Send Us Feedback

Interactive Technology Portal
From Food Tech Innovation Portal

Vision system for robot guidance and food inspection


Key words Vision, food, 3D, food industry, NIR, inspection, quality, processing, control
Latest version 2011/12/08
Completed by SP

How does it work?

Primary objective 1. Non-contact product inspection.

2. Product location for robot handling.

Working principle A vision system, or machine vision system, generally consists of a light source, a camera, an image capture hardware and computer hardware and software. The hardware and the software can be incorporated in the camera housing or in an external PC. To enable product images capturing, the camera unit and the light source are generally mounted above or at the sides of a product conveyer, only requiring a relatively small space.

When a product passes the vision system an image, or a series of images, is captured and processed by the hardware/software. The image consists of a set of data (pixels) representing the magnitude of captured voltage from the sensors. Image processing and image analysis is then used to distinguish shapes and features in the image. The software generally starts with extracting continous areas with similar colour or grayscale in an image. These objects are then examined to ensure that it is a product by e.g. measuring its size, diameter, elongation or colour. When a product is identified it can be inspected to its specification regarding shape, size, colour etc. and its coordinates can be extracted (1, 6, 8).

For vision systems, two main application areas can be distinguished: product inspection, and product location for robot handling. These are however often inter-connected as the vision system used for product location potentially also can be used for product inspection to some extent, or vice versa.
Product inspection can involve measurements of the product size, shape, stains, miscolourations, quality grading, variety classification (5), blemishes, marks or colour on the sample, to measurements of inner defects (3). Systems combining vision and laser can be used to create a 3D image of an object, enabling both volume measurements and to some extent an estimation of weight. In a robot food handling system the vision system is used to extract product parameters such as position and orientation to enable accurate gripping.

If the light source is placed above or at the side, it is the reflected light that is captured by the camera. A light source can also be placed under a semitransparent, white, conveyer, resulting in a silhouette image of a product. With reflected light information of the colour and the texture of the products surface can be extracted, as well as shape and position. A silhouette image is used to extract position data.

Some application uses multiple cameras at various angles to ensure complete product inspection whilst others only need an overhead or side view. For vision system inspection various light sources and suitable sensors can be used. The light sources that are the one most commonly researched and to various extent commercialised are using wavelength in the visual spectra (often LED or fluorescence), X-ray and NIR, selected depending on what is to be analysed. The visual spectrum is the predominant technique. With a “normal” vision system black and white or colour images are captured and processed.

Additional effects Product images captured for product inspection can also be stored to increase traceability.

An automated quality inspection can be used as feedback data in a process control system. With a functional product inspection system a company is not dependent on highly experienced workers to maintain product quality. A running vision system never tires or gets bored and performs the programmed inspections 24h with very little need for maintenance. This technology will always give an objective inspection of the process.

Important process parameters
  • Selecting an appropriate technique depending on use.
  • Correct and stable illumination.
  • Suitable vision system algorithms (software).
  • Available development software.
  • Processing speed (hardware).
  • Choosing correct aperture - lens.
  • IP rating.
Important product parameters

What can it be used for?

Products Vision systems can be used for all solid products.
Operations Operations where vision systems can be used are: identification, quality inspection, checking lable position, checking amount, packaging, robot handling, re-orientation, robotic pick and place.


- Detection of uneven browning on cookies (2).

- Measurement of stains on pistachio nuts.

- Grading of pork colour (7).

- Distribution of tomato sauce on pizza (12).

- Checking if a tray is full or empty.

- Distribution of colour or examination of areas covered by topping.

- Detection of internal watercore in apples (4).

- Grading of aubergin and oranges (5).

- Sorting of pistachio nuts with closed shells (10).

- Identification of a product type in a mixed infeed of different fruit and vegetable products enebeling the use of a predefined grip force to avoid product handling damage (13).

- Finding the least overlapped hamburger from a pile of overlapping burgers (9).

- 3D vision for bin picking applications.

- Stereoscopic vision for fruit harvesting and fish bone detection.

Solutions for short comings Process and product inspection is important to order to ensure food quality, product consistency and safety. For most food products it is an advantage to be able to perform fast and non-contact measurements of the products to minimize contamination sources and to avoid product damage. Both can be achieved by the use of vision systems.

What can it NOT be used for?

Products Various features and products can be more or less difficult to analyse and thus make the use of vision sytems practically impossible.
Operations This will depend on how well a parameter can be captured and measured with a vision system. Overlapping products can be difficult to measure, but for simpler geometries this is possible today. Product singulation in bins is an area of research.
Other limitations Depending on camera resolution and conveyer speed there will be a limitation of how small details or how fast products that can be analysed.

Today inspection or identification software algorithms can relatively straightforwardly be developed for standardised shapes, or products with relatively small variations in size e.g. cookies, sausages, pancakes, packed products etc. For natural products the variation in shape, color and size can be problematic for the system to handle.

Risks or hazards There are no clear risks with vision systems. X-ray should be properly shielded.


Maturity Many vision system applications, such as quality inspection and recording of robot handling parameters, are used today in the food industry. There is research on new applications and development is continuous where some applications are on lab scale and others closer to commercialisation.

For handling of products with a relatively consistent shape and size variation, e.g. cookies, sausages, pralines etc., standardised commercial solutions are available. Industrial examples are automated handling of pancakes, sausages, coockies, pork shops, chicken fillets, chocolate, or foreign objects (1, 6, 8).

Modularity /Implementation A vision system can often be fitted over a relatively small section of e.g. a product conveyer. This is of course dependent on the use. For example it only requires a small space to place a scanner to check if a box is empty, to make an NIR measurements the instrument requires a section of the conveyer. To fit a device that e.g. redirects products that fail to pass a quality check might require some space.
Consumer aspects There should be no consumer aspects on vision technique as it is a non contact measurement and often used to inspect and improve quality, see X-ray for non-invasive food quality control.
Legal aspects Please check local legislation.
Environmental aspects None

Further Information

Institutes SP, IRTA, Wageningen UR - FBR
Companies JAI, E2v technologies, Cognex
References 1. ABB.

2. Abdullah, Z. M., Aziz, A. S., & Dos-Mohamed, A. M. (2000). Quality inspection of bakery products using a colour-based machine vision system. Journal of Food Quality, 23(1), 39–50.

3. Brosnan, T., & Sun, D. (2004). Improving quality inspection of food products by computer vision-a review. Journal of Food Engineering, Vol. 61, pp. 3-16

4. Kim, S., & Schatzki, T. F. (2000). Apple watercore sorting using X-ray imagery: I. Algorithm development. Transactions of the ASAE, 43(6), 1695–1702.

5. Kondo, N. (2009). Automation on fruit and vegetable grading system and food traceability. Trends in Food Science & Technology, in press

6. KUKA,

7. Lu, J., Tan, J., Shatadal, P., & Gerrard, D. E. (2000). Evaluation of pork color by using computer vision. Meat Science, 56, 57–60.

8. Marel,

9. Muhammedali, B., Abdullah, M. Z., & Mohd Azemi, M. M. N. (2004). Food Handling and Packaging Using Computer Vision and Robot. International Conference on Computer Graphics, Imaging and Visualization (CGIV'04) , pp.177-182

10. Pearson, T., & Toyofuku, N. (2000). Automated sorting of pistachio nuts with closed shells. Applied Engineering in Agriculture, 16(1), 91–94.

11. Pettersson, A., Ohlsson, T., Davis., Gray, J. O., & Dodd, T. J. (2010). A hygienically designed force gripper for flexible handling of variable and easily damaged natural food products. Submitted to Innovative Food Science and Emerging Technologies (IFSET)

12. Sun, D, -W. & Brosnan, T, (2003). Pizza quality evaluation using computer vision––part 1 Pizza base and sauce spread. Journal of Food Engineering, 57, 81–89.

13. N.J.J.P. Koenderink, M. Wigham, F. Golbach, G. Otten, R. Gerlich and H.J. van de Zedde, MARVIN: High speed 3D imaging for seedling classification, Precision agriculture '09, Contribution in proceedings, p. 279

  • Selecting an appropriate technique depending on use.
  • Correct and stable illumination.
  • Suitable vision system algorithms (software).
  • Available development software.
  • Processing speed (hardware).
  • Choosing correct aperture - lens.
  • IP rating.warning.png"<span></span>
  • Selecting an appropriate technique depending on use.
  • Correct and stable illumination.
  • Suitable vision system algorithms (software).
  • Available development software.
  • Processing speed (hardware).
  • Choosing correct aperture - lens.
  • IP rating." cannot be used as a page name in this wiki.

Sensors and Indicators 2.2.5 physical packaging, other ICT ScienceDirekt, Search terms: Vision system food, 3D vision system food, vision system food industry, vision food processing, vision binpicking, NIR food inspection, x-ray food, food inspection , food quality control WikiSysop :Template:Review document :Template:Review status

Translate this page with Google Translator (automatic translation)
Created by Evelina on 9 March 2012, at 18:17