| 摘要 |
The present application concerns the visual identification of materials or documents for tracking purpose. The invention proposes a method to automatically identify an object by the following steps :
- digitally acquiring a two-dimensional image through sampling on a uniformly spaced orthogonal grid of at least one color component of an area of interest of its surface illuminated by a non-coherent light which location, resolution, imaging angle and lighting angle are optimized for a given image matching metric,
- comparing the acquired image to stored digital images of the same area, such comparison including the steps of :
o preprocessing the image,
o flattening the image,
o applying a rotation compensation procedure,
o applying a translation compensation procedure,
o masking the areas of interest
o computing a metric according to a deterministic convergence behavior when the mean square error between the acquired and recorded image tends towards a constant, such metric being dependent on the acquired and stored image resolutions and enabling the identification from the recorded images of a matching set of images for a given object image, the metric and threshold having the property that the cardinality of said matching set converges towards singleton when the resolution increases. |
