AR Feature Point Classification for Real-Time 3D Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Markerless-based augmented reality techniques face challenges in implementing real-time augmented reality due to increased data processing requirements when recognizing 3D objects, which affects accuracy and efficiency.
Innovation Solution
A method and system that classify feature points from images captured while changing capture position or orientation, using a linearity index to distinguish between recognition reference and dummy feature points, reducing data processing and enhancing object recognition accuracy by excluding feature points with high uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If markerless-based augmented reality technique is used to recognize 3D objects, then object recognition capability is improved, but data processing amount increases and real-time performance deteriorates
Solution Approach 1:
The patent segments feature points into two categories: recognition reference feature points (with low position uncertainty) and dummy feature points (with high position uncertainty). This segmentation allows the system to use only the necessary subset of feature points for recognition, reducing data processing amount while maintaining recognition capability.
Solution Approach 2:
The patent applies local quality by treating different feature points differently based on their position uncertainty. Recognition reference feature points are used for accurate object recognition, while dummy feature points are excluded from recognition processes. This selective usage optimizes processing efficiency without compromising recognition reliability.
2Reliability
If all extracted feature points are used for object recognition, then recognition completeness is improved, but processing complexity and time increase
Solution Approach 1:
The patent applies partial action by using only a subset of feature points (recognition reference feature points with low position uncertainty) for object recognition, rather than processing all extracted feature points. This reduces processing complexity while maintaining sufficient recognition completeness through selective feature point usage.
3Quantity of substance
If feature points with high position uncertainty are included in recognition, then feature point quantity is improved, but recognition accuracy deteriorates
Solution Approach 1:
The patent changes the parameter of feature point selection by introducing position uncertainty as a criterion. Feature points are classified based on their position uncertainty values, and only those with low uncertainty (recognition reference feature points) are used for recognition. This parameter-based filtering maintains sufficient feature point quantity while significantly improving recognition accuracy.
Data Source
AI summary
An augmented reality content display method and an apparatus and system for performing the same are disclosed. The method includes extracting, by a feature point classification device, feature points from a plurality of images captured while changing a capture point such that at least one of a capture position and a capture orientation is changed, tracking, by the feature point classification device, feature points extracted from the same area of the plurality of images to associate the tracked feature points with one another, calculating, by the feature point classification device, uncertainty about a position of the associated feature points, and classifying and storing, by the feature point classification device, the associated features as a recognition reference feature point group or a dummy feature point group according to the position uncertainty.