Statistical, Structural, Hybrid, and Graph Theoretical Features to Measure Land Development

Abstract

Extracting information on a developing region from its sequential satellite images has many benefits. Therefore, in a previous study, we introduced graph theoretical and conditional statistical features to measure land development in a predefined region. There, we only used the grayscale information from the satellite image at hand. Here, we extend that work by introducing novel statistical, hybrid, and graph theoretical features using multispectral information. We also introduce novel structural features based on three different structure extraction methods. We test our new features on a diverse data set and report their performances in measuring land development.