Deep Learning based Detection of Fishing Vessels and Fishing Monitoring using Nightlight Images

This study presents a novel approach for detecting small-scale fishing vessels using nighttime light (NTL) imagery from the SDGSAT-1 satellite, combined with deep learning techniques to enhance fishing monitoring awareness along the western coast of India. A dual-branch YOLO11 architecture was developed to exploit both the 10-meter panchromatic and 40-meter RGB imagery from SDGSAT-1. The custom model architecture was specifically optimized for small object detection in NTL imagery, featuring parallel convolutional backbones that process both modalities before concatenation for enhanced feature extraction. The dual-branch YOLO11 model demonstrated optimal performance with a precision of 0.99, recall of 0.93, F1-score of 0.96, and mAP@50 of 0.96, significantly outperforming single-branch implementations of YOLOv5s, YOLOv8s, and standard YOLO11s architectures. When applied to the study area covering the waters off Maharashtra, Goa, and Karnataka, the model detected 31,525 vessel instances across the temporal dataset spanning from 1st January 2022 to 31st December 2023. Cross-matching analysis with AIS data revealed that only 7,146 (22.7%) of detected vessels had corresponding AIS transmissions, while 24,379 (77.3%) were identified as potential dark vessels. Spatio-temporal analysis showed peak fishing activity during January-April, with a primary activity corridor parallel to the coastline within 50-100 km, corresponding to productive continental shelf areas. The continental shelf regions of Maharashtra, Karnataka, and Goa were identified as major fishing hotspots on the Western coast of India. This research contributes to maritime surveillance capabilities by highlighting the effectiveness of nighttime lights satellite imagery for fishing vessel detection and provides valuable insights into fishing patterns and potential regulatory compliance issues in Indian waters. The findings have important implications for fisheries management, maritime security, and sustainable ocean resource utilization.

Improvements to AI Systems:

1. Multi-Modal Fusion for Small Object Detection: Integrate a dual-branch convolutional backbone (e.g., YOLO11-based) that processes panchromatic (10m) and RGB (40m) inputs in parallel, with feature concatenation before detection heads. This improves detection of sub-20m objects in low-light, high-noise satellite imagery.

2. Dark Vessel Identification Pipeline: Add a post-processing module that cross-references detected vessel coordinates and timestamps with AIS (Automatic Identification System) data. The AI can automatically classify vessels as dark (non-transmitting) or compliant based on spatio-temporal matching thresholds, enabling real-time anomaly alerts.

3. Temporal Activity Pattern Recognition: Train the system to output spatio-temporal density maps (e.g., monthly fishing activity heatmaps) by aggregating detections over time. This allows prediction of peak fishing seasons (e.g., January–April) and identification of high-activity corridors (e.g., 50–100 km from coast) for dynamic resource allocation.

4. Continental Shelf Hotspot Detection: Implement a clustering algorithm (e.g., DBSCAN or grid-based) on detection coordinates to automatically delineate fishing hotspots over continental shelves. The AI can rank regions by vessel density and flag regulatory compliance risk zones for targeted enforcement.

5. Model Selection and Optimization: Use the paper’s performance metrics (precision 0.99, recall 0.93, F1 0.96, mAP@50 0.96) as a benchmark to auto-tune hyperparameters (e.g., anchor sizes, NMS thresholds) for small-object NTL datasets, avoiding manual trial-and-error in future deployments.

What the Improved AI System Can Do:

• Real-time maritime surveillance: Detect small fishing vessels at night from satellite NTL imagery, even in coastal waters with high background light, and distinguish them from larger ships or land-based lights.

• Automatic dark fleet monitoring: Continuously compare detected vessels with AIS broadcasts, flagging non-cooperative vessels within minutes, and generating alerts for fisheries authorities or maritime security agencies.

• Predictive fishing activity forecasting: Use historical detection data to forecast weekly or monthly fishing intensity along specific coastlines, enabling preemptive patrol scheduling and quota enforcement.

• Hotspot-aware resource management: Automatically generate maps of fishing hotspots over continental shelves, helping governments prioritize marine protected areas or regulate overfishing in sensitive zones.

• Scalable compliance auditing: Process large temporal datasets (e.g., 2 years of imagery) to produce statistical reports on compliance rates (e.g., 22.7% AIS-matched vs. 77.3% dark), supporting policy decisions on mandatory vessel tracking systems.

The demand for maritime surveillance has given rise to the need for monitoring fishing vessel activities, particularly in addressing the challenge of "dark vessels" that operate without Automatic Identification System (AIS) transmission. This study presents a novel approach for detecting small-scale fishing vessels using nighttime light (NTL) imagery from the SDGSAT-1 satellite, combined with deep learning techniques to enhance fishing monitoring awareness along the western coast of India. A dual-branch YOLO11 architecture was developed to exploit both the 10-meter panchromatic and 40-meter RGB imagery from SDGSAT-1. The custom model architecture was specifically optimized for small object detection in NTL imagery, featuring parallel convolutional backbones that process both modalities before concatenation for enhanced feature extraction. The dual-branch YOLO11 model demonstrated optimal performance with a precision of 0.99, recall of 0.93, F1-score of 0.96, and mAP@50 of 0.96, significantly outperforming single-branch implementations of YOLOv5s, YOLOv8s, and standard YOLO11s architectures. When applied to the western coast of India, the model detected 31525 vessel instances across the temporal dataset spanning 2022-23. Cross-matching analysis with AIS data revealed that only 7146 (22.7%) of detected vessels had corresponding AIS transmissions, while 24379 (77.3%) were identified as potential dark vessels. Spatio-temporal analysis showed peak fishing activity during January-April, with a primary activity corridor parallel to the coastline within 50-100 km, corresponding to productive continental shelf areas. This research contributes to maritime surveillance capabilities by highlighting the effectiveness of nighttime lights satellite imagery for fishing vessel detection and provides valuable insights into fishing patterns and potential regulatory compliance issues in Indian waters.