Line Segment End Point Estimation via Luminance Modeling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current line segment detection methods, such as those using the Hough transform technique, are unable to accurately detect the start and end points of a line segment in an image, which is crucial for identifying and tracking space debris in orbit.

Innovation Solution

A line segment detection method that generates a virtual function model based on luminance distribution to probabilistically determine whether luminance values near the end points belong to the foreground or background, allowing for the efficient and accurate estimation and correction of end points using a maximum likelihood method.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the Hough transform technique is used to detect linear objects in an image, then the detection of straight lines is achieved, but the linear end points (start point and end point) cannot be detected

Engineering Contradiction:
Improveend point detection accuracyVSAvoiddetection method capability
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent segments the line detection problem into two distinct parts: first detecting the straight line using Hough transform, then separately detecting the end points along that line. This segmentation allows each sub-problem to be solved with appropriate methods, overcoming the limitation of Hough transform which only solves the line detection part

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary virtual function model (likelihood function, ramp function, or sigmoid function) that represents the expected luminance distribution along the line. This virtual model serves as a mediator to compare with actual image data, enabling end point detection by finding where the actual luminance deviates from the expected distribution

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If a method is needed to accurately detect end points of line segments in images of orbiting debris, then detection accuracy must be improved, but the complexity of the detection method increases

Engineering Contradiction:
Improvedebris position accuracyVSAvoiddetection system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary action by first detecting the straight line trajectory of the debris using Hough transform before attempting end point detection. This preliminary line detection provides a foundation and constraint for the subsequent end point detection, reducing the search space and simplifying the overall process

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent replaces complex mechanical or computational search methods with a statistical approach using virtual function models. Instead of exhaustive search or complex optimization, the method uses probability-based luminance distribution models to efficiently estimate end point positions, reducing computational complexity while maintaining accuracy

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3309752B1Line segment detection method
Publication Date: 2021.01.27 IHI CORP
  • EP3309752B1 patent drawingFigure 1
  • EP3309752B1 patent drawingFigure 2(a)~2(b)
  • EP3309752B1 patent drawingFigure 3(a)~3(c)

AI summary

A line segment detection method includes acquiring an image (1) (Step 1), extracting from the image (1) a straight line (L) that includes a line segment (P) (Step 2), acquiring a luminance distribution (2) of the straight line (L) (Step 3), generating a virtual function model (Lm) that contains a distribution of a foreground and a background of the image (1) by using the luminance distribution (2) (Step 4), and determining whether luminance values near end points (start point (s) and end point (t)) of the line segment (P) contained in the virtual function model (Lm) probabilistically belong to the foreground or the background, and estimating end points (start point (s') and end point (t')) in the virtual function model (Lm) (Step 5).