Extracting rural impervious surface from LANDSAT 8 OLI imagery using K-Nearest neighbor algorithm
Khoa Trắc địa - Bản đồ và Quản lý đất đai, Trường Đại học Mỏ - Địa chất, Việt Nam
- Received: 15th-Aug-2017
- Revised: 18th-Oct-2017
- Accepted: 30th-Oct-2017
- Online: 30th-Oct-2017
- Section: Geomatics and Land Administration
The Impervious surface area in rural areas are difficult to extract from satellite imagery, especially for medium-resolution images such as Landsat. There have been many studies using image classification algorithms based on the basis of Pixel-based values. However, the problems are the estimation of errors during the time of classification of each pixel. The main contribution of this study is that it utilizes of K-Nearest Neighbor (K-NN) algorithm with Landsat 8 OLI imagery to detect a rural impervious surface area in Giao Thuy district. This paper discusses the uses of K-NN rules and its error estimation for classification of each object images in the medium spatial image. In order to achieve the best accuracy using the K-NN algorithm, the standard samples need to have the following criteria: 1. the sample size is large enough, 2. the distribution of samples is in the study area, 3. Maximum separation between standard sets. Results showed that K-NN algorithm was enough accurate for practical applicability for mapping rural impervious surface areas.