Visual Odometry Fusion for Monocular Camera Target Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately tracking targets using monocular camera systems due to the lack of depth information, especially in low-speed maneuvers and complex surface conditions, which affects odometry calculations and vehicle maneuvering.
Innovation Solution
The implementation of a target tracker apparatus that combines Ackermann odometry with visual odometry methods like ground feature tracking and global motion vector estimation, characterizing the surface type and quality to select the most effective method, and fusing data to improve odometry parameter calculation and vehicle command generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If Ackermann odometry is used for vehicle maneuvering, then vehicle control is simple, but odometry accuracy deteriorates in low-speed maneuvers and complex surface conditions
Solution Approach 1:
The patent combines Ackermann odometry with visual odometry methods (ground feature tracking and global motion vector estimation) to create a hybrid system. The Ackermann odometry provides basic vehicle control simplicity while visual odometry compensates for accuracy losses in low-speed and complex surface conditions, resolving the contradiction between ease of operation and measurement precision.
Solution Approach 2:
The patent introduces visual odometry as an intermediary component that mediates between the simple Ackermann odometry model and the actual vehicle motion. The visual odometry system processes camera data to provide correction factors and alternative motion estimates, acting as a bridge that maintains computational simplicity while improving accuracy without requiring complex hardware changes.
2Measurement precision
If visual odometry is added to correct Ackermann odometry errors, then odometry accuracy is improved, but device complexity increases
Solution Approach 1:
The patent makes the existing monocular camera system perform multiple functions: it continues to provide basic visual feedback for driver awareness while simultaneously executing visual odometry algorithms to correct Ackermann odometry errors. This multi-functionality approach improves measurement precision without adding dedicated hardware, thereby limiting the increase in device complexity.
Solution Approach 2:
The system uses the vehicle's existing camera infrastructure to serve the additional function of visual odometry. By repurposing already-deployed sensors rather than adding new hardware, the patent achieves improved odometry accuracy while minimizing the increase in device complexity and avoiding additional computational resource requirements.
3Measurement precision
If surface characterization is performed to select visual odometry method, then odometry accuracy in diverse conditions is improved, but computational resources increase
Solution Approach 1:
The patent implements dynamic selection of visual odometry methods based on real-time surface characterization. The system analyzes surface features from camera data and adaptively switches between ground feature tracking and global motion vector estimation approaches, optimizing accuracy for diverse surface conditions while managing computational resources through conditional processing rather than continuous execution of all algorithms.
Solution Approach 2:
The patent changes operational parameters (algorithm selection, processing intensity) based on detected surface conditions. By monitoring surface characteristics and adjusting the computational approach accordingly, the system improves odometry accuracy in diverse conditions while avoiding unnecessary computational expenditure in scenarios where simpler methods suffice, thus balancing precision with energy consumption.
Data Source
AI summary
Methods, apparatus, systems, and articles of manufacture are disclosed to perform visual odometry using a monocular camera. An example apparatus includes a method fusioner to determine a visual odometry method using a surface characterization to calculate first odometry data, a displacement determiner to calculate second odometry data based on sensor information, an odometry calculator to calculate an odometry parameter of a target object with respect to a vehicle based on the first and the second odometry data, and a vehicle command generator to generate a vehicle command based on the odometry parameter.