Technological developments create opportunities for machine vision

Paul Boughton
and the following year saw more mobile telephones with built-in cameras sold than digital cameras.
"In the realm of amateur photography, digital cameras have been reported to be out-selling their...  

In the realm of amateur photographydigital cameras have been reported to be out-selling their traditional film-carrying counterparts since 2003and the following year saw more mobile telephones with built-in cameras sold than digital cameras. With these rapid advances in digital imagingit is not surprising that machine vision has also developed considerably in the last five years – even though some people might say the rate of progress has been disappointingconsidering the remarkable speed with which consumer products have evolved in this time.
Machine vision is employed in a variety of manufacturing environments for inspectiondata collection and process control tasks. ‘Smart’ cameras are now popularas they incorporate all the necessary image processing hardware and software to enable the camera to be used as a sensorwith an output signal that can be used as an input to a PLC (programmable logic controller)PAC (programmable automation controller) or PC-based monitoring and/or control system. In common with other products in the industrial automation marketsuppliers are under pressure to deliver greater functionality at lower cost – and without compromising reliability.
Although there are a number of established manufacturers of smart camerasa company from outside this field has recently entered the market with two new products. The National Instruments NI1722 and NI1742 smart cameras combine an image sensor with an industrial controller and NI vision softwaremaking them suitable for applications such as locating and inspecting parts and assembliesas well as reading 1D barcodes and 2D codes such as Data Matrix (Fig.1).
Perhaps the most notable feature of these cameras is that they are shipped with National Instruments’ Vision Builder for Automated Inspection (AI) software for configuringbenchmarking and deploying machine vision applications without programming. Using this intuitivemenu-driven softwareengineers can build complex machine vision applications incorporating not only vision algorithms but also state-based execution with looping and branching. For more advanced applicationsthe cameras also integrate with National Instruments’ Labview software and the full NI library of image processing and machine vision algorithms such as edge detectionpattern matchingcode reading and optical character recognition (OCR).
Both cameras are suitable for use in harsh industrial environmentswith the NI1722 featuring a 400MHz PowerPC processor and the NI1742 having a 533MHz processor. They incorporate a monochrome VGA (640x480) image sensor and built-in inputs and outputsincluding two

opto-isolated digital inputs and two

opto-isolated digital outputsone RS232 serial port and two gigabit Ethernet ports. In additionthe NI1742 includes quadrature encoder support and a built-in controller with NI direct drive lighting technology.

Robot vision
Today’s machine vision systems can be used for a wide range of industrial applicationsas outlined abovebut there is one in particular that is receiving increasing attention. Vision-guided robots have the potential to broaden the application scope of industrial robotsmaking them less reliant on a defined working environment and set of tools and components with which they interact. While such applications clearly require cameras with a suitably high specificationit is the software that makes the biggest difference.
Stemmer ImagingKuka and the Swiss Interstaatliche Hochschule für Technik (Interstate College for Technology) in Buchs (NTB) have been collaborating to develop a ‘seeing robot’ (Fig.2). The software runs directly on the Kuka robot controller so thatin just a few steps and without any programmingusers can now set up the systemlink the robot and image processing and teach the robot to recognise objects. After thatthe robot can be left to run fully automatically.
The core of the system is Stemmer Imaging’s Common Vision Blox (CVB)an image processing librarywhich is the first application of the Shapefinder2 tool for robots. Other CVB tools can also be integrated for additional functions.
A useful feature of the software is that it provides a high degree of flexibility in camera selectionso users can select a monochrome or colour camera with the required resolution and other characteristics to match the application. It is anticipated that the system will be used for object location and recognitionrobot handling and pick-and-place tasks. Currently the system is only available to customers in Switzerland.
Howeverthere are other options for companies wishing to use vision-assisted robots. Scorpion Vision’s latest software package includes a 3D add-on module for stereo vision and 3D modelling. The software operates in conjunction with two cameras in the same way that a human uses a pair of eyes. Images of the subject are taken from two different anglesthen the Scorpion software creates a third image that is a combination of the twobut with the perspective removed. The image therefore appears to be a top-down view of the subject. 3D vision algorithms extrapolate the vertical distance from the subject – and can additionally be used to make 3D measurements.
While Scorpion software has been used for robot vision applications for a number of yearsthe company says its 3D vision capability can be used for picking components that are arranged randomly in a pile. This means that the robot does not have to be presented with components arranged in trays or in random orientation on a common plane – as might be the case on a conveyor. With Scorpion’s 3D technologya robot can pick components that are supplied in a boxbin or other container. There is therefore no need for an operative to handle the components or for any mechanical sorting by means offor examplea vibratory bowl feeder.
Infrared spectrum
While robots are one niche application for machine visioninfrared is often seen as another. Howeverthe breadth of applications for infrared or thermal imaging is greater than many engineers realiseand prices for infrared cameras are fallingwhich makes them feasible for applications where the cost was traditionally prohibitive. Obvious applications are those where temperature differentials can indicate underlying problemssuch as testing of electrical and electronic equipment. Howevera little lateral thinking can lead to infrared cameras being used for less obvious applicationssuch as monitoring fill levels when cold drinking yogurts are added to opaque plastic pots that have only just been moulded – and are therefore still warm.
Other applications for infrared cameras include quality control processes where the deliberate application of controlled heat can be used to highlight sub-surface defects through unevenly absorbed heat. One example of this is inspecting for inadequately bonded laminations in fibre-reinforced composite components; another is checking for sub-surface bubbles in moulded automotive dashboards. Infrared cameras used in conjunction with infrared light-emitting diode (LED) illumination can also be used successfully in applications where ambient light or other factors cause difficulties for machine vision systems operating in the visible spectrum.
A further way in which infrared cameras can be used is in the imaging of stresses in materials and structuresand damage evaluation. Cedipa company in which Flir Systems has recently purchased a major stakehas developed the Altair LI system that is based upon a

high-performance focal plane array camera and digital image processing software. It is capable of producing high-quality images of stress fields in materials and structures under dynamic loading conditions. The Altair LI provides full field stress images in real time by using the thermoelastic effectin which there is a linear relationship between the stress at the material surface and the temperature changes induced by loading. To measure stress with a resolution of 1MPa requires a vision system with a thermal resolution in the order of 1mK for steel and 2mK for aluminium.

Traditionally infrared imaging has cost more than machine vision operating in the visible spectrumbut costs are falling and manufacturers are building cameras with a variety of specifications to suit different application requirements and budgets. Impac’s HS-vision series cameras (Fig.3)for exampleare suitable for automatic detection of thermal anomalies (hot or cold spots)or the company offers TG-vision series cameras for research and development and light industrial applications. For harsh industrial applicationsthe ID-vision series feature a system for lens purging as well as a special front plate for air- or water-cooling.
Simple implementation
Flir Systems says that its affordable Thermovision A-series infrared cameras are aimed at making it as simple to implement an infrared vision system as a visual camera (Fig.4). FurthermoreFlir claims that these are the first thermal cameras to be compatible with Gigevision and Genicam standards and to feature a trigger/synchronisation capability. This flexibility makes it easy to use the cameras with

third-party hardware and software. All that is needed to receive high-qualityreal-time radiometric images is to connect the Thermovision A320G camera to a PC and install the software. The fixed-installation camera provides excellent live images with 320x240pixel resolution.

In some applications it can be beneficial to work with both the visible and infrared spectra – which is now possible using a single camera from e2v. This company’s EliixaUC8 line scan camera has four 4096-pixel linescan elements on a single sensor chipwhich enables both colour RGB (redgreen and blue) and either a monochrome channel or near-infrared signal to be acquired at the full 18kHz line rate (Fig.5). Being a line scan camerathe device relies on the inspected product being moved across the field of view on a conveyor or by other means. Neverthelessthe camera can be used for a variety of inspection applications – such as websfoodpharmaceuticals and printed circuit boards – as well as scanning of documentslabels or film.
If truly high-speed imaging is requiredthere are specialist cameras available that can capture thousands of images every second. One example of such a product is the Optronis Camrecord5000 high-speed camera that benefits from a very

high-sensitivity sensorwhich makes the camera suitable for applications with low light level or in areas where it is not possible to install an additional light source (Fig.6). Capable of capturing 5000 frames per second (fps)the Camrecord 5000 is designed for use in both industrial and research and development environments. Typical applications include analysis of combustion or explosionsprocess analysis on high-speed machinery (including packaging machinery)and analysis of products subjected to drop tests.

The Camrecord5000 has a resolution of 512x512pixels and is controlled via a standard Gigabit Ethernet interface. Howeverif the application is suitable for using windowing to sample a smaller areas of 512x8pixelsthe frame rate can be increased to 320000fps.
Safety applications
In the April 2007 issue of
In the near future Pilz will be introducing an upgraded version of the PSENvip system that can measure the angle of the bend being formedas well as performing its safety function. This process data can then be fed back into the press brake’s control systemthereby helping to make the forming operation as accurate and efficient as possible.
Sick has also developed a camera-based safety systembut this time as a replacement for conventional safety light curtains. The advantages of the SickV300 are said to be a lower purchase costand easierfaster installation and commissioning. Suitable for monitoring rectangular openingsthe V300 is mounted in one corner and a reflector strip is installed along the two opposite sides of the aperture. For larger openings or machines requiring guarding on two adjacent sidesa pair of V300 units can be synchronised together.
ompared with what was availablesay10 years agoindustrial machine vision systems have changed enormously. Thanks to developments in hardware and softwarethey cost significantly lessare easier to implementand are more reliable. In some cases today’s vision systems can perform functions that their predecessors were simply incapable ofwhereas in other cases machine vision systems are now a cost-effective alternative to multiple conventional sensors. Either waythey are worthy of serious consideration for a wide range of qualityprocess controland research and development applications.
European Design Engineer we explained how Pilz had developed the Safetyeye machine safeguarding system based on a three-dimensional vision system. More recently the company has launched a vision-based press brake monitoring system that uses a green LED light source and a single camera of the same type as the three in the Safetyeye. Easy to configurethe new PSENvip system creates protected fields that are constantly monitored and represented on the camera unit’s integral display (Fig.7). Should any finger-sized object breach the protected fieldsthe press is stopped immediately.


Recent Issues