Identifying ginkgo trees using spectrum and texture time series from very high resolution satellite data

The identification of ginkgo is of great importance for the intelligent use and management of ginkgo resources. A texture and spectrum time series analysis (TeSTA) model was used to identify and map ginkgo trees as the vegetation spectral and textural properties change regularly during the growing period. The texture and spectrum time series were constructed from temporal remotely sensed images and then projected into a multidimensional measurement space in which the vector difference was used to determine the threshold to identify ginkgo trees. Using this approach, ginkgo trees in Beijing’s Olympic Forest Park were extracted using six Gaofen-2 satellite images. In order to assess the accuracy of this method, two other classification methods were used for ginkgo identification. One method uses only the spectrum time series and the other method uses the support vector machine (SVM) algorithm. The results showed that the TeSTA model performed better than the SVM and spectrum time series model with lower commission errors and higher user’s accuracy. Thus, the TeSTA model can be used for identifying ginkgo trees and providing an important tool for forestry management.