Coral Reefs Detecting with Artificial Neural Network Classification and PlanetScope Imagery in Cu Lao Xanh Island, Binh Dinh province

Abstract

Abstract Coral reefs are a vital component of coastal and marine ecosystems. They are now under strong environmental challenges and are being harmed by human activities and rising sea surface temperatures, which are reducing the living coral cover. The purpose of this research was to evaluate the mapping accuracy of coral covers using PlanetScope satellite pictures with the Artificial Neural Network (ANN) method surrounding Cu Lao Xanh Island in Binh Dinh province. To adjust for the sunglint effect, the bands were corrected using the Hedley technique. After that, the Depth-Invariant Index technique was utilized to reduce the influence of the water column, and the ANN algorithm was employed for mapping. Hard coral, soft coral, seagrass, deep water, and bare bottom were identified as the five kinds of benthic habitat. The accuracy of the classification results was assessed using field data collected on May 10 and 11, 2022. The results indicated that the artificial neural network (ANN) technique had a higher accuracy, with a total classification accuracy of 89.55% and a kappa value of 0.87. Cu Lao Xanh’s coral area is around 68 hectares, with soft corals mostly found west and southwest of the island and hard corals in the east. This finding demonstrates that Planetscope satellite imagery is effective at monitoring shallow coral reefs on small islands, providing a scientific foundation and reliable data for the development of a more comprehensive coral reef ecological monitoring and management.