Author(s): Siyoon Kwon; Hyoseob Noh; Il Won Seo; Yun Ho Lee And Byungman Yoon
Linked Author(s): Il Won Seo
Keywords: Sediment Monitoring Remote Sensing CCTV Imaging Spectroscopy Light Source Types
Abstract: Effective sediment monitoring is essential for understanding and managing river environments, particularly in regions prone to sediment-related challenges. However, traditional measurement methods are labor-intensive, and remote sensing is typically limited to daytime use under allowable weather conditions. To address these limitations, we propose a novel CCTV-type hyperspectral camera system integrated with an optimized machine learning framework for continuous monitoring of suspended sediment concentration (SSC) during both day and night. The system combines hyperspectral imaging capabilities with low-light adaptability, enabling the detection of fine spectral variations in sediment under various lighting conditions, including sunlight and halogen light sources. We conducted field tests in the outdoor river-scale channel by injecting bright silt and low-visibility sand. We categorized experimental conditions into day, transition, and night cases based on the light source types. During the night case, the number of halogen light sources was adjusted to evaluate the system’s sensitivity to varying light intensities. Our machine learning framework, by combining a Random Forest classification and regression, demonstrated the robust classification of light sources and good agreement with in-situ SSC measurements. Further, we cross-validated the CCTV-type hyperspectral data with spectrometer and laser diffraction sensor, highlighting its reliability and potential for long-term deployment. By integrating imaging spectroscopy into a CCTV platform, the system proposed in this study shows potential to connect advanced remote sensing technologies with the growing demand for continuous sediment monitoring in the face of climate change.
Year: 2025