Computer Vision - Business area

Cameras

Depending on your application, the most suitable type of camera could be an area scan camera, line scan camera, thermal imaging camera, smart camera or something else. OEM Automatic have over 20 years of experience helping customers finding the right camera. We offer cameras from several of the leading camera manufactures in the world such as Basler, Matrox Imaging, Sony, and FLIR (thermal imaging). We have cameras with all kinds of specifications including resolutions up to 45 Mega Pixel.    

• Area Scan Cameras
• Line Scan Cameras
• Smart Cameras
• Thermal Imaging Cameras
• Network Cameras
• Embedded Cameras

3d products

OEM Automatic offers a broad portfolio of 3D-products using different techniques for acquiring the images. Apart from being excellent stand alone products for acquiring 3D data, the products can also be combined with 2D cameras and vision software such as Matrox Design Assistant and Pekat vision for even greater possibilities. 

• Laser Triangulation
• Structured light
• Stereo Vision
• Time of flight

Industrial Computers

In today’s industries, reliability in production is very important. By using a purpose-built industrial computer without moving parts as for example traditional hard drives (HDD) or fans, you minimize the risk of computer failure. We offer computers from Neousys and Matrox, made for round the clock use in environments with temperatures from -25 to 70 °C and even IP67 protection.

• Compact
• Surveillance
• Vehicle
• ShoeBox
• Ultra Compact
• Weather Protected
• Computer Accessories

interface solutions

Whether you are looking for alternatives to replace parts of your current vision system or expanding the capabilities of it, an interface solution may be required. We offer internal and external frame grabbers, expansion cards and interface cards that will make the upgrade easier. 

• Internal frame grabbers/Interface boards
• External frame grabbers
• Embedded boards

Software

Whether you are planning for image analysis, bin picking or AI-driven vision applications we have the software needed.  For traditional machine vision tasks such as detection, measuring and pattern matching the softwares from Matrox are the best choice. Available for integrations with a large variety of cameras – both 2D and 3D. For bin picking applications using a structured light 3D camera we recommend one of the options from Photoneo together with their own 3D scanners. We have dedicated an own chapter to Artificial Intelligence where the software from Pekat Vision is presented.

• Incoil
• Matrox Design Assistant
• Matrox Imaging Library
• Photoneo Bin Picking Studio
• Photoneo Locator
• Photoneo 3D Localization C++ SDK
• Photoneo Anypick
• Photoneo Depalletization

TAILOR-MADE SYSTEMS
Proof of concept is all about testing and verifying the solution before the final buying decision is made. In our vision lab we have access to many different cameras, lenses and lights for solving customer cases. We often ask our customers to provide us with real world samples to ensure an accurate result. If necessary, proof of concept studies can also be performed at the customer’s site for an even greater accuracy in the study. 

MATROX DESIGN ASSISTANT
For non-AI imaging cases, we use the Matrox Design Assistant vision software based on Matrox Imaging Library. Using Matrox Design Assistant, we can set up a custom vision application for all kinds of advanced measurement. Paired with selection of cameras, lenses and lighting we can provide a proof of concept that can be trusted. 

PHOTONEO BIN PICKING STUDIO
Photoneo’s Bin picking studio software is the natural choice for visual guided bin picking solutions. Setting up the picking solution all begins with scanning and identifying the part. As a proof of concept we can scan and localize the parts and optimize for your application or provide point clouds for further evaluation.

PEKAT VISION
When traditional machine vision solutions using measurements are not sufficient, for example when analyzing fabric or organic materials we use the AI-driven Pekat Vision software. A feasibility study here typically consists of training the software with an image set of the objects to be analyzed.  All of which can be done in our lab facilities.  

HARDWARE FOR TRAINING
Training the system with image and data sets is a very hardware resource consuming task. To do this efficiently, a top-of-the-line PC with fast CPU and GPU is always recommended. We partner up with Neousys to be able to offer a broad range of computers, all rugged and of industrial grade. For AI training, the typical choice would be a shoe box style computer with lots of expansion capabilities such as the Nuvo-8240GC which supports dual NVIDIA® Tesla T4 GPU and Intel® Xeon® E CPU.

ON THE EDGE
While you can use the same PC for inference as for training, there are more dedicated alternatives available for on the edge applications. One reason for selecting dedicated PC’s is to lower the cost in multi-instance applications, another is to have a computer more suited for tough and/or constricted spaces. The NRU-120S, which is powered by NVIDIA® Jetson AGX Xavier™, is a compact and cost efficient, yet very capable choice for on the edge inference. For even more demanding applications the IP67 protected SEMIL series which can be equipped with NVIDIA® Tesla T4 or Quadro P2200 GPU is up to the task.

SOFTWARE
For vision based industrial AI applications, we offer the excellent software from Pekat Vision. With ease of use in focus, it uses its own proprietary focused-learning algorithm which goes beyond deep-learning. Focus-learning works like a human eye - It can focus on specific details and needs only a few images for learning and understanding the task. This means that even from a small data set of images, it can recognize when an object or material is flawless and when it is defective. You can use Pekat Vision to find anomalies, detect and classify defects and check surfaces on materials and objects where traditional machine vision systems cannot.

In the unsupervised mode, the software detects defects which does not have to be specified on beforehand. This means that it is enough to train the software only using images of defect-free objects. In supervised mode the software can be trained to search for and categorize specific defects such as scratches, rust, holes etc.
The inspection results can continuously be used to take further actions using for example an in-line robot. For this, there are multiple options such as command line scripts, HTTP requests, Profinet/TCP connections to PLC’s. For image acquisition, a GenICam compliant camera can be directly connected to the Pekat Vision software.