Efficient Coral Survey Using Aerial Remote Sensing and Multi-modal Segmentation for Large-Scale Ecological Assessment

Due to the ecological pressures of global warming and human activities in coastal regions, coral reef ecosystems, predominantly located in shallow marine areas, are facing severe threats to their survival. Scientists and governmental managers are eager to leverage novel aerial remote sensing technologies to address the challenges of acquiring accurate, comprehensive, and timely data on coral reef health, structural complexity, and spatial distribution. This study aims to tackle these challenges by using precise and accurate aerial remote sensing data to support the restoration and sustainable prosperity of coral reef systems. Specifically, this study develops and applies an efficient coral survey method based on aerial remote sensing. The method integrates aerial imagery and bathymetric lidar (light detection and ranging) point cloud data and uses advanced photogrammetric computer vision and deep learning algorithms. Using a state-of-the-art multi-modal neural network segmentation technique, the proposed method enables high-precision and intelligent identification of coral reefs, facilitating detailed habitat mapping. Furthermore, by accurately delineating the habitat range and geometric structures of reefs, this approach allows for precise measurements of coral biomass production and skeletal calcification. These metrics help assess coral reef structural complexity and their adaptability to environmental stressors, providing robust scientific data for conservation strategies and policy making. The use of advanced multi-modal aerial remote sensing data not only enhances monitoring reliability and accuracy but also offers a cost-effective and flexible tool for coral reef ecological mapping. This approach effectively addresses challenges encountered in coastal ecological surveys, particularly in areas where direct human access or boat entry is difficult.