Maritime Small Object Detection from UAVs using Deep Learning with Altitude-Aware Dynamic Tiling

2025-11-24

AI Summary

In maritime UAV search-and-rescue (SAR) operations, detecting small targets—such as persons in water—is challenging due to low pixel occupancy and complex sea conditions, especially at high altitudes. To address this, this paper proposes a height-aware dynamic tiling detection method: image tiling size and scaling factor are adaptively adjusted based on flight altitude, and inference is performed efficiently by integrating YOLOv5 with the SAHI framework. The core innovation lies in leveraging flight altitude as prior knowledge to guide tiling strategy, thereby preserving detection accuracy while significantly reducing computational redundancy. Evaluated on the SeaDronesSee dataset, the method achieves a 38% improvement in small-object mAP and operates at 2.1× the inference speed of conventional static-tiling approaches. These gains markedly enhance the real-time performance and practical applicability of long-range maritime SAR missions.

Technology Category

• Computer Vision: Object Detection & Categorization
• Search and Optimization: Algorithm Configuration
• Machine Learning: Hardware-aware ML

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Unmanned Aerial Vehicles (UAVs) are crucial in Search and Rescue (SAR) missions due to their ability to monitor vast maritime areas. However, small objects often remain difficult to detect from high altitudes due to low object-to-background pixel ratios. We propose an altitude-aware dynamic tiling method that scales and adaptively subdivides the image into tiles for enhanced small object detection. By integrating altitude-dependent scaling with an adaptive tiling factor, we reduce unnecessary computation while maintaining detection performance. Tested on the SeaDronesSee dataset [1] with YOLOv5 [2] and Slicing Aided Hyper Inference (SAHI) framework [3], our approach improves Mean Average Precision (mAP) for small objects by 38% compared to a baseline and achieves more than double the inference speed compared to static tiling. This approach enables more efficient and accurate UAV-based SAR operations under diverse conditions.

Problem

• Detecting small maritime objects from high UAV altitudes
• Improving low object-to-background pixel ratio challenges
• Enhancing detection efficiency while maintaining accuracy

Innovation

• Altitude-aware dynamic tiling for small object detection
• Integrates altitude-dependent scaling with adaptive tiling factor
• Improves detection precision and doubles inference speed