Articulated Object Part Estimation via Edge Pair Likelihood Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing posture estimation techniques struggle with high accuracy due to difficulties in extracting parallel lines from images affected by wrinkles, shadows, and object shapes, especially when estimating parts of articulated objects like humans or robots.
Innovation Solution
A part estimation apparatus and method that generate edge pair likelihood maps, continuity likelihood maps, and integrated likelihood maps to identify candidate regions for articulated object parts, using edge pair likelihood calculations and predefined conditions to filter out noise and accurately pinpoint part locations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parallel lines are extracted from captured images to estimate part locations, then part estimation can be performed, but extraction accuracy deteriorates due to wrinkles, shadows, and object shapes interfering with parallel line detection
Solution Approach 1:
The patent segments the image processing into multiple independent stages: edge detection, parallel line extraction, continuity evaluation, and part estimation. Each stage processes specific features separately, allowing the system to handle complex images by breaking down the detection task into manageable components that can be optimized individually.
Solution Approach 2:
The patent introduces an intermediary continuity evaluation mechanism that acts as a mediator between edge detection and part estimation. This intermediary layer evaluates the continuity of extracted parallel lines and filters out spurious detections caused by wrinkles and shadows, thereby improving the accuracy of subsequent part estimation without directly interfering with the edge detection process.
2Adaptability or versatility
If multiple candidate regions are generated for part estimation, then coverage is improved, but noise and false positives increase due to wrinkles and shadows being misinterpreted as part boundaries
Solution Approach 1:
The patent implements a feedback mechanism through the continuity evaluation step, where the extracted parallel lines are evaluated for continuity and used to feedback into the candidate region selection. This feedback loop allows the system to iteratively refine the candidate regions by eliminating those that do not satisfy continuity constraints, thereby reducing false positives while maintaining comprehensive coverage.
Solution Approach 2:
The patent changes the parameters used for evaluation by introducing continuity as a new criterion alongside traditional edge detection parameters. By evaluating both the presence of parallel lines and their continuity, the system can distinguish between valid part boundaries and spurious edges caused by wrinkles and shadows, improving the reliability of candidate regions.
Data Source
AI summary
With this device, an edge pair likelihood map generation unit (120) computes edge pair likelihood which denotes the plausibility that a pixel of a pair is an edge of an estimation subject site, and generates an edge pair likelihood map which denotes the edge pair likelihood for each pixel. A continuity likelihood map generation unit (130) evaluates, upon the edge pair likelihood map, the continuity for the edge pair likelihood of a pixel which is included in a region wherein the estimate subject site is presumed, and generates a continuity likelihood map which denotes an edge pair likelihood which has continuity as a candidate region of the estimate subject site. An integrated likelihood map generation unit (140) generates an integrated likelihood map which denotes the candidate region which the continuity likelihood map denotes, refined based on a predetermined condition.