Hardware-Accelerated Image Feature Matching for AR Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image feature matching methods in augmented reality and computer vision face challenges with high processing time and power requirements, especially in textured and non-textured environments, and lack flexibility and distinctiveness in feature descriptors, leading to reduced matching quality and robustness.
Innovation Solution
Implementing a method on an integrated circuit with hardwired or configurable logic to determine similarity measures between current and reference feature descriptors, enabling faster and more robust image feature matching by optimizing hardware and adapting software for real-time processing with reduced power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If exhaustive search or approximate nearest neighbor search is used for feature matching, then matching completeness is improved, but processing time increases significantly
Solution Approach 1:
The patent segments the feature matching process into two distinct stages: a fast hardware-accelerated stage that computes similarity measures for all reference features using dedicated logic circuits, and a subsequent software stage that performs selective refinement. This segmentation allows the system to quickly eliminate non-matching features in hardware while maintaining comprehensive matching through the subsequent software verification stage.
Solution Approach 2:
The patent introduces a hardware intermediary component that acts as a bridge between the input feature descriptors and the matching algorithm. This hardware block pre-computes similarity measures and filters candidate matches before passing results to the software processor, effectively mediating between the need for comprehensive searching and the constraint of processing time.
2Reliability
If more reference features are used to improve matching distinctiveness, then matching quality is improved, but processing complexity increases
Solution Approach 1:
The patent divides the processing workload by segmenting reference features into those that can be efficiently evaluated in hardware based on simple similarity measures, and those requiring more complex software-based evaluation. This allows the system to handle large numbers of reference features without overwhelming the processing system.
Solution Approach 2:
The patent applies partial action by computing similarity measures for all reference features in hardware but only performing full verification and selection for promising candidates. This excessive computation in the hardware stage ensures no potential match is missed, while the software stage selectively processes only the most promising candidates, balancing completeness with efficiency.
3Adaptability or versatility
If feature matching is performed in software only, then flexibility is maintained, but processing speed decreases
Solution Approach 1:
The patent segments the processing system into hardware and software components, each handling different aspects of the matching task. The hardware segment provides fast, fixed-functionality processing for similarity measure computation, while the software segment maintains flexibility for algorithm adjustments and complex verification, achieving both speed and adaptability.
Solution Approach 2:
The patent substitutes the mechanical/software-based feature matching process with a hardware-accelerated system for the computationally intensive similarity measure computation. This substitution replaces general-purpose software processing with dedicated hardware logic, dramatically improving speed while the software layer retains algorithmic flexibility.
4Productivity
If hardware acceleration is implemented for feature matching, then processing speed is improved, but power consumption increases
Solution Approach 1:
The patent applies partial action by using hardware acceleration only for the specific task of computing similarity measures, rather than accelerating the entire feature matching pipeline. This selective hardware intervention provides speed improvement for the most computationally intensive operation while keeping power consumption moderate by leaving other operations in software.
Solution Approach 2:
The patent extracts and isolates the similarity measure computation task from the overall feature matching process, implementing it in hardware as a separate, dedicated function. This extraction allows the system to accelerate only the necessary computation while avoiding the power overhead of hardware-accelerating the entire pipeline, including verification and selection operations.
Data Source
AI summary
The invention is related to a method of matching image features with reference features, comprising the steps of providing a current image captured by a capturing device, providing reference features (r), wherein each of the reference features comprises at least one reference feature descriptor (d(r)), determining current features (c) in the current image and associating with each of the current features at least one respective current feature descriptor (d(c)), and matching the current features with at least some of the reference features by determining a respective similarity measure (D(c, r)) between each respective current feature descriptor (d(c)) and each respective reference feature descriptor (d(r)). According to the invention, the determination of the similarity measure is performed on an integrated circuit by hardwired logic or configurable logic which processes logical functions for determining the similarity measure. The invention is also concerned with an integrated circuit for matching of image features with reference features.