Means for using microstructure of materials surface as a unique identifier

• US 8,180,174 B2
• Filed: 03/04/2008
• Issued: 05/15/2012
• Est. Priority Date: 09/05/2005
• Status: Active Grant

• Current Assignee: AlpVision S. A.
• Original Assignee: AlpVision S. A.
• Inventors: Di Venuto, Celine, Kutter, Martin, Jordan, Frederic
• Primary Examiner(s): Liew, Alex
• Application Number:
  US12/041,790
• Publication Number:
  US 20080219503A1
• Time in Patent Office: 1,533 Days
• Field of Search: 382181-231
• US Class Current: 382/276
• CPC Class Codes: G06V 10/242 by image rotation, e.g. by 90 degrees
• CPC Class Codes: G06V 20/80 Recognising image objects characterised by unique random patterns
• CPC Class Codes: G06V 30/2504 Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches
• CPC Class Codes: G07D 7/121 Apparatus characterised by sensor details
• CPC Class Codes: G07D 7/20 Testing patterns thereon G07D7/004, G07D7/005 take precedence

Abstract

A method and apparatus for the visual identification of materials for tracking an object comprises parameter setting, acquisition and identification phases. The parameter setting phase comprises the steps of defining acquisition parameters for the objects. The acquisition phase comprises the steps of digitally acquiring two-dimensional template image of an object, applying a flattening function and generating downsampled template version of the flattened template and storing it in a reference database with the flattened template. The identification phase comprises the steps of digitally acquiring a snapshot image, applying the flattening function and generating one downsampled version, cross-correlating the downsampled version of the flattened snapshot with the corresponding downsampled templates of the reference database, and selecting templates according to the value of the signal to noise ratio, for the selected templates, cross-correlating the flattened snapshot image with the reference flattened template, and identifying the object by finding the best corresponding template.

Claims

1. Method to identify an object comprising a parameter settings phase, an acquisition phase and an identification phase, the parameter setting phase comprising the steps of:

defining for a given set of objects, a resolution, a type of non-coherent light illumination and a location, called region of interest, for the acquired image for which the object'"'"'s microstructure image contains noise,

the acquisition phase comprising the following steps, for each object to be later identified;

digitally acquiring a two-dimensional image of the object according to parameter settings through sampling on a uniformly spaced orthogonal grid of at least one color component,

applying a flattening function on said template in order to produce a flattened template by removing macroscopic color variations,

generating at least one downsampled template version of the flattened template,

storing in a reference database the downsampled template version and the flattened template, the identification phase comprising the following steps, for the object to be identified;

digitally acquiring a two-dimensional snapshot image according to the same parameters as the template image,

applying to the snapshot image the same flattening function as the one applied to the template image in order to produce a flattened snapshot image,

generating at least one downsampled version of the flattened snapshot image,

cross-correlating the downsampled version of the flattened snapshot image with the corresponding downsampled templates version stored in the reference database, and selecting the templates according to the value of the signal to noise ratio of said cross-correlation,

for the selected templates, cross-correlating the flattened snapshot image with the flattened template stored in the reference database, and thus identifying the object by finding the best corresponding template which signal to noise ratio value of said cross-correlation is above a predefined threshold.

2. Method of claim 1, wherein the flattening function is calculated using the difference between the image and the low-pass version of said image.

3. Method of claim 1, wherein the acquisition or the identification phase comprises a prior step of applying rotation compensation by acquiring another image of the object and applying a Fourier transform to the said image in order to measure anisotropic information and use it to compute the rotation compensation angle.

4. Method of claim 1, wherein the signal to noise ratio is calculated using the cross-correlation of the template and the snap-shot image, the signal to noise ratio being the ratio between the peak value of the cross-correlation and the mean value of said cross-correlation.

5. Method of claim 1, wherein it comprises the step of converting the two-dimensional template in one-dimensional template by summing the intensity along each line, and converting the snapshot image into one-dimensional in one-dimensional template by summing the intensity along each line image by summing the intensity along each line.

6. Method of claim 1, wherein the parameter settings phase and the identification phase are modified as follows:

the parameter settings phase comprise an additional step consisting in defining for a given set of objects at least one visual feature whose distance to the X/Y position of the region of interest is known,

creating a mask image by acquiring a two-dimensional image of feature points,

the identification phase comprising preliminary steps of;

digitally acquiring a two-dimensional image of the feature points of the object,

cross-correlating the acquired image with mask image in order to retrieve the location of the region of interest where the snapshot image will be acquired.

7. Method of claim 1, wherein the template identified can serve to authenticate a plurality of objects having the same microstructure.

8. Method of claim 1, wherein no template identified allows detecting counterfeited object.

9. Method of claim 1, wherein the acquisition and identification phases are modified, the acquisition phase generating at least two downsampled versions of the template having different resolutions,

the identification phase including the following steps:

generating the same number of downsampled versions of the snapshot having different resolutions

iteratively selecting templates by a cross-correlation between a downsampled version of the snapshot image and the template of the same resolution according to the value of the signal to noise ratio of said cross-correlation, starting from the lowest resolution and ending at higher resolutions.

10. Method of claim 1, wherein the parameter settings, acquisition and identification phases are modified as follow in order to successfully match corresponding template and snapshot images having different angles of rotation:

the parameter settings phase having an additional step of defining the shape and location of a circular donut inside the region of interest,

the acquisition phase comprising the steps of;

extracting from the acquired image the circular donut region,

unwarping the donut in a rectangular shape in order to define the template image,

the identification phase comprising the steps of;

extracting from the acquired image the circular donut region,

unwarping the donut in a rectangular shape in order to define the snapshot image.

11. A device to identify an object comprising:

image acquisition means;

storage means, said storage means comprising a plurality of template entries, at a predefined resolution, type of non-coherent light illumination and location, for an acquired image for which the object'"'"'s microstructure image contains noise, each template comprising at least two versions of the template, a full resolution flattened template and a downsampled version of the flattened template, said device comprising

means to produce a flattened snapshot of the acquired image,

means to produce at least one downsampled version of the flattened snapshot,

means to compare the downsampled version of the flattened snapshot with the downsampled version of the flattened template using cross-correlation means to produce a first signal to noise ratio; and

selection means to select the downsampled flattened template according to the value of the first signal to noise ratio;

wherein the means to compare further performs cross-correlations of the full resolution flattened snapshot with the full resolution flattened templates corresponding to the previously selected downsampled version templates to produce a second signal to noise ratios and to use the second signal to noise ratios to identify one template as corresponding to the acquired image.