Authentication and identification of physical objects using machine vision protocols

A system for authenticating a physical object using computer vision (CV) and machine learning (ML) algorithms, the system comprises:

• a device camera of a frontend module configured to capture a plurality of images of the physical object in a plurality of conditions, wherein the plurality of conditions include the physical object with a plurality of materials, the physical object with a plurality of shapes, a plurality of cameras used in different lighting condition to capture the plurality of images, and/or the physical object with a plurality of alignments;
• a backend module including machine learning (ML) models configured to create a plurality of datasets of the plurality of images of the physical object captured in the plurality of conditions, wherein:
• an image capture toolkit of the frontend module configured to capture a plurality of analysis images of a focus area of the physical object using a ML algorithm from the ML models, wherein the focus area is auto-selected based on one or more features of the physical object;
• an image processing unit of the backend module configured to:
• create a machine vision agnostic protocol using the plurality of datasets and the ML models, wherein the machine vision agnostic protocol identifies the physical object in the plurality of conditions based on unique similar patterns identified from the focus area of the physical object;
• and
• receive an image of a query object, the image is captured using the device camera, wherein a query focus area of the image is auto-focused based on a pre-stored focus area,
• wherein the device camera of the frontend module is configured to capture a plurality of query images of the query focus area of the query object;
• and
• a physical code toolkit of the backend module is configured to:
• select a query image from the plurality of query images based on an image quality;
• create a plurality of feature vectors based on a plurality of features of the selected query image;
• compress the plurality of feature vectors to create a query physical code to represent the query object, wherein the query physical code is created using data extracted from the machine vision agnostic protocol;
• compare the query physical code with a pre-stored physical code;
• and
• determine whether the query object is authentic based on a result of the comparison.

A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause a data processing apparatus to perform operations, for machine authentication of a physical object using a computer vision (CV) model and machine learning (ML) algorithms for comparing attributes of the physical object, the operations comprising:

• capturing a plurality of images of the physical object with a computer vision in a plurality of conditions, wherein the plurality of conditions include the physical object with a plurality of materials, the physical object with a plurality of shapes, a plurality of cameras used in different lighting condition to capture the plurality of images, and/or the physical object with a plurality of alignments;
• creating a plurality of datasets of the plurality of images of the physical object captured in the plurality of conditions;
• capturing a plurality of analysis images of a focus area of the physical object using a ML algorithm, wherein the focus area is auto-selected based on one or more features of the physical object;
• creating a machine vision agnostic protocol using the plurality of datasets for identifying the physical object in the plurality of conditions based on unique similar patterns, wherein the unique similar patterns are identified from the focus area of the physical object;
• receiving an image of a query object, the image is captured using a camera, wherein a query focus area of the image is auto-focused based on a pre-stored focus area;
• capturing a plurality of query images of the query focus area of the query object;
• selecting a query image from the plurality of query images based on an image quality;
• creating a plurality of feature vectors based on a plurality of features of the selected query image;
• compressing the plurality of feature vectors to create a query physical code to represent the query object, wherein the query physical code is created using data extracted from the machine vision agnostic protocol;
• comparing the query physical code with a pre-stored physical code;
• and
• determining whether the query object is authentic based on a result of the comparison.