The present invention relates to systems for selecting containers, comprising steps for detecting aberrations in the exterior of the container. Such systems are important since, in the market for containers such as for instance bottles or crates, such containers with a modern appearance impart a commercial added value to the products supplied therein. Damage or imperfections on such containers will however detract greatly from the high perceived value of the products.
Containers with a damaged exterior greatly impair the image of the product for the modern consumer. In order to prevent such impairment, it is important that the containers are properly inspected.
In order to enable such an inspection, the present invention provides a method for composing a robust reference image, or a reference image with permissible deviation values of the exterior of a series of practically identical containers, such as bottles, crates and the like, comprising steps for:
Such a robust reference image is preferably obtained by determining, preferably per pixel, a mean value of the image and a mean deviation (standard deviation) thereof; the robust reference image then consists of two parts. It is preferably built up by starting with an image, whereafter subsequent images are automatically added thereto until, for instance after 1000 bottles, the robust reference image is complete. This can hereby take place during normal use of the installation, while the images of bottles in different positions show coherence.
Such a robust reference image, or a reference image with permissible deviation values, is important since printed bottles for instance differ slightly from each other due to for instance the production method. Printed or adhered labels can for instance be shifted somewhat in height or be printed in slightly deformed manner. Such deviations are however no reason for rejection. An apparatus which would not take into account such deviations would however select bottles for rejection if such deviations are observed. According to the present invention a robust reference image, or a reference image with permissible deviation values, is therefore composed in which acceptable deviations are taken into account. Such deviations may comprise damage, size, features or colour differences.
An example of such a method is that a reference image can be formed on the basis of data obtained on the basis of the containers used in practice. In order to manufacture the robust reference image, image recordings are made of for instance at least a thousand containers. Such a quantity gives sufficient deviation information for approving the normally occurring, acceptable deviations during a selection process of containers.
The above described method for composing a robust reference image or reference image with permissible deviation values for the exterior of a series of practically identical containers preferably comprises steps for applying statistical analysis methods in determining the robust reference image. Using such steps it is possible to use standard deviations, or for instance to apply factors to such standard deviations in order to obtain an optimal ratio between acceptance and rejection of containers in a selective process. Such an optimal ratio can for instance be determined on the basis of the cost of the containers in relation to a market value assessment based on the value of high-quality containers as perceived by consumers.
A further aspect of the present invention relates to a method for selecting a container, such as a crate or bottle and the like, comprising steps for detecting deviations on the exterior of the container, comprising steps for:
Processing of the recorded image preferably comprises steps for transforming an aspect view of a curved surface to a faithful flat rendition. Using a method according to such a preferred embodiment, the exterior of a container, such as for instance a label arranged thereon or a printing applied thereto, can be compared to the robust reference image or the reference image with the permissible deviation values of the exterior of a series of practically identical containers.
Image recordings are herein made of a part of the side wall of for instance round containers. In practice a readily processable recording is made of for instance 60° of the periphery. A wider recording of for instance 120° is also possible. The theoretically feasible recording of 180° however produces in practice less useful results close to the edges. In the case of a container with straight walls such as a square or rectangular container, recordings are made of a whole side wall.
In the case of a round container such as a bottle, the robust reference image comprises a faithful flat rendition of the whole side wall over 360°. The recording of for instance 60° which is made in the present embodiment, is fitted onto the corresponding part of the robust reference image after being processed to a faithful flat rendition. A comparison is then made as to whether the similarities fall within the set limit values of the robust reference image, which results in approval of this part-image and thereby the bottle, or that the present image falls outside the tolerances of the robust reference image, which results in rejection of the present bottle.
In a further embodiment use is made of the second representation to compose the robust reference image. This embodiment has the advantage that in the reference image, or in the deviation values included therein, account is taken of data of containers of which recordings are made during the method for selection. During the selection process, the robust reference image is hereby more and more adapted to a series of deviation values to be detected. In statistical terms this means that the sample size on the basis of which deviation values are determined becomes increasingly larger.
In order to obtain a sufficiently high speed (for instance 16 bottles/sec), use is preferably made of a coarse and a fine fitting.
Approval or rejection preferably takes place using distinguishable, parameterized differential measurements, such as colour difference, shape difference (non-roundness and the like) and so on.
The parameters preferably comprise data relating to the container, threshold values relating to colour differences or formats of printing and the like. It is important for the appearance of the container that for instance a printing is placed somewhat at the correct height; variations in the height of the printing will however, within limits, detract little from the appreciation of a container by an end user. Variations therein are therefore allowed. Too rigid an assessment of the height of the printing would therefore result in undesirable rejection. Such undesirable rejection is prevented by taking this into account using this method. The same sort of reasoning applies in relation to for instance colour differences or differences in printing format. Small aberrations herein can occur in large series of containers. By taking account of data relating to large quantities of containers of a series, undesirable rejection, which involves high costs, is prevented.
In order to be able to work in simple manner with containers from different series, a preferred embodiment of the method comprises the possibility of composing profiles comprising the above described data for the purpose of carrying out selections on mutually alternating series of containers matching respective profiles. If such a method is performed using an apparatus for selecting the containers, it is recommended to process different series of containers in efficient manner using one apparatus. These profiles enable simple switching of series, wherein the correct selections are made by applying the profile matching the series.
In a further preferred embodiment the container is illuminated while the recording is made. This has the advantage for instance that a usable recording can be made by applying sufficient light. By applying light of the same intensity and colour in different recordings, it is further possible to make recordings wherein data relating to the colour of the exterior of the container, such as a printing, is kept constant, whereby the forming of a robust reference image or the assessment on the basis of colour aspects of the exterior of the container can be performed with constant quality.
The container is preferably a bottle or a crate. Such containers as bottles or crates are very important in a distribution channel for the value perceived by consumers.
In the preferred embodiment the processing speed of the containers is higher than ten containers per second, preferably higher than sixteen containers per second. A high processing speed is important for the manufacturers of products sold in the containers. In practice it is very practical to carry out the method subsequent to filling of containers. Such containers, such as bottles, are filled in a filling line. The method for selecting and building up the robust reference image must then be performed at the same speed as the filling of the bottles. In practice, this speed is becoming increasingly higher.
Particularly in the case of bottles, use is preferably made of skimming light on the top and bottom, wherein the background remains as dark as possible.
A further aspect of the present invention relates to an apparatus for selecting containers, comprising:
Such an apparatus preferably further comprises composing means for composing, on the basis of first and second representations and/or predetermined parameters, a robust reference image or a reference image with permissible deviation values, on the basis of which image acceptable and/or natural differences between individual containers within a series of containers can be taken into account during selection of containers. Such apparatus embodiments are able to perform an above described method and have at least the above described advantages.
Such an apparatus preferably further comprises ejecting means for ejecting selected containers. Selected containers are in this respect containers which do not conform to the robust reference image, or the reference image with permissible deviation values.
The use of illuminating means and screening means while making the recording increases the quality of the recording.
In the further preferred embodiment the apparatus comprises actuating means for actuating making of a recording when a container is correctly positioned relative to the recording means. An advantage hereof is that a recording is made which is centred as accurately as possible relative to the container. A further advantage is that it is possible to keep track of which container is being assessed.
The apparatus preferably comprises rotating means for rotating the container between the making of two successive recordings. Such an embodiment enables making of a plurality of recordings of one container, whereby on the basis of complete information relating to the exterior of the container it is possible to assess whether it falls within the correct tolerances.
A significant advantage of the present invention relates to the possibility, using differential images, of detecting contaminants such as cigarette ends and the like in bottles with printed, i.e. partly transparent, labels, since the image information of the labels can be filtered out.
Further advantages, features and details of the present invention will be elucidated on the basis of the following description of preferred embodiments thereof with reference to the annexed figures, in which:
A first embodiment (
The bottles are transported past the lighting and the camera by means of a conveyor 7 (shown schematically). Bottles B are situated in a random orientation relative to the camera. In this embodiment a recording is made by the camera which covers ⅙ of the periphery of the bottle.
In the image processing computer 6 this recording is processed such that an image is obtained as if the periphery of the bottle had been rolled out. For this purpose the image information is stretched to some extent towards both sides of the image recording relative to the image information in the centre of the recording. The thus obtained flat representation is compared to a stored reference image or robust reference image. The method for obtaining this robust reference image is discussed in greater detail hereinbelow. A robust reference image comprises data relating to a plurality of images of bottles. Another manner of describing the robust reference image is that it comprises data relating to one graphic representation of a periphery of a bottle, together with data concerning maximum desired deviations from this ideal reference image. These permitted deviations are related to the desired degree of rejection and the desired quality of the images on the periphery of the container.
In this embodiment the comparing takes place on the basis of creating a differential image, by subtracting the reference image from the flat representation of the recorded image of the bottle for testing. The result of this subtraction is a so-called first differential image D. Using data relating to the mean differential image correction E, a second differential image F is then calculated on the basis of the first differential image D minus the mean differential image correction E.
A top view of the embodiment of
A further embodiment according to the present invention is shown in
In a further embodiment (
In this embodiment the resolution of the camera is 640 by 480 pixels. Depending on the desired selection quality, the image quality can be increased to for instance 600 by 800, 1024 by 768 or for instance 1280 by 960 pixels. The resolution of 640 by 480 pixels enables rejection of the bottle on the basis of defects of 2 by 2 mm in the exterior of the bottle. If higher camera resolutions are used, it becomes possible to reject smaller deviations. The desirability hereof depends for instance on the cost of rejecting bottles.
Making a recording of the bottle is preferably further actuated by means of a sensor which measures the arrival of the bottle at the position of the neck. The embodiment is further at least suitable for the so-called longneck (XLN) bottle. However, through the above stated use of profiles, an unlimited number of different bottles can be detected using this embodiment.
For operation by means of a computer of the above described system, a screen display (
In a further embodiment (
An embodiment of the inspection method is further described hereinbelow on the basis of the inspection of a crate. In step 51 (
In step 60 an image mask 61 is further produced. In step 68 the shrinkage is extracted from transformation vector 66, which results in shrinkage data 80. This shrinkage data is exported in 85 and 86. On the basis of the shrinkage data and image mask 61, the image mask is compensated for shrinkage 80 in step 69, which results in a compensated label mask 72.
Image data 59 of crate and handgrip further resulting from step 56 are subjected to a positioning measurement in step 65, which results in data relating to a horizontal and/or vertical shift and/or rotation relative to camera 70. This data 70 is exported in 71. This data 70 is further used in the compensating in 69. The compensated image mask 72 is compared to a blanking mask 76. This blanking mask 76 is the reference mask of the label. The resulting arrangement position data of a label on a crate 74 is used in 75 together with transformation vector 66 and the horizontal shift relative to camera 70 to calculate the blanking position 92. This data 92 is exported in 93. In step 89 a reference label is cut out on the basis of the mean reference image 62 and blanking mask (is reference mask) 76, in order to produce reference cut-out label 93.
The operation of the inspection of the crate colour and crate handgrip is explained with reference to the method of
In step 100 image data is compensated for shrinkage and horizontal shifting, which results in a stretched crate and handgrip image 101. In step 102 profile data is selected for the crate and handgrip. In step 105 profile data 104 is used together with data relating to the reference profile of crate 108 in order to measure a colour difference. Colour difference data 106 is exported in 107. Profile data relating to the handgrip resulting from step 102 is segmented to the handgrip in step 109. Handgrip image data 110 resulting from step 109 is processed in step 111 by means of a differential measurement on the basis of reference data 112 of the handgrip. Differential image data 113 relating to the handgrip is exported in 114.
The validation of crate and label is described with reference to
The present invention is not limited to the above described preferred embodiment thereof, the rights sought being rather defined by the following claims, within the scope of which many modifications can be envisaged.
Number | Date | Country | Kind |
---|---|---|---|
1019381 | Nov 2001 | NL | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/NL02/00750 | 11/18/2002 | WO |