Vision Probe Scoring for Deformation-Tolerant Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vision system techniques for aligning trained models with runtime images face challenges due to deformation and optical issues, requiring computationally expensive operations like feature extraction and smoothing, which can lead to unstable and inaccurate scoring.
Innovation Solution
A system and method that uses pre-generated probes with normalized directions and offsets to score candidate poses in runtime images, accommodating local deformation without explicit feature extraction or downsampling, allowing for rapid and accurate alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature extraction and neighborhood search are performed to match probes in deformed images, then measurement precision is improved, but computing time increases significantly
Solution Approach 1:
Gradient values and angles are pre-computed and stored in a lookup table during image acquisition, eliminating the need for real-time gradient calculations. This preliminary action allows the matching process to directly query pre-computed values, significantly reducing computing time while maintaining alignment accuracy
Solution Approach 2:
The patent extracts only the essential gradient information (magnitude and angle) at probe locations and stores it in a compact lookup table, rather than performing full feature extraction and neighborhood search. This selective extraction removes unnecessary computational steps while preserving the critical information needed for accurate alignment scoring
2Device complexity
If a hard contrast threshold is used to identify featurelets, then device complexity is reduced, but reliability deteriorates due to unstable behavior near threshold
Solution Approach 1:
Instead of using a fixed hard threshold, the patent changes the approach by pre-computing gradient values and angles for all pixels and storing them in a lookup table. This parameter transformation allows the system to access gradient information without applying thresholding operations during runtime, eliminating the instability associated with hard thresholds while maintaining simplicity
3Measurement precision
If gradient computation and non-peak suppression are performed at every location, then measurement precision is improved, but use of energy increases due to computational expense
Solution Approach 1:
Gradient computations are performed as a preliminary action during image acquisition and the results are cached in a lookup table. This eliminates the need to re-compute gradients during the matching process, significantly reducing energy consumption while maintaining gradient accuracy for precise alignment scoring
Data Source
AI summary
A system and method for scoring trained probes for use in analyzing one or more candidate poses of a runtime image is provided. A set of probes with location and gradient direction based on a trained model are applied to one or more candidate poses based upon a runtime image. The applied probes each respectively include a discrete set of position offsets with respect to the gradient direction thereof. A match score is computed for each of the probes, which includes estimating a best match position for each of the probes respectively relative to one of the offsets thereof, and generating a set of individual probe scores for each of the probes, respectively at the estimated best match position.