2D to 3D Part Identification via Pose Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing approaches face difficulties in accurately identifying parts in photographs due to variations in orientation and lack of sufficient information, especially when dealing with complex assemblies.
Innovation Solution
A system and method that utilize a database of reference photographs from different orientations to determine the pose of an assembly in a query photograph, relating two-dimensional pixels to a three-dimensional representation for part identification, and transferring part information from engineering data to the query photograph.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If image processing approaches are used to recognize objects in photographs, then object recognition capability is provided, but accuracy deteriorates due to variations in orientation and insufficient information
Solution Approach 1:
The patent transitions from two-dimensional image processing to three-dimensional spatial reasoning by establishing correspondences between 2D photograph locations and 3D assembly space. This dimensional extension enables accurate part identification despite orientation variations, as the system can map photographed views to their corresponding three-dimensional positions and identify parts based on spatial relationships rather than just 2D image patterns.
2Measurement precision
If reference photographs from different orientations are used to determine pose, then part identification accuracy is improved, but device complexity increases
Solution Approach 1:
The system pre-establishes a database of reference photographs from multiple orientations along with their corresponding three-dimensional spatial information and part labels. This preliminary preparation allows the system to quickly match incoming query photographs against the pre-organized reference data, determining pose and identifying parts without complex real-time computation, thus reducing operational complexity while maintaining high accuracy.
3Loss of information
If two-dimensional pixels are related to three-dimensional representation, then part information transfer is enabled, but processing time increases
Solution Approach 1:
The patent creates a virtual copy or mapping between the two-dimensional photograph space and three-dimensional assembly space. By establishing a correspondence relationship where each 2D pixel location can be mapped to its corresponding 3D spatial position, the system efficiently transfers part information without requiring complex real-time processing. The pre-computed spatial mapping acts as a copy that enables rapid information transfer from engineering data to the photograph.
Data Source
AI summary
A database stores reference photographs of an assembly. The reference photographs are from different orientations relative to the assembly. By matching the query photograph to one or more of the reference photographs, the pose of the assembly in the query photograph is determined. Based on the pose, the pixels of the two-dimensional query photograph are related to a three-dimensional representation from engineering data. Using labeled parts from the engineering data, the parts represented in the query photograph are identified, and part information (e.g., segmentation, number, or other metadata) is provided relative to the query photograph.