3D Object Invariant Feature Extraction for Efficient Database Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for comparing and searching three-dimensional objects are inefficient due to the need for full representation storage and alignment, which consumes significant computational resources and memory, especially when dealing with large databases of similar objects.
Innovation Solution
A computerized system and method for generating translation and rotation invariant representations of three-dimensional objects using spherical harmonics and irreducible representations of the rotation group, allowing for compact and accurate representation and search, with features that require less storage and computation, utilizing low-frequency components and matrix-vector multiplication on GPUs for efficient comparison.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full representation storage and alignment methods are used for comparing three-dimensional objects, then comparison accuracy is maintained, but storage requirements and computational resources consume significantly more
Solution Approach 1:
The patent extracts only the essential invariant features from the complete 3D object representation. By taking out only the necessary components (invariant features under rotation and translation) rather than storing the full object representation, the system achieves accurate comparison while dramatically reducing storage requirements and computational load.
Solution Approach 2:
The patent segments the 3D object representation into invariant features that capture the essential geometric characteristics. This segmentation allows the system to work with simplified feature representations rather than complete object data, resolving the contradiction between accuracy and storage efficiency.
2Measurement precision
If full representation storage and alignment methods are used for comparing three-dimensional objects, then comparison accuracy is maintained, but computational resources and processing time increase significantly
Solution Approach 1:
The patent extracts only the essential invariant features from the complete 3D object representation. By taking out only the necessary components (invariant features under rotation and translation) rather than processing full object representations, the system achieves accurate comparison while dramatically reducing computational load and processing time.
Solution Approach 2:
The patent performs preliminary computation to establish invariant feature representations before actual comparison operations. By pre-computing and storing only the invariant features rather than complete object data, the system enables fast retrieval and comparison operations without the computational burden of processing full 3D representations in real-time.
3Measurement precision
If traditional representation methods are used for searching three-dimensional objects in databases, then search accuracy is maintained, but the system becomes inefficient for large databases
Solution Approach 1:
The patent extracts only the essential invariant features from complete 3D object representations. By storing and searching only these compressed invariant features rather than full object data, the system maintains search accuracy while achieving linear scalability and efficiency even for very large databases.
Solution Approach 2:
The patent segments object representations into invariant features that are sufficient for accurate searching. This segmentation creates a simplified search space that maintains the essential discrimination capabilities while dramatically improving system efficiency and scalability for large-scale databases.
Data Source
AI summary
A non-transitory computer readable medium that stores instructions that once executed by a computer cause the computer to execute the stages of: calculating a first function that represents an object that is three dimensional; calculating a second function that is a convolution or an approximated convolution of (a) the first function applied on points of the object, and (b) an other function that is the first function composed with a function that sends points of the object to opposite points; wherein the second function is translation invariant; and calculating the translation and rotation invariant features of the query object, based on the second function.