Knowledge-based land use/cover classification of remote sensing image in Kenli, Shandong Province, China

Knowledge-based land use/cover classification of remote sensing image in Kenli, Shandong Province, China

Quantitative remote sensing methods have been widely adopted for land use/cover information extraction. However, in the Yellow River delta, spectral confusions between different surface features are common, which always introduce errors into the ordinary automatic spectral classification. A knowledge-based land use/cover classification is tried. Based on a systematic analysis of the spectral characteristics of different land use/cover types in the study area, several index models derived from TM digital data of Kenli County are built, which can make pixels of certain land use/cover types enhanced. With the combination of the index models, the expert knowledge of the area and some ancillary data, such as land use maps and soil maps based on ground surveys, rules of every class are determined to extract the land use/cover information. Accuracy evaluation shows that without any manual modification, the land use/cover information determined with remote sensing is 81.8% and accuracy for the spatial distribution of land use/cover cross checked with a random sample is 84.5%. Therefore, this approach provides an effective means for land use/cover information extraction in the Yellow River delta and can be used as reference for other areas where spectral confusions commonly exist.