Watching food in 3D
09 March 2009
As the food and packaging industry comes under increasing pressures to reduce waste and save money a factory and process automation specialist says can offer a solution. David Hannaby of Sick UK explains 3D vision…
Biscuit application: Sick camera tracks height and shape of biscuits on production line
Three-dimensional vision cameras can measure height, depth and volume, and examine the relative contours of an uneven surface profile. By comparison, two dimensional systems are only able to generate a flat profile from above and cannot measure the depth of any defect, or calculate the volume of the product. A 3D system can be set up to reject products with air pockets or sinkage outside certain tolerances for example - a 2D camera would detect only the outline shape of the defect, and not its depth.
Laser-based triangulation forms the basis of 3D vision. A single camera is set at an angle on the axis of an object moving along a conveyor. This camera is integrated with a laser that is mounted vertically above the object. As the object moves through the laser line, the camera builds up a series of profiles to create a digitised 3D image which can follow the surface contours.
Some developments of this technology have used so-called 3D Smart cameras with integrated processing capability. The smart 3D cameras have a fixed angle between camera and laser, which means distortions can be calibrated out at the factory. As the relationship between the camera and laser cannot be changed, any further perspective calibration is unnecessary.
In many cases a 2D vision system is all that is needed for robotic cell pick and place applications, but two-dimensional systems have their limitations. In essence, they can see only ‘from above’ the length, width, orientation and position; no height detail is available.
Quality inspection with a 3D system can be enhanced because it can detect the depth of a surface defect. A 2D inspection system may pick up defects in a product that result in a colour or grey scale differences but it cannot detect the depth of the defect. So for example, a 3D system could be set up to reject baked products with air pockets or sinkage outside a certain tolerance, whereas a 2D system simply wouldn’t ‘see’ such defects.
3D visualisation of a biscuit box
Sick UK were challenged to solve a problem in tracking the height and shape of biscuits prior to wrapping. The manufacturer had experienced difficulty when the biscuits strayed outside of tolerance, accumulating an oversize which made them difficult or impossible to wrap. Unsatisfactory packs could easily escalate into major line downtime, as the packs often burst under handling and whole batches of product were being wasted as a result.
In addition, checking back on all the production tolerances up the line to determine the exact problem, recalibrating the process machinery – moulds, depositors, extruders, mix recipes, oven cooking times and so on – could cause expensive and time-costly hold ups in production.
Typical parameters for a round biscuit could be to find the height profile, find the centres (for exactly positioning) or find the edges; measure the dimensions, or look at ‘ovality’. The volume of the biscuit can then be calculated and compared against a check weight and known volume to determine dough consistency.
Lid Integrity Control
Bottles containing liquids can cause significant production problems and lead to returns if the lids are not properly sealed. Using a 3D camera, the lids can be inspected at high speed. The camera records the 3D shape of each lid as it passes through the conveyor. The angle of each lid is measured and the height of the lid surface is measured and compared to a horizontal reference surface.
Inspecting the naan bread for pitted surfaces
Faulty bottles are rejected and the measured values are communicated through a digital I/O signal to an external unit for analysis and statistics accumulation purposes.
For plant bakers, one exciting new application of the technology is to measure the volume of double-dough in bread pans prior to baking. Too much dough in the tins can lead to halting of the production line whilst corrections are made – a potentially highly costly error.
Previous 2D technology could only establish whether dough in the tin was too ‘low’ or ‘high’, but the measurement was unreliable because it could not account for the uneven profile of the dough in tins. The 3D application can plot the x, y and z co-ordinates of the dough’s surface profile and the data can then be used to accurately calculate the volume of the dough in each tin.
This is one of the most important features of 3D vision: It leads to an ability to make corrections in the process much more quickly ‘in real time’ when integrated with other production line systems.
Contact Details and Archive...
Most Viewed Articles...