Dynamic Weighting Position Estimation for Visual SLAM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The reliability of position estimation in autonomous systems is compromised by environmental factors such as brightness and calibration errors, leading to low accuracy in estimating the current position of moving objects like robots and cars.
Innovation Solution
A position estimation device that creates multiple virtual positions and images, compares them with actual images, and calculates weights based on acquired information and position errors to correct the current position, ensuring accurate matching even in challenging conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed weight is assigned to the matching result of the intensity image and the matching result of the parallax image, then the calculation process is simple, but the reliability of information acquired by the imaging device is lowered due to brightness of traveling environment or calibration error
Solution Approach 1:
The patent applies dynamics by making the weight coefficients dynamic rather than fixed. The weight for the intensity image matching result and the weight for the parallax image matching result are adjusted based on detection results from the imaging device, allowing the system to adapt to varying environmental conditions such as brightness levels and calibration accuracy, thereby improving information reliability without excessive complexity
Solution Approach 2:
The patent changes parameters by adjusting the weight coefficients based on detection results. When the imaging device detects good lighting conditions and low calibration error, the system assigns higher weight to the intensity image matching result. Conversely, in poor lighting or high calibration error conditions, it increases the weight of the parallax image matching result, optimizing position estimation accuracy under different conditions
2Ease of operation
If fixed weight is used for matching results, then the system is simple to operate, but the current position accuracy estimated by matching becomes low in challenging environments
Solution Approach 1:
The system applies self-service by automatically adjusting weight coefficients based on its own detection capabilities. The imaging device continuously monitors environmental conditions and calibration status, and the control device automatically recalibrates weights without user intervention, maintaining both operational simplicity and high position accuracy in varying conditions
3Adaptability or versatility
If SLAM method is used to create map and estimate relative position, then position estimation can be performed without map information or GPS, but position error is accumulated requiring frequent correction
Solution Approach 1:
The patent applies feedback by using the imaging device to detect surrounding information and compare it with map information, then feeding back correction data to adjust the accumulated position error. The system continuously monitors the matching result between detected objects and map data, and automatically corrects drift accumulation by recalculating position based on visual landmarks, maintaining long-term position accuracy
Data Source
AI summary
Provided is a position estimation device capable of highly accurate position estimation. A position estimation device 1 of the present invention is the position estimation device 1 which estimates a current position of a moving object 100 equipped with an imaging device 12, estimates the current position of the moving object 100, create a plurality of virtual positions based on the current position, creates virtual images at the plurality of virtual positions, respectively, compares the plurality of virtual images with an actual image to calculate a comparison error, calculates a weight based on at least one of information acquired by the imaging device 12 and information of a current position error of the moving object, performs weighting on the comparison error using the weight, and corrects the current position based on the comparison error to be weighted.