Related Experiment Video
Updated: May 5, 2026
Investigating the Relationship between Sea Surface Chlorophyll and Major Features of the South China Sea with Satellite Information
Published on: June 13, 2020
A semantic segmentation network for red tide detection based on enhanced spectral information using HY-1C/D CZI
Kunpeng Sun1, Guanghao Jiang2, Ning Wang1
1North China Sea Marine Forecasting and Hazard Mitigation Center, Ministry of Natural Resources, Qingdao 266061, China; Shandong Key Laboratory of Marine Ecological Environment and Disaster Prevention and Mitigation, Qingdao 266061, China.
This study introduces a new deep learning model, SIC-RTNet, for accurate red tide monitoring using satellite data. The model effectively detects red tides, improving marine ecological early warning systems.
Area of Science:
- Marine Science
- Remote Sensing Technology
- Ecological Monitoring
Background:
- Traditional red tide monitoring relies on low-spatial-resolution sensors, hindering detection of small events and early outbreaks.
- Existing methods are incompatible with medium-to-high spatial resolution, low-spectral-resolution satellite data, limiting early detection capabilities.
Purpose of the Study:
- To develop an advanced satellite remote sensing model for efficient red tide monitoring.
- To improve the detection accuracy of red tide events, especially in early stages.
- To enable monitoring using high spatial resolution and wideband satellite data.
Main Methods:
- Proposed the Residual Neural Network Red Tide Monitoring Model based on Spectral Information Channel Constraints (SIC-RTNet).
- Incorporated residual blocks to preserve weak red tide signal features.
- Utilized spectral information channels derived from wideband data to enhance spectral differences.
- Implemented an improved loss function to handle imbalanced datasets of red tide and seawater.
Main Results:
- SIC-RTNet achieved high performance metrics: 85.5% precision, 95.4% recall, 0.90 F1-Score, and 0.90 Mean Intersection over Union (MoU).
- The model demonstrated superior accuracy compared to other existing methods.
- Successfully identified red tides using high spatial resolution and wideband remote sensing data.
Conclusions:
- SIC-RTNet offers an effective solution for automated red tide identification using readily available satellite data.
- The model significantly enhances marine ecological disaster monitoring and early warning systems.
- This approach expands the utility of medium-to-high spatial resolution satellite sensors for red tide detection.
