Hierarchical Multiscale Denoising Method for Spaceborne Photon-Counting LiDAR Based on Minimum Spanning Tree

Space-borne photon-counting light detection and ranging (LiDAR) can obtain high-precision bathymetry information for nearshore waters. However, the signals obtained by the space-borne photon-counting LiDAR contain a large amount of noise due to atmospheric scattering, solar radiation and instrumental noise. Therefore, the denoising of the space-borne photon-counting LiDAR is a key component in the acquisition of bathymetric information. With respect to the difficulty of accurately denoise the density heterogeneity photons in different water depths, this paper proposes a hierarchical multiscale denoising method for space-borne photon-counting LiDAR based on minimum spanning tree. First, sea surface photons and water-column photons are obtained separately based on the kernel density estimation. Second, the minimum spanning tree is constructed and split based on the spatial distribution of water-column photons. In the process of splitting the minimum spanning tree, the length threshold of the interrupted branches is adaptively calculated. Then, water-column photons are denoised based on the hierarchical multiscale subtree node judgment strategy. Finally, water-column photon denoising results are refined using a box plot outlier detection method. The proposed method is compared with some state-of-the-art photon denoising algorithms using six trajectory data under various seafloor terrains. The experimental results demonstrate that the proposed method is better than other photon denoising algorithms, and can obtain accurately denoising results for different density distributions of photons under different seafloor terrains.