XR Object Focus Detection Using Segmentation And Reticle Overlap
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing extended reality (XR) systems struggle to accurately determine whether an object of interest is in the center of focus, particularly in mixed reality (MR) and augmented reality (AR) environments, where virtual objects interact with the real world, leading to challenges in precise targeting and image analysis.
Innovation Solution
A computer system that includes a processor and storage system to obtain images and pose data, perform target detection and segmentation, and determine if a reticle overlaps with the object of interest, using a two-stage hierarchical algorithm combining supervised object detection and unsupervised foreground-background segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional target detection methods are used in XR systems, then the system can identify objects of interest, but the accuracy of determining whether the object is in the center of focus deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the target detection process into two distinct stages: (1) initial target detection to identify the object of interest and generate a bounding box, and (2) reticle overlap analysis to determine if the object is in the center of focus. This segmentation allows each stage to optimize for its specific function, improving overall measurement precision while preserving center of focus information.
Solution Approach 2:
The patent introduces an intermediary reticle element as a mediator between the camera view and the target object. The reticle serves as a reference frame that overlays the bounding box, enabling the system to determine whether the object is centrally positioned without losing focus determination accuracy. This intermediary element bridges the gap between detection and focus assessment.
2Measurement precision
If a two-stage hierarchical algorithm is implemented, then targeting accuracy is improved, but device complexity increases
Solution Approach 1:
The algorithm is segmented into two hierarchical stages: a first stage using a trained machine learning model for initial target detection and bounding box generation, and a second stage performing reticle overlap analysis. This segmentation allows computational complexity to be distributed appropriately, with the first stage handling object identification and the second stage handling precision focus determination, thereby improving targeting accuracy while managing device complexity through structured processing.
Solution Approach 2:
The first stage of the algorithm performs preliminary action by detecting the target object and generating a bounding box before the second stage conducts the reticle overlap analysis. This preliminary detection prepares the data structure and identifies the region of interest, reducing the computational burden on the second stage and allowing the system to achieve high targeting accuracy without overwhelming device complexity.
3Productivity
If rapid target acquisition is implemented, then user interaction speed is improved, but measurement precision may deteriorate
Solution Approach 1:
The system performs preliminary target detection and bounding box generation in the first stage, preparing the necessary data structures and identifying the object of interest before conducting the reticle overlap analysis. This preliminary action enables rapid target acquisition by having the foundational detection work completed first, while the subsequent precision focus determination maintains measurement accuracy through the structured two-stage approach.
Solution Approach 2:
The segmented algorithm structure allows the system to rapidly complete the first stage of target detection using efficient machine learning models, achieving quick target acquisition. The second stage then performs the more computationally intensive reticle overlap analysis to ensure measurement precision is maintained. This segmentation enables the system to balance speed and accuracy by optimizing each stage for its specific requirement.
Data Source
AI summary
Techniques for determining whether an object of interest (OOI) is the center of focus are described herein. An image of an environment is obtained. This image includes pixel content representative of the OOI. Pose data for the camera that generated the image is obtained. A bounding element is generated and surrounds the pixel content that represents the OOI. Subsequently, a segmentation process is performed on the pixels surrounded by the bounding element. This segmentation identifies pixels that represent the OOI and pixels that do not represent the OOI. This segmentation also identifies a silhouette of the OOI. A determination is made as to whether the pixels that represent the OOI are at the center of focus.