Land cover classification has long been primarily focused on automated image analysis applications and there is ongoing search for new classifiers that can yield improvements in results. This study shows the method of combining unsupervised classification and Artificial Neural Network (ANN) to the land cover classification of whole China and the time series National Oceanic and Atmospheric Administration (NOAA) advanced very high resolution radiometer (AVHRR) 1-kilometer (km) data is used. Some factors related to the effect on accuracy of land cover classification are discussed. The research involves the following steps: (1) Production of monthly maximum normalized difference vegetation index (NDVI). (2) Land cover classification system of China is proposed. (3) Unsupervised clustering of monthly NDVI data using ISOCLASS algorithm. (4) The preliminary identifying with the addition of digital elevation, ecoregions data and other land cover/vegetation reference data and extraction of the training data. (5) Land cover classification of China using Neural network. The results indicate that the accuracy of classification is much improved comparing with the common classification method.

Date Published: 19 August 1998
PDF: 8 pages
Proc. SPIE 3504, Optical Remote Sensing for Industry and Environmental Monitoring, (19 August 1998); doi: 10.1117/12.319548

Published in SPIE Proceedings Vol. 3504:
Optical Remote Sensing for Industry and Environmental Monitoring
Upendra N. Singh; Huanling Hu; Gengchen Wang, Editor(s)