A Linear Feature-Based Method for Signal Photon Extraction and Bathymetric Retrieval Using ICESat-2 Data

Zhenwei Shi, Jianzhong Li, Ze Yang, Hui Long, Hongwei Cui, Shibin Zhao, Xiaokai Li, Qiang LiRemote 2025 · DOI

TL;DR

This paper introduces a novel Linear Feature-Based Signal Photon Extraction (LFSPE) method to improve the accuracy of bathymetric mapping using ICESat-2 satellite data. By leveraging the unique linear characteristics of seafloor photons and employing adaptive techniques, LFSPE significantly enhances signal photon identification, especially in challenging shallow-water environments with low photon density and complex terrain, outperforming traditional methods like DBSCAN.

What Problem Does This Paper Address?

Accurately mapping shallow-water bathymetry is crucial for marine development and environmental protection. While the ICESat-2 satellite's photon-counting LiDAR offers unprecedented potential for this task due to its high sensitivity and broad coverage, a major hurdle is distinguishing true seafloor "signal" photons from various sources of noise (e.g., sunlight, atmospheric scattering). This challenge is amplified in low-photon-density areas and complex underwater terrains, where signal photons become sparse and difficult to isolate. Existing methods, often relying on density-based clustering, struggle with parameter tuning and can be less effective in these demanding scenarios.

Methods

The LFSPE algorithm tackles this challenge through a three-stage approach:

  1. Adaptive Resolution Adjustment: This initial step standardizes the photon data by adjusting vertical and horizontal resolutions. It involves grouping photons by laser pulse, applying Euclidean clustering, and selecting a representative "retained point" for each cluster to reduce redundancy and improve data quality.
  2. Segmented Gaussian Fitting for Coarse Segmentation: To separate underwater photons from non-underwater ones, the algorithm uses elevation histograms and Gaussian fitting. Crucially, it employs a segmented approach, fitting Gaussian models to smaller segments of the data rather than a single global model. This allows for better adaptation to varying terrain and water surface elevations, improving the distinction between sea surface and seafloor photons, especially in complex or shallow areas.
  3. Linear Feature-Based Signal Photon Extraction: This core stage identifies signal photons by analyzing their spatial distribution. For non-underwater photons, a fixed neighborhood search is used. For underwater photons, where signal strength decreases with depth, a depth-dependent adaptive neighborhood search radius is employed. This radius dynamically expands in deeper regions to ensure sufficient photons are captured for analysis. Two key linear features are calculated: point density (how many points fall within a defined linear structure) and point-to-line distance (how close points are to a fitted linear segment). Signal photons are identified by setting thresholds on these features, effectively separating them from noise.

Datasets

The study utilizes eight diverse ICESat-2 ATL03 datasets, covering various geographical locations, terrains, and water depths. These datasets were chosen to represent a range of challenges, including daytime data with higher noise levels and nighttime data with clearer signals.

For validation, the study uses the Continuously Updated Digital Elevation Model (CUDEM), a high-resolution land-sea integrated digital elevation model, to assess the accuracy of the extracted bathymetric depths.

Key Insights and Ideas

  • Linearity is Key: The fundamental insight is that true seafloor photons exhibit a distinct linear or gently curved spatial continuity along the laser track, unlike the more random distribution of noise photons. This characteristic is the bedrock of the LFSPE algorithm.
  • Depth-Dependent Adaptability: The adaptive neighborhood search radius for underwater photons is a critical innovation. It directly addresses the signal attenuation problem in deeper waters, allowing the algorithm to maintain performance where fixed-radius methods would fail.
  • Segmentation for Robustness: The segmented Gaussian fitting approach, rather than a single global fit, is crucial for accurately separating underwater from non-underwater photons across varied landscapes and water conditions.
  • Parameter Simplicity: Compared to methods like DBSCAN, which require extensive parameter tuning for each dataset, LFSPE's approach, particularly its adaptive nature, leads to more consistent performance with fewer dataset-specific adjustments.

Key Results

The LFSPE method demonstrated strong performance across the eight experimental datasets:

  • Signal Photon Extraction Accuracy: Achieved average precision of 0.977, recall of 0.958, F1 score of 0.967, and overall accuracy of 0.972. Precision exceeded 0.96 for all datasets, and recall surpassed 0.90 in most.
  • Bathymetric Accuracy: When compared to CUDEM data for two regions, the extracted bathymetric depths showed excellent agreement, with R² values up to 0.995 and RMSE as low as 0.718 m.
  • Comparison with DBSCAN: LFSPE significantly outperformed DBSCAN, showing a 30.0% increase in precision, 17.3% improvement in recall, 24.3% increase in F1 score, and 22.2% improvement in overall accuracy. This improvement was particularly notable in sparse photon regions and complex terrains.
  • Depth Performance: The algorithm maintained high performance across different water depth intervals (<5m, 5-10m, >10m), with only minor degradation in deeper waters, highlighting its robustness.

Limitations and Future Work

  • Parameter Sensitivity: While generally robust, the study notes that datasets with lower signal-to-noise ratios or more complex topography can be more sensitive to parameter changes (density_thr and dist_thr). Future work could explore more automated or adaptive threshold determination.
  • Refraction Correction: The paper mentions the need for refraction correction for accurate depth measurements, employing a specific method. Further research could explore more advanced or universally applicable refraction correction techniques.
  • Computational Cost: While not explicitly detailed, complex algorithms like this can have computational demands. Future work might focus on optimizing efficiency for large-scale applications.
  • Beyond Shallow Water: The current study focuses on shallow-water bathymetry. Extending the LFSPE method to deeper water scenarios, where signal attenuation is more severe, could be a valuable future direction.

Summary updated

Abstract

The ATL03 data from the photon-counting LiDAR onboard the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) holds substantial potential for shallow-water bathymetry due to its high sensitivity and broad spatial coverage. However, distinguishing signal photons from noise in low-photon-density and complex terrain environments remains a significant challenge. This study proposes an adaptive photon extraction algorithm based on linear feature analysis, incorporating resolution adjustment, segmented Gaussian fitting, and linear feature-based signal identification. To address the reduction in signal photon density with increasing water depth, the method employs a depth-dependent adaptive neighborhood search radius, which dynamically expands into deeper regions to ensure reliable local feature computation. Experiments using eight ICESat-2 datasets demonstrated that the proposed method achieves average precision and recall values of 0.977 and 0.958, respectively, with an F1 score of 0.967 and an overall accuracy of 0.972. The extracted bathymetric depths demonstrated strong agreement with the reference Continuously Updated Digital Elevation Model (CUDEM), achieving a coefficient of determination of 0.988 and a root mean square error of 0.829 m. Compared to conventional methods, the proposed approach significantly improves signal photon extraction accuracy, adaptability, and parameter stability, particularly in sparse photon and complex terrain scenarios. In comparison with the DBSCAN algorithm, the proposed method achieves a 30.0% increase in precision, 17.3% improvement in recall, 24.3% increase in F1 score, and 22.2% improvement in overall accuracy. These findings confirm the effectiveness and robustness of the proposed algorithm for ICESat-2 shallow-water bathymetry applications.
BibTeX
@inproceedings{shi2025linear,
  author = {Zhenwei Shi and Jianzhong Li and Ze Yang and Hui Long and Hongwei Cui and Shibin Zhao and Xiaokai Li and Qiang Li},
  title = {{A Linear Feature-Based Method for Signal Photon Extraction and Bathymetric Retrieval Using ICESat-2 Data}},
  booktitle = {Remote},
  year = {2025},
  doi = {10.3390/rs17162792},
}
Figure