Metal Sheet Edge Recognition for Precise Position and Orientation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machines and methods for working and moving metal sheets or plates face challenges in precisely and automatically determining the position and orientation of pieces due to inadequate contour extraction algorithms, especially when there is low contrast with the background, surface finish, material type, coloration, or lighting conditions, and lack of available reference figures.
Innovation Solution
Employing a Deep Learning algorithm to process images of metal sheets or plates, combined with mask means such as elliptically shaped masks, to accurately identify and isolate contour stretches, enabling precise geometric information extraction without manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional contour extraction algorithms are used to process images of metal sheets, then the processing speed is relatively fast, but the measurement precision of contour recognition deteriorates due to low contrast with background, surface finish, material type, coloration, and lighting conditions
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into multiple stages: first processing the entire image to identify potential contour regions, then applying mask means to specific portions containing contour stretches that are not clearly distinguishable. This segmentation allows the system to focus computational resources only on problematic areas, improving contour recognition accuracy without proportionally increasing overall processing complexity.
Solution Approach 2:
The patent implements preliminary action by first processing the complete image to identify regions with indistinct contour stretches before applying mask means. This preliminary identification step allows the system to pre-determine where enhanced processing is needed, rather than attempting to process the entire image with maximum complexity from the outset.
2Measurement precision
If manual intervention is used to select points on processed images for contour identification, then the measurement precision of geometric information improves, but the productivity deteriorates due to laborious and time-consuming operations
Solution Approach 1:
The patent implements self-service by enabling the artificial vision system to automatically identify and process indistinct contour stretches using mask means, eliminating the need for manual operator intervention. The system autonomously determines which contour stretches require enhanced processing and applies the appropriate masks, thereby maintaining high measurement precision while significantly improving productivity by removing the bottleneck of manual point selection.
3Measurement precision
If reference figures are required for contour comparison, then the measurement precision of piece alignment improves, but the adaptability deteriorates when dealing with scraps or skeletons for which reference figures are not available
Solution Approach 1:
The patent applies universality by creating a reference figure from the processed image itself, making the system applicable to both standard pieces with pre-existing reference figures and scraps or skeletons without reference figures. The mask means can identify contour stretches and generate reference data universally, allowing the same processing methodology to work across different material types and conditions.
Solution Approach 2:
The patent implements copying by creating a reference figure from the processed image of the piece itself. Instead of requiring external reference figures, the system generates a digital copy or representation from the captured image, which can then be used for comparison and alignment determination. This copying approach enables the system to handle scraps and skeletons that lack pre-existing reference documentation.
Data Source
AI summary
A machine for working and/or moving metal pieces has an operating system, an artificial vision system with a camera to acquire an image of at least one piece, a control unit to control the operating system, and a processing unit for processing the image with a Deep Learning algorithm or a contour recognition algorithm depending on whether the contour stretch is precisely and completely delineated, extracting geometric information of the piece based on the contour stretch with reference to a reference system of the machine and sending the geometric information to the control unit to configure operating parameters of the operating system.