Man Overboard Alert Verification Using Motion Compensated Image Summation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current man overboard detection systems on marine vessels rely on wearable tags, which can be removed accidentally or intentionally, reducing their reliability, and are not optimized for efficient detection and classification of human bodies in real-time, leading to delayed and costly search and rescue operations.
Innovation Solution
A method and apparatus that analyze sequential frames of image data to detect and classify movement, using motion compensation, statistical analysis, and filtering to identify clusters of pixels corresponding to a human body, enabling faster and more accurate detection without the need for wearable tags.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If wearable tags are used for man overboard detection, then detection capability is provided, but reliability deteriorates because tags can be removed accidentally or intentionally
Solution Approach 1:
The patent removes the wearable tag component entirely from the detection system. Instead of requiring tags on passengers, the system uses onboard sensors (radar, LIDAR, cameras) to directly detect objects in the water, eliminating the reliability issue of tag removal while maintaining detection capability
Solution Approach 2:
The patent creates a digital representation of the monitored area using sensor data and generates a virtual model of objects detected in water. This allows the system to track and analyze potential man-overboard events without physical tags, using computational copies of spatial information instead
2Measurement precision
If video capture apparatus is used to provide visual confirmation, then detection accuracy improves, but processing time and complexity increase
Solution Approach 1:
The system continuously captures and pre-processes video data in the background before events occur, maintaining a buffer of analyzed visual information. When an alert is triggered, pre-processed frames are immediately available for verification, eliminating the need for real-time processing during critical rescue windows
Solution Approach 2:
The video processing is divided into separate stages: continuous background capture, periodic frame analysis, and event-specific verification. Only relevant frames surrounding alert events undergo detailed processing, while other frames are processed at lower priority or skipped, reducing overall processing time while maintaining accuracy for critical moments
3Measurement precision
If high imaging resolution is used to identify human bodies, then detection precision improves, but system complexity and computational requirements increase
Solution Approach 1:
The system applies different processing qualities to different regions of the video feed. High-resolution analysis is concentrated on specific zones of interest (areas where objects are detected near or in water), while other areas receive minimal or no processing. This localized approach maintains detection precision where needed while reducing overall computational complexity
Solution Approach 2:
The system uses multiple sensors with varying resolution capabilities (radar, LIDAR, standard video cameras) rather than relying solely on high-resolution imaging. The combination of lower-resolution sensors for broad coverage and selective high-resolution verification provides sufficient detection precision without the complexity and cost of deploying high-resolution cameras across the entire monitoring area
Data Source
AI summary
A method of verifying a triggered alert in a man overboard detection system comprises: receiving (600) a plurality of sequential frames of image data associated with the triggered alert followed by motion compensating (616) difference frames formed by generating differences (608) between successive frames. A summation of the motion compensated difference frames is then generated (618). The summation image is then analysed (630, 632) and the detection associated therewith is classified (634, 636) in response to identification of a cluster of pixels corresponding to a predetermined size range.