This website uses cookies primarily for visitor analytics. Certain pages will ask you to fill in contact details to receive additional information. On these pages you have the option of having the site log your details for future visits. Indicating you want the site to remember your details will place a cookie on your device. To view our full cookie policy, please click here. You can also view it at any time by going to our Contact Us page.

A machine vision evolution

07 October 2017

Paul Wilson looks at some key developments in machine vision which have offered benefits for the food industry. 

This year is the 25th anniversary of the UK Industrial Vision Association (UKIVA) which was formed in 1992 to help to educate manufacturers and processors in many different industrial sectors about the benefits vision could bring. The technology can offer a competitive advantage for food manufacturers. It opens up possibilities in quality assurance that were previously impossible to implement. 

The use of vision in the food industry stretches back to the mid 1980s, where one early application was for muffin inspection. Using a camera with a resolution of just 30 x 32 pixels, the system was used to reject rogue oversize muffins that would have jammed up the packaging machines. Machine vision technology has moved on dramatically since then, not only in terms of camera technology, but also in computer processing power, software and illumination techniques. Some of the key developments that have had a major impact in food manufacture include: 

• General imaging technology developments
• Line scan cameras
• Smart cameras
• Machine vision standards
• Imaging outside the visible spectrum
• 3D imaging
• Robot pick-and-place

By automating in-line inspection/quality control processes, machine vision has had an immediate impact on productivity, efficiency, and product quality in the food industry.

Imaging developments
Machine vision systems typically consist of a camera, lens and lighting for imaging and a PC where imaging software makes the required measurements. One big and on-going development is in camera sensor technology, most recently with the latest generation of CMOS sensors. This has brought continuous enhancements in terms of image resolution, imaging speed and sensitivity, which means faster processes can be imaged with greater detail and often at lower cost. This improves throughput and may mean the difference between the ability to inspect 100% of the product and using a sampling regime. An important early development was the emergence of colour cameras. This meant that items could either be identified for measurement according to colour or that colour comparisons could be made. For example, to ensure the correct level of ‘browning’ on a baked product or to grade fruit and vegetables according to colour. Increases in the levels of computer processing power have significantly increased image processing and measurement capabilities. Easy to set-up and implement software has expanded the range of applications and means that imaging systems can readily accommodate different product types on a manufacturing or processing line. 

Ease of use has also made machine vision much more accessible for non-specialist personnel. Faster frame rates allow faster throughput, but less light falls on the sensor per frame, so developments in LED illumination technology have been important in delivering sufficient light intensity. The appropriate illumination configurations can also minimise any effects from the natural light that may be present in food manufacturing plants. Vision systems can offer much more than simple ‘pass/fail’ decisions on product quality. By integrating results from 100% inspection into a statistical process control system, trends can be monitored which can lead to early remedial action. For example, oven temperature could be adjusted before any overcooked product is produced, avoiding waste and any downtime of the process.

Line scan cameras
The vision systems described so far make use of area scan cameras, However the introduction of line scan cameras opened up more opportunities in the food industry. Line scan cameras only have a single row of light-sensitive pixels, which build an image line by line as the objects move past them. This makes them ideal where moving, continuous material needs to be inspected. Applications could include inspection of cereal grains or similar, which are allowed to cascade past the camera like a curtain. The system identifies imperfect grains or foreign bodies and triggers compressed air jets to automatically eject the unwanted material.

Smart cameras 
One of the problems with early vision systems was transferring images from the camera to the PC over long distances. This led to limitations in the allowable camera positions in the factory. That changed with the introduction of the intelligent or ‘smart’ camera. Here, all of the image processing hardware and software is contained within the camera body itself. The camera can make the required measurement and simply send a pass/fail signal to the processing line control system. While this opened up many more inspection possibilities, there was a cost implication if a significant number of cameras were needed. The introduction of dedicated machine vision data transfer standards helped to solve this. 

Before 1995 there were no machine vision transmission standards. Between 1995 and 2005, however, the first dedicated standard, CameraLink, was developed and extensive use was also made of two consumer interfaces: IEEE1394 (Firewire) and USB 2.0 for lower end applications. The development of the GigE Vision standard was introduced in 2005. This had a number of advantages. Firstly it allowed the transfer of image data over distances of up to 100 m, meaning cameras located all over a factory could be connected to a single PC. Secondly industry standard Ethernet cables and switches, already in use in many factories, could be used, and thirdly, GigE Vision compliant cameras from all manufacturers were interchangeable. Since then, new standards have evolved, primarily to handle the ever-increasing volumes of data produced as cameras achieve increasing spatial resolution and frame rates. 

Increasing processing power has also fuelled the emergence of affordable 3D imaging systems which allow the accurate measurement of product volume. Probably the most popular 3D imaging technique in the food industry is that of laser triangulation. The object to be measured passes through a line of laser light and a camera mounted at a known angle to the laser records the resulting changing profile of the laser line. These 3D profiles produce a point cloud which can be analysed to produce a 3D reconstruction of the object. The ability to make real time measurements in the X, Y and Z axes at production line speeds not only allows volumes of product to be calculated and defects to be detected but also pass/fail decisions to be made on far more parameters than would be possible for traditional 2D measurements. Decisions can be made on product shape, proportions and even surface quality. 

3D measurements have been used in the baking industry on products such as pizza, pies, bread, pasties, cakes and biscuits to check for shape, size, edge defects and thickness. 3D vision can further enhance manufacturing automation by being used to guide robots for pick- and-place applications. 

Near infrared light (NIR) extends beyond the visible spectrum from wavelengths of around 750nm to 1µm. NIR radiation allows subsurface features of an object to be viewed, revealing otherwise unobservable defects. NIR inspection of foods such as fruit, vegetables, nuts and meat allows checks to be made for subsurface signs of decay, mechanical bruising and pest damage. Short Wave (0.9 – 1.7 µm) and Long Wave (8 – 13 µm) IR can be used for thermal imaging and thermography to allow temperature variations in food products to be visualised and actual temperatures to be measured.

Vision in action
Traditional inspection applications involve checking parameters such as shape, size, position, edge defects, holes, and the correct presence and distribution of fillings or toppings etc, using both 2D and 3D imaging. Other applications include use in slicing equipment for portion control for products such as bacon, cheese and ham to maximise the on-weight percentages and minimise giveaway. New methods, such as hyperspectral imaging, combine spectroscopy and machine vision to create images which are colour coded according to the actual chemical composition of the material being imaged. Application examples include the detection of fat and bones on chicken; identifying bruising in oranges, apple sorting; and coffee bean inspection. There are also countless applications in food labelling; checking of packaging for defects; checking fill levels and the integrity of the final packaging for product purity and shelf-life considerations.

In the last 25 years we have seen a myriad of technological developments that have transformed the machine vision landscape and this is set to continue. UKIVA remains committed to raising awareness both of current and future capabilities.

Paul Wilson is chairman of the UKIVA. Thanks to UKIVA members Alrad Imaging, Allied Vision, Multipix Imaging, Scorpion Vision, Sick (UK) and Stemmer Imaging for contributions to this article.

Contact Details and Archive...

Print this page | E-mail this page