Technological Innovation for Agricultural Statistics by Asian Development Bank;

Author:Asian Development Bank;
Language: eng
Format: epub
Publisher: Asian Development Bank Institute
Published: 2018-08-15T00:00:00+00:00

Land cover classification

To identify paddy rice areas from other types of land cover, ADB researchers classified the fused satellite data covering Thai Binh into six categories using the International Geosphere-Biosphere Programme classification scheme (Friedl et al. 2002) and a random forest classifier algorithm (Breiman 2001). The principle behind this kind of classification can be explained as a three-step process:

Step 1: Sample pixels from the fused satellite images were selected as evenly as possible across the spatial extent of Thai Binh. The land cover of these pixels were validated through ground observations and visual interpretation of high-resolution images such as those derived from Google Earth.

Step 2: Once a sufficient number of training pixels were obtained for each of the six categories, the threshold values for each of the land cover categories were established and applied to the whole study area using a machine learning algorithm. This generated a first prototype for a land-cover classification map.

Step 3: A random number of pixels were again selected across the six categories (different from the training pixels) and again verified through ground observations or high-resolution satellite data to construct various classification accuracy statistics. This study exhibited close to 91% accuracy for paddy rice areas.

The classification map created for Thai Binh province is shown in Figure 8.

Figure 8: Classified Land Cover Map for Thai Binh Province, Viet Nam

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.