As maritime transportation and human activities at sea continue to grow, ensuring the safety of offshore infrastructure has become an increasingly pressing research focus. However, traditional high-precision sensor systems often involve prohibitive costs, and the Automatic Identification System (AIS) faces signal loss or data manipulation problems, highlighting the need for a complementary, affordable, and reliable supplemental solution. This study introduces a monocular vision-based safety monitoring framework for offshore infrastructures. By combining advanced computer vision techniques such as Grounded SAM and horizon-based self-calibration, the proposed framework achieves accurate vessel detection, instance segmentation, and distance estimation. The model integrates open-vocabulary object detection and zero-shot segmentation, achieving high performance without additional training. To demonstrate the feasibility of the framework in practical applications, we conduct several experiments on public datasets and couple the proposed algorithms with the Leaflet.js and WebRTC libraries to develop a web-based prototype for real-time safety monitoring, providing visualized information and alerts for offshore infrastructure operators in our case study. The experimental results and case study suggest that the framework has notable advantages, including low cost, convenient deployment with minimal maintenance, high detection accuracy, and strong adaptability to diverse application conditions, which brings a supplemental solution to research on offshore infrastructure safety.

Stereo matching technology, enabling the acquisition of three-dimensional data, holds profound implications for marine engineering. In underwater images, irregular object surfaces and the absence of texture information make it difficult for stereo matching algorithms that rely on discrete disparity values to accurately capture the 3D details of underwater targets. This paper proposes a stereo method based on an energy function of Markov random field (MRF) with 3D labels to fit the inclined plane of underwater objects. Through the integration of a cross-based patch alignment approach with two label optimization stages, the proposed method demonstrates features akin to segment-based stereo matching methods, enabling it to handle images with sparse textures effectively. Through experiments conducted on both simulated UW-Middlebury datasets and real deteriorated underwater images, our method demonstrates superiority compared to classical or state-of-the-art methods by analyzing the acquired disparity maps and observing the three-dimensional reconstruction of the underwater target.