3D Object Identification Using Multi-Angle Feature Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for identifying objects using three-dimensional digital models in factories suffer from accuracy issues due to interference caused by different view angles.
Innovation Solution
A method and system that generate synthesized images with different view angles, extract eigenvectors using CNN, fuse them, and train a classifier, followed by acquiring and fusing eigenvectors of actual images with controlled view angles to reduce angle-related interference, utilizing AutoML and neural architecture search for optimal fusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional methods use three-dimensional digital models for object identification, then the system can handle various objects, but the identification accuracy deteriorates due to interference from different view angles
Solution Approach 1:
The patent segments the view angle parameter by generating synthesized images at specific, discrete view angles (e.g., 0°, 45°, 90°, 135°) rather than attempting to handle all possible angles. This segmentation allows the system to capture essential characteristics at key angles while avoiding interference from intermediate angles, thereby resolving the contradiction between versatility and accuracy.
Solution Approach 2:
The patent performs preliminary action by pre-generating synthesized images at controlled view angles before actual object identification. These pre-generated images serve as reference data that encode characteristic information at specific angles, allowing the system to compare actual images against these references and eliminate angle-related interference in advance.
2Reliability
If synthesized images with multiple view angles are generated to embody more characteristics, then identification capability is improved, but the complexity of the system increases
Solution Approach 1:
The patent changes parameters by selectively controlling view angles rather than using all possible angles. Instead of generating synthesized images across the entire angular range, the system identifies and uses only the critical view angles that provide the most discriminative information, thereby maintaining high identification capability while reducing system complexity.
Solution Approach 2:
The patent achieves universality by creating a view angle control mechanism that can adapt to different objects and identification requirements. The same synthesized image generation system serves multiple functions: it can generate images at any required angle, select optimal angles for specific objects, and provide reference data for various classification tasks, thereby reducing overall system complexity through multi-functionality.
Data Source
AI summary
The present disclosure provides methods and/or systems for identifying an object. An example method includes: generating a plurality of synthesized images according to a three-dimensional digital model, the plurality of synthesized images having different view angles; respectively extracting eigenvectors of the plurality of synthesized images; generating a first fused vector by fusing the eigenvectors of the plurality of synthesized images; inputting the first fused vector into a classifier to train the classifier; acquiring a plurality of pictures of the object, the plurality of pictures respectively having same view angles as at least a portion of the plurality of synthesized images; respectively extracting eigenvectors of the plurality of pictures; generating a second fused vector by fusing the eigenvectors of the plurality of pictures; and inputting the second fused vector into the trained classifier to obtain a classification result of the object.