Vehicle Pose Determination Using Machine Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle navigation systems face challenges in accurately determining the six degree of freedom (DoF) pose of vehicles in complex environments, which is crucial for safe and efficient autonomous or semi-autonomous operation, especially in scenarios where traditional sensors like GPS and fiducial markers may be unreliable or incomplete.
Innovation Solution
The method involves using machine vision techniques, such as edge detection and the perspective-n-points algorithm, combined with convolutional neural networks trained on images of vehicles at various poses, to determine the six DoF pose of a vehicle based on CAD data and physical measurements, transforming this data into global coordinates for precise navigation without the need for fiducial markers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensors like GPS and fiducial markers are used for vehicle pose determination, then the system has simple implementation, but the measurement precision and reliability deteriorate in complex environments
Solution Approach 1:
The patent replaces traditional mechanical and optical marker-based measurement systems (GPS, fiducial markers) with a computer vision-based system using cameras and machine learning algorithms. The camera captures images of the vehicle, and a trained neural network model processes these images to determine the six-degree-of-freedom pose, eliminating the need for physical markers or complex sensor arrays while improving accuracy in complex environments.
Solution Approach 2:
The patent uses a pre-trained convolutional neural network model that has been trained on synthetic rendered images to process real camera images. The model learns to map image features to vehicle pose parameters through training on large datasets of rendered vehicle images at various poses, enabling accurate pose estimation without requiring fiducial markers or complex calibration procedures.
2Measurement precision
If machine vision techniques and neural networks are used to determine vehicle pose, then the measurement precision improves, but the loss of time increases due to complex processing
Solution Approach 1:
The patent performs pose determination in advance by capturing images and processing them through the trained neural network model before the vehicle needs to execute navigation actions. The system determines the six-degree-of-freedom pose (position and orientation) of the vehicle relative to the roadway ahead, allowing the autonomous driving system to plan and execute maneuvers based on this pre-computed pose information.
3Reliability
If fiducial markers are used for pose determination, then the ease of operation is maintained, but the reliability deteriorates when markers are unreliable or incomplete
Solution Approach 1:
The patent replaces the fiducial marker-based system with a camera-based computer vision system that detects natural features of the vehicle and roadway. The trained neural network model identifies vehicle geometry, lane markings, and road surface features to determine pose, eliminating the need for artificial markers while improving reliability in environments where markers may be obscured or unavailable.
Data Source
AI summary
A computer, including a processor and a memory, the memory including instructions to be executed by the processor to determine a vehicle six degree of freedom (DoF) pose based on an image where the six DoF pose includes x, y, and z location and roll, pitch, and yaw orientation and transform the vehicle six DoF pose into global coordinates based on a camera six DoF pose. The instructions can include further instructions to communicate to the vehicle the six DoF pose in global coordinates.