Method and apparatus for determining authenticity of an information bearing device
AssigneeInfotoo International Limited
InventorsTak Wai Lau, Wing Hong Lam
Life of this patent
Estimating the term…
Abstract
An authentication apparatus and a method to devise an authentication tool is provided for facilitating determination of authenticity or genuineness of an article with reference to a captured image or a purported primary image of an information bearing device on the article. The authenticity or genuineness of an article is determined with reference to whether a captured image is a primary image of an authentic information bearing device using a trained neural network. The neural network is trained using training images comprising imperfect images having varying degrees of image imperfections which are introduced under controlled conditions.
Claims20
An authentication apparatus configured for determining whether a captured image is a primary image or a non-primary image of an authentic information bearing device, the authentication apparatus comprising a microprocessor, a display, a data storage device, and an authentication tool which is resident in the data storage device;
- wherein the authentic information bearing device comprises a data-embedded image pattern which is encoded with a set of discrete data, wherein each discrete data of the set of discrete data has a characteristic signal strength and correlates with a set of spatially distributed pattern defining elements of the data-embedded image pattern by a domain transformation function, the spatially distributed pattern defining elements being spread in the data-embedded image pattern to form a spread image pattern;
- wherein the authentication tool comprises a trained neural network having a neural network structure comprising an input layer, an output layer and a plurality of neural network layers intermediate the input layer and the output layer, and each neural network layer has a set of filters comprising a plurality of learned filters;
- wherein each learned filter comprises a plurality of filter elements and each filter element has a filter value, and wherein the learned filters and filter values were composed or constituted by computer based-deep-learning using a plurality of training images;
- wherein the training images comprises a plurality of primary images of the data-embedded image pattern and a plurality of non-primary images of the data-embedded image pattern of the data-embedded image pattern, and wherein a primary image and a corresponding non-primary image have different characteristic signal strengths;
- and
- wherein the trained neural network was trained to determine whether a captured image is a primary image or a non-primary image with reference to the characteristic signal strengths of the training images.
A method of devising an authentication tool which is to reside on a mobile apparatus as a set of computer executable instructions in the form of a trained neural network, wherein the mobile apparatus is to operate as an authentication apparatus to determine whether a captured image is a primary image or a non-primary image of a data-coded image pattern of an authentic information bearing device upon execution of the instructions of the authentication tool, and wherein the trained neural network comprises a number of channels and each channel has a plurality of feature data;
- wherein the method comprises:
- feeding a computer-based neural network with a plurality of training files, wherein each training file is a data file of a training image comprising a plurality of image-defining elements, wherein the training images comprises a plurality of primary images of the data-coded image pattern and a plurality of non-primary images of the data-coded image pattern, and wherein the data-coded image pattern comprises a set of spatially distributed pattern defining elements which correlates to a coded data by a domain transformation function and has a characteristic signal strength relating to magnitude of the coded data;
- and
- training the neural network to differentiate between a primary image and a non-primary image with reference to the characteristic signal strengths of the primary images and the non-primary images to obtain the feature data to form the trained neural network.
Dependencies are inferred from the claim text.