SLAM Position Estimation Using Dual-Path Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SLAM techniques face challenges in achieving both quickness and accuracy in self-position estimation of objects, such as head-mounted displays, with existing methods either requiring long times for accurate results without accumulated errors or fast but less accurate results with accumulating errors.
Innovation Solution
A system that combines two SLAM techniques, one generating results without accumulated errors and another with accumulated errors, where the latter's results are corrected using the former's to achieve both quick and accurate self-position estimation by comparing and correcting the estimation results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If SLAM tracking is performed using only feature point movement for quick estimation, then processing speed is improved, but accumulated error increases over time
Solution Approach 1:
The patent segments the SLAM system into two distinct processing paths: a fast path using only feature point tracking for quick estimation, and a slow path incorporating additional constraints (ground plane, vanishing point) for accurate estimation. This segmentation allows the system to maintain both speed and accuracy by selecting appropriate processing levels based on operational needs.
Solution Approach 2:
The system dynamically switches between different estimation modes. The correcting unit activates correction processing when accumulated error exceeds a threshold, transitioning from fast-only mode to a mode that incorporates accurate estimation. This dynamic adjustment optimizes the balance between processing speed and estimation accuracy based on real-time error accumulation.
2Measurement precision
If accurate estimation without accumulated error is performed continuously, then measurement precision is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by implementing correction processing only when necessary (when accumulated error exceeds threshold) rather than continuously. The correcting unit selectively activates the accurate estimation path (slow path) to correct errors, while relying on the fast path for normal operation, thus reducing overall processing time while maintaining accuracy when needed.
Solution Approach 2:
The system performs self-correction by monitoring its own accumulated error and autonomously activating correction processing when error thresholds are exceeded. The correcting unit uses the accurate estimation results to correct the fast estimation results, allowing the system to self-regulate its accuracy without continuous external intervention.
3Measurement precision
If multiple sensors are used to improve estimation accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent makes the single camera perform multiple functions by extracting different types of information from it: feature points for fast tracking, ground plane information for accurate estimation, and vanishing point data for orientation correction. This multi-functional use of a single sensor achieves accurate estimation without increasing hardware complexity or adding multiple sensors.
Solution Approach 2:
The system changes the processing parameters and depth of analysis rather than adding hardware. By adjusting which image features are extracted and how deeply they are analyzed (feature points only vs. feature points plus ground plane plus vanishing point), the system achieves variable estimation accuracy using the same camera hardware, avoiding increased device complexity.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A first estimating unit 44 estimates at least one of a position and an attitude of a predetermined object on the basis of an image of a periphery of the object, the image being obtained from an imaging device, and generates an estimation result not including an accumulated error. A second estimating unit 46 estimates at least one of the position and the attitude of the object on the basis of the image, and generates an estimation result including an accumulated error. A correcting unit 50 compares the estimation result of the first estimating unit 44 and the estimation result of the second estimating unit 46 with each other, and corrects, on the basis of a result of the comparison, a subsequent estimation result of the second estimating unit 46, the subsequent estimation result being subsequent to the estimation result of the second estimating unit 46 which estimation result is used for the comparison. An App executing unit 38 performs predetermined data processing on the basis of the estimation result of the second estimating unit 46 which estimation result is corrected by the correcting unit 50.