3D Point Group Alignment via Segmented Translation and Rotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for aligning three-dimensional point groups using iterative closest point (ICP) methods face high processing loads due to brute-force searches for translation and rotation parameters, which reduces efficiency in recognizing object posture accuracy.
Innovation Solution
A method involving multiple sensors to extract point groups from distance images, where the position of one sensor is translated to match the object's position, and rotation parameters are calculated to align point groups in a rotational direction, reducing the need for extensive translation parameter calculations by executing alignment in the translation direction only once.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If iterative closest point (ICP) method with brute-force search is used to align three-dimensional point groups, then alignment accuracy is improved, but processing load increases
Solution Approach 1:
The patent applies preliminary action by performing translation alignment before rotation alignment. The translation parameters are calculated and applied first to eliminate positional offsets between point groups, creating a pre-aligned state. This preliminary translation step simplifies the subsequent rotation search space, allowing the ICP algorithm to focus only on rotational parameters, thereby reducing overall processing load while maintaining alignment accuracy
Solution Approach 2:
The patent segments the alignment process into two distinct phases: translation alignment and rotation alignment. By dividing the originally unified alignment problem into separate sequential steps, the complex brute-force search is broken down into manageable portions. The translation phase handles positional differences, while the rotation phase handles orientational differences, reducing the computational complexity of each individual phase compared to solving both simultaneously
2Measurement precision
If multiple sensors are used to detect three-dimensional point groups from multiple directions, then object posture recognition accuracy is improved, but calibration complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating translation parameters that transform sensor coordinates to object coordinates before performing rotation alignment. This preliminary coordinate transformation establishes a common reference frame for all sensors, simplifying the subsequent calibration process. By preparing the translation parameters in advance, the patent reduces the complexity of integrating data from multiple sensors while maintaining high posture recognition accuracy
Data Source
AI summary
An information processing device (100) extracts a first point group constituting an object from a first distance image acquired by a first sensor. The information processing device (100) extracts a second point group constituting an object from a second distance image acquired by a second sensor. The information processing device (100) generates a first translated point group by translating the first point group by first translation coordinates. The information processing device (100) generates a second translated point group by translating the second point group by second translation coordinates. The information processing device (100) calculates a rotation parameter and a translation parameter based on the first translated point group and the second translated point group.