Camera Re-Localization With Prior Model Confidence Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera re-localization methods in augmented reality applications often output inaccurate pose estimates, leading to incorrect virtual content placement and delayed application initiation due to false matches or insufficient confidence scores.
Innovation Solution
Fusing computer vision algorithm outputs with a prior model of the geographic area to generate updated confidence scores by mapping pose estimates to a heatmap-style histogram, combining location probabilities and algorithmic scores for accurate re-localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computer vision algorithms are used to estimate pose from camera frames, then pose estimation can be performed, but the accuracy of the pose estimates deteriorates leading to false matches
Solution Approach 1:
The patent combines computer vision algorithm outputs with a prior model (heatmap-style histogram) to generate updated confidence scores. This merging of two independent data sources (algorithmic confidence and prior location probability) resolves the contradiction by using their complementary strengths to achieve both reliable detection and precise measurement.
Solution Approach 2:
The prior model acts as an intermediary that mediates between the computer vision algorithm and the final pose determination. It transforms the algorithm's confidence score into an updated confidence score by incorporating prior knowledge about likely user locations, thereby improving both reliability and precision.
2Reliability
If a high confidence score threshold is used to ensure accurate pose estimation, then false matches are reduced, but the time to initiate the application increases due to repeated failed attempts
Solution Approach 1:
The system performs preliminary action by using the prior model to predict likely user locations before final pose determination. This preliminary assessment allows the system to prepare and prioritize candidate poses, reducing the time needed to achieve accurate pose estimation while maintaining high reliability.
Solution Approach 2:
The updated confidence score provides feedback that combines algorithmic confidence with prior model probability. This feedback mechanism allows the system to make more informed decisions faster, reducing the number of failed attempts and the time to initiate the application while maintaining accurate pose estimation.
3Measurement precision
If the system makes multiple attempts at camera re-localization to achieve accurate pose estimation, then pose accuracy improves, but application initiation is delayed frustrating the user
Solution Approach 1:
The prior model performs preliminary action by pre-computing location probabilities and preparing a heatmap-style histogram of likely user positions. This preliminary preparation enables the system to quickly evaluate candidate poses without requiring multiple time-consuming attempts, thus achieving precise pose estimation rapidly.
Solution Approach 2:
The system changes the parameter of confidence scoring by introducing updated confidence scores that incorporate prior model probabilities. This parameter transformation allows the system to achieve precise pose estimation in fewer attempts by using more informative confidence metrics.
Data Source
AI summary
The present disclosure describes approaches to camera re-localization that improve the speed and accuracy with which pose estimates are generated by fusing output of a computer vision algorithm with data from a prior model of a geographic area in which a user is located. For each candidate pose estimate output by the algorithm, a game server maps the estimate to a position on the prior model (e.g., a specific cell on a heatmap-style histogram) and retrieves a probability corresponding to the mapped position. A data fusion module fuses, for each candidate pose estimate, a confidence score generated by the computer vision algorithm with the location probability from the prior model to generate an updated confidence score. If an updated confidence score meets or exceeds a score threshold, a re-localization module initiates a location-based application (e.g., a parallel reality game) based on the associated candidate pose estimate.