3D Article Recognition via Sensor Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for recognizing the three-dimensional position and orientation of articles using range sensors are costly, require high-performance processors, and have extended measurement times, making them unsuitable for practical robot operations, while combining three-dimensional and two-dimensional sensors increases calculation costs and processing time.
Innovation Solution
A recognition device and method that utilize a camera for two-dimensional imaging and a range sensor for three-dimensional information, with an information processor setting search conditions based on three-dimensional data to efficiently detect and calculate the two-dimensional and three-dimensional positions and orientations of articles, reducing calculation costs and processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional range sensor methods are used for three-dimensional recognition, then measurement capability is improved, but calculation cost and processing time increase
Solution Approach 1:
The patent segments the recognition process into two distinct stages: first using a range sensor to obtain rough three-dimensional position and orientation information, then using a camera to capture detailed two-dimensional images only of the identified article. This segmentation allows the system to leverage the strengths of each sensor type while avoiding their individual weaknesses, thereby reducing overall processing time and calculation cost.
Solution Approach 2:
The patent introduces an information processor as an intermediary that coordinates between the range sensor and camera. This intermediary first processes range sensor data to identify the article's rough position and orientation, then uses this information to guide the camera's detailed imaging process. The information processor acts as a mediator that optimizes the workflow between two different sensing modalities.
2Measurement precision
If conventional range sensor methods are used for three-dimensional recognition, then measurement capability is improved, but calculation cost increases
Solution Approach 1:
The patent segments the recognition process into two distinct stages: first using a range sensor to obtain rough three-dimensional position and orientation information, then using a camera to capture detailed two-dimensional images only of the identified article. This segmentation allows the system to leverage the strengths of each sensor type while avoiding their individual weaknesses, thereby reducing overall processing time and calculation cost.
Solution Approach 2:
The patent uses the range sensor to create a rough three-dimensional model or representation of the article's position and orientation, which then serves as a guide for the camera's detailed imaging. This rough copy or approximation from the range sensor allows the system to avoid performing computationally intensive three-dimensional analysis on all captured images, thereby reducing calculation cost.
3Measurement precision
If three-dimensional and two-dimensional sensors are combined, then recognition accuracy is improved, but processing time and calculation cost increase
Solution Approach 1:
The patent performs preliminary action by using the range sensor to first identify the rough three-dimensional position and orientation of the article before the camera captures detailed images. This preliminary information from the range sensor pre-preps the system by eliminating the need to process all possible articles, thereby reducing subsequent processing time while maintaining high recognition accuracy through the combined use of both sensors.
Data Source
AI summary
A recognition device and method capable of recognizing 3D position and orientation of an article at low calculation cost. A 2D image of a region, where articles are randomly located, is obtained by a camera, and 3D information of generally the same region is obtained by a range sensor. A space, where an article to be taken out is considered to exist, is roughly limited. Based on the limited space, a search condition for searching the article by 2D image processing is set, and 2D positional information on the image of the article is obtained. Then, 3D point data used to recognize the 3D position and orientation of the article is selected, and a view line in the 3D space, extending from the camera to the article, is calculated, whereby the 3D position and orientation of the article is calculated.