Quasi-circular Vegetation Patch Mapping with Multitemporal Kauth-Thomas Transformation of the mIHS Pansharpened GF-2 Images

The quasi-circular vegetation patches (QVPs) are better objects for studying the ecosystem evolution, functioning, and maintenance in the Yellow River Delta, China. Remote sensing with linear change detection technique is an effective approach for mapping the vegetation dynamics. In this paper, the multitemporal Kauth-Thomas transformation (MKT) change detection technique with the decision tree classifier was used to map the QVPs based on the modified intensity-hue-saturation pansharpened April and August Gaofen-2 images. Results indicated that mapping the QVPs could be performed well using the approaches used in this paper. The precision, recall rate, and F-measure were 66.7%, 52.9%, and 59.0%, respectively. In the future, patch splitting techniques and more test areas should be tested for improving the detection accuracy of the QVPs. In addition, an assessment on the possibility of change in brightness and greenness between multitemporal images for mapping the dominant communities of the QVPs should be performed, which were important for the establishment, evolution, and disappearance of the QVPs.