Machine Vision Technology Forum 2017 - Schedule & presentations
Here you find an overview of all sessions offered at our Nordic Machine Vision Technology Forum in Sweden, Stockholm.
Get an overview about all presentations and choose your presentation program for the event at the online registration.
SCHEDULE for 24th October 2017 in Stockholm
Fast machine vision solution development for IIoT-based smart factories
In the arena of manufacturing, emerging Internet of Things (IoT) technology has ushered in the Industry 4.0 Initiative, migrating from conventional automated production to IoT-based intelligent automation, replacing semi-automated or standalone automatic machining with network-connected processes based on M2M (machine to machine) and M2P (machine to person) communication. This, in combination with corporate information systems and analytics, creates endless possibilities for the smart factory model.
As the IIoT-based smart factory initiative encourages manufacturers to actively implement smart automation, machine vision has become indispensable to the quality control of automated production. A solution providing fast and easy development of machine vision applications is a key factor in empowering IIoT-based smart factories.
ASIC – Utilization in camera design for machine vision
Before the ALVIUM technology was launched all machine vision cameras had relied on FPGAs for their brains. ASICs are widely used in many electronic devices, so why had they never been used in a machine vision camera?
This presentation will explain the difference between FPGAs and ASICs and will detail the advantages and disadvantages of both technologies relating to the design and manufacture of machine vision cameras.
Data from comparisons of real cameras will be shown to give a practical demonstration of the differences in real world applications.
MIPI CSI-2 – A new camera interface for embedded machine vision systems
Before 2017 there were no machine vision cameras on the market using the MIPI CSI-2 interface and many applications have been successfully solved with the established interfacing methods (GigE, USB3, etc.). So, why do we think there is the need for a new interface? Where has this “new” interface come from? What are the advantages? What kind of applications is it good for? Who should use it? Who shouldn’t use it?
This presentation will provide an introduction to MIPI CSI-2 and will show a comparison to other commonly used machine vision interfaces. It will show the prevalence of CSI-2 on embedded processing boards and will explain the engineering design considerations that will help users decide if they should develop their next system with this interface technology.
Optimized design of 3D laser triangulation systems
As 3D laser triangulation is used more and more by system integrators and OEMs in the development of industrial inspection systems. Questions arise on how to optimally design the scanning setup to fulfill the application requirements.
The presentation provides a brief guideline for the selection of components (camera, lens, laser), setting up the right geometry (triangulation angle, working distance), the application of suitable algorithms for 3D line detection in the camera image and the estimation of 3D scan properties such as precision and profile speed. Issues like the use of camera Scheimpflug adapters, the choice of laser wavelength and the calibration of the 3D setup will be part of the discussion. Furthermore, the all-in-one 3D compact sensors will be presented as an alternative to the discrete camera-laser triangulation setup and their specific characteristics will be compared.
Challenges of installing and operating machine vision components in diverse industries
Machine vision applications advance to markets with increased requirements for the protection of cameras, lights, lasers and other machine vision components: Applications areas such as the automotive industry, metal, glass and paper industries, food, medical, pharmaceutical industries, sport, entertainment sectors and traffic or solar industry, all have different requirements.
This presentation discusses the technical and legal requirements for machine vision applications in diverse industries. It also covers technical solutions that enable safe and compliant installation and operation of cameras and other components in challenging environments.
Getting the best image for your vision application with computational imaging
By creating an output image focused on the image properties that are most important to a particular machine vision task, computational imaging (CI) offers powerful advantages over traditional one shot imaging. Relying on data extracted and computed from a series of input images captured under different lighting or optical conditions, computational imaging techniques outperform traditional image acquisition.
This presentation covers six practical examples of CI solutions for machine vision applications, including photometric stereo (also known as shape from shading), super-resolution colour, high dynamic range (HDR), extended depth of field (EDOF), bright field/dark field, and multi-spectral imaging. CCS computational imaging solutions simplify the hardware, timing and acquisition to easily bring the benefits of CI to any application.
Colour identification, colour quantification and hyperspectral imaging – VIS and NIR inspection
The lecture deals with the basics and technologies for colour differentiation of objects. Starting from simple colour recognition with conventional colour cameras and evaluation in the RGB colour space, we will discuss the advantages of other methods such as the HSV colour space or the spectral quantification in the CIE XYZ colour space.
How does the human eye perceive colours and levels of brightness, and what is colour colourimetry? What are the identification capabilities of RGB colour cameras, multi camera systems and hyperspectral imaging systems? What can be inspected and detected in the visible and in the IR range? The lecture also addresses the chemical-physical effects that are used for evaluation purposes.
Colour inspection, Infrared and UV – tips, special features, requirements
Colour, IR- and UV illumination can be used in combination with monochrome or colour cameras in order to visualise the diverse inspection features. In addition to the illumination technology basics we address topics such as the spectral sensitivity of camera sensors, suitable lenses, optical filters, and other subjects such as the screening of extraneous light.
Inspecting transparent objects
Transparent materials such as glass, film, plastics, adhesive film or liquids have proven again and again to be difficult test objects.
However, the right optical set-up, an appropriate lighting method or inspection technology can solve these problems, in order to locate objects or detect surface defects, cracks, edge chipping and impurities.
Five reasons to use a lighting controller
Do you want to improve the reliability and repeatability of your machine vision applications? Do you need to increase the speed? Will your system need maintenance when the lighting becomes less bright?
Lighting controllers give more stable light output and can give remote control of the lighting so that the brightness can be maintained as the light gets older and less bright. Overdriving is a powerful technique to get more brightness from LED lighting. High speed synchronisation of lighting enables multi-light applications whilst maintaining high line speeds.
This presentation explains the features available in lighting controllers and how they can be used to improve the capability, speed and reliability of your machine vision system. It then describes the latest techniques and features available, including how you can save cost and complexity by reducing the number of inspection positions in a machine.
Intel RealSense™ depth camera technology review for acquisition of 3D data
Colour imaging: Getting the best out of multi-sensor prism cameras
Multi-sensor prism cameras offer significant advantages to colour imaging. Bayer, multi-linear colour line scan and hyperspectral sensors featuring colour filters on top of each single pixel block-out most of the light falling onto the sensor. In contrast, multi-sensor cameras don’t block but separate the light with the use of dichroic prisms: Thus (almost) no light gets lost. Advantages of this design include better signal-to-noise ratio, higher colour contrast, much lower crosstalk between colour channels, lower of colour interference effects as often seen in Bayer images as well as a lack of halo effects as often seen in images taken with multi-linear line scan cameras.
Furthermore, due to the nature of the multi-sensor camera design, every colour channel can be adjusted in gain and exposure time separately. As a result, images show much higher dynamic range, contrast and signal-to-noise ratios across the whole colour bandwidth. For multi-spectral applications, channel separation can be customised by adjusting the dichroic coating of the prisms.
Smart customisation of 3D sensors with application specific algorithms
Solving specific application needs is one of the challenges in today’s 3D sensor market. A modern factory cannot only care about speed, but also has to reach a high precision in quality control through accurate and reliable measurement data. In order to achieve this, LMI will demonstrate different options that allow the user to adapt 3D smart sensors to solve their unique inspection problems.
With the help of real life examples, LMI will explain how software developers can test their own applications in a safe offline environment without the need for a physical sensor. Additionally working with large 3D point clouds can be simplified by adding the data-processing power of one or more PCs to an inspection solution.
Additionally, the presentation will highlight how users can develop custom measurement algorithms that run directly on 3D smart sensors. This extends the functionality of the sensors and allows for the flexibility needed in a fast changing environment.
Multi-spectral, SWIR and hyperspectral, next generation of LED illumination
The next generation of LED Illumination is upon us! Metaphase Lighting is bringing the next generation of LEDs, optics, and driver controls technology to machine vision today. This presentation will cover the technology behind the latest LED Illumination that allows vision systems to extract more information than ever before. New driver and optic blending technologies allow Multispectral LED solutions to incorporate more wavelengths in a single light source. High Powered SWIR (short-wave Infrared) LEDs and optics expand the inspection capabilities for Line Scan & Area Scan applications. We also go into how LED technology is playing a role in Hyperspectral Imaging.
Filters for machine vision by machine vision
Optical filters are critical components of machine vision systems. They’re used to maximise contrast, improve colour, enhance subject recognition and control the light that’s reflected from the object being inspected. Learn more about the different filter types, what applications they’re best used for and the most important design features to look for in each. Not all machine vision filters are the same. Learn how to reduce the effects of angular short-shifting. Discover the benefits of filters that emulate the bell-shaped spectral output curve of the LED illumination being used. And find out more about the importance of a high-quality inspection process that limits the possibility for imperfections and enhances system performance.
(When in Trouble) Go Small
The trend towards miniaturization is entering a new and crucial phase. Improved performance and reduced power requirements make embedded devices an attractive alternative to computers. How does the machine vision industry deal with the change? Come discover MVTec’s take on this subject as a leading software developer.
Correct combination of liquid lenses with endocentric and telecentric optics
Liquid lenses are a great technology for fast and reliable focusing. As the vast diversity of options for sensors and optics can make the selection of components challenging, the goal of this talk is to provide simple design guidelines and examples so that you can make liquid lenses a practical part of your imaging toolbox.
Big image data - Smart image data
High-speed cameras are devices that generate extremely high data volume per unit of time. In the recording of single events, typical for slow-motion, data is predominantly unstructured (Big Image Data); on the other hand there are machine vision tasks that work mainly with structured image data (Smart Image Data) that leads to an image result after certain image processing. Due to the high datastream a new approach is required when selecting machine vision components during the design-in phase (to be able to use classical image processing tools during deployment).
Chemical Colour Imaging … makes hyperspectral cameras ready for the factory floor
Vibrational spectroscopy is based on the fact, that molecules reflect, absorb or ignore electromagnetic waves of certain wavelengths. Hyperspectral sensors measure those responses and return a spectrum per spatial point as the chemical fingerprint of a material. This data requires extensive processing to be useable for vision systems.
In this presentation, Perception Park explain how the hyperspectral camera technology and image processing can be combined/used in a HSI solution. Chemical colour imaging methods transform hyperspectral data into image streams. These streams can be configured to highlight chemical properties of interest and are sent to image processing systems via protocols like GigE Vision. Vision systems are extended to see additional chemical properties without the need of further development. Thus, objects, which only differ in their chemical properties, can now be separated from each other. Applications: Recycling, food safety, Quality Assurance (e.g. Pharma, Food and Packaging), colour measurement etc.
The industrial application of hyperspectral imaging is especially challenging in terms of sensor correction, repeatability and processing speed. Due to this, highly configurable tools are required which can fully utilise the sensor's capacity, due to GPU accelerated processing.
Microscope systems and their application possibilities in image processing
Microscopes are designed for use in the laboratory to look at small structures and objects. But for machine vision such systems are inefficient." Once upon a time this was the case, but this thought still haunts people today. In fact, there are now microscope systems that were specially developed for machine vision and, compared to lenses with finite imaging, not only offer higher resolutions but also open up new approaches to inspection procedures.
The importance of wavelengths on optical designs
Today's machine vision applications call for a much greater specialization of lenses for different wavelength ranges. Requirements range from monochrome illumination for metrology applications and daylight or NIR illumination for outdoor applications up to hyperspectral technologies in SWIR. Lenses to be used in such applications therefore require different kinds of colour correction to be considered during the design phase of a lens.
Illumination sequences for surface inspection: High-speed mechanisms for image acquisition and pre-processing
In several industry sectors surface inspection is utilizing industrial machine vision. A variety of algorithms is applied to the image data in order to solve this inspection task.
The detection of tiny surface errors is enabled by using high resolution sensors and leads to high bandwidth demands. Especially the control and acquisition of illumination sequences is a central element in this field of machine vision. Additionally several pre-processing steps need to be performed in an efficient way.
Different mechanisms for image creation are implemented and based on robust image acquisition and processing. How this is realised in practice – especially for high-speed processes – will be covered in this talk.
Use of telecentric lenses for applications beyond standard metrology setups
Why choose a telecentric lens? The advantages of telecentric measurements are obvious: High precision, constant magnification and low distortion, even for objects with a certain depth.
But the range of applications goes beyond that. Telecentric coaxial illumination through the lens improves surface texture evaluation. A telecentric lens combines with a focus tunable lens enables a variable working distance and z-scan. Moreover, telecentric imaging of a tilted object plane results in considerably lower distortion. Furthermore, special sensors and projection tasks need a telecentric beam path. All cases require verification of different factors before system development.
The presentation gives an overview of special optical setups as well as the possibilities and limitations of certain solutions.
IEC 62471 photobiological safety standards for LED lighting products
The machine vision technology forum will provide information to help users understand how to test LED lighting products in accordance to IEC 62471. The information presented will explain the IEC 62471 photobiological safety of lamps and lamp systems for LED lighting. A practical approach will be covered regarding the testing and the classification of an LED light into the proper risk group classification. An overview of the testing procedures, equipment and the IEC 62471 report will also be discussed.
Push broom hyperspectral imaging – advantages explained
Hyperspectral imaging (HSI) is a new vision technology for industrial on-line quality control, inspection and process monitoring. Many different HSI technologies have appeared as potential alternatives. When selecting instrumentation for real life application there are many functional and non-functional quality attributes to consider. Typically, spectral range, resolution, speed and cost (return of investment) set by application are the first ones to think of. However, one should also consider how selected technology affects the need for illumination, light collection efficiency, purity of spectral data, accuracy and simplicity of data processing.
This presentation explains major technical differences between push broom and competing approaches and how they affect practical usability in industry.
3D image processing – From challenge to achievement
3D image processing is now a widely accepted part of the automated quality inspection in industrial applications and still experiencing strong growths. It‘s used whenever the limits of classical procedures of the 2D image processing have been exhausted or when highly complex camera systems can simply be replaced by one 3D sensor.
This presentation gives an overview about the current state of 3D image processing. Furthermore, it describes the workflow from the selection of the hardware to the evaluation of received 3D data. Hereby, a particular focus is on the calibration of raw sensor data and the processing of 3D point clouds using variance analysis from a golden sample.
Image processing through the years
This year Common Vision Blox is 20 years old, what are the changes that have come to the machine vision market over those years? How has the hardware developed and what effect did it have on the software and the systems?
This talk tracks the changes as they affected Common Vision Blox as an independent programming library, from Intel Pentium 2 CPUs running at 233MHz to embedded ARM boards running at 2.4GHz. From TV standard interlaced analogue cameras to 260 megapixels. Learning tools, GPUs and distributed processing. But the biggest change? Standards
CVB++, CVB.Net and pyCVB – new approaches to state of the art application development with Common Vision Blox
Over the past 20 years, programming languages, runtime environments and programming techniques have continued to evolve. Today, software developers have a much broader range of tools, platforms and methods than when Common Vision Blox was first launched. Thanks to its C-based procedural API, Common Vision Blox can still be used with virtually every common programming language and on any standard platform, 20 years after its design.
On the other hand, many patterns are tied to high-linguistic resources, and the Common Vision Blox API now needs to be adapted by almost every customer. For this reason, three new APIs are being introduced, so that Common Vision Blox is easier to use with C ++, C# and Python languages, in particular for the creation and debugging of complex applications.
The design of the three language-specific APIs will be presented in a basic and comparative manner. Knowledge of at least one of the three languages is an advantage. The improved possibilities for troubleshooting and connection to common runtime libraries (Qt, WPF, Windows Forms, NumPy) are also discussed. An outline of the current state of the efforts is given as well as an outlook on further development steps.
Hardware and software for embedded machine vision
The generic term “Embedded Vision“ tells us very little about the hardware and software that encapsulates it. But there are huge differences in the hardware and operating systems involved. Whether it’s Windows IoT on Intel platforms, Linux on TX1 graphics processors or Android on simple ARM architectures, everything is lumped together and labelled “Embedded Vision“.
This presentation covers the diversity in hardware platforms, each with their pros and cons, and provides information on the possible tools for cross-platform application development. In addition, it presents the image acquisition possibilities for such platforms.
Industry 4.0 (IoT) – communication via OPC-UA
Open Platform Communications Unified Architecture (OPC UA) is a new interface standard for future industry 4.0 projects. Unlike other standards, OPC UA’s focus lies on creating new connections between levels that have, until now, been separated. Thus extending enterprise communication into the vertical dimension.
This means that units from different layers of the automation pyramid will be able to communicate with machine vision systems (MV systems). Thus, abstract results and raw data can be transmitted to a PLC, MES or ERP and these, in turn, may configure and control MV systems. Currently, MV systems use many different proprietary interfaces which increase complexity and development costs. A cross-system interface could facilitate the fusion of MV systems with further sensors.
This presentation shows the advantages of OPC UA for machine vision, the applications that can be solved, and the potential for more complex scenarios that can be considered.
Machine vision classifiers – Advantages and challenges of selected methods
Machine vision classifiers can be used for judging objects’ condition or appearance and assigning them class labels. In this presentation, we will discuss the use of ridge regression and convolutional neural networks (deep learning) for classification tasks as well as their advantages and challenges.
We will look at the basic theory necessary to get an insight on how these approaches work, why they perform well for certain use cases and where potential impasses lie. Different data sets have been trained and the results will be analyzed regarding classification accuracy and training- and classification times. We will investigate the requirements regarding dedicated hardware and training set sizes. Lastly, we will briefly explore how one of these approaches can also be used for robust scale and rotation invariant object detection.
Why is my system not working? Troubleshooting in machine vision systems
What to do when nothing works? Machine vision systems become more and more complex and manifold. It gets more difficult to classify problems as the cause of trouble and the symptoms might be far away from each other.
This presentation shows methods for troubleshooting and gives hints how to avoid mistakes or detect errors early.
Embedded vision for industry – A look at embedded vision solutions for industrial applications
The term “Embedded Vision” can take on a different meaning depending on who is using it and what the end application is. To some, embedded vision can simply mean the integration of a sensor for the purpose of digitizing an image, whereas to others it could mean deploying smart vision solutions for manufacturing quality control, robotic guidance or logistical movement and tracking of product.
This presentation explores some of these differences and focuses in on embedded vision for industry, including a brief tour of the different industrial embedded solutions available on the market today, reviewing their applicability and comparing their strengths and weaknesses. It will also discuss some of the common integration challenges that users face and take a look at some important software requirements that should be considered before deciding on which embedded vision solution is right for you.
It’s not just black and white anymore – How multi-modal imaging is changing machine vision
The world of imaging has moved well beyond the realm of the monochrome and colour imaging. In the past years imaging was used to generate pass/fail criteria from images collected from measurement and verification applications. But today image data must provide sufficient enough information to understand the root causes and improve yield. Furthermore, image data is now being associated with many invisible product quality attributes.
New techniques from research consortiums and universities continue to identify new relationships between material properties and the electromagnetic spectrum, such that non-destructive or non-invasive techniques can be employed to gather real-time information. However, the additional information desired often requires multiple data streams and synthesis of the raw image. Many of the most basic applications now utilize multiple methods of imaging and this session will explore some of these techniques.
Challenges and trends in microbolometer design
Due to the availability of smaller, more sensitive and cheaper solutions, long-wave infrared technologies are of increasing interest in a very wide variety of application areas. The aim of the session is to illustrate the challenges during the development and manufacturing process of uncooled LWIR detectors, inform you about the next big trend like wafer-level packaging, and highlight different techniques that allow the implementation of specific feature sets to meet the different requirements in industrial applications. Finally, an outlook on future designs and market trends will be given to the audience.
Why and for which applications do we need 155 Mpixel sensors/cameras?
Since the creation of the first digital sensor/camera in 1975 by Kodak (with only 0.1 megapixels) the resolution of sensors has increased tremendously, while their weight has dropped. There is an almost insatiable need for ever higher resolutions and speed. This need is driven from the Automated Optical Inspection (AOI) systems that are used in electronic, semiconductor and flat panel display (FPD) manufacturing.
AOI systems ensure consistent production quality in high-speed manufacturing processes. The camera with increased resolution monitors devices under inspection for failures and quality.
10 GigE and NBASE-T – Image processing made easy thanks to Cisco
PC hardware interfaces have gained popularity in the world of image processing due to the shift from analogue to digital cameras. Whereas the early digital cameras used Firewire interfaces, the current generation uses interfaces such as USB and Gigabit Ethernet, which are available on any standard PC and CoaXPress or CameraLink when high data rates are required.
CoaXPress and CameraLink represent the underlying trend of machine vision applications due to increasing sensor resolutions and speeds. The last few years have seen key players such as Cisco roll-out 10 GigE und NBASE-T interfaces in Ethernet backbones and high-performance Wi-Fi routers. In 2017, the increased acceptance of this standard in the consumer marketplace has seen the arrival of 10GbE Network Attached Storage (NAS) devices, 10GbE integration into Motherboards and even the first Network Interface Controller (NIC) cards for less than 100 USD.
The industrial image processing community benefits from the work of Cisco, Intel and others: The first 10gbE cameras for image processing are available and applications engineers profit from simplified component selection, long cable lengths and low latency times – and all this with low cost standard network infrastructure.
High-speed InGaAs SWIR line arrays & cameras for OCT and machine vision systems
IIndium Gallium Arsenid (short InGaAs) photo diodes conquer more and more industrial opto-electronic applications in the short wave infrared spectrum from 0.9 to 1.7µm (so called SWIR).
This lecture provides a short historical overview on the material development up to the modern state of the flip-chip detector architecture. Furthermore, practical machine vision and optical coherence tomography examples will be presented based on the newest high speed InGaAs line arrays and cameras.
New laser developments and how you can benefit from them
Lasers have become well-established as light source in 3D measurement applications such as triangulations sensors over the last 20 years. They are an indispensable part of today’s automated production lines. Nowadays, various wavelengths and output powers are available to provide the optimum foundations for the inspection of various materials under different conditions.
All components in a machine vision setup are being confronted with new challenges since the era of Industry 4.0 arrived. Due to new production processes and sophisticated drivers in electronics, the laser modules offer more than just the projection of a pattern. This presentation will show you new developments in the field of laser modules and how you can benefit from them.
Why the F-Mount is obsolete?
Since the 1980s people have been crying out computerized automatic image processing and analysis. The technology drivers until now have mainly been progresses in computer technology, sensor technology, illumination and the algorithms. The optics used were mostly those that were available from the video market (C-Mount) or the photo market (F-Mount).
Today F-Mount lenses are used in large numbers of industrial applications despite their widely known drawbacks and the fact that better alternatives are available. Are there still reasons to use F-Mount lenses in industrial applications?