Reproduce “A Fast Horizon Detector and a New Annotated Dataset for Maritime Video Processing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yassir Zardoua, Boulaala Mohammed, Mhamed El Mrabet, Astito Abdelali
An efficient and robust algorithm for sea horizon detection in RGB videos is presented, which effectively filters noisy line segments to identify the horizon and incorporates temporal consistency for improved stability. The work also contributes a new, annotated Tangier Maritime Dataset, designed to test algorithms under diverse and challenging maritime conditions not well-represented in existing datasets.
Key insights
- Vectorized segment-based filtering, specifically the Length-Slope Filter and Region of Interest Filter, can efficiently isolate true horizon elements from noisy maritime scenes at low computational cost.
- Intelligently incorporating temporal information through an Outlier Handler Module significantly enhances detection stability and robustness against transient false positives in video sequences.
- The primary computational bottleneck in this line segment-based approach resides in the initial line segment detection, indicating areas for future speed optimizations of the overall system.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.