3D Point Set Article Recognition for Bulk Storage Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing article take-out systems require preparation of 2D or 3D model patterns for each type of article, which is labor-intensive and impractical for large numbers or irregularly shaped articles, and often fail to recognize positions and postures accurately due to poor lighting and slanting conditions.
Innovation Solution
A system using a 3D measuring device to acquire 3D points, connecting nearby points to form connected sets, and a robot control unit to identify and manipulate the positions and postures of articles without pre-defined patterns, enabling efficient recognition and retrieval of articles in bulk storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If model patterns are prepared for each type of article, then article recognition can be performed, but tremendous work is required when there are a large number of article types
Solution Approach 1:
The patent creates a 3D point set that serves as a digital copy of the article's surface geometry. Instead of preparing model patterns for each article type, the system directly processes the 3D point cloud data obtained from the article surface, enabling recognition without repetitive pattern preparation work while maintaining recognition accuracy.
Solution Approach 2:
The patent transforms the article recognition approach from 2D image pattern matching to 3D point set analysis. By changing the dimensional parameter from 2D to 3D and using geometric features of the point cloud (such as surface curvature, point distribution, and spatial relationships) instead of 2D patterns, the system eliminates the need for extensive pattern preparation while preserving recognition precision.
2Measurement precision
If model patterns are prepared for each type of article, then article recognition can be performed, but it is impossible to create model patterns for irregularly shaped articles
Solution Approach 1:
The patent transitions from 2D pattern-based recognition to 3D point set-based recognition. This parameter change allows the system to handle irregularly shaped articles effectively, as the 3D point cloud naturally captures complex geometries without requiring pre-defined model patterns. The geometric features extracted from the point cloud adapt to any shape, providing both precision and versatility.
Solution Approach 2:
The patent moves the recognition process from 2D image space to 3D spatial domain. By analyzing the article surface as a 3D point set with depth information, the system can accurately represent and recognize irregular shapes that cannot be adequately captured by 2D patterns, thereby expanding adaptability while maintaining recognition accuracy.
3Productivity
If 2D pattern matching is used on images, then article positions can be recognized, but recognition accuracy deteriorates due to poor lighting and slanting conditions
Solution Approach 1:
The patent transitions from 2D image analysis to 3D point set analysis. By capturing the article surface in three dimensions, the system obtains geometric information that is independent of lighting conditions and viewing angles. This dimensional change enables accurate position and posture recognition even when the article is slanted or lighting is poor, as the 3D point cloud preserves the true geometric structure of the article surface.
Data Source
AI summary
An article take-out apparatus including: a 3D measuring device measuring surface positions of a plurality of articles stored in bulk in a 3D space so as to acquire position information of a plurality of 3D points; a connected set processing unit determining connected sets made by connecting 3D points which are close to each other, from the plurality of 3D points acquired by the 3D measuring device; an article identifying unit identifying positions and postures of the articles, based on position information of 3D points belonging to the connected sets; a hand position and posture processing unit determining positions and postures of the hand capable of taking out the identified articles; and a robot control unit controlling a robot to move the hand to the positions and postures determined by the hand position and posture processing unit and take out the articles.