Virtual Binocular Image Generation for Monocular Pose Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Monocular visual odometry struggles with scale loss and accuracy in harsh conditions like linear motion and illumination changes, often requiring additional sensors, which increase costs and complexity, and end-to-end learning methods provide only probabilistic pose estimation that does not meet actual requirements.
Innovation Solution
An image-based pose determination method that acquires current and historical image frames from a camera on a moving object, determines the relative camera pose, generates a virtual binocular image, and uses preset visual odometry to determine the camera's pose, effectively achieving binocular visual odometry without a binocular camera, thereby reducing costs and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If monocular visual odometry is used, then cost is reduced, but measurement precision deteriorates due to scale loss and inability to obtain real physical scale
Solution Approach 1:
The patent creates a virtual binocular image by copying and transforming the monocular image through projection and back-projection operations. This virtual copy simulates binocular vision effects, enabling scale recovery and improved measurement precision while maintaining the cost advantage of monocular hardware
Solution Approach 2:
The patent introduces a virtual binocular image as an intermediary between the monocular input and the pose estimation process. This intermediary enables the system to achieve binocular-like measurement precision by providing depth information and scale recovery without requiring actual binocular hardware
2Measurement precision
If additional sensors such as IMU or binocular camera are added, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent enables the monocular camera to serve multiple functions by processing its own output image through virtual binocular image generation. The single camera performs both monocular capture and, through computational processing, achieves binocular-like measurement capabilities, eliminating the need for additional sensors and reducing device complexity
3Ease of operation
If end-to-end learning method is used, then ease of operation improves, but measurement precision deteriorates due to probabilistic pose estimation
Solution Approach 1:
The patent replaces the probabilistic end-to-end learning approach with a deterministic geometric computation system. By using projection and back-projection operations based on camera poses and intrinsic parameters, the system achieves precise pose estimation through mathematical calculations rather than probabilistic predictions, improving measurement precision while maintaining ease of operation
Data Source
AI summary
Embodiments of the present disclosure disclose an image-based pose determination method and apparatus, a computer readable storage medium, and an electronic device. The method include: acquiring a current image frame captured by a camera disposed on a moving object and a historical image frame captured before the current image frame; determining a first relative camera pose between the current image frame and the historical image frame; determining a virtual binocular image based on the first relative camera pose; and determining current pose information of the camera based on preset visual odometry and the virtual binocular image. According to the embodiments of the present disclosure, the virtual binocular image may be generated based on a monocular image, thus achieving effects of binocular visual odometry without using a binocular camera, thereby reducing costs. Moreover, monocular odometry may be enabled to obtain real physical scale of a space where the moving object is located, thereby improving accuracy of the monocular odometry in determining a position and a pose of the camera.