Low Feature Object Detection Using Two-Stage Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality systems face challenges in accurately detecting and estimating the pose of low feature objects due to difficulties in detecting stable keypoints and computing distinctive appearance-based descriptors, leading to false detections in regions with uniform intensity or intensity variations.
Innovation Solution
The method involves performing first template-matching using intensity-related data sets to generate response maps and then second template-matching using shape-related feature data sets to derive the pose of low feature objects, enhancing accuracy in object detection and pose estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stable keypoints and appearance-based descriptors are used for object detection, then detection accuracy is improved for rich feature objects, but detection fails for low feature objects with uniform intensity
Solution Approach 1:
The patent changes the detection parameters from appearance-based descriptors to intensity-based and shape-based descriptors. This allows the system to detect low feature objects by comparing intensity distributions and shape characteristics rather than relying on texture or appearance features that don't exist on uniform objects.
Solution Approach 2:
The patent segments the detection process into two distinct stages: first performing intensity-based template matching to generate response maps and identify candidate regions, then performing shape-based template matching on those candidates. This segmentation allows each stage to specialize in what it does best.
2Productivity
If traditional template matching is used on low feature objects, then processing speed is maintained, but false detections increase in uniform intensity regions
Solution Approach 1:
The patent divides template matching into two sequential stages: intensity-based matching first to generate response maps and filter candidate regions, then shape-based matching on reduced candidates. This segmentation maintains speed by avoiding exhaustive shape matching on all possible locations while improving reliability through the two-stage verification process.
Solution Approach 2:
The patent performs preliminary intensity-based template matching to generate response maps and identify candidate regions of interest before performing the more computationally intensive shape-based matching. This preliminary action filters out false detection locations early, improving reliability without significantly impacting overall processing speed.
Data Source
AI summary
A method includes acquiring, from a camera, an image data sequence of a real object in a real scene and performing a first template-matching on an image frame in the image data sequence using intensity-related data sets stored in one or more memories to generate response maps. The intensity-related data sets represent an intensity distribution of a reference object from respective viewpoints. The reference object corresponds to the real object. A candidate region of interest is determined for the real object in the image frame based on the response maps, and second template-matching is performed on the candidate region of interest using shape-related feature data sets stored in one or more memories to derive a pose of the real object. The shape-related feature data sets represent edge information of the reference object from the respective viewpoints.