Block feature usage in new construction land change detection supplemented by land use data

Abstract

Aimed at solving the problems existent in the change detection of new construction land based on high- resolution optical images and land use data,this paper summed up the workflow of new construction land change detection supplemented by land use,analyzed the potential utility of spectra vegetation index,texture,gradient, and variance for land block in detecting new construction land,followed by a comparison of feature uses of unsupervised detection. Tests were carried out on 2 SPOT5 imageries covering rural- urban fringe areas located in Changping District of Beijing,with the utilization of 4 groups of features,which were spectra and vegetation index, spectra,vegetation index and texture,spectra,vegetation index and gradient,and spectra,vegetation index and variance. The results show that the performance using spectra and vegetation index feature is the best,and the addition of the other features would reduce the detection accuracy. Therefore,in order to achieve a good result in practical application,we should mainly use the result of spectra and vegetation index feature,aided by other re- sults.