Water Surface Object Detection with Close-Range Map Overlap Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing water area object detection systems inaccurately represent objects near the imager due to excessive reduction in object presence range, leading to gaps between adjacent probability distribution ranges, especially when the imager is close to the object.
Innovation Solution
A water area object detection system that includes an imager on a hull and a controller programmed to set a lower limit for the object presence range when the imager is close to the object, ensuring overlap of object presence ranges and preventing gaps, while adjusting range sizes based on distance measurement errors and using Bayesian estimation for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the object presence range is reduced as the distance from the imager to the feature point decreases, then the measurement precision is improved, but gaps occur between adjacent probability distribution ranges when the imager is close to the object
Solution Approach 1:
The patent changes the parameter of object presence range size based on the distance from the imager to the feature point. When the distance is greater than a predetermined threshold, the object presence range is set according to distance measurement error (larger range). When the distance is within the threshold, the object presence range is set to a predetermined fixed size (smaller range), preventing excessive reduction and gaps in the probability distribution map.
2Manufacturing precision
If the probability distribution range is reduced according to distance measurement error, then the manufacturing precision of the map is improved, but the real-time characteristics of control deteriorate due to increased processing complexity
Solution Approach 1:
The patent simplifies the parameter calculation by using a predetermined fixed size for the object presence range when the distance is within a threshold, rather than calculating based on complex distance measurement error formulas. This reduces computational complexity while maintaining adequate precision for close objects, thereby improving real-time control responsiveness.
3Measurement precision
If the object presence range is excessively reduced when the imager is close to the object, then the measurement precision is improved, but the object presence range may not include portions where the object is originally present
Solution Approach 1:
The patent prevents excessive reduction of the object presence range by setting a predetermined fixed size when the distance from the imager to the feature point is within a threshold. This ensures that the object presence range maintains sufficient area to include the entire object even when the imager is close, while still achieving better precision than a uniformly large range would provide.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
A water area object detection system includes an imager (1a) configured to capture an image around a hull (101), and a controller configured or programmed to perform a control to detect a feature point (F) corresponding to an object (O) in the image together with a distance to the feature point based on the image captured by the imager to create a water area map (M) in which an object presence range (F1) including a likelihood that the object is present is set around the feature point. The controller is configured or programmed to reduce the object presence range as the distance from the imager (1a) to the feature point (F) decreases, and set a size of the object presence range to a lower limit (r1) when the distance from the imager to the feature point is equal to or less than a predetermined distance (d1).