A noise removal algorithm for photon counting LiDAR data: the ConvDS model

The Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2), a satellite focused on changes in the heights of ice caps, clouds, and the land surface, provides valuable data resources for global Earth science research through its advanced laser altimetry technology. Since the ICESat-2 satellite transmits weak pulses that are susceptible to solar radiation and other factors, there is a large amount of noise in the data. In this paper, a new denoising model (ConvDS) is proposed. We use a combination of discrete convolution algorithms and cluster analysis algorithms to reduce the noise photon points under the ground to some extent. Comparing the ConvDS model with the other two methods, in steep terrain, the ConvDS model is 2.76% better than the improved DBSCAN algorithm and 1.48% better than the improved OPTICS algorithm. In hilly terrain, the denoising accuracy of the ConvDS model is 2.09% higher than the improved DBSCAN algorithm and 0.48% higher than the improved OPTICS algorithm. It can be concluded that the denoising effect of the ConvDS model is superior.