Today all Vendors have unified the main design for all-around (360º) CCD-camera equipped inspection of bottles’ sidewall, necks and finishes (labels, fill level and closure), in a stainless cabinet which can be opened for maintenance on both lateral sides and by the top with bottles passing through


                        Introduction

A real jump-up on Label Inspection performances is obtained when changing technology from Photoscanners to Machine Vision.   Reason is the much greater amount of Information about the label properties that it results possible to register with a camera.  Amount established correlating a label with hundredths of thousands (or, millions) of individual photo-detectors, rather than just one.   The figure above shows an example of the equipments’ for final inspection using CCD- or CMOS-technology cameras to inspect the labels’ presence and position.   At left side, a typical example of labelling where Machine Vision becomes nearly unavoidable.


Label Inspection

       Flash animation, 49 MB, ZIP

   After download click index.html 



       

            70 pages, 37 MB


What performances

The performances to expect depend on:

  • the optic configuration, 
  • amount of cameras, 
  • eventual adoption of Artificial Intelligence algorithms,
  • amount of CPUs devoted to the task or, availability of GPUs.




 In the fast Labeller Machines >50000 bottles-per-hour the Collars missing and mal-positioned, account typical rates as high as ~80 % of the total rejects, with missing and malpositioned Body and Black labels summing up to the remaining ~20 % of the detected defects (  CARLSBERG Group, 2014)





As an example, for the final inspection equipment depicted above, they are close to:

  • (1.0 ± 0.1) mm malpositioned labels, detected at >96 % with associated false reject ratio < 0.01 %,
  • missing- or flagging-labels, detected at >99.5 % with associated false reject ratio < 0.01 %.

To detect the entire external sidewall of the container, means to use two CCD- or CMOS- cameras equipped with some megapixels each.   Notoriously, each pixel emulates an elementary (passive-only, no light source) analog photoscanner, whose dynamical range following its model ranges today (4096 - 65535) grey levels.   By these digits, millions of individual analog photo detectors, and the fact that from each one we are capable to discriminate thousands of grey levels, it is easy to infer why the performances jump-up from the mere presence or absence of a label detectable with a single photoscanner, to positions whose accuracy level rounds 0.1 mm.




Transparent Labels, Collars, Neck and Foils: 

Machine Vision shows its best

Collars, like those visible in the figure below shot at a SABMiller® PLC Brewery in Colombia, when flagging are a typical challenge fronted daily in the breweries.  impossible to counter with a single Label Presence photoscanner in-the-Labeller Machine.  On the opposite, a truly easy  application where the same defect is detected by mean of Machine Vision in an external final inspection system at-the-Conveyor.          In today’s fast (>50000 bottles-per-hour) rotary Labeller Machines, the missing plus malpositioned collars account typically for ~80 % of the total rejects, with missing and mal positioned Body and Black labels summing up to the remaining ~20 % of the detected defects.   One of the reasons lies in the difficulty to label the most curved  surface of the bottles’ external sidewall.  Also, a non-cylindrical shape, rather a conical section.    Same is true also for the Label Presence inspection in-the-Labeller Machine.  It’ll be established a correlation of the shortest duration between the single photoscanner and label, in an area of the bottle so curved to result unavoidably that one featuring the highest luminous diffusion, due to known laws of Geometric Optic.  Then resulting a minimum for the luminous re-emission infeeding photoscanner's phototransistor. 

   Transparent self-adhesive labels.  Their reflective area reduced to a small text and logo.  An application prone to cause false rejects yet adopting the best and most modern photoscanners today existing.  A natural field of application for camera-systems




 Flagging Collars are a typical challenge impossible to counter with a single Label Presence photoscanner in-the-Labeller Machine.  On the opposite, an easy application where the same defect is detected by mean of Machine Vision in an external final inspection system at-the-Conveyor, in this image shot at a SABMiller PLC Colombia, Grupo Bavaria (   Tom Parker/OneRedEye/2012)


What is a “good Label inspection” ?

The correlated Label-Pixel State










As we saw with plenty of details elsewhere in this web site, as an example we’ll try to determine if a bottle is labelled by mean of a system like a camera, including millions of Pixels.   Pixels accumulate one-by-one the photons reflected by the sidewall, neck and closure of the bottle.   Later discharging the potential accumulated to an A/D converter whose outfeed is sequentially read by following circuits to deduce what to a level of illumination has been subjected each individual pixel during the limited exposure Time.  [There are particularly fast Electronic Inspectors, adopted to inspect plastic empty bottles, having the exposure Time of the Pixels in their cameras forcedly limited to tens of microseconds…]    Whatever physical system, label included, is represented by its wave function or state vectors.    The physical meaning of the state vector becomes apparent when making a measurement.   Then the state of the system assumes one of the eigenstates, with probability given by the Born rule, and the result of the measurement is the corresponding eigenvalue.


 To establish a relation between a label (“Label”) and a Photoscanner (“Photoscanner”), both differentially related with the Environment (“Environment1”, Environment2”), is necessary Time.  Time to transform the previous state, in which all possible kinds of correlation of the Photoscanner coexist, in a following state in which the Photoscanner is “aware” to be correlated to a label, because having recorded eigenvalues for the eigenfunction ΦiS1 describing a cap.   The vertexes represent interferences


To definetely perceive a correlated Label-Pixel State, they are necessary:

  1. Timeto transform the previous state, in which all possible kinds of correlation of the pixel coexist, in a following state in which a camera inspection is “aware” to be correlated to a Label, because having recorded eigenvalues for the eigenfunction ΦiS1 describing a Label.  The correlation between the two systems, camera inspection and Label, is progressively established during interaction and proportional to the natural logarithm (ln t) of the interaction time t.   An ideal correlation, one corresponding to a maximised information of the inspection about the Label, can only be reached allowing an infinite time.   The fact we cannot wait for an infinite time causes the measurements’ fluctuations, a synonimous of the spectrum of the eigenvalues, resulting in the Electronic Inspector's false positives (false rejects).    Time, for what ?    To transform the previous state, in which all possible kinds of correlation (superpositions) of each one of the many pixels composing a camera sensor coexist, in a following state in which the inspection is aware to be correlated to a Label, because having recorded eigenvalues for the eigenfunction ΦiS1 describing a Label by the camera pixels;
  1. Interaction between the systems such that the Information in the marginal distribution of the object inspected is never decreased.  In a probability distribution deriving by two random variables, we remember that marginal distribution is where we are only interested in one of them.   Otherwise, we’d have forced a reduction in the sample space of one of the random variables and then, we could not have any more repeatability of the following measurements.   As an example, this should be the case if to interact with the Label we’d erroneously try to use a beam of high energy neutrons, rather than strobo flashers’ or LED-matrices's low energy photons.   The neutrons should modify the molecular structure of the Label, modifying its eigenstates and then the eigenvalues we expected to derive by the measurement.


 To detect a Label they are necessary Time and a kind of interaction which do not reduce the information in the marginal distribution of the Label.  The second requirement is always fulfilled in the Electronic Inspectors but a real problem exists in relation to the first, depending on design choices of the Labeller and of the Bottling Line layout (  CARLSBERG Group/2014)










  SABMiller PLC, Kompania Piwowarska, at Poznań, Poland (  Tom Parker/OneRedEye/2012)









Links to other pages:



Contact


















                                                                                                            Copyright Graphene Limited 2013-2019