Defect detection tasks that rely on human sensibility are a challenge to automate. Fortunately, powerful new AI technology can match the skills and capabilities of experienced inspectors.
Vision System
last update: July 27, 2020
Defect detection tasks that rely on human sensibility are a challenge to automate. Fortunately, powerful new AI technology can match the skills and capabilities of experienced inspectors.
The AI Scratch Defect Filter learns by means of images in which human inspectors noticed defects. Whereas previous inspection methods found the unexpected size, shape or color of a particular defect to be a barrier to automation, AI successfully extracts abnormalities by judging their features without definition. The learned data facilitates defect detection on processed surfaces and other uneven backgrounds that previously posed an insurmountable challenge.
*1. The FH-UMAI1 Scratch Detect AI Software Installer is required to use AI Scratch Detect Filter.
Regardless of material type, color, or size, defects can be extracted reliably without previously required definition and adjustment.
Sensory inspection requires a certain tolerance for variations that don't pass a certain threshold.
Determining what variations are acceptable is a key capability of expert inspectors and poses a challenge for automated inspection systems.
AI Fine Matching identifies a future that is not included in good products as a defect.
AI learns images of good products with variations, and generates an AI model.
Every time an inspection is carried out, AI reconstructs a model that is presumed to be a good product.
AI extracts a difference between the reconstructed good product image and a captured image to identify a defect, reducing overdetection.
Three quick steps on the settings screen guide the user through the process of creating the good product model with the minimum number of images.
*1. “Patent pending” means that we applied for a patent in Japan, and “Patented” means that we obtained a patent in Japan. (as of May 2020)
With the FH Series, there's no need for high-end hardware or specialized engineers who can configure the system to suit your needs. Our general-purpose vision system makes it easier than ever to introduce AI into production sites.
Artificial intelligence has traditionally required a high-end environment, but our lightweight creative solution comes in the form of user-friendly processing items that have been integrated into our popular FH Series hardware.
*1. The FH-5550 Controller is compared with the FH-3050 Controller.
We offer a range of cameras that can capture high-resolution images suitable for sensory inspection at high speeds.
There was a trade-off between high-resolution image capture like the human eye and inspection processing speed. We use new CMOS image elements and dual transfer technology to capture high-resolution images while transferring images at high speeds. This facilitates applications that previously required multiple cameras or a mechanism to move a camera.
This light can be adjusted to defects by combining the illumination colors and angles like humans do.
Even if new objects or inspection items are added after installation, there is no need to add or change the light—just change the illumination pattern.
The illumination patterns can be registered as settings, facilitating duplicating production lines.
Multi-Trigger Imaging captures long objects at high speeds
The Multi-Trigger Imaging function can capture images and process them in parallel, leveraging the speed of the multi-core processor to capture long objects at high speeds.
Camera Image Input HDR optimizes contrast
Camera Image Input HDR helps create optimized HDR images under variable ambient conditions. Once you specify the optimum area to capture on the image, the FH Series automatically adjusts the shutter speed while capturing images and combining the images.
A single controller can perform inspections at different points at different timings.
Controllers installed for each process can be integrated into one, reducing initial costs and saving space.
A single controller that can control each line saves initial costs and space.
Four cameras can be connected to one controller, enabling simultaneous inspection of dents and scratches from four directions.
Shape Search III is robust against shape variations
High-precision and robust positioning is possible even under the adverse conditions, such as changes in environments and materials.
*1. The value measured under our specified conditions is provided for reference.
A 20.4 Mpix camera can search a positioning mark in as fast as 12 ms *2, and a 5 Mpix camera, widely used for alignment applications, in as fast as 2 ms.
*2. The value measured under our specified conditions is provided for reference. 20.4 Mpix camera.
Circular Scan Edge Position accurately estimates the center and radius of a circle
The new algorithm accurately detects a whole circle from a part of the circle.
Scan Edge Position removes noise to detect edges
This algorithm accurately estimates lines even when the edges are unclear due to variations in objects or disturbance.
Stage Data calculates for various stages
The popular single axis + θ axis stages as well as UVW stages can be used. The use of the same axis for both handling and positioning simplifies machine configuration.
Communication programs to connect robots from various vendors and FH flowcharts required for robot applications are provided free of charge. You can quickly set up robot vision applications.
The FH Series can associate a unique ID with the inspection image and result, and then output them to the host device.
You can immediately find required inspection images and quickly identify causes of fails.
The amount of inspection image data required for defect cause analysis can be so large that conventional controllers are unable to store it given their storage time and capacity constraints.
The high-speed, large-capacity controller has USB 3.0 ports and the improved algorithm to compress image data at high speeds, enabling all images to be stored to meet increasing needs in quality control.
The times in the following figure provided for reference only and their accuracy cannot be guaranteed. They are measured under the following conditions:
•FH-5050 Controller
•5 Mpix monochrome images
•Size of converted JPEG file: 0.6 MB
2D Code II provides powerful code reading
The FH Series incorporates a dedicated algorithm for reliable and fast 2D code reading even under variable ambient brightness or adverse conditions such as after processing or washing.
OCR reliably reads difficult-to-read characters
OCR can reliably read characters printed too close to each other or on curved surfaces. Also plus signs can be read.
Character Inspection reads special fonts
Character Inspection recognizes special fonts and non-alphanumeric characters based on pattern search using the dictionary set up by the user.
Sysmac Studio is a unique environment that integrates logic, motion and drives, robotics, safety, visualization, and information technologies in a single project, thus reducing the learning curve and the intra-operative software costs.
You can use EtherCAT® to connect NJ/NX Machine Automation Controllers and 1S/G5 AC Servo System to increase the control speed of everyday communications protocols from position detection to starting axis motion.
This design interface includes pre-installed screens for all phases, from design through to setting and operation.
Just select processing items and determine the order to manage variables. Time-consuming calculations and inputs are no longer required.
All the common settings of multiple scenes can be made at once. Simplified inspection flowcharts reduce setting errors and prevent from forgetting to change settings.
To inspect aligned parts, the FH Series can repeat the same measurements while shifting the measurement region within the same image. This reduces setting times.
Showing only necessary screens for production makes the interface easier to use. Screen layout can be customized just by selecting and placing objects, without programming.
last update: July 27, 2020