Benthic habitat sediments mapping in coral reef area using amalgamation of multi-source and multi-modal remote sensing data
Xue Ji,Bisheng Yang,Zheng Wei,Mingchang Wang,Qiuhua Tang,Wenxue Xu,Yanhong Wang,Jingyu Zhang,Lin Zhang
DOI: https://doi.org/10.1016/j.rse.2024.114032
IF: 13.5
2024-02-17
Remote Sensing of Environment
Abstract:Scientific management of coral reefs is a global research topic due to their latent economic value and ecological significance. This research benefits from optical and acoustic remote sensing technology with high spatial resolution, sufficient coverage, and stable repeatability. However, each sensor with its own set of limitations, such as insufficient penetration ability, limited sensor resolution and range, which often hinder their applications in the precise classification of coral reef habitats. To address these issues, a novel scheme integrating airborne laser bathymetry (ALB), multibeam echo sounder (MBES) and the multispectral image is designed to serve large-scale and high-precision coral reef habitat sediments research. This scheme focuses on valuable attribute mining, inconsistent coverage, feature default, and redundancy according to multi-source data used in engineering demands. After per-system pretreating, numerous properties, including band, waveform, intensity, terrain, and derivative characteristics, are extracted and quantitatively analyzed to be responsible for habitat description. Subsequently, a complete seabed terrain is generated using the LightGBM regression model to establish the conversion relationship between satellite images and depth. Eventually, a new benthic habitat sediments classification method called RF-SAPSO-LightGBM is introduced with a powerful generalization ability. The unique functions of automatic default feature completion and optimal parameter search improve the discrimination ability of complex sediments. The results indicate that the new method outperforms five other approaches in sediment classification accuracy, and the combination of multi-source data provides more extensive coverage and higher accuracy with an overall of 87. 9%. In comparison, the accuracy values for multispectral image and terrain data alone are 74.5% and 54.7% respectively.
environmental sciences,imaging science & photographic technology,remote sensing