Vehicle Image Position Correction via 3D Model and Aerial Reference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for improving position information in panoramic images captured by vehicle-mounted cameras, especially in urban areas, face challenges due to inaccurate GPS data and the requirement for significant overlap and availability of individual images, which limits their applicability and accuracy.
Innovation Solution
A method that constructs a 3D model using overlapping images taken by the same camera at different positions along a trajectory, with positional and orientation data, and corrects positional offsets using aerial images, allowing for a large continuous overlapping region without requiring specific object knowledge or full image availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS data is used for positioning in urban areas, then position information can be obtained quickly, but accuracy deteriorates due to multipath phenomena and signal blocking
Solution Approach 1:
The patent uses aerial images as an intermediary reference to correct ground-based image positions. Instead of relying directly on inaccurate GPS data, the method introduces aerial images (obtained from drone or aircraft) as a mediator that provides accurate positional reference, eliminating the harmful effects of urban multipath phenomena on positioning accuracy
Solution Approach 2:
The patent replaces the GPS-based positioning system with an image-matching-based positioning system. By substituting the mechanical/electronic GPS positioning mechanism with an optical/image processing approach (comparing aerial and ground images to determine positions), the system achieves higher accuracy in urban environments where GPS fails
2Measurement precision
If a stereoscopic camera setup is used to capture road marks, then depth information can be obtained, but device complexity increases and requires specific object visibility
Solution Approach 1:
The patent extracts the essential function of depth measurement from the complex stereoscopic camera system. By using a single moving camera to capture images at different positions and using image matching with aerial references, the method extracts depth and position information without requiring a complex fixed stereoscopic setup
Solution Approach 2:
The patent makes the camera system universal by using a single camera that can capture both ground-level images for panoramic reconstruction and, through movement and image matching, derive depth information. This multi-functional approach replaces the specialized stereoscopic setup, reducing device complexity while maintaining measurement capabilities
3Measurement precision
If significant overlap between images is required for 3D model construction, then measurement accuracy improves, but productivity decreases due to longer data collection time
Solution Approach 1:
The patent performs preliminary action by obtaining aerial images before or during the ground-based image collection. These aerial images serve as pre-established reference data that enables accurate positioning without requiring extensive overlap in the ground images themselves, thus reducing data collection time while maintaining accuracy
Solution Approach 2:
The patent uses partial action by requiring only minimal overlap between ground images, rather than the significant overlap needed in traditional methods. The aerial images provide the additional constraint information needed for accurate 3D reconstruction, allowing the ground-based collection to be faster and more efficient
Data Source
AI summary
A method and system for improving position information associated with a collection of images is disclosed, more in particular, a method and system for improving position information of images obtained using a camera system including at least one camera mounted on or to a vehicle. A 3D model is constructed from an overlapping region in which at least two images obtained at different positions of the vehicle among the collection of images at least partially overlap. By comparing the 3D model to an aerial image, the position of the 3D model can be updated.