Coral reef classification using drone screening in Teluk Segadas, Pulau Pangkor

Coral reefs are complex and diverse marine ecosystems that provide numerous benefits to the environment and to humans. However, they face multiple threats that jeopardise their well-being, necessitating effective conservation efforts. Traditional methods for assessing coral reefs are time-consuming and resource-intensive, limiting the scale and frequency of data collection. This study aims to assess the potential of drone technology for studying coral reefs in Teluk Segadas, Pulau Pangkor, where spatial information on benthic habitats is limited. The primary objectives are to determine the presence of coral reefs, and the density of coral cover in the study area. A total of 330 aerial images of Teluk Segadas were taken using a DJI Pro Mavic 2. The aerial photos were combined into a single orthophoto image. Later, segmentation and classification were performed using Structure-from-Motion (SfM) and a supervised classification technique, K-Nearest Neighbour (KNN). The orthomosaic image of Teluk Segadas is presented in this study and five classes were identified, that is live coral, dead coral, sediment, coral rubble and rock. Overall, the shallow water in Teluk Segadas was dominated by live coral (38947.38 m2) along with sediment (9273.9 m2). The rest of the area was covered by dead coral (4946.08 m2), rubble (3709.56 m2) and rock (1854.78 m2). Furthermore, coral coverage of Teluk Segadas, Pulau Pangkor was 66% and was dominated by massive coral. The overall accuracy was 73% with producer accuracy (PA) and user accuracy (UA) values ranging from 60-80% and the misclassification rate ranging between 20-30%. This study demonstrates that images captured by drone in any environment setting can be processed, classified, and assessed for accuracy. In addition, this study provides a different perspective in understanding coral reef well-being, and aids in monitoring and management efforts.