Bottom-Up 2D Imaging for Industrial Robot Part Picking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing part picking systems for industrial robots are difficult and time-consuming to teach, requiring extensive user knowledge and multiple inputs to achieve robust results, often taking several days to manually tune the vision system for specific applications.
Innovation Solution
A method and system that utilize a 2D image taken from below to define the positions of pickable parts on a known plane, with a reference image and surface image processing to determine recognizing features and grasp configurations, simplifying the teaching process by assuming all parts rest on a known plane and using a 2D image scanner for image acquisition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional vision systems with overhead cameras and complex algorithms are used, then part positioning accuracy is achieved, but the teaching process becomes extremely time-consuming and complex
Solution Approach 1:
The patent inverts the traditional overhead camera approach by using a bottom-up imaging method. Instead of placing the camera above the picking surface, the imaging device is positioned below the transparent picking surface, capturing images from underneath. This inversion simplifies the teaching process while maintaining positioning accuracy, reducing teaching time from days to minutes.
Solution Approach 2:
The patent replaces complex mechanical vision systems with a simpler optical solution. Instead of using overhead cameras with sophisticated algorithms for 3D reconstruction and part orientation detection, the system uses a bottom-up transparent surface imaging approach that directly captures part positions and orientations through the transparent picking surface, eliminating the need for complex image processing algorithms.
2Reliability
If manual teaching of vision algorithms is performed, then robust part recognition is achieved, but extensive user knowledge and multiple days of tuning are required
Solution Approach 1:
The patent enables the system to automatically determine part positions and orientations without requiring manual teaching or user intervention. The bottom-up imaging approach through the transparent picking surface allows the system to self-calibrate and automatically recognize parts, eliminating the need for users to spend days tuning vision algorithms.
Solution Approach 2:
The patent creates a simplified 2D copy of the picking surface view by imaging from below through the transparent surface. This 2D image copy contains all necessary information about part positions and orientations, replacing the need for complex 3D reconstruction and manual teaching processes while maintaining recognition accuracy.
3Loss of information
If overhead cameras and complex lighting systems are used, then complete part information is captured, but system complexity and setup time increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for part picking by imaging from below through the transparent surface. Instead of using overhead cameras that capture complex 3D information requiring sophisticated processing, the system extracts 2D position and orientation data directly from the bottom view, simplifying the overall system architecture.
Solution Approach 2:
The transparent picking surface serves multiple functions: it acts as the physical surface for holding parts, allows optical transmission for bottom-up imaging, and provides a clear view of part positions and orientations. This multi-functionality eliminates the need for separate imaging components and complex lighting systems required in traditional overhead approaches.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces the complexity and time required to teach an industrial robot to pick parts, making the process faster and more accessible by using a 2D image scanner to capture reference and surface images, allowing for efficient determination of part positions and grasp configurations.
Implementation Method 1
The picking surface is, for example, a surface of a conveyor belt... The vision system makes use of a camera that is mounted directly overhead of the picking surface... a 2D image from below
Data Source
Figure 1~6
Figure 3
Figure 4~5
AI summary
A method for teaching an industrial robot (3) to pick parts (120, 122) comprises the steps of: placing a reference part (7) on a picking surface (2); providing a reference image comprising information about the reference part (7); placing a gripping tool (5) of the industrial robot (3) in relation to the reference part (7) so that the gripping tool (5) is in a grasp configuration in relation to the reference part (7); and storing the grasp configuration. The reference image is a 2D image from below. When the reference part (7) is resting on a known plane a 2D image taken from below is enough for defining the position of the same.