Mobile Robot Position Recognition Using Edge-Based Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile robots face challenges in accurately recognizing their position due to errors in dead reckoning navigation, especially when using monocular cameras, and the high cost of laser scanners, limiting their mobility and service capabilities.
Innovation Solution
An apparatus and method using edge-based refinement, which projects seeds onto stereo images to calculate the sum of edge sizes and update camera rotation and movement parameters, estimating the position based on visual and internal odometry information, and performing dense stereo alignment to obtain obstacle distance information without expensive sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If monocular camera is used for position recognition, then device cost is reduced, but measurement precision deteriorates due to inability to determine distance to feature points
Solution Approach 1:
The patent combines monocular camera imaging with inertial measurement unit (IMU) data and edge-based refinement algorithms to achieve accurate position recognition. By merging multiple data sources (camera images, acceleration data, gyro data) and processing them through edge-based refinement, the system overcomes the limitation of monocular cameras in determining absolute distance while maintaining low device cost.
Solution Approach 2:
The patent replaces traditional mechanical/laser-based ranging systems with a computational approach using monocular vision combined with inertial measurement. Instead of using expensive laser scanners or stereo cameras, the system uses algorithmic processing of monocular images combined with IMU data to achieve precise position and orientation estimation.
2Measurement precision
If laser scanner is used for position recognition, then measurement precision is improved, but device cost increases significantly
Solution Approach 1:
The patent replaces expensive laser scanners with inexpensive monocular cameras and IMUs. The system uses disposable-like computational resources (processing power for edge-based refinement) instead of expensive hardware, achieving similar or better performance at a fraction of the cost.
Solution Approach 2:
The patent substitutes mechanical laser scanning systems with a computational vision system. Instead of using physical laser beams to measure distances, the system uses image processing algorithms combined with inertial measurement to estimate position and orientation, eliminating the need for expensive laser scanners.
3Device complexity
If dead reckoning navigation is used, then device complexity is reduced, but reliability deteriorates due to accumulated errors
Solution Approach 1:
The patent implements feedback by continuously comparing edge-based observations with predicted positions from dead reckoning. The system uses the difference between observed and predicted positions to correct accumulated errors, creating a closed-loop navigation system that maintains reliability while keeping device complexity low.
Solution Approach 2:
The patent introduces edge-based refinement as an intermediary mechanism between dead reckoning and final position estimation. The edge features extracted from images serve as intermediate reference points that help correct drift accumulation, acting as a mediator that bridges the gap between simple dead reckoning and complex sensor fusion.
Data Source
AI summary
Disclosed are an apparatus of recognizing a position of a mobile robot using edge based refinement and a method thereof. An apparatus of recognizing a position of a mobile robot includes two wide angle cameras which obtain one pair of stereo images on a region of interest including a vertical direction and a horizontal direction in accordance with movement of a mobile robot; and a position recognizing unit which extracts an edge from the obtained stereo images, projects each seed in a key frame selected based on the extracted edge onto the stereo images, updates a status parameter indicating rotation and movement of an optical camera along a direction in which the calculated cost is increased in accordance with the projected seed, and estimates a current position using the updated result.