Accurate Interpretation of Urban Mangrove Stand Structure Based on Low-altitude Unmanned Aerial Vehicle System

Abstract

The effective management of the mangrove forests is especially important, which requires precise and timely monitor of the stand structure of mangrove forest. Meanwhile, stand structure is a key feature that indicates the health status of mangrove forests. Therefore, we developed an effective method to interpret the stand structure of urban mangrove forests using visible light remote sensing images from Unmanned Aerial Vehicles (UAV). The results showed that: (1) Pixel-based deep natural networks under the constraint of physical parameters (tree height) have been explored for mangrove species classification, and the overall classification accuracy of the whole study site was 88.29%. (2) Based on watershed algorithm and the relationship between tree height and tree crown, single tree crown segmentation in non-closed mangrove area was realized, with recognition accuracy of 78.57%~ 88.89%. (3) Based on the obtained stand structure parameters, including the size ratio, angle scale, mixing degree, etc., a stand structure health evaluation model was established to determine the health level of a mangrove forest. Stand structure evaluation score (A) in Futian Mangrove National Nature Reserve was derived that the core area (A: 0.9315~0.9391) was better than the non-core area (A: 0.8622~0.9127). The proposed method will reduce the time and effort required for future mangrove plant observations and contribute to future assessments of the health of mangrove forests.