose a land-cover classification method based on a modified weighted k-nearest neighbor (MWkNN) method, which uses feature and observation space information. The present method performs a good classification by employing a pair of the spatial coordinate in the observation space and its corresponding feature vector in the feature space. The classification performance of the proposed method is compared with the conventional normalized difference vegetarian index, multiple density slicing and nearest neighbor-based methods. The effectiveness and validity of the proposed method are confirmed by applying it to the real remote sensing images.