3D Scene Geometry from Laser and Image Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Panoramic images alone do not provide a sufficiently realistic visual experience for users, as they lack three-dimensional features, and existing methods struggle to accurately reconstruct 3D geometry, especially for surfaces orthogonal to the vehicle trajectory.
Innovation Solution
A method that combines panoramic images with distance measurements from both side-facing and front-facing devices to generate 3D scene geometry, merging data from multiple vehicle trajectories to create a complete 3D model, and includes a system for removing spurious objects by segmenting and classifying depth measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If panoramic images are used alone to represent scenes, then the system is simple and easy to operate, but the visual realism and three-dimensional experience are insufficient
Solution Approach 1:
The patent combines panoramic images with laser distance measurements to create a hybrid system that delivers both visual realism and three-dimensional information. By merging photogrammetric data with LIDAR range data, the system achieves enhanced visual fidelity without requiring a complete system overhaul
Solution Approach 2:
The patent transitions from two-dimensional panoramic images to three-dimensional scene representations by integrating depth information from laser measurements. This dimensional enhancement provides authentic 3D visual experience while maintaining compatibility with existing panoramic imaging systems
2Loss of information
If distance measurements from multiple vehicle trajectories are collected to improve 3D geometry completeness, then the completeness of scene geometry improves, but the complexity of data processing and registration increases
Solution Approach 1:
The patent introduces a reference trajectory as an intermediary framework to which all other vehicle trajectories are registered. This mediator enables systematic alignment of multiple trajectories through relative pose estimation, making the integration of multi-trajectory data tractable and organized
Solution Approach 2:
The patent divides the scene into multiple segments captured from different vehicle trajectories, each processed independently before being integrated into the complete 3D model. This segmentation allows parallel processing and reduces the computational burden of handling all data simultaneously
3Reliability
If laser distance measurements are used to generate 3D geometry, then three-dimensional features are achieved, but spurious objects and measurement errors are introduced
Solution Approach 1:
The patent implements consistency checks and validation mechanisms that compare laser measurements against panoramic image data and geometric constraints. This feedback loop identifies and filters spurious measurements, ensuring that only geometrically consistent data contributes to the final 3D model
Solution Approach 2:
The patent uses the presence of spurious objects and measurement anomalies as opportunities to refine the processing algorithm. By identifying and correcting these errors, the system develops more robust validation rules that improve overall measurement accuracy and filter out future spurious data
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the realism of panoramic views by generating accurate 3D models, including orthogonal surfaces, and improves the completeness of scene geometry by integrating data from multiple perspectives, while removing spurious objects to focus on primary features.
Implementation Method 1
a laser device that measures a distance to an object in the scene
Data Source
AI summary
Photographic images can be used to enhance three-dimensional (3D) virtual models of a physical location. In an embodiment, a method of generating a 3D scene geometry includes obtaining a first plurality of images and corresponding distance measurements for a first vehicle trajectory; obtaining a second plurality of images and corresponding distance measurements for a second vehicle trajectory, the second vehicle trajectory intersecting the first vehicle trajectory; registering a relative vehicle position and orientation for one or more segments of each of a first vehicle trajectory and a second vehicle trajectory; generating a three-dimensional geometry for each vehicle trajectory; mapping the three-dimensional geometries for each vehicle trajectory onto a common reference system based on the registering; and merging the three-dimensional geometries from both trajectories to generate a complete scene geometry.