Over the past several decades, moderate-resolution remote sensing data have been used to obtain regional-scale land-cover information. Nowadays Change detection has been widely used in many applications. Change detection techniques may capture the changes of phenology information which are quite useful to distinguish a certain kind of crop from other land-cover types. In order to extract single late rice through phenology characteristic, a change detection model based on neighborhood correlation images (NCIs) and supervised classification was proposed.
In this study, Landsat 8 OLIdigital images acquired on July 19, 2013 and November 8, 2013 respectively were used to extract single late rice. Single late rice was chosen as the study object because of its unique phenology information changes, from steeping field to vegetation. The method proposed includes 5 parts: (1) Preprocessing data, (2) Creating NCI images, (3) Collecting training/test samples, (4) Extracting single late rice and (5) Classifying images and assessing accuracy. The overall accuracy of single late rice was 90.3%, indicating that the change detection method using NCIs is an effective way to extract crops with useful phenology information.