Hybrid Neural Network and Corner Detector for Head-Worn Displays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current head-worn image display devices face challenges in accurately detecting features, such as corners, in images, especially in high noise conditions and low light environments, due to their local nature and high computational resource requirements, which affect the creation of localization maps and user immersion in mixed reality scenarios.
Innovation Solution
A hybrid approach combining a neural network with a corner detector for real-time feature extraction, where the neural network identifies feature candidates and a corner detection algorithm selects good features, leveraging a larger image context and reducing computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a corner detection algorithm is applied to the entire image, then corner detection accuracy is improved, but computational resource consumption increases
Solution Approach 1:
The patent divides the image processing task into two stages: first, a neural network identifies candidate regions of interest; second, the corner detection algorithm is applied only to these smaller candidate regions rather than the entire image. This segmentation reduces computational load while maintaining detection accuracy in the identified regions.
Solution Approach 2:
The patent applies different processing strategies to different parts of the image: the neural network provides a coarse global analysis to identify promising regions, while the corner detection algorithm provides fine-grained local analysis only where needed. This local quality approach optimizes resource allocation by applying intensive processing only to relevant image portions.
2Measurement precision
If feature detection is performed in high noise and low light conditions, then localization accuracy is improved, but detection reliability deteriorates
Solution Approach 1:
The patent merges two different detection approaches: a neural network that can identify potential feature locations even in noisy conditions, and a traditional corner detection algorithm that provides reliable corner identification. By combining these methods, the system achieves both robustness to noise and reliable corner detection.
Solution Approach 2:
The neural network acts as an intermediary that pre-processes the image data to identify candidate regions before the corner detection algorithm is applied. This intermediary step filters out noisy areas and directs the reliable corner detector to focus on promising regions, thereby maintaining detection reliability in challenging conditions.
3Productivity
If a neural network is used for feature identification, then feature candidate identification speed is improved, but device complexity increases
Solution Approach 1:
The patent implements a dynamic two-stage processing system where the neural network operates first to quickly identify candidate regions, and then the corner detection algorithm is dynamically applied only to those candidates. This dynamic approach optimizes processing speed while managing complexity by avoiding the need to run both full algorithms on the entire image.
Data Source
AI summary
An apparatus configured to be head-worn by a user, includes: a screen configured to present graphics for the user; a camera system configured to view an environment in which the user is located; and a processing unit coupled to the camera system, the processing unit configured to: obtain locations of features for an image of the environment, wherein the locations of the features are identified by a neural network; determine a region of interest for one of the features in the image, the region of interest having a size that is less than a size of the image; and perform a corner detection using a corner detection algorithm to identify a corner in the region of interest.