Extraction of mangrove in Hainan Dongzhai Harbor based on CART decision tree

Mangroves play an important part of coastal ecosystem. However, in recent years, many mangroves were damaged. Therefore, the monitoring of mangrove forests timely and accurately becomes of up. This paper selects the Northeast Hainan Dongzhai Harbor Mangrove Wetland as the study area, based on OLI images through the image spectral information, vegetation indices, and texture and texture auxiliary information. The CART decision tree model which analyzed the training data from the test variables and objective variables to constitute a stable Binary Tree form, and then finally extracted the mangrove. We employed maximum likelihood classification to compare the classification results using the same sample points. The results show that: the overall accuracy and Kappa from the CART were both higher than the maximum likelihood, with the total accuracy 82.06%, 9.56% higher than the latter; the Kappa 0.7713, 0.0886 higher than the latter, illustrating that the extraction of mangrove was feasible through the CART.