Point Cloud Identification via Rotation Invariant Feature Orthogonalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing feature representing device for point clouds experiences errors in positioning due to nearest neighbor distance calculations, leading to inaccurate identification of point clouds with the same shape being classified differently, resulting in low accuracy.
Innovation Solution
A point cloud identifying device that acquires point cloud information, performs orthogonalization of basis vectors, calculates rotation invariant features, and uses these features along with a learned model to improve the accuracy of point cloud identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If nearest neighbor distance is used for positioning point clouds, then the method is simple to implement, but positioning errors occur and identification accuracy decreases
Solution Approach 1:
The patent changes the parameter used for point cloud comparison from nearest neighbor distance to rotation invariant features. By computing features that remain unchanged under rotation (such as geometric moments or shape descriptors), the system eliminates positioning errors while maintaining computational feasibility through standardized feature extraction algorithms.
Solution Approach 2:
The patent replaces the mechanical/geometric approach of measuring distances between points with a computational approach using rotation invariant feature extraction. This substitution transforms the positioning problem into a feature recognition problem, eliminating the sensitivity to rotation and translation that plagues distance-based methods.
2Ease of operation
If nearest neighbor distance is used for point cloud alignment, then the process is straightforward, but point clouds with the same shape are identified as different
Solution Approach 1:
The patent introduces asymmetry in the feature extraction process by computing rotation invariant features that capture the essential shape characteristics while being insensitive to orientation. This allows the system to correctly identify that rotated versions of the same point cloud should be classified as identical, resolving the reliability issue while maintaining operational simplicity through automated feature computation.
Solution Approach 2:
The patent makes the feature extraction process universal by designing rotation invariant features that work for point clouds of any orientation. The same feature extraction algorithm correctly handles point clouds in different rotations, making the system universally applicable without requiring complex alignment procedures before classification.
3Measurement precision
If rotation invariant features are computed through orthogonalization of basis vectors, then identification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments the computational process into distinct stages: (1) extracting point cloud data, (2) computing rotation invariant features through orthogonalization, and (3) performing classification. By separating these operations, the system manages computational complexity through modular processing, where each stage can be optimized independently and the most computationally intensive orthogonalization is performed only when necessary for feature extraction.
Data Source
AI summary
A point cloud identifying device includes a point cloud acquiring unit to acquire point cloud information indicating N (N≥2) points in k (k≥2) dimensions, a model acquiring unit to acquire a model having a learning parameter, a rotation invariant converting unit to perform orthogonalization of each of basis vectors for each of points indicated in the point cloud information, and calculate a rotation invariant feature by using data after the orthogonalization, an inference unit to identify a point cloud indicated in the point cloud information by using the rotation invariant feature and the model, and a result output unit to output a classification result by identification of the inference unit.