Scholarly record

LAND COVER CHANGE DETECTION BASED ON THE FALSE COLOR COMPOSITE OF NDBSI DERIVATIVES

Xiaoyang Liu, Ying Sun

Abstract

We proposed an approach to detected land cover change based on the false color composite of the combination of bare soil index and building index (NDBSI) derivatives. NDBSI is an index that can distinguish building and vegetation well. In this paper, we first obtained a false color composite, named RGB-NDBSI, based on the NDBSI image of previous phase, NDBSI image of latter phase and NDBSI different image between the two phases. Based on the data generated, we identified the land cover change by BP neural network classification method. We applied the developed method to Landsat images in 2011 and 2020 covering the area of Guangzhou, China. According to the urban development, five types of land cover change were concerned: “unchanged urban land”, “urban land to Vegetation”, “Vegetation to urban land”, “unchanged Vegetation” and “unchanged Water”. RGB-NDBSI image showed distinct color feature among these change types in the form of light cyan, navy blue, bright yellow, earthy red and khaki. For comparison, NDBSI difference images and RGB-NDBSI images were both classified by another two methods, i.e., K-means clustering, object-oriented classification. The combination of RGB-NDBSI image and BP neural network achieved the best accuracy, with the overall accuracy of about 96% and kappa coefficient 0.93. The experimental results show that the RGB-NDBSI image is an effective solution for change detection

Publication Impact Profile

• Citations
• Scopus - Citation Indexes: 1

• Captures
• Mendeley - Readers: 3

PLUMX see details

DOI resolver: doi.org/10.5593/sgem2021/2.1/s10.69

Publication details

Back to publication list