Computer Vision-Based Early Detection of Container Loss at Sea Thanks to Singapore Maritime Institute (SMI) for supporting this work.
Vishakha Lall 1, Capt. Stanley S Pinto 2, Capt. Chu Xing Peng 2, Wu Kaiwen 1 · 2026
Read Computer Vision-Based Early Detection of Container Loss at Sea Thanks to Singapore Maritime Institute (SMI) for supporting this work. by Vishakha Lall 1, Capt. Stanley S Pinto 2, Capt. Chu Xing Peng 2, Wu Kaiwen 1 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-24193.txt.
Introduction
Lall and three colleagues build a low-cost vision system for early detection of destabilized containers at sea. Vessel motion, wind, and heavy seas can slowly loosen stacks despite securing manuals, and new IMO reporting rules make early evidence urgent. The system segments container stacks in onboard camera footage, tracks them with optical flow, and extracts residual motion per container to quantify relative movement. Tests on real ship footage isolate container-level motion across sea states and visibility, giving early crew alerts and course adjustments.
Catalog record
| CREATOR | Vishakha Lall 1, Capt. Stanley S Pinto 2, Capt. Chu Xing Peng 2, Wu Kaiwen 1 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33257 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.24193
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Computer Vision-Based Early Detection of Container Loss at Sea Thanks to Singapore Maritime Institute (SMI) for supporting this work. Vishakha Lall 1, Capt. Stanley S Pinto 2, Capt. Chu Xing Peng 2, Wu Kaiwen 1 Abstract Containerised shipping underpins global trade, yet container loss at sea remains a persistent safety, environmental, and economic challenge. Despite compliance with Cargo Securing Manuals, dynamic maritime conditions such as vessel motion, wind loading, and severe sea states can progressively destabilise container stacks, leading to overboard losses. With the new International Maritime Organisation’s (IMO) mandatory reporting requirements for lost containers, there is an urgent need for a reliable, evidence-based early detection solution for destabilised containers. This study showcases a low-cost, retrofittable computer vision–based system for early detection of destabilised containers using existing onboard cameras. The framework integrates object segmentation to isolate container stacks, temporal object tracking using optical flow and individual objects’ residual motion extraction to quantify relative movement. Experimental evaluation on real onboard ship footage demonstrates that the proposed pipeline effectively isolates container-level motion under challenging conditions of varying sea states and visibility conditions. By enabling early alerts for crew intervention and navigational adjustment, the proposed approach enhances cargo safety, operational resilience, and regulatory compliance. 1 Introduction Container loss at sea is principally driven by two interrelated factors: deficiencies in cargo securing practices as described in the vessel’s Cargo Securing Manual (CSM) and the dynamic environmental loads encountered during a voyage [ 9 ] . The CSM sets out mandatory requirements for container stack configuration, permissible stack weights, and the application of securing devices such as twist locks and lashing systems, and compliance with these provisions can be effectively regulated and enforced through established port, class, and shipboard procedures. In contrast, the environmental conditions encountered at sea, characterised by strong winds, large waves, and dynamic phenomena such as parametric and synchronous rolling, are inherently beyond operational control and can impose dynamic loads on container stacks that exceed the assumptions underpinning CSM design limits [ 5 ] . Empirical analyses of incident data between 2011 and 2023 reveal that adverse weather conditions are the predominant cause of container loss, accounting for 57.14% of incidents [ 14 ] , underscoring that even with effective enforcement of the CSM, environmental forcing remains an inherent and intensifying risk factor in global container ship operations. The loss of containers at sea has wide-ranging economic, safety, regulatory, and environmental consequences. Each container loss event represents a substantial financial burden, encompassing the direct loss of cargo, often valued in the millions of dollars per incident, as well as secondary costs related to insurance claims, salvage operations, voyage delays, and potential liability for damage to third parties. At a systemic level, these losses contribute to increased insurance premiums and operational costs across the global container shipping industry. From a navigational safety perspective, lost containers, particularly those that remain partially submerged, pose a serious collision hazard to other vessels. Such containers are difficult to detect visually and by conventional radar, increasing the risk of hull damage, propulsion failure, or catastrophic accidents [ 1 ] . The vastness of the ocean, coupled with historically incomplete reporting and limited tracking capabilities, further complicates efforts to locate and recover lost containers, allowing many to remain unaccounted for over extended periods. In addition to navigational hazards, container loss presents significant environmental risks. Containers carrying hazardous or polluting cargoes may rupture, releasing chemicals, plastics, or other harmful substances into the marine environment, with long-term consequences for marine ecosystems and coastal regions [ 10 , 11 ] . Even non-hazardous cargo can contribute to marine debris, exacerbating pollution and posing ingestion and entanglement risks to marine life. These environmental impacts reinforce the need for improved monitoring, reporting, and mitigation strategies to address container loss. The frequency of container loss, therefore, remains a persistent concern. As per the World Shipping Council’s Container Loss Report - 2025 Update, the most notable spike occurred in 2013, when 5,578 containers were lost, by far the highest. Other years with elevated loss figures include 2020 (3,924) and 2021 (2,301), which were periods marked by several large-scale incidents. In contrast, 2023 saw a significant reduction with 221 containers lost. But in 2024, the total rose to 576 containers, highlighting that container loss is not an isolated or rare phenomenon but a recurring operational risk in global maritime transport. Recognising the implications of underreporting, the International Maritime Organisation (IMO) has adopted amendments to the Safety of Life at Sea (SOLAS) Convention, mandating the reporting of container losses from January 2026 onwards. Existing research on container loss at sea has predominantly focused on post-loss analysis, encompassing incident investigations, environmental impact assessments, and methods for detecting and mitigating hazards posed by containers after they have entered the marine environment. Representative studies, as summarised in [ 7 ] , examine local detection systems, including radar, sonar, and camera-based observation, for identifying nearby floating or partially submerged containers, as well as remote monitoring systems that rely on embedded communication devices to transmit container location and status information. While these approaches enhance collision avoidance, situational awareness, and long-term tracking, they are inherently reactive, operating only after container loss has occurred. Related work by [ 8 ] advances this post-loss paradigm by integrating container-mounted tracking units (CTDs) equipped with sensors and communication modules to detect when a container goes overboard automatically and broadcast alerts to the host vessel, nearby ships, and shore-based authorities via existing maritime communication standards, including mobile and satellite links. Although such systems improve the timeliness and effectiveness of post-incident response, they do not address the conditions leading up to loss events. Recent studies have also explored the application of advanced artificial intelligence techniques for post-loss container detection. In particular, [ 6 ] proposes a submerged container detection framework that combines the You Only Look Once (YOLO) object detection model with the Segment Anything Model (SAM) to identify containers in water-column imagery acquired from sonar or related sensing modalities. Experimental results demonstrate high accuracy in container localisation and segmentation, highlighting that modern deep learning and segmentation models can reliably detect containers in complex and cluttered marine environments, including scenarios involving partial submersion. While these results clearly motivate the use of state-of-the-art computer vision models for accurate container detection, the approach remains post-analytical, relying on data collected after containers have already been lost and offering no mechanism for early warning or loss prevention. In contrast to these reactive approaches, limited prior work has investigated pre-loss monitoring of container stability during voyage. An early example is presented in [ 2 ] , which explores the use of wireless vibration sensor tags affixed to containers to continuously monitor dynamic responses associated with stability degradation, such as excessive sway or tipping that may precede cargo shifts or stack destabilisation. While this study demonstrates the feasibility of using wireless sensing to provide early indicators of potential stability loss, it does not address challenges related to large-scale deployment, integration with shipboard systems, or operational scalability. Based on the insights from domain experts from Pacific Interational Lines (PIL), early notification of container instability is consistently identified as a critical enabler of effective corrective action. Experts emphasised that the time window between the onset of container instability