Image-Based Position Estimation With Distance-Aware Azimuth Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing position estimating devices for moving bodies, such as robots or vehicles, face errors in estimating the azimuth due to the assumption of constant feature point movement in the far region of images, which is not always accurate, especially when three-dimensional objects near the moving body appear in the far region.
Innovation Solution
A position estimating device that includes an image acquiring unit, a moving body movement amount estimating unit, and a moving body position estimating unit, which extracts feature points from images at different timings, tracks these points, calculates their distances, and estimates the azimuth and movement based on these distances, allowing for accurate position estimation by considering the distance of feature points from the moving body.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the amount of movement of far feature points is assumed to be constant for azimuth estimation, then the processing load is reduced, but the measurement precision of azimuth estimation deteriorates when three-dimensional objects near the moving body appear in the far region
Solution Approach 1:
The patent changes the parameter used for azimuth estimation from a constant value to a variable parameter (amount of movement of far feature points). By dynamically calculating the actual amount of movement instead of assuming it is constant, the system adapts to different environmental conditions where three-dimensional objects may appear in the far region, thus improving measurement precision while maintaining reasonable processing efficiency.
2Quantity of substance
If feature points from three-dimensional objects in the far region are used for azimuth estimation, then more feature points are available for processing, but the measurement precision of position estimation deteriorates due to incorrect distance assumptions
Solution Approach 1:
The patent extracts and removes feature points that belong to three-dimensional objects from the set of feature points used for odometry calculation. By identifying and excluding these problematic feature points, the system prevents them from degrading the position estimation accuracy, while still utilizing the remaining valid feature points for robust estimation.
Solution Approach 2:
The patent applies different processing quality to different feature points based on their characteristics. Far feature points that belong to three-dimensional objects are identified and handled differently (removed or weighted differently) compared to legitimate far feature points, allowing the system to maintain high measurement precision by treating each feature point according to its actual nature.
Data Source
AI summary
A position estimating device is provided with an image acquiring unit that acquires images around a moving body, a moving body movement amount estimating unit that estimates an amount of movement of the moving body based on the images, and a moving body position estimating unit that estimates the position of the moving body based on the estimated amount of movement. The moving body movement amount estimating unit includes a tracking unit that tracks a first feature point to a second feature point, a distance calculating unit that calculates the distance from the moving body to the second feature point, an azimuth estimating unit that estimates the azimuth of the moving body based on the distance from the moving body to the second feature point, and a movement amount estimating unit which estimates the amount of movement of the moving body based on the azimuth of the moving body.