A field-based crop and land use map over Sanjiang plain in Heilongjiang derived from multi-temporal imagery of Landsat7-ETM+

Abstract

Multitemporal Landsat7-ETM+ imagery and supplementary field survey data was used to establish a crop and land use map for an area of ±20 000 km² of Sanjiang Plain in China’s north-eastern province Heilongjiang. The “field-based” map is based on data of the year 2002 and is created by combining the results of a supervised Maximum Likelihood classification with the estimated field boundaries extracted by an automatic segmentation procedure. No sufficient ground data was available to perform an independent validation, and some classes are certainly uncertain. Nevertheless, given the minimal investments and the unexplored potentials for improvement, the approach seems very promising and cost-effective. By superimposing the classification with the map of the administrative counties, regional statistics were derived with the acreage distributions of the concerned crops and land use types.

Keywords

• segmentation
• post-classification
• per-field filter
• CGMS
• maximum likelihood classification