Methods for identifying the authenticity of objects

Abstract

The present invention relates to a method for identifying the authenticity of an object, the method comprising: maintaining a reference image of an original object in an identification server system (50), the reference image being provided to represent all equivalent original objects; receiving one or more input images of an object to be identified in the identification server system (50); and generating a target image by the identification server system (50) based on the one or more input images. The method further comprises: aligning the target image with the reference image by the identification server system (50); and analyzing the target image relative to the aligned reference image by the identification server system (50) for identifying the authenticity of the object.

Description

Technical Field

[0001] This invention relates to a method for identifying the authenticity of an object, and more specifically to a method for identifying the authenticity of an object, the method comprising the steps of: a) maintaining a reference image of an original object in an identification server system, the reference image being provided to represent an equivalent original object; b) receiving one or more input images of the object to be identified in the identification server system; c) generating a target image by the identification server system based on at least one of the one or more input images; d) aligning the target image with the reference image by the identification server system by distorting the target image to match the reference image; and e) analyzing the aligned target image relative to the reference image by the identification server system for identifying the authenticity of the object. Background Technology

[0002] Counterfeit products are a global problem affecting individuals, companies, and society. Counterfeit products cause economic losses to companies that manufacture the original product. However, there are even more significant problems associated with counterfeit products. For example, counterfeit food or counterfeit medicine can cause serious health problems for humans, society, and animals. In the prior art, two different methods have been used to identify the authenticity of objects. According to a first alternative, each original object is photographed or otherwise analyzed or recorded during manufacturing. Then, later throughout the supply chain, such as in a store, the object being sold can be photographed or analyzed again. The photos or analysis results from the store are then matched with the photos or analysis results or records from manufacturing to find exact matches and evidence. Thus, in this first alternative, each object is identified individually. According to a second alternative, the object or its packaging is marked with an identifier or label indicating the authenticity of the object. The identifier or label can then be analyzed to identify the authenticity of the object.

[0003] One of the problems associated with the first alternative to the prior art is that each object must be matched one-to-one with the real object. This may be a feasible solution for products with a limited number manufactured. However, when the number of similar products manufactured is large—e.g., hundreds of thousands or millions—or when the supply chain is complex, this solution becomes very cumbersome and difficult. One of the problems associated with the second alternative to the prior art is that identifiers or labels can also be forged, making it difficult to identify the original product. Furthermore, a drawback of both the first and second alternatives is that they both require changes or additional process steps in the manufacturing and supply of the product. This is undesirable in efficient manufacturing processes. Summary of the Invention

[0004] One object of the present invention is to provide a method to address or at least mitigate the disadvantages of the prior art. This object is achieved by a method comprising the following steps: a) maintaining a reference image of an original object in an identification server system, the reference image being provided to represent an equivalent original object; b) receiving one or more input images of the object to be identified in the identification server system; c) generating a target image by the identification server system based on at least one of the one or more input images; d) aligning the target image with the reference image by the identification server system by distorting the target image to match the reference image; and e) analyzing the aligned target image relative to the reference image by the identification server system for identifying the authenticity of the object, wherein the alignment in step d) includes image alignment: - associating a reference image grid with the reference image by the identification server system, the reference image grid including reference grid points; - The recognition server system associates a target image grid onto the target image, the target image grid including target grid points; and the recognition server system aligns the target image with the reference image by moving the target grid points of the target image grid relative to each other and relative to corresponding reference grid points of the reference image grid to distort the target image.

[0005] Preferred embodiments of the present invention are disclosed in other methods for identifying the authenticity of an object.

[0006] This invention is based on the concept of providing a method for identifying the authenticity of an object. The method includes the following steps:

[0007] a) Maintaining a reference image of the original object in the identification server system, the reference image being provided to represent an equivalent original object;

[0008] b) Receive one or more input images of the object to be identified in the recognition server system;

[0009] c) The recognition server system generates a target image based on at least one of the one or more input images;

[0010] d) The recognition server system aligns at least one of the one or more input images with the reference image by distorting the target image; e) The recognition server system analyzes the aligned target image relative to the reference image to identify the authenticity of the object.

[0011] According to the present invention, the target image is an image that is aligned with the reference image such that the target image is modified to match the reference image. The image alignment of the target image is performed relative to the reference image by distorting the target image. This allows for accurate and reliable comparison or analysis of the target image relative to the reference image for identifying the authenticity of the object.

[0012] In the context of this application, alignment by distortion means compensating for distortions and / or defects in the target image by aligning the target image with a reference image, so that the target image corresponds to the reference image as well as possible. Therefore, during alignment, distortions caused by capturing the input image and defects in the object itself are compensated for.

[0013] Typically, distortion means generating a deviation from the projection of straight lines in an image. Therefore, image distortion means altering the spatial relationships between parts of an image.

[0014] When an image is captured, it may appear distorted relative to the object. Therefore, the resulting image differs from what the eye sees. This is called image distortion. In digital imaging, distortion is the deviation between the observed pixels and their predicted coordinates in a 2D plane. This results in an unnatural, more curved, or distorted appearance that is difficult for the eye to perceive.

[0015] Therefore, distortion is a term used to define a series of aberrations that occur under certain optical choices and shooting angles. There are generally two types of distortion: optical and perspective. Each contains various aberrations. Optical distortion occurs when it produces distortion in an image based on lens type. Optical distortion includes at least barrel distortion, pincushion distortion, mustache distortion, and lens-type distortion.

[0016] Barrel distortion occurs when straight lines in an image begin to curve outward from the center, creating a barrel-like effect. Pincushion distortion occurs when straight lines in an image begin to curve inward from the edge of the frame. Mustache distortion is a combination of pincushion and barrel distortion, where straight lines in the image curve outward from the center and then inward from the corner of the frame. Lens-type distortion caused by straight-line lenses—such as wide-angle prime lenses—tends to make lines in the image appear straight. On the other hand, curved lenses—such as fisheye lenses—make straight lines in the image appear curved.

[0017] Perspective distortion means that objects in an image may appear larger or smaller than they actually are, depending on their distance from the lens.

[0018] As mentioned above, in the context of this application, alignment means distorting an input or target image to compensate for image distortion caused by or generated during the capture of the input image. Therefore, aligning an input or target image includes changing the spatial relationships between portions, positions, regions, or pixels of the input or target image.

[0019] In one implementation, in step d), aligning the target image with the reference image by distortion by the recognition server system includes adjusting the target image to match the reference image. Therefore, adjusting the target image or distorting the target image relative to the reference image so that the target image matches the reference image.

[0020] For example, in one implementation, in step d), aligning the target image with the reference image by distorting the target image by the recognition server system includes: adjusting the size of the target image to match the size of the reference image. Therefore, adjusting the size of the target image relative to the reference image such that the size of the target image matches the size of the reference image. Size refers to the internal dimensions of the target image.

[0021] The target image then has a size that matches the reference image, thereby enabling accurate analysis and comparison of the target image and the reference image.

[0022] In one embodiment, the method further includes step f), performed prior to step e). Step f) includes defining one or more sub-portions of the aligned target image by the recognition server system, and step e) includes analyzing at least one of the one or more sub-portions of the aligned target image by the recognition server system relative to at least one corresponding sub-portion of the reference image for identifying the authenticity of the object. Therefore, in this embodiment, one or more sub-portions of the target image can be analyzed relative to one or more corresponding sub-portions of the reference image. Thus, it is not necessary to analyze the entire target image, and the analysis can be performed efficiently and quickly.

[0023] In another embodiment, the method further includes step f), performed prior to step e). Step f) includes defining one or more sub-portions in the target image by the recognition server system. Step d) includes aligning at least one of the one or more sub-portions of the target image with at least one corresponding sub-portion of the reference image by the recognition server system. Furthermore, step e) includes analyzing at least one of the one or more aligned sub-portions of the target image relative to the at least one corresponding sub-portion of the reference image by the recognition server system for identifying the authenticity of the object. Therefore, it is not necessary to align the entire target image, but only the sub-portion to be analyzed.

[0024] In one implementation, step f) includes the recognition server system dividing the aligned target image into two or more sub-parts.

[0025] In another embodiment, step f) includes maintaining the reference image of the original object in the recognition server system. The reference image is pre-divided into two or more sub-parts. Step f) also includes dividing the aligned target image into two or more sub-parts by the recognition server system based on the pre-divided sub-parts of the reference image. The target image is aligned in step d) prior to step f). Thus, the aligned target image is divided into two or more sub-parts based on the pre-divided reference image and therefore similarly to the reference image.

[0026] In yet another embodiment, step f) includes: the recognition server system dividing the reference image into two or more sub-parts, and the recognition server system dividing the aligned target image into two or more sub-parts based on the sub-parts of the reference image. In this embodiment, preferably, the reference image is first divided into two or more sub-parts based on the situation or as needed, and similarly or based on the sub-parts of the reference image, the aligned target image is divided into two or more sub-parts.

[0027] In one implementation, step e) includes: the recognition server system comparing the aligned target image with the reference image using statistical methods to identify the authenticity of the object. The statistical methods used in step e) are fast and efficient analysis methods requiring moderate computational power.

[0028] In another embodiment, step e) includes: the recognition server system comparing at least one of the one or more aligned sub-parts of the target image with at least one corresponding sub-part of the reference image using statistical methods to identify the authenticity of the object. Because the amount of data to be analyzed is reduced, using statistical methods to analyze sub-parts enables more efficient analysis.

[0029] In yet another implementation, step e) includes: maintaining a machine learning recognition algorithm or a recognition neural network in the recognition server system, and having the recognition server system compare the aligned target image with the reference image using the machine learning recognition algorithm or the recognition neural network. The machine learning recognition algorithm or the recognition neural network is trained to perform the analysis. Furthermore, the machine learning recognition algorithm or the recognition neural network can be continuously trained to enhance the accuracy of the analysis.

[0030] In yet another embodiment, step e) includes: maintaining a machine learning recognition algorithm or recognition neural network in the recognition server system; and having the recognition server system compare at least one of the one or more aligned sub-parts of the target image with at least one corresponding sub-part of the reference image by utilizing the machine learning recognition algorithm or the recognition neural network. The machine learning recognition algorithm or the recognition neural network can be trained to analyze one or more sub-parts of the target image. Therefore, the machine learning recognition algorithm or the recognition neural network can perform the analysis on the sub-parts efficiently and more accurately. Furthermore, in some embodiments, the machine learning recognition algorithm or recognition neural network can be targeted at a specific sub-part, or each sub-part can include a machine learning recognition algorithm or recognition neural network configured for it.

[0031] In one implementation, step e) includes providing a first machine learning recognition algorithm and a second machine learning recognition algorithm to the recognition server system. The first machine learning recognition algorithm is trained to determine differences between the aligned target image and the reference image. The second machine learning recognition algorithm is trained on the reference image to analyze the differences determined by the first machine learning recognition algorithm relative to the reference image. Step e) includes the recognition server system processing the aligned target image and the reference image using the first machine learning recognition algorithm, and the recognition server system generating a first difference vector by utilizing the first machine learning recognition algorithm. Step e) further includes the recognition server system processing the first difference vector using the second machine learning recognition algorithm for the reference image to identify the authenticity of the object. Not all differences in the target image and the reference image are severe, or they are not relevant to the object authenticity assessment. This implementation allows for the analysis of authenticity by utilizing two machine learning algorithms, taking into account different types of differences. Dividing the analysis into two machine learning algorithms enables efficient analysis.

[0032] In an alternative implementation, step e) includes providing a first recognition neural network and a second recognition neural network in the recognition server system. The first recognition neural network is trained to determine differences between the aligned target image and the reference image, and the second recognition neural network is trained on the reference image to analyze the differences determined by the first recognition neural network relative to the reference image. Step e) includes the recognition server system processing the aligned target image and the aligned reference image using the first recognition neural network, and the recognition server system generating a first difference vector by utilizing the first recognition neural network. Step e) further includes the recognition server system processing the first difference vector using the second recognition neural network for the reference image to identify the authenticity of the object. This implementation enables the analysis of authenticity by utilizing two neural networks, considering different types of differences. Dividing the analysis into two neural networks enables efficient analysis.

[0033] In one implementation, step e) includes providing a first machine learning recognition algorithm and one or more second machine learning recognition algorithms in the recognition server system. The first machine learning recognition algorithm is trained to determine the difference between an aligned sub-part of the target image and a corresponding sub-part of the reference image. The one or more second machine learning recognition algorithms are trained for the one or more sub-parts of the reference image and are trained to analyze the difference determined by the first machine learning recognition algorithm relative to each of the one or more sub-parts of the reference image. Step e) includes the recognition server system processing the one or more aligned sub-parts of the target image and the one or more corresponding sub-parts of the reference image respectively using the first machine learning recognition algorithm, and the recognition server system generating one or more difference vectors by utilizing the first machine learning recognition algorithm, each of the one or more difference vectors being for a sub-part of the reference image. Step e) further includes the recognition server system processing the one or more difference vectors respectively using the one or more second machine learning recognition algorithms for the one or more sub-parts of the reference image for the purpose of identifying the authenticity of the object. This implementation, by utilizing two machine learning algorithms and focusing on one or more sub-parts of the target image, enables the analysis of authenticity while considering different types of differences. Dividing the analysis into two machine learning algorithms and one or more sub-parts enables even more efficient analysis.

[0034] In another embodiment, step e) includes providing a first recognition neural network and one or more second recognition neural networks in the recognition server system. The first recognition neural network is trained to determine the difference between an aligned sub-part of the target image and a corresponding sub-part of the reference image. The one or more second recognition neural networks are trained for the one or more sub-parts of the reference image and are trained to analyze the difference determined by the first recognition neural network relative to each of the one or more sub-parts of the reference image. Step e) includes the recognition server system processing the one or more aligned sub-parts of the target image and the one or more corresponding sub-parts of the reference image respectively using the first recognition neural network, and the recognition server system generating one or more difference vectors by utilizing the first recognition neural network, each of the one or more difference vectors being for a sub-part of the reference image. Step e) further includes the recognition server system processing the one or more difference vectors respectively using the one or more second recognition neural networks for the one or more sub-parts of the reference image for identifying the authenticity of the object. This embodiment enables the analysis of authenticity while considering different types of differences by utilizing two neural networks and focusing on one or more sub-parts of the target image. Dividing the analysis into two neural networks and the one or more sub-parts enables more efficient analysis.

[0035] In some implementations, the machine learning algorithm may include a network-based machine learning algorithm, a model-based machine learning algorithm, or a non-parametric machine learning algorithm. The neural network may be any suitable artificial neural network.

[0036] In one embodiment of the invention, the alignment in step d) includes image alignment, which involves the recognition server system identifying corresponding positions in the target image and corresponding positions in the reference image, and the recognition server system aligning the corresponding positions in the target image and the corresponding positions in the reference image to each other by distorting the target image. Therefore, the method can include aligning one or more of the corresponding positions, and the alignment can be performed accurately and efficiently to provide high-quality realism analysis.

[0037] In another embodiment, step d) includes: identifying, by the recognition server system, a corresponding position in at least one of the one or more sub-parts of the target image and a corresponding position in at least one corresponding sub-part of the reference image; and aligning the corresponding positions in at least one of the one or more sub-parts of the target image and the corresponding positions in at least one corresponding sub-part of the reference image by distorting the at least one of the one or more sub-parts of the target image. Therefore, the method may include aligning one or more corresponding positions of one or more sub-parts, and the alignment can be performed accurately and efficiently to provide high-quality realism analysis.

[0038] In one implementation, the alignment in step d) includes image alignment and associating a reference image grid onto the reference image by the recognition server system; associating a target image grid onto the target image by the recognition server system; and aligning the target image with the reference image by the recognition server system adjusting the target image grid relative to the reference image to align the target image with the reference image. Aligning the target image with the reference image using both the target image grid and the reference image grid enables alignment of the target image with the reference image in a simplified manner that provides efficient processing.

[0039] In another embodiment, the alignment in step d) includes image alignment and associating a reference image grid onto a reference image by the recognition server system; associating a target image grid onto a target image by the recognition server system; and aligning the target image with the reference image by the recognition server system through distorting the target image grid relative to the reference image for aligning the target image with the reference image. Aligning the target image with the reference image using both the target image grid and the reference image grid enables alignment of the target image with the reference image in a simplified manner that provides efficient processing.

[0040] In one embodiment of the invention, the reference image grid includes reference grid points, and the target image grid includes target grid points. Step d) aligning the target image with the reference image by the recognition server system further includes: adjusting the target grid points of the target image grid relative to corresponding reference grid points of the reference image by the recognition server system to distort the target image. Alignment is performed by moving the target grid points of the target image grid individually relative to each other and relative to reference grid points of the reference image grid. Therefore, the target image and the target image grid are distorted and become an image aligned with the reference image.

[0041] In another embodiment, step d) aligning the target image with the reference image by the recognition server system further includes: adjusting the target grid points of the target image grid relative to each other and relative to corresponding reference grid points of the reference image by the recognition server system by moving the target grid points of the target image grid to distort the target image.

[0042] In one implementation, step d) alignment by the recognition server system includes: iteratively aligning the target image with the reference image by iteratively moving the target grid points of the target image grid to distort the target image. Each target grid point of the target image grid can be moved individually relative to other target grid points of the target image grid and relative to a reference grid point of the reference image grid. This is performed during the iteration process by moving the target grid points of the target image grid two or more times to provide accurate alignment.

[0043] In another embodiment, step d) aligning the target image with the reference image by the recognition server system further includes: aligning the target image with the reference image by the recognition server system moving the target grid points of the target image grid to distort the target image. Step d) further includes: verifying the alignment of the target image with the reference image by the recognition server system; and iteratively aligning the target image with the reference image by the recognition server system iteratively moving the target grid points of the target image grid to distort the target image. The iterative movement of the target grid points of the target image grid may include verifying or evaluating the alignment of the target image with the reference image and repeating the movement or iterative movement based on the verification. The verification may include performing step e) for analysis. Step e) may be performed between iterative alignment steps d).

[0044] In one implementation, step d) aligning the target image with the reference image by the recognition server system includes: scaling down the reference image along with the reference image grid by the recognition server system to form a scaled-down reference image with the scaled-down reference image grid; and aligning the target image grid with the scaled-down reference image grid by the recognition server system to provide an initial image alignment between the target image and the reference image. Step d) further includes scaling up the scaled-down reference image along with the scaled-down reference image grid by the recognition server system to form a first scaled-up reference image with a first scaled-up reference image grid; and aligning the target image with the first scaled-up reference image grid by the recognition server system to provide a first image alignment between the target image and the reference image. Iterative alignment is performed by scaling down the reference image and the reference image grid and scaling up the reference image and the reference image grid, and successively aligning the target image with the scaled-up reference image between successive alignments two or more times.

[0045] In another embodiment, step d) aligning the target image with the reference image by the recognition server system includes: scaling down the reference image along with the reference image grid by the recognition server system to form a scaled-down reference image with the scaled-down reference image grid; and aligning the target image with the scaled-down reference image by the recognition server system to provide an initial image alignment between the target image and the reference image. Step d) further includes: scaling up the scaled-down reference image along with the scaled-down reference image grid by the recognition server system to form a first scaled-up reference image with a first scaled-up reference image grid; aligning the target image with the first scaled-up reference image by the recognition server system to provide a first image alignment between the target image and the reference image; and repeating the scaling up of the reference image along with the reference image grid and aligning the target image with the scaled-up reference image one or more times by the recognition server system to provide the first image alignment between the target image and the reference image. Therefore, alignment is provided more accurately and gradually.

[0046] In one implementation, step d) aligning the target image with the reference image by the recognition server system further includes: scaling down the reference image along with the reference image grid by the recognition server system to form a scaled-down reference image with the scaled-down reference image grid; and aligning the target image with the scaled-down reference image by the recognition server system moving the target grid points of the target image grid. Step d) further includes: scaling up the scaled-down reference image along with the scaled-down reference image grid by the recognition server system to form a first scaled-up reference image with a first scaled-up reference image grid; and aligning the target image with the first scaled-up reference image by the recognition server system moving the grid points of the target image grid. Scale-down and then scale-up progressively and iteratively provides a more accurate alignment between the target image and the reference image. Therefore, high-quality alignment is effectively achieved by moving the target grid points of the target image grid.

[0047] In another embodiment, step d) aligning the target image with the reference image by the recognition server system further includes: scaling down the reference image along with the reference image grid by the recognition server system to form a scaled-down reference image having the scaled-down reference image grid; and aligning the target image with the scaled-down reference image by the recognition server system moving the target grid points of the target image grid. Step d) further includes: scaling up the scaled-down reference image along with the scaled-down reference image grid by the recognition server system to form a first scaled-up reference image having a first scaled-up reference image grid; aligning the target image with the first scaled-up reference image by the recognition server system moving the target grid points of the target image grid; and repeating the scaling up of the reference image along with the reference image grid and aligning the target image with the scaled-up reference image once or multiple times by the recognition server system to provide a first image alignment between the target image and the reference image. Therefore, more accurate alignment is gradually provided by scaling up the reference image and the reference image grid and moving the target grid points of the target image grid to distort the target image.

[0048] In some implementations, the analysis step (e) can be performed between scaling up and aligning the reference image. Furthermore, each alignment with a reference image that is not scaled to a different size can include alignment by moving or iteratively moving target grid points.

[0049] In one implementation, step b) includes receiving two or more input images of the object to be identified in the recognition server system, and step c) includes the recognition server system selecting one of the two or more input images as the target image. Therefore, one of the two or more input images is selected as the target image.

[0050] In another embodiment, step b) includes receiving two or more input images of the object to be identified in the recognition server system; step c) includes the recognition server system comparing the two or more input images with the reference image; and step c) further includes the recognition server system selecting one of the two or more input images as the target image based on the comparison. Therefore, the selection is performed by comparing the input image with the reference image and selecting the input image with the best match to the reference image as the target image. Thus, the most promising input image for identifying the authenticity of the object is selected to obtain a high-quality recognition result.

[0051] In one implementation, step b) includes receiving two or more input images of an object to be identified in the recognition server system; step c) includes the recognition server system aligning at least two of the two or more input images relative to the reference image, and the recognition server system generating the target image by combining at least two of the two or more aligned input images into the target image. Therefore, a target image is generated from the two or more input images by combining two or more target images. Thus, a high-quality target image can be generated by utilizing the most promising portion of each of the two or more input images.

[0052] In another embodiment, step b) includes receiving two or more input images of the object to be identified in the recognition server system; and step c) includes the recognition server system selecting one of the two or more input images as a primary input image, the recognition server system aligning at least one of the two or more input images relative to the primary input image, and the recognition server system generating the target image by combining at least one of the two or more aligned input images with the primary input image to form the target image. Therefore, a target image is generated based on two or more aligned input images. Thus, a high-quality target image can be generated by utilizing the most promising portion of each of the two or more input images.

[0053] The alignment of at least one input image with the primary input image is performed in a manner similar to that disclosed above or by any of the methods and alternatives mentioned above for aligning an input image or target image with a reference image.

[0054] In one implementation, step b) includes receiving two or more input images of an object to be identified in the recognition server system. Step c) includes: aligning at least two of the two or more input images with one or the main input image by the recognition server system; dividing at least two of the two or more input images into two or more sub-parts by the recognition server system; selecting at least two of the two or more sub-parts by the recognition server system; and generating the target image by the recognition server system by combining the selected sub-parts into the target image. Therefore, each input image can be divided into two or more sub-parts. One or the most promising sub-part of the corresponding sub-parts of the two or more input images is selected as the target image. Therefore, the target image consists of the most promising sub-part of the two or more input images.

[0055] In another embodiment, step b) includes receiving two or more input images of the object to be identified in the recognition server system. Step c) includes: the recognition server system aligning at least two of the two or more input images with one or the main input image; the recognition server system dividing the at least two of the two or more input images into two or more sub-parts; the recognition server system comparing the two or more sub-parts of the input images with corresponding sub-parts of the reference image; the recognition server system selecting at least two of the two or more sub-parts based on the comparison; and the recognition server system generating the target image by combining the selected sub-parts into the target image. Therefore, a target image is generated based on the two or more input images by combining them. Therefore, a high-quality target image can be generated by utilizing the most promising portion of each of the two or more input images, based on the sub-parts corresponding to the sub-parts of the input images and / or the sub-parts corresponding to the reference image. Therefore, each input image can be divided into two or more sub-parts. Therefore, the selected sub-parts are merged together to form the target image.

[0056] In one implementation, step c) or d) involves the recognition server system extracting an input object image from the one or more input images or from the target image, the target image being composed of the input object image conforming to the outline of the object. Therefore, the generated target image includes only the image of the object, and other portions of the input image are removed. Thus, excessive portions of the input or target image do not affect the accuracy of the recognition.

[0057] In another embodiment, the recognition server system extracts an input object image from the one or more input images or from the target image, and removes the background to form the target image, which is composed of the input object image that conforms to the outline of the object. Therefore, the generated target image includes only the image of the object, and the background of the input image or the target image is removed. Thus, the background of the input image or the target image does not affect the accuracy of the recognition.

[0058] In one embodiment of the invention, step c) includes the recognition server system aligning the size of the one or more input images with that of the reference image.

[0059] In another implementation, step c) or d) includes the recognition server system aligning the target image with the reference image size.

[0060] Size alignment refers to changing, increasing, or decreasing the pixel size of an input or target image so that its size matches that of a reference image.

[0061] In one implementation, step c) includes the recognition server system color-aligning the one or more input images with the reference image.

[0062] In another implementation, step c) or d) includes the recognition server system aligning the target image with the reference image in color.

[0063] Color alignment refers to changing the colors of an input or target image to match each other, or to match a reference image, or to make each other and the reference image match.

[0064] In one embodiment, the method further includes, prior to step e), excluding a predetermined region of the aligned target image from the analysis of the aligned target image relative to the reference image by the recognition server system. Therefore, unwanted or altered portions or regions of the target image can be excluded from the authenticity analysis. Thus, regions including information or features that have been altered, such as manufacturing date, are excluded from the analysis. This altered information does not affect the analysis.

[0065] In another embodiment, the method further includes, prior to step e), applying a mask by the recognition server system to a predetermined region of the aligned target image and a corresponding predetermined region of the reference image to exclude the predetermined region from the analysis of the aligned target image relative to the reference image. Thus, the mask excludes the predetermined region from the analysis.

[0066] One advantage of this invention is that the authenticity of the identification is based on analyzing the image of the object to be identified relative to a reference image of the original object. Furthermore, in this invention, the target image to be analyzed relative to the reference image is an image aligned with the reference image such that the analysis or comparison is as accurate as possible. Therefore, in this invention, a target image to be analyzed relative to the reference image is provided to correspond to the reference image as closely as possible. This means that the target image is provided to match the reference image as well as possible, thereby achieving reliable and high-quality authenticity identification. Attached Figure Description

[0067] The invention is described in detail with reference to the accompanying drawings and specific embodiments, in which...

[0068] Figure 1 An embodiment of an identification system for performing the method of the present invention is illustrated schematically;

[0069] Figure 2 The user equipment is shown schematically;

[0070] Figure 3A and Figure 3B The input image and the target image are shown schematically, respectively.

[0071] Figure 4A and Figure 4B The reference image and target image of the original object are shown schematically, respectively;

[0072] Figure 5 and Figure 6 Flowcharts illustrating different embodiments of the method according to the present invention are shown schematically;

[0073] Figure 7A , Figure 7B and Figure 7C It schematically shows how to make Figure 4A and Figure 4B An image aligned with a reference image and a target image;

[0074] Figures 8 to 12 Flowcharts illustrating different embodiments of the method according to the present invention are shown schematically;

[0075] Figure 13A and Figure 13B The reference image and the target image, each with a grid, are shown schematically.

[0076] Figure 14 This schematically illustrates how to use a grid to make Figure 13A and Figure 13B Image alignment is performed between the reference image and the target image; and

[0077] Figure 15 A flowchart illustrating one embodiment of the method according to the present invention is shown schematically;

[0078] Figure 16A , Figure 16B , Figure 16C and Figure 16D The scaled-down and scaled-up views of the reference image and the reference image grid are schematically shown.

[0079] Figure 17 This schematically illustrates how to align scaled-down reference and target images using a grid.

[0080] Figure 18 Flowcharts illustrating different embodiments of the method according to the present invention are shown schematically;

[0081] Figure 19 The movement of the target grid points is illustrated schematically;

[0082] Figure 20A and Figure 20B The illustration schematically demonstrates how moving target grid points distorts the target image grid; and

[0083] Figure 21 A flowchart illustrating one embodiment of the method according to the present invention is shown schematically. Detailed Implementation

[0084] This invention and its embodiments are not directed at any specific information technology system, communication system, or access network. However, it should be understood that this invention and its embodiments have applications in many system types, and for example, in the circuit-switched domain (e.g., in GSM (Global System for Mobile Communications) digital cellular communication systems), in the packet-switched domain (e.g., in UMTS (Universal Mobile Telecommunications System) systems, LTE (Long Term Evolution) or the 5G NR (New Radio) standard standardized by 3GPP (3G Partners Project), and for example in networks according to the IEEE 802.11 standard—WLAN (Wireless Local Area Network), HomeRF (Radio Frequency), or BRAN (Broadband Radio Access Network) specifications (HIPERLAN1 and 2, HIPERACCESS). This invention and its embodiments can also be applied in ad hoc communication systems—such as IrDA (Infrared Data Association) networks or Bluetooth networks. In other words, the basic principles of the present invention can be used in combination with, between, and / or within any of the following systems: any mobile communication systems of the second, second, fifth, third, fourth, and fifth generations (and thereafter), such as GSM, GPRS (General Packet Radio Service), TETRA (Terrestrial Trunking Radio), UMTS systems, HSPA (High-Speed ​​Packet Access) systems—for example in WCDMA (Wideband Code Division Multiple Access) technology, and PLMN (Public Land Mobile Network) systems.

[0085] Communication technologies using the IP (Internet Protocol) protocol can include, for example, GAN (General Access Network), UMA (Unlicensed Mobile Access), VoIP (Voice over Internet Protocol), peer-to-peer networking, ad hoc networking, and other IP protocol technologies. Different versions of the IP protocol or combinations thereof can be used.

[0086] exist Figure 1 The illustration shows the architecture of a communication system to which embodiments of the present invention can be applied. Figure 1 A simplified system architecture is illustrated, showing only some components and functional entities—all logical units, the implementation of which may differ from those shown. The connections shown in the figures are logical connections; actual physical connections may differ. It will be apparent to those skilled in the art that the system also includes other functions and structures.

[0087] As mentioned above, the present invention is not limited to any known or future system, device or service, but can be utilized in any system by the methods described below according to the present invention.

[0088] Figure 1An identification system is illustrated, in which a user can connect to an identification server system 50 via a communication network 100 using a user device 10. It should be noted that... Figure 1 A simplified version of the identification system is presented, and in other embodiments, an unlimited number of users may be able to connect to the identification server system 50 via the communication network 100.

[0089] The communication network 100 may include one or more wireless networks, which may be based on any mobile system (such as GSM, GPRS, LTE, 4G, 5G and beyond) and wireless local area networks (such as Wi-Fi). Additionally, the communication network 100 may include one or more fixed networks or the Internet.

[0090] The identification server system 50 may include at least one identification server connected to the identification database 58. The identification server system 50 may also include one or more other network devices (not shown), such as terminal devices, servers, and / or database devices. The identification server system 50 is configured to communicate with one or more user devices 10 via a communication network 100. The identification server system 50, or the server and identification database 58, may form a single database server; in other words, a combination of data storage (database) and data management system, as in... Figure 1 The data storage can be a single entity, or it can be a conventional or future data repository managed by any suitable data management system—including distributed and / or centralized data storage, cloud-based storage in a cloud environment (i.e., a computing cloud). Specific implementations of the data storage are not relevant to this invention and are therefore not described in detail. In addition to or in place of identification database 58, other parts of identification server system 50 may also be implemented as a distributed server system comprising two or more individual servers or as a computing cloud comprising one or more cloud servers. In some embodiments, identification server system 50 may be a fully cloud-based server system. Furthermore, it should be understood that the location of identification server system 50 is irrelevant to this invention. Identification server system 50 may be operated and maintained using one or more other network devices within the system or via communication network 100 using terminal devices (not shown). Identification server system 50 may also include one or more user devices.

[0091] In some implementations, the identification server system 50 is integrated into the user equipment 10 and configured as an internal server system within the user equipment 10. Furthermore, in these implementations, the communication network 100 is implemented as an internal communication network or component within the user equipment 10, such as a wireless or wired communication network or connection within the user equipment.

[0092] The identification server system 50 may further include a processing module 52. The processing module 52 is coupled to or otherwise has access to the memory module 54. The processing module 52 and the memory module 54 may form an identification server or at least a portion thereof. The identification server, or the processing module 52 and / or the memory module 54, has access to the identification database 58. The processing module 52 may be configured to execute instructions of an identification application or unit by utilizing instructions of the identification application. The identification server system 50 may include an identification unit 56, which may be an identification application. The identification application 56 may be stored in the memory module 54 of the identification server system 50. The identification application or unit 56 may include instructions for operating the identification application. Therefore, the processing module 52 may be configured to execute the instructions of the identification application.

[0093] Processing module 52 may include one or more processing units or a central processing unit (CPU) or similar computing unit. The invention is not limited to any kind or number of processing units. Memory module 54 may include a non-transitory computer-readable storage medium or a computer-readable storage device. In some embodiments, memory module 54 may include temporary memory, meaning that the primary purpose of memory module 54 may not be long-term storage. Memory module 54 may also refer to volatile memory, meaning that memory module 54 does not maintain its stored contents when it is not receiving power. Embodiments of volatile memory include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory known in the art. In some embodiments, memory module 54 is used to store program instructions for execution by processing module 52, such as identifying an application. In one embodiment, memory module 54 may be used by software (e.g., an operating system) or an application (such as software, firmware, or middleware). Memory module 54 may include, for example, an operating system or software application—identifying an application that includes at least a portion of instructions for performing the methods of the invention. Therefore, the identification unit 56 of the identification server system 50 includes an identification application, and it can be a separate application unit, such as... Figure 1 What is shown, or alternatively, can be a separate application unit.

[0094] The processing module 52, the memory module 54, and the identification unit 56 together form the identification module 55 in the identification server system 50.

[0095] It should be noted that the identification database 58 can also be configured to include a software application—an identification application—which includes at least a portion of instructions for performing the methods of the present invention.

[0096] The identification database 58 can maintain information about one or more original objects and one or more reference images of the original objects. The identification database 58 can also maintain information about one or more user accounts of multiple users and / or information uploaded to the server system 50 via said user accounts or user devices 10. The identification database 58 may include one or more storage devices. The storage devices may also include one or more transient or non-transient computer-readable storage media and / or computer-readable storage devices. In some embodiments, the storage devices may be configured to store a larger amount of information compared to the memory module 54. The storage devices may also be configured for long-term storage of information. In some embodiments, the storage devices include non-volatile storage elements. Examples of such non-volatile storage elements include magnetic hard disks, optical disks, solid-state drives, flash memory, various forms of electrically programmable memory (EPROM) or electrically erasable programmable memory (EEPROM), and other forms of non-volatile memory known in the art. In one embodiment, the storage device may include a database, and the memory module 54 includes instructions and an operation identification application for executing the method according to the invention using the processing unit 52. However, it should be noted that the storage device can also be omitted, and the identification server system 50 may only include a memory module 54, which is also configured to maintain the identification database 58. Alternatively, the memory module 54 can be omitted, and the identification server system 50 may only include one or more storage devices. Therefore, the terms memory module 54 and identification database 58 can be used interchangeably in embodiments where neither is present. By utilizing instructions stored in the memory module 54 and executed by the processing unit 52, the identification database 58 can operate together with other components and data of the identification server system 50 via the communication network 100.

[0097] The identification database 58 may be provided together with the identification server, or the identification server may include the identification database 58, such as... Figure 1 As shown in the figure. Alternatively, identification database 58 may be provided as an external database 58 outside of identification server 50, and identification database 58 may be accessible to and connected to identification server directly or via communication network 100.

[0098] The storage device may store one or more identification databases 58 for maintaining identification information, object information, or reference image information. These different information items may be stored in different database blocks within the identification database 58, or alternatively, they may be grouped differently, for example, based on each individual object or reference image.

[0099] Users can, for example Figure 1The user equipment 10 shown is used to utilize the methods and systems of the present invention. User equipment 10 can be configured to connect to or access identification server system 50 via communication network 100. User equipment 10 can be a personal computer, desktop computer, laptop computer, or user terminal, as well as a mobile communication device, such as a mobile phone or tablet computer suitable for performing web browsing or able to access and interact with identification server system 50. However, user equipment 10 can also be a personal digital assistant, thin client, e-notebook, or any other device with a display and interface suitable for performing web browsing or able to access and interact with identification server system 50.

[0100] Furthermore, user equipment 10 can refer to any portable or non-portable computing device. Possible computing devices include wireless mobile communication devices that operate with or without a subscriber identification module (SIM) in hardware or software form.

[0101] like Figure 2 As shown, user equipment 10 includes a user interface 12. User interface 12 can be any suitable user interface for a human user to utilize and interact with the identification server system 50 and identification module 55 to perform or interact with the identification server system 50 and methods of the present invention. User interface 12 can be a graphical user interface (GUI) provided by a web browser or dedicated application on the display 11 of user equipment 10 (e.g., a monitor screen, LCD display, etc.) along with information provided by the identification server system 50 or other systems or servers. User interface 12 can, for example, enable a user to input messages or data, upload and / or download data files, input images and information, and submit requests for object identification processes.

[0102] Access to user interface 12 is determined by the user detecting the accessibility of an input device through an input device (not shown) (such as a touchscreen, keyboard, mouse, touchpad, keypad, trackball, or any other suitable manually operated input device) or some other type of input device (such as a voice-operated user device) or human gestures (such as hand or eye gestures). The input device can be configured to receive input from the user. A user interface (typically an API, application programmable interface) may be provided in conjunction with the systems and methods of the present invention to enable the user to interact with the recognition server system 50.

[0103] User interface 12 may also be an interface facing and accessible to the imaging device or camera 14 of user device 10. Therefore, the identification server system 50 may be accessible to the camera 14 of user device 10 and is arranged to receive one or more images obtained using the camera 14 of user device 10.

[0104] In some implementations, user equipment 10 may include user applications 17, such as software applications, stored in user equipment memory 16 and executed by user equipment processor 15.

[0105] Furthermore, the system and method of the present invention can be configured to interact with external or third-party services via a communication network 100 and a suitable communication protocol. In the present invention, external services may be, for example, external map services, official databases, etc.

[0106] In this invention, the authenticity of an object is identified by analyzing one or more images of the object to be identified—meaning the input image—relative to reference images representing the original object or all equivalent original objects.

[0107] In the context of this application, one or more input images represent objects whose authenticity will be identified. Therefore, the input images are photographs taken from objects whose authenticity will be identified using the system and method of this invention. The input images are image data files that provide digital images of the objects.

[0108] In one implementation, one or more input images are generated or captured using camera 14 of user equipment 10. Alternatively, input images can be provided using user application 17, and recognition server system 50 can receive input images directly from user equipment 10 via a communication network. Alternatively, input images provided using camera 14 of user equipment 10 can be stored on user equipment 10, or stored in device memory 16 of user equipment 10, and sent to recognition server system 50 via communication network 100 using a separate application (such as a messaging application or email application).

[0109] In yet another alternative implementation, the input image is generated or captured using a separate imaging device or camera. The input image is further transmitted to, received in, or stored in the user equipment 10. The input image is then sent to the recognition server system 50 via the communication network 100 using the user application 17 or using a separate application (such as a messaging application or email application).

[0110] The imaging device or camera 14 of user equipment 10, or a separate imaging device or camera, can be any known imaging device or camera, such as a digital camera, digital video camera, analog camera, or analog video camera. In the case of an analog camera or analog video camera, the image is digitized into a digital input image.

[0111] In the context of this application, a reference image can be any kind of image, video, or image file, or a data file representing the appearance of an original object. The reference image in this invention represents an original object, with respect to which the authenticity of the object is identified or compared. Therefore, a reference image represents one or more or all original objects.

[0112] In some implementations, the reference image is a digital image of the original object or a digital video of the original object.

[0113] In another implementation, the reference image is a digital model or digital model file of the original object, such as a 2D model, a 3D model, or a similar model of the original object.

[0114] In yet another implementation, the reference image is a manufacturing file of the original object, such as a print file or a 3D print file. The print file could be, for example, a print file of packaging.

[0115] The reference image and the input image can be in different data formats. Therefore, the recognition server system 50 can be configured to convert the input image into a file format corresponding to the reference image. Alternatively, the recognition server system 50 can be configured to analyze the reference image and the input image or target image in different file formats, or to compare the reference image and the input image or target image with each other.

[0116] It should be noted that the present invention is not limited to any file type or image type, but the file and image types can vary, and any kind of image and data file can be used in the method and system of the present invention.

[0117] Figure 3A An original input image 90 is shown to represent the object to be identified. The original input image 90 includes an object input image 70 representing the object and a background image or additional image 91 that is not part of the object and object image 70. The object in the input object image 70 is a packaging surface of a package.

[0118] It should be noted that the object whose authenticity is being verified can be any object.

[0119] Figure 3B An input object image 70 is shown, which merely represents the object whose authenticity will be recognized. Therefore, the input object image 70 represents or conforms to the outline or appearance of the object.

[0120] It should be noted that there may be one or more input images 70 that will be used together in a similar manner to identify the authenticity of an object.

[0121] Therefore, in one implementation, the background image or additional image 91 of the original input image 90 is removed.

[0122] Alternatively, the input object image 70 is extracted from the input image 90.

[0123] Alternatively, the input object image 70 is identified only in the original input image 90, and the original input image 90 is used in such a way that recognition is performed only relative to the identified input object image 70. Therefore, only the identified input object image 70 is used for recognition, but the background 90 of the original input image 90 is not removed or the input object image 70 is not extracted.

[0124] Removing the background 91 and extracting the input object image 70 also includes recognizing the input object image 70 in the original input image 90. Therefore, recognizing the input object image 70 in the original input image 90 is performed before removing the background 91 or extracting the input object image 70 from the original input image 90.

[0125] Recognizing the input object image 70 may include recognizing the input object image 70 or the corner 79 of the object in the original input image 90. Recognizing the corner 79 may include recognizing the shape of the input object image 70 or the object in the original input image 90. Recognizing the input object image 70 or the corner 79 and the shape of the object in the original input image 90 may also be different steps and performed sequentially. Shape recognition may be performed before corner recognition, or vice versa. Recognizing the input object image 70 includes recognizing both corners and shapes, or only one of them.

[0126] The removal of background 91, extraction of input object image 70 and / or recognition of input object image 70 are performed using recognition server system 50 or recognition application 56, or alternatively by user device 10 or user application 17.

[0127] The input object image 70 is used to form the target image 70 or during the generation of the target image. The target image 70 is an image of the object being analyzed or compared relative to a reference image.

[0128] In the context of this application, the term input image 70 means input object image 70.

[0129] Figure 4A A reference image 60 of the original object is shown. Figure 4AReference image 60 represents a surface or side of the original packaging. Reference image 60 includes a reference packaging surface 62 having reference contours or edges 68 and reference corners 69 that form the shape of reference image 60 and the original object. Reference packaging surface 62 includes a reference symbol area 64 with a reference symbol 65 representing a sun with multiple rays extending from its center. Reference packaging surface 62 also includes a reference text area 66 containing characters 67 representing the word "Product".

[0130] It should be noted that the reference text region 64 and reference symbol region 66 of the reference image 60 may vary depending on the original object. Therefore, the reference text region 66 and / or reference symbol region 64 may be absent, or there may be two or more reference text regions 66 or reference symbol regions 64.

[0131] Figure 4B It shows Figure 3B The input image 70 of the object to be identified, and its authenticity. Figure 4B The input image 70 represents a surface or side of a package that has been identified as authentic by the method and system of the present invention. The input image 70 includes a package surface 72 having contours or edges 78 and corners 79 that form the shape of the input image 70 and the object. The package surface 72 includes a symbol area 74 with a symbol 75 representing a sun with multiple rays extending from its center. The package surface 72 also includes a text area 76 containing characters 77 representing the word "Product".

[0132] In this invention, the system and method are configured to analyze an input image 70 relative to a reference image 60 to identify the authenticity of objects in the input image 70. Therefore, the input image is compared with the reference image 60. Furthermore, the shapes or contours 68, 78, symbol regions 64, 75 and / or symbols 65, 75 and / or text regions 66, 76 and / or characters 67, 77 of the reference image 60 and the input image 70 are compared with each other. Therefore, the system and method of this invention evaluate whether the input image matches or corresponds to the reference image 60, and further evaluate whether the object matches or corresponds to the original object.

[0133] As from Figure 4A and Figure 4B As can be seen, a ray is missing from symbol 75 in symbol region 74 of input image 70, and there is a symbol deviation 80 in input image 70. Furthermore, relative to reference image 60, there is a character deviation 81 in the word "Produ ct" in text region 77 of input image 70.

[0134] In this invention, one or more input images 60 of the object to be identified are prepared to correspond as closely as possible to a reference image. This ensures that variations in the obtained input images do not affect the accuracy of the identification and allows for high-quality identification results.

[0135] Therefore, at least one of the one or more input images 70 is used to generate a target image analyzed relative to a reference image. Thus, the target image is generated based on at least one or more input images 70.

[0136] Figure 5 An embodiment of the method for identifying the authenticity of an object according to the present invention is illustrated schematically.

[0137] The method includes maintaining a reference image 60 of the original object in the recognition server system 50 in step 300. The reference image 60 is provided to indicate that the object to be recognized is equivalent to the original object in terms of its authenticity. The method also includes receiving one or more input images of the object to be recognized in the recognition server system 50 in step 350. The input images may be the original input image 90 or the input object image 70.

[0138] Input images 90 and 70 are received from user equipment 10.

[0139] The method further includes, in step 400, the recognition server system 50 aligning at least one of the one or more input images with a reference image by distorting at least one of the input images. Alignment through distortion means changing the spatial relationships between parts, positions, regions, or pixels of the image, thereby altering the relative sizes within the image. Therefore, image alignment means adjusting the size and / or internal dimensions of the input image 70 or the target image to match the corresponding size of the reference image 60. Furthermore, alignment can be considered as arranging the target image on top of the reference image such that identical points, positions, or content in the target and reference images correspond to each other. Therefore, distorting the target image to correspond to the reference image allows for accurate analysis of their differences.

[0140] Therefore, the purpose of alignment and distortion is to compensate for at least one of the following: lens distortion caused by the camera, curves and bulges in the object to be identified, perspective, reflection, etc. Thus, the input image is aligned to the best possible version and used to provide reliable recognition.

[0141] Therefore, in step 400, at least one of the one or more input images 70 is distorted relative to the reference image 60 so that it corresponds to or matches the reference image 60 for alignment. This allows for accurate analysis of the details of at least one of the one or more input images relative to the reference image 60. Thus, at least one aligned input image is generated based on at least one of the one or more input images 70.

[0142] Following alignment step 400, the method includes, in step 450, generating a target image by the recognition server system 50 based on at least one aligned input image from the one or more input images. Therefore, in this step, an aligned target image is formed from at least one of the one or more aligned input images provided in step 400. Because at least one of the one or more input images is aligned with the reference image 60 in step 400, the target image generated in step 450 is also aligned.

[0143] In step 450, a target image may be generated based on one or alternatively based on two or more aligned input images 70.

[0144] Generating the target image in step 450 may include combining two or more aligned input images 70. Furthermore, generating the target image in step 450 may include preprocessing at least one of the aligned input images 70 through image preprocessing. When only one aligned input image 70 is used to generate the target image, step 450 may consist solely of image preprocessing.

[0145] Before step 500, the generated target image can be further aligned with the reference image.

[0146] also, Figure 5 The method includes analyzing an aligned target image relative to an aligned reference image 60 by a recognition server system 50 to identify the authenticity of an object. Therefore, in step 500, the target image aligned with the reference image 60 is analyzed relative to the reference image 60, or the target image aligned with the reference image 60 is compared with the reference image 60, to identify the authenticity of the object. Authenticity is identified by analyzing whether the target image corresponds to or matches the reference image 60.

[0147] Figure 6 An alternative embodiment of the method according to the invention is shown. In this embodiment, steps 300, 350, and 500 correspond to... Figure 5 The same steps 300, 350, and 500 are used in the implementation scheme.

[0148] exist Figure 6In one implementation, step 450, in which the target image is generated by the recognition server system 50, is performed before the alignment step 400.

[0149] In step 450, a target image is generated based on at least one of one or more input images 70.

[0150] When the target image is generated using only one aligned input image 70, step 450 may consist of only image preprocessing.

[0151] When two or more input images are used in step 450, generating the target image in step 450 may include combining the two or more input images 70 to form the target image. Furthermore, generating the target image in step 450 may include preprocessing at least one of the one or more input images 70 by image preprocessing or preprocessing the combined target image.

[0152] Combining two or more input images 70 in step 450 to form a target image may also include aligning the two or more input images 70 with each other or with a primary input image selected from the input images.

[0153] In one implementation, this is achieved by selecting one of the input images as an alignment reference image—meaning the primary input image—and aligning the other input images to the selected alignment reference image through distortion. The alignment in step 450 enables the generation of a high-quality target image by combining two or more input images.

[0154] Alternatively, combining two or more input images 70 to form a target image in step 450 may also include aligning the two or more input images 70 with a reference image 60. The aligned input images 70 are then combined to form the target image. The alignment in step 450 enables the generation of a high-quality target image by combining two or more input images. Then, in step 400, the generated target image is further image-aligned with the reference image 60 by the recognition server system 50 or the recognition application 56.

[0155] In one embodiment, the method includes receiving two or more input images of an object to be identified in an identification server system 50 in step 350, and aligning at least two of the two or more received input images relative to each other or relative to a reference image.

[0156] Step 450 of generating a target image from two or more aligned input images includes: dividing at least two of the two or more input images into two or more sub-parts by the recognition server system 50; selecting at least two of the two or more sub-parts by the recognition server system 50; and generating the target image by the recognition server system 50 by combining the selected sub-parts into a target image.

[0157] Preferably, two or more input images are divided into two or more sub-parts in a similar manner, such that for each sub-part, there are two or more corresponding sub-parts. Therefore, for each sub-part, one of the corresponding sub-parts is selected for the target image. The target image is formed from the selected sub-parts. This allows for the formation of the best possible target image.

[0158] In another embodiment, step 450 of generating a target image from two or more aligned input images includes: dividing at least two of the two or more input images into two or more sub-parts by an identification server system 50; comparing the two or more sub-parts of the input images with corresponding sub-parts of a reference image by the identification server system 50; selecting at least two of the two or more sub-parts based on the comparison by the identification server system 50; and generating the target image by combining the selected sub-parts into a target image. Therefore, for each sub-part of the target image, there are two or more corresponding alternative sub-parts. These corresponding alternative sub-parts are compared with corresponding sub-parts of the reference image, and one of the two or more alternative sub-parts is selected for the target image based on the comparison. This operation is repeated for all corresponding alternative sub-parts of the two or more input images. The target image is formed from the selected sub-parts. This makes it possible to form the best possible target image.

[0159] When only one input image 70 is received in step 350, the input image 70 forms a shape for... Figure 5 and Figure 6 The target image in steps 400 and 450 of the implementation scheme.

[0160] Therefore, when only one input image is received, step 400 can be performed before step 450, or alternatively, step 450 can be performed before step 400.

[0161] When two or more input images 70 are received in step 350, the recognition server system 50 can select one input image as the one used for recognition. Figure 5 and Figure 6The target image in steps 400 and 450 of the implementation scheme. This selection can be performed by the recognition server system 50 in step 350, or in step 400, or in step 450. Furthermore, the selection can include first comparing two or more input images 70 with a reference image 60, and selecting one of the input images 70 as the target image based on the comparison. The selection performed by the recognition server system 50 can be performed such that the one of two or more input images 70 that provides the best match with the reference image 60 is selected.

[0162] Also in this embodiment, step 400 may be performed before step 450, or alternatively, step 450 may be performed before step 400.

[0163] It should be noted that in all implementation schemes, in Figure 6 In some implementations, step 450 may also be performed by user device 10 or user application 17 prior to step 350. Therefore, in step 350, the generated target image is received in the recognition server system 50.

[0164] In step 400, the target image, which is a combination of two or more input images, can be further aligned with the reference image. Alternatively, the two or more input images can be aligned with the reference image and then combined into the target image, without further alignment. However, in the latter case, the combined target images can also be aligned.

[0165] Figure 7A , Figure 7B and Figure 7C The step 400 of aligning the target image 71 with the reference image 60 is illustrated schematically. One or more input images 70 can be aligned with or with the reference image 60 in the same manner.

[0166] like Figure 7A As shown, the target image is aligned with the reference image 60 such that corresponding positions or points are aligned to form the aligned target image 71. In this application, image alignment means deforming and distorting the target image to match the reference image 60. When the target image 70 and the reference image 60 are aligned, image comparison can be performed efficiently and accurately because the same features in the aligned target image 71 and the reference image 60 are compared or analyzed against each other.

[0167] Due to environmental factors during the provision of the input image 70, or due to variables in the imaging or capturing of the image and the imaging device, the input image 70 often includes deviations relative to the original object and the reference image 60. These variables cause deviations in the input image, making the identification of its authenticity difficult or unreliable.

[0168] Figure 7B An input image 70, or input object image, representing the object to be identified as real, is shown. The input image 70 is captured using a camera and includes outer edges or contours 78, symbols 75, and text 77. (As shown from...) Figure 7B As can be seen, due to variable factors in the environment where the image was captured and the imaging device itself, the input image 70, or its outer edges 78, symbols 75, and / or text, are slightly distorted or do not perfectly match the appearance of the original object. During the alignment step 400, the input image 70, or a target image generated from one or more input images, is aligned with the reference image 60 to form the aligned target image 71. During the alignment step, the input image 70 or the target image is distorted so that similar positions or points or content match the corresponding positions or points or content in the reference image.

[0169] Therefore, the purpose of alignment and distortion is to compensate for at least one of the following: lens distortion caused by the camera, curves and bulges in the object to be identified, perspective, reflection, etc. Thus, the input image is aligned to the best possible version and used to provide reliable recognition.

[0170] Distorting the input image or target may include at least scaling the input image 70 or the target relative to a reference image, and keystone correction to generate an aligned target image. Distortion may be performed globally on the image or locally on one or more sub-parts, points, locations, or pixels of the input image 70 or the target image.

[0171] Trapezoidal distortion correction refers to the distortion of an image caused by projecting it onto an angled surface or vice versa—projecting an image taken at an angle onto a straight surface.

[0172] In some implementations, image alignment may also include rotating and / or shifting the input image 70 or the target image relative to the reference image 60.

[0173] Figure 7C The aligned target image 77 is shown. As can be seen, the aligned target image 71, along with the aligned outline or outer edge 78', the aligned symbol 75', and the aligned text 77', is aligned to match the reference image 60. Therefore, the aligned target image 71 is provided to correspond to the reference image 60 as closely as possible before the realism analysis in step 500.

[0174] Figure 8 A flowchart of one embodiment of the method according to the present invention is shown. This can be seen as described above regarding... Figures 3A to 7CSteps 300, 350, 400, and 450 are performed as described. However, in this embodiment, in step 400, one or more input images 70 are aligned with a reference image 60, and then in step 450, an aligned target image 71 is generated. Step 450 can be performed as disclosed above.

[0175] Figure 8 The implementation also includes step 550, performed prior to step 500. Step 550 includes defining one or more sub-portions in the aligned target image 71 by the recognition server system. Then, step 500 includes analyzing at least one of the one or more sub-portions of the aligned target image 71 by the recognition server system 50 relative to at least one corresponding sub-portion of the aligned reference image 60 for the purpose of identifying the authenticity of the object.

[0176] In this embodiment, step 450 may further include aligning the generated target image with the reference image 60 to form an aligned reference image.

[0177] Alternatively, step 550 or step 500 may include aligning at least one of one or more sub-parts of the target image with a corresponding sub-part of the reference image 60 before performing the analysis.

[0178] Therefore, in this embodiment, each sub-part of the aligned target image 71 can be analyzed individually relative to the corresponding sub-part of the reference image 60. Thus, it is not necessary to analyze the entire aligned target image.

[0179] The reference image 60 can be pre-divided into two or more sub-parts, and the aligned target image 71 can be similarly divided into corresponding sub-parts based on the pre-divided sub-parts of the reference image 60.

[0180] Alternatively, the aligned target image 71 can be divided into two or more sub-parts. Then, the reference image 60 is similarly divided into two or more sub-parts based on the divided aligned target image 71.

[0181] Alternatively, the attention algorithm can be used to analyze one or more corresponding sub-parts of the target image 71 to be aligned and the reference image 60 to be compared, which are analyzed. Again, alternatively, the attention algorithm can be used to analyze one or more corresponding sub-parts of the target image 71 to be aligned and the reference image 60 to be compared, which are analyzed.

[0182] The input image can be aligned with the main input image or reference image in a similar way to generate the target image.

[0183] Figure 9 A flowchart of another embodiment of the method according to the invention is shown. This can be seen as described above regarding... Figures 3A to 7C Steps 300, 350, and 450 are performed as described. However, in this embodiment, the target image is generated in step 450 based on one or more input images 70 before aligning the target image 70 with the reference image 60.

[0184] Therefore, in step 450 of generating a target image based on at least one of one or more input images, in some embodiments, it may include combining two or more input images to form the target image. Furthermore, this may include aligning the two or more input images to each other, to one of the input images, or to a reference image before combining them. In an alternative embodiment, one of the one or more input images is selected as the target image, as disclosed above.

[0185] The method then includes defining one or more sub-regions in the target image in step 550. The definition of the sub-regions in the target image can be performed in any of the manner disclosed above.

[0186] After defining one or more sub-parts in the target image, in step 400 at least one of the one or more sub-parts of the target image is aligned with at least one corresponding sub-part of the reference image 60.

[0187] Therefore, one or more sub-parts of the target image can be aligned with corresponding sub-parts of the reference image 60 independently or separately. Furthermore, not all sub-parts of the target image need to be aligned and analyzed.

[0188] In step 500, at least one aligned sub-part of one or more sub-parts of the target image is further analyzed relative to the corresponding sub-part of the reference image 60.

[0189] Alignment of a target image with a reference image, or at least one sub-part of a target image with a corresponding sub-part of a reference image, can be performed in various ways. The same applies to aligning one or more input images with each other, with one of the input images, or with a reference image or a sub-part thereof.

[0190] In one implementation, for example, the alignment in step 400 includes identifying corresponding positions in the target image and corresponding positions in the reference image in step 402. Then, in step 404, the corresponding positions in the target image and the corresponding positions in the reference image are aligned to align the target image with the reference image, such as... Figure 11 As shown in the flowchart.

[0191] In one implementation, identifying corresponding positions in the target image and corresponding positions in the reference image in step 402 and aligning the target image and the reference image in step 404 are performed in the recognition server system 50 by an alignment algorithm. The alignment algorithm may be, for example, a machine learning alignment algorithm, such as an alignment neural network, which is trained to identify corresponding positions and align the target image and the reference image or their corresponding sub-parts.

[0192] In an alternative implementation, step 402, which identifies corresponding locations in the target image and corresponding locations in the reference image, may include identifying corresponding points or content in the reference image. This can be performed in the recognition server system 50 using image recognition algorithms, machine learning recognition algorithms, etc. Then, in step 404, the identified corresponding locations, points, or content in the target image and reference image are aligned by distorting the target image. Step 404 can be performed using an alignment algorithm, such as a machine learning alignment algorithm or an alignment neural network, which is trained to align the target image and the reference image or their corresponding sub-parts based on the identified corresponding locations, points, or content.

[0193] In some implementations, step 402 involves first identifying one or more locations, points, or contents in the reference image. The mentioned locations in the reference image may also be identified beforehand, or an alignment or identification algorithm trained on a reference image having the mentioned locations, points, or contents may be used. Step 402 then includes identifying locations, points, or contents in the target image that correspond to the mentioned locations, points, or contents in the reference image.

[0194] In an alternative implementation, step 402 is performed such that one or more locations, points, or contents are first identified in the target image. The mentioned locations or contents in the target image can be identified using an alignment algorithm or identification algorithm trained to identify locations, points, or contents in the target image. Step 402 then includes identifying locations, points, or contents in a reference image that correspond to the mentioned locations, points, or contents in the target image.

[0195] Then, step 404 is performed to align the corresponding positions, points, or contents of the target image with the corresponding positions, points, or contents of the reference image as discussed above.

[0196] Figures 12 to 17 An embodiment is shown that aligns the target image with a reference and performs step 400. It should be noted that regarding... Figures 12 to 17 The disclosed method can also be used to align input images with each other or with a reference image.

[0197] Figure 12A flowchart illustrating the alignment of a target image with a reference image is shown schematically. It should be noted that this alignment method can also be used to align input images to each other or to a reference image.

[0198] Alignment includes associating or arranging the reference image grid 100 to the reference image 60 in step 410. The reference image 60, having the reference image grid 100, is shown in... Figure 13A In the middle, the reference image grid 100 is locked or fixed to the reference image 60, so that the reference image grid 100 is immovable or unable to move relative to the reference image 60.

[0199] exist Figure 13A In one embodiment, a reference grid 100 is associated with a reference image 60 that exposes a reference packaging surface 62. The reference image grid includes first grid lines 101 extending in a first direction and second grid lines 102 extending in a second direction. In this embodiment, the first grid lines 101 and the second grid lines 102, and therefore the first and second directions, extend perpendicularly to each other. However, in other embodiments, they may extend in an alternative lateral direction or at an angle relative to each other. The reference image grid 100 includes two or more first grid lines 101 and two or more second grid lines 102 for forming the reference image grid 100.

[0200] The first grid line 101 and the second grid line 102 of the reference image grid 100 intersect at an intersection point or grid point 103. Furthermore, the first grid line 101 and the second grid line 102 of the reference image grid 100 provide two or more grid regions 104. Therefore, each grid region 104 is defined by two adjacent first grid lines 101 and two adjacent second grid lines 102, and by grid points 103 between these first grid lines 101 and second grid lines 102.

[0201] It should be noted that, according to the present invention, the reference image grid 100 may include any number of first grid lines 101 and second grid lines 102, depending on the embodiment and depending on the reference image 60.

[0202] Furthermore, in some embodiments, the reference image grid 100 is predefined to have a predetermined number of first grid lines 101 and second grid lines 102 and thus a predetermined number of reference grid points 103 and reference grid regions 104.

[0203] Alternatively, in step 410, when associating the reference image grid 100 with the reference image 60, the number of the reference image grid 100, the number of the first grid lines 101 and the second grid lines 102, and the number of reference grid points 103 and reference grid regions 104 are determined. In this embodiment, the reference image grid 100 may be determined based on the target image, based on the reference image, or based on both the reference image and the target image.

[0204] like Figure 12 As shown, the alignment includes associating or arranging the target image grid 110 to the target image 70 in step 412. The target image 70 having the target image grid 110 is shown in... Figure 13B In the middle. The target image grid 110 is locked or fixed to the target image 70, so that the target image grid 110 is immovable or unable to move relative to the target image 70.

[0205] exist Figure 13B In one embodiment, a target grid 110 is associated with a target image 70 that exposes a target packaging surface 72. The target image grid 110 includes first grid lines 111 extending in a first direction and second grid lines 112 extending in a second direction. In this embodiment, the first grid lines 111 and the second grid lines 112, and therefore the first and second directions, extend perpendicularly to each other. However, in other embodiments, they may extend in an alternative lateral direction or at an angle relative to each other. The target image grid 110 includes two or more first grid lines 111 and two or more second grid lines 112 for forming the target image grid 110.

[0206] The first grid line 111 and the second grid line 112 of the target image grid 110 intersect at an intersection point or grid point 113. Furthermore, the first grid line 111 and the second grid line 112 of the target image grid 110 provide two or more grid regions 114. Therefore, each grid region 114 is defined by two adjacent first grid lines 111 and two adjacent second grid lines 112, and by grid points 113 between these first grid lines 111 and second grid lines 112.

[0207] It should be noted that, according to the present invention, the target image grid 110 may include any number of first grid lines 111 and second grid lines 112, depending on the embodiment and depending on the target image 70.

[0208] Furthermore, in some embodiments, the reference image grid 100 is predefined as having a predetermined number of first grid lines 101 and second grid lines 102, and thus a predetermined number of reference grid points 103 and reference grid regions 104. The target image grid 110 is then provided to correspond to, be similar to, or be identical to the reference image grid 100.

[0209] Alternatively, in step 412, when associating the target image grid 110 with the target image 70, the number of the target image grid 110, the number of the first grid lines 111 and the second grid lines 112, and the number of target grid points 113 and target grid regions 114 are determined. In this embodiment, the target image grid 110 can be determined based on the target image, based on a reference image, or based on both the reference image and the target image. Furthermore, in this embodiment, step 412 can be performed before step 410, and the target image grid 110 can be determined and associated before the reference image grid 100.

[0210] The alignment then includes step 414, in which the target image grid 110 and the reference image grid 100 are used to align the target image 70 with the reference image 60 and provide an aligned target image 71.

[0211] Due to factors such as the angle at which the image is captured and distortion caused by the imaging device, the input image is typically distorted relative to the reference image. Furthermore, because the target image grid 110 is associated with the target image 70, the target grid point 113 associated with the target image 70 is not exactly the same as the reference grid point 103 in the reference image 60 associated with the reference image 60.

[0212] The reference image grid 110 and the target image grid 110 include corresponding reference grid points 103 and target grid points 113, and the alignment step 414 includes the recognition server system 50 aligning the target image 70 with the reference image 60 by moving at least one of the target grid points 113 of the target image grid 110 to distort the target image 70. Therefore, moving the target grid points 113 relative to each other or moving at least one target grid point 113 relative to other target grid points 113 of the target image grid 110 distorts the target image. Therefore, aligning each or at least one of the target grid points 113 individually relative to a corresponding one of the reference grid points 103 of the reference image grid 100 is used to align the target image 70 with the reference image 60.

[0213] Therefore, by moving the target grid point 113 relative to the reference grid point 103 of the reference image grid 100, the target image is distorted relative to the reference image 60.

[0214] As mentioned above, because each target grid point 113 is individually aligned relative to its corresponding reference grid point 103, the target image 70 and the target image grid 110 become distorted. Therefore, the aligned target image 71 is distorted relative to the target image 70 generated before alignment. Furthermore, the target image grid 110 and the target grid region 114 become distorted relative to the target image grid 110 and target grid region 114 provided before alignment. Thus, the distortion provides an aligned target image 71 aligned with the reference image 60, and the distortion of the target image 70 relative to the reference image 60 is eliminated or substantially eliminated.

[0215] Figure 19 The movement of a target grid point 103a relative to other target grid points 113 and reference image grid 100 is schematically shown.

[0216] Figure 20A and Figure 20B Further illustration shows moving several target grid points 113a, 113b, 113c, 113d relative to each other and other target grid points to distort target image grid 110 and target image 70 for aligning the target image with the reference image.

[0217] Figure 14 A schematic view of the aligned target image grid 110 and reference image grid 100, as well as the aligned target image 71 and reference image 60, is shown.

[0218] In some implementations, step 414, which aligns the target image 70 with the reference image 60, is performed iteratively.

[0219] In one implementation, step 414 includes the identification server system 50 iteratively aligning the target image 70 with a reference image by iteratively moving target grid points 113 of the target image grid 110 to distort the target image 70. Therefore, during the alignment in step 414, at least one of the target grid points 113 is moved two or more times. This can be performed such that in a first movement cycle, the target grid points 113 are moved relative to each other for the first time with a first predetermined movement or movement distance. Then, in the next and subsequent possible successive movement cycles, the target grid points 113 are successively moved relative to each other with successive predetermined movement or movement distances and aligned with corresponding reference grid points 103, said successive predetermined movement or movement distances being smaller than the first predetermined movement or movement distance and arranged to decrease successively. Thus, iterative and accurate alignment is achieved. Analysis step 500 can be performed between successive movement cycles.

[0220] In one embodiment, the iterative alignment in step 414 further includes aligning the target image with the reference image by the recognition server system 50 through moving the target grid points 113 of the target image grid 110 to distort the target image 70. Between successive movement cycles, the alignment includes: verifying the alignment of the target image with the reference image by the recognition server system 50; and iteratively aligning the target image with the reference image by the recognition server system 50 through iteratively moving the target grid points 113 of the target image grid 110 in successive movement cycles to distort the target image 70.

[0221] In one implementation, this verification is performed by step 500 analysis, such as... Figure 21 As shown in the figure.

[0222] Figure 15 A flowchart of one embodiment is schematically shown, in which the process of iteratively aligning the target image grid 110 with the reference image grid 100 is performed in more detail.

[0223] In this embodiment, alignment includes step 410 of associating target image grid 110 with target image 70 and step 412 of associating reference image grid 100 with reference image 60.

[0224] In this embodiment, in step 416, the reference image 60 and the reference image grid 100 are scaled down to form a scaled-down reference image 60 with an associated and scaled-down reference image grid 100.

[0225] Then, step 418 is performed, in which the recognition server system 50 aligns the target image with a scaled-down reference image to provide initial image alignment between the target image 70 and the reference image 70. A scaled-down reference image 60 with a scaled-down reference image grid 100 is shown in... Figure 16A middle.

[0226] Then, step 420 includes the recognition server system 50 scaling up the scaled-down reference image 60 together with the scaled-down reference image grid 100 to form a first scaled-up reference image 60 having the first scaled-up reference image grid 100. The first scaled-up reference image 60 having the first scaled-up reference image grid 100 is shown in... Figure 16B middle.

[0227] Then, the recognition server system 50 aligns the target image with a first scaled-up reference image to provide a first image alignment between the target image 70 and the reference image 60 or the first aligned target image 71.

[0228] Steps 420 and 422 may be repeated once or multiple times to provide subsequent or successive image alignment between the target image 70 and the reference image 60, such as... Figure 16C and Figure 16D As shown in the figure.

[0229] Therefore, the initial alignment of the target image with the scaled-down reference image provides a coarse alignment. Then, the reference image 60 and the reference image grid 100 are gradually scaled up, and in step 422 between the scaling up step 420, the target image is aligned with the scaled-up reference image to provide an iterative alignment in which the alignment of the target image 70 with the reference image 60 gradually becomes more and more accurate.

[0230] In one embodiment, the reference image 60 and the reference image grid 100 are first scaled down to 1 / 24 of the original size of the reference image 60, then scaled up to 1 / 12 of the original size of the reference image 60, then scaled up to 1 / 4 of the original size of the reference image 60, and then further scaled up to the original size of the reference image 60.

[0231] Figure 17 The diagram schematically illustrates an alignment of the target image 70 with a reference image 60 scaled down relative to its original size. This can be illustrated as follows: Figure 14 Perform the final alignment of the target image with the original-sized reference image as shown.

[0232] In one implementation, the alignment of the target image with the reference image includes, for example: Figure 15 The steps 410, 412, and 416 are shown and disclosed above. Step 418 then includes the recognition server system 50 aligning the target image with a scaled-down reference image by moving target grid points 113 of the target image grid 110 to provide initial alignment. Next, step 420 is performed by the recognition server system 50 scaling up the scaled-down reference image 60 along with the scaled-down reference image grid 100 to form a first scaled-up reference image having the first scaled-up reference image grid 110. Then, step 422 includes the recognition server system 50 aligning the target image with the first scaled-up reference image by moving target grid points 110 of the target image grid 100.

[0233] Steps 420 and 422 may be repeated once or multiple times to provide subsequent or successive image alignment between the target image 70 and the reference image 60.

[0234] Therefore, the initial alignment of the target image with the scaled-down reference image provides a coarse alignment. Then, the reference image 60 and the reference image grid 100 are gradually scaled up, and in step 422 between the scaling up step 420, the target image is aligned with the scaled-up reference image to provide an iterative alignment, in which the alignment of the target image 70 with the reference image 60 gradually becomes more and more accurate.

[0235] The above relates to image alignment between a target image and a reference image. In some embodiments, step 400 of aligning the target image with the reference image may further include color alignment of the target image, or the aligned target image, with the reference image by the recognition server system 50. Therefore, the color of the target image is aligned with or corresponds to the color of the reference image. This may include analyzing the color of the reference image and color aligning the target image based on that analysis. Color alignment may be performed, for example, using a color alignment machine learning algorithm, a color alignment neural network, or a color alignment filter. Furthermore, in addition to or instead of the target image, color alignment with the reference image may be performed on one or more input images. The color alignment of the input images is performed before generating the target image.

[0236] Furthermore, in some embodiments, the step 400 of aligning the target image with the reference image may also include the recognition server system 50 aligning the size of the target image or the aligned target image with that of the reference image. Therefore, the size of the target image is aligned with or corresponds to the size of the reference image. This may include analyzing the size of the reference image and aligning the size of the target image based on that analysis. Size alignment may be performed, for example, using size alignment machine learning algorithms, size alignment neural networks, etc. In addition, besides or instead of the target image, one or more input images may also be aligned with the reference image. The input image size alignment is performed before generating the target image.

[0237] Aligned target image 71 and reference image 60, or similarly, at least one aligned sub-part of target image 71 and a corresponding sub-part of reference image 60 are analyzed relative to each other to determine the authenticity of the object to be identified, such as... Figure 5 , Figure 6 , Figure 8 and Figure 9 As defined in step 500.

[0238] In some implementations, step 500 includes the recognition server system 50 comparing the aligned target image 71 with the aligned reference image 60 using statistical methods to identify the authenticity of the object.

[0239] In an alternative implementation, step 500 includes the recognition server system 50 comparing at least one of one or more aligned sub-parts of the target image 71 with at least one corresponding sub-part of the reference image 60 using statistical methods to identify the authenticity of the object.

[0240] In some implementations, the statistical methods or statistical analysis algorithms used to compare the aligned target image 7 with the reference image include one of the following: linear regression analysis, resampling methods, subset selection, shrinkage, dimensionality reduction, nonlinear models, tree-based methods, standard deviation, etc. Statistical methods may include linear methods, nonlinear methods, or numerical algorithms.

[0241] In some implementations, the statistical method includes performing pixel-to-pixel statistical analysis between the reference image and the input image.

[0242] Furthermore, pixel-to-pixel analysis involves analyzing pixel-to-pixel distances in both the input and reference images. A threshold can be used to determine the number of erroneous pixels in the input image relative to the reference image, for the purpose of authenticity verification. Additionally, a product-specific threshold for an acceptable number of erroneous pixels can be predetermined. Therefore, the analysis results are compared to the threshold to determine authenticity.

[0243] In an alternative implementation, step 500 of analyzing the aligned target image 71 relative to the aligned reference image 60 includes: maintaining a machine learning recognition algorithm or recognition neural network in the recognition server system 50; and having the recognition server system 50 compare the aligned target image with the aligned reference image by utilizing the machine learning recognition algorithm or recognition neural network.

[0244] In yet another alternative implementation, step 500 of analyzing the aligned target image 71 relative to the aligned reference image 60 includes: maintaining a machine learning recognition algorithm or recognition neural network in the recognition server system 50; and having the recognition server system 50 compare at least one of one or more aligned sub-parts of the target image with at least one corresponding sub-part of the reference image by utilizing the machine learning recognition algorithm or recognition neural network.

[0245] Figure 10 An embodiment of step 500 is shown, in which a machine learning algorithm is used to identify the authenticity of an object.

[0246] exist Figure 10In one implementation, the analysis includes step 502, which provides a first machine learning recognition algorithm in the recognition server system 50, the first machine learning recognition algorithm being trained to determine the difference between an aligned target image and an aligned reference image. In step 504, the recognition server system 50 processes the aligned target image and the aligned reference image using the first machine learning recognition algorithm, and then in step 506, the recognition server system 50 generates a first difference vector by utilizing the first machine learning recognition algorithm. Figure 10 The implementation also includes a step 508 of providing a second machine learning recognition algorithm for a reference image, which is trained to analyze the differences determined by the first machine learning recognition algorithm relative to the reference image. In step 510, the recognition server system 50 processes the first difference vector using the second machine learning recognition algorithm for the reference image to identify the authenticity of the object.

[0247] Therefore, the differences between the aligned target image and the aligned reference image are first analyzed, and a first difference vector is generated. Then, it is analyzed whether the differences, defined by the first difference vector, are significant or important to the reference image in question. Some differences may be acceptable, while others may not.

[0248] exist Figure 10 In an alternative implementation, the analysis in step 500 includes step 502 providing a first recognition neural network in the recognition server system 50, the first recognition neural network being trained to determine the difference between the aligned target image and the aligned reference image. In step 504, the recognition server system 50 processes the aligned target image and the aligned reference image using the first recognition neural network, and then in step 506, the recognition server system 50 generates a first difference vector by utilizing the first recognition neural network. Figure 10 The implementation also includes a step 508 of providing a second recognition neural network, which is targeted at the reference image and trained to analyze differences determined by the first recognition neural network relative to the reference image. In step 510, the recognition server system 50 processes the first difference vector using the second recognition neural network for the reference image to identify the authenticity of the object.

[0249] exist Figure 10In another alternative implementation, analysis step 500 includes step 502 of providing a first machine learning recognition algorithm in the recognition server system 50, the first machine learning recognition algorithm being trained to determine the difference between an aligned sub-part of the target image and a corresponding aligned sub-part of the reference image. In step 504, the recognition server system 50 processes the aligned sub-part of the target image and the corresponding aligned sub-part of the reference image using the first machine learning recognition algorithm, and then in step 506, the recognition server system 50 generates a first difference vector by utilizing the first machine learning recognition algorithm. Figure 10 The implementation also includes a step 508 of providing a second machine learning recognition algorithm for one or more sub-parts of a reference image, and being trained to analyze the differences determined by the first machine learning recognition algorithm relative to each of the one or more sub-parts of the reference image. In step 510, the recognition server system 50 processes the first difference vector using the second machine learning recognition algorithm, respectively for one or more sub-parts of the reference image, for the purpose of identifying the authenticity of the object.

[0250] exist Figure 10 In another alternative embodiment, analysis step 500 includes step 502 of providing a first recognition neural network in the recognition server system 50, the first recognition neural network being trained to determine the difference between an aligned sub-part of the target image and a corresponding aligned sub-part of the reference image. In step 504, the recognition server system 50 processes the aligned sub-part of the target image and the corresponding aligned sub-part of the reference image using the first recognition neural network, and then in step 506, the recognition server system 50 generates a first difference vector by utilizing the first recognition neural network. Figure 10 The implementation also includes a step 508 of providing a second recognition neural network for one or more sub-parts of a reference image, and being trained to analyze differences determined by the first recognition neural network relative to each of the one or more sub-parts of the reference image. In step 510, the recognition server system 50 processes the first difference vector using the second recognition neural network, respectively for one or more sub-parts of the reference image, for the purpose of identifying the authenticity of the object.

[0251] The method of the present invention may further include excluding certain regions of the target image and / or reference image from the analysis and step 500. This exclusion of certain regions is performed before the target image is analyzed and aligned relative to the reference image. Therefore, this exclusion can be performed either in the alignment step 400 or in the analysis step 500, prior to the analysis.

[0252] In one embodiment, the method includes: prior to analysis, performing an analysis by a recognition server system of the aligned target image relative to an aligned reference image to exclude a predetermined region of the aligned target image. Thus, unwanted or altered portions or regions of the target image can be excluded from the analysis of authenticity.

[0253] In another embodiment, the method includes: prior to analysis, performing a mask application by a recognition server system on a predetermined region of the aligned target image and a corresponding predetermined region of a reference image to exclude the predetermined regions from the analysis of the aligned target image relative to the aligned reference image. Thus, the mask excludes the predetermined regions from the analysis. Preferably, the masks applied to the aligned target image and the reference image are identical, such that the analysis is unaffected by the mask.

[0254] Figure 18 This illustrates another embodiment or some other features of the invention. The method includes step 600, which generates a recognition output based on analysis of an aligned target image relative to an aligned reference image. Step 600 is performed after step 500. Step 600 also provides different recognition outputs based on whether the aligned target image matches the reference image (meaning whether the object to be recognized matches the original object). If the aligned target image matches the reference image, a positive recognition output is generated in step 710. If the aligned target image does not match the reference image, a negative recognition output is generated in step 720.

[0255] The recognition output can be a simple "yes" or "no" output. Alternatively, the recognition output can be a matching score determined by steps 500 and 600, which define how well the aligned target image matches the reference image.

[0256] The invention has been described above with reference to the embodiments shown in the accompanying drawings. However, the invention is by no means limited to the above embodiments, but can be varied within the scope of the claims.

Claims

1. A method for identifying the authenticity of an object, the method comprising the following steps: a) A reference image of the original object is maintained in the identification server system (50), the reference image being provided to represent an equivalent original object; b) Receive one or more input images of the object to be identified in the identification server system (50); c) The recognition server system (50) generates a target image based on at least one of the one or more input images; d) The recognition server system (50) aligns the target image with the reference image by distorting the target image to match the reference image; as well as e) The target image, aligned with the reference image, is analyzed by the recognition server system (50) to identify the authenticity of the object. The feature is that the alignment in step d) includes image alignment: - The recognition server system (50) associates a reference image grid (100) onto the reference image, the reference image grid (100) including reference grid points (103). - The recognition server system (50) associates a target image grid (110) onto the target image, the target image grid (110) comprising target grid points (113); and - The recognition server system (50) aligns the target image with the reference image by moving the target grid points (113) of the target image grid (110) relative to each other and relative to the corresponding reference grid points (103) of the reference image grid (100) to distort the target image.

2. The method according to claim 1, characterized in that: The method further includes step f), which is performed prior to step d). f) One or more sub-parts of the aligned target image are defined by the recognition server system (50), and Step e) includes analyzing at least one of the one or more sub-parts of the aligned target image by the recognition server system (50) relative to at least one corresponding sub-part of the reference image for the purpose of identifying the authenticity of the object; or The method further includes step f), which is performed prior to step d). f) One or more sub-parts of the target image are defined by the recognition server system (50), and Step d) includes the recognition server system (50) aligning at least one of the one or more sub-parts of the target image with at least one corresponding sub-part of the reference image; as well as Step e) includes the identification server system (50) analyzing at least one of one or more aligned sub-parts of the target image relative to at least one corresponding sub-part of the reference image to identify the authenticity of the object.

3. The method according to claim 2, characterized in that, Step f) includes: - The alignment target image is divided into two or more sub-parts by the recognition server system (50); or - The reference image of the original object is maintained in the identification server system (50), the reference image being pre-divided into two or more sub-parts; and - The alignment target image is divided into two or more sub-parts by the recognition server system (50) based on the pre-divided sub-parts of the reference image; or - The recognition server system (50) divides the reference image into two or more sub-parts; and - The identification server system (50) divides the aligned target image into two or more sub-parts based on the sub-parts of the reference image.

4. The method according to claim 1, characterized in that, Step e) includes: - The recognition server system (50) compares the aligned target image with the reference image using statistical methods to identify the authenticity of the object; or - The recognition server system (50) uses statistical methods to compare at least one of the one or more aligned sub-parts of the target image with at least one corresponding sub-part of the reference image to identify the authenticity of the object; or - Maintaining machine learning recognition algorithms or recognition neural networks in the recognition server system (50); and - The alignment target image is compared with the reference image by the recognition server system (50) using the machine learning recognition algorithm or the recognition neural network; or - Maintaining machine learning recognition algorithms or recognition neural networks in the recognition server system (50); and - The recognition server system (50) compares at least one of the one or more aligned sub-parts of the target image with at least one corresponding sub-part of the reference image by utilizing a machine learning recognition algorithm or the recognition neural network.

5. The method according to claim 4, characterized in that, Step e) includes: - The recognition server system (50) provides a first machine learning recognition algorithm and a second machine learning recognition algorithm, the first machine learning recognition algorithm being trained to determine the difference between the aligned target image and the reference image, and the second machine learning recognition algorithm being trained on the reference image to analyze the difference determined by the first machine learning recognition algorithm relative to the reference image; - The aligned target image and the reference image are processed by the recognition server system (50) using the first machine learning recognition algorithm, and a first difference vector is generated by the recognition server system (50) using the first machine learning recognition algorithm; and - The first difference vector is processed by the recognition server system (50) using the second machine learning recognition algorithm for the reference image to identify the authenticity of the object; or - The recognition server system (50) provides a first recognition neural network and a second recognition neural network, the first recognition neural network being trained to determine the difference between the aligned target image and the aligned reference image, and the second recognition neural network being trained for the reference image to analyze the difference determined by the first recognition neural network relative to the reference image; - The alignment of the target image and the reference image is processed by the recognition server system (50) using the first recognition neural network, and a first difference vector is generated by the recognition server system (50) using the first recognition neural network; and - The first difference vector is processed by the recognition server system (50) using the second recognition neural network for the reference image to identify the authenticity of the object.

6. The method according to claim 4, characterized in that, Step e) includes: - The recognition server system (50) provides a first machine learning recognition algorithm and one or more second machine learning recognition algorithms, the first machine learning recognition algorithm being trained to determine the difference between an aligned sub-part of the target image and a corresponding sub-part of the reference image, the one or more second machine learning recognition algorithms being trained for the one or more sub-parts of the reference image and to analyze the difference determined by the first machine learning recognition algorithm relative to each of the one or more sub-parts of the reference image; - The recognition server system (50) processes one or more aligned sub-parts of the target image and one or more corresponding sub-parts of the reference image respectively using the first machine learning recognition algorithm, and the recognition server system (50) generates one or more difference vectors by utilizing the first machine learning recognition algorithm, each of the one or more difference vectors corresponding to a sub-part of the reference image; and - The recognition server system (50) processes the one or more difference vectors respectively using the one or more second machine learning recognition algorithms for the one or more sub-parts of the reference image to identify the authenticity of the object; or - The recognition server system (50) provides a first recognition neural network and one or more second recognition neural networks, the first recognition neural network being trained to determine the difference between an aligned sub-part of the target image and a corresponding sub-part of the reference image, the one or more second recognition neural networks being trained for the one or more sub-parts of the reference image and to analyze the difference determined by the first recognition neural network relative to each of the one or more sub-parts of the reference image; - The recognition server system (50) processes one or more aligned sub-parts of the target image and one or more corresponding sub-parts of the reference image respectively using the first recognition neural network, and the recognition server system (50) generates one or more difference vectors by utilizing the first recognition neural network, each of the one or more difference vectors corresponding to a sub-part of the reference image; and - The recognition server system (50) processes the one or more difference vectors using the one or more second recognition neural networks for the one or more sub-parts of the reference image to identify the authenticity of the object.

7. The method according to claim 1, characterized in that, The alignment mentioned in step d) includes image alignment: - The recognition server system (50) identifies the corresponding position in the target image and the corresponding position in the reference image; as well as - The recognition server system (50) distorts the target image to align the corresponding position in the target image with the corresponding position in the reference image; or - The recognition server system (50) identifies the corresponding position in at least one of the one or more sub-parts of the target image and the corresponding position in at least one corresponding sub-part of the reference image; as well as - The recognition server system (50) aligns the corresponding positions in the at least one of the one or more sub-parts of the target image with the corresponding positions in the at least one corresponding sub-part of the reference image by distorting the at least one of the one or more sub-parts of the target image, so as to perform image alignment between the at least one of the one or more sub-parts of the target image and the at least one corresponding sub-part of the reference image.

8. The method according to claim 1, characterized in that: - Step d) Alignment by the recognition server system includes: iteratively aligning the target image with the reference image by iteratively moving the target grid points (113) of the target image grid (100) to distort the target image; or - Step d) Aligning the target image with the reference image by the recognition server system (50) further includes: - The recognition server system (50) aligns the target image with the reference image by moving the target grid points (113) of the target image grid (110) to distort the target image; - The alignment of the target image with the reference image is verified by the recognition server system (50); and - The recognition server system (50) iteratively aligns the target image with the reference image by moving the target grid points (113) of the target image grid (110) iteratively to distort the target image; or - Step d) Aligning the target image with the reference image by the recognition server system (50) further includes: - The recognition server system (50) aligns the target image with the reference image by moving the target grid points (113) of the target image grid (110) to distort the target image; - Step e) is performed by the recognition server system (50), in which the recognition server system (50) analyzes the aligned target image relative to the reference image; - In step d), the recognition server system (50) iteratively aligns the target image with the reference image by iteratively moving the target grid points (113) of the target image grid (110) to distort the target image.

9. The method according to claim 1, characterized in that, Step d) Aligning the target image with the reference image by the recognition server system (50) includes: - The recognition server system (50) scales down the reference image along with the reference image grid (100) to form a scaled-down reference image with the scaled-down reference image grid (100); - The recognition server system (50) aligns the target image grid (110) with the scaled-down reference image grid (100) to provide initial image alignment between the target image and the reference image; - The recognition server system (50) scales up the reduced reference image along with the scaled-down reference image grid (100) to form a first scaled-up reference image having a first scaled-up reference image grid (100); and - The recognition server system (50) aligns the target image grid (110) with the first scaled-up reference image grid (100) to provide a first image alignment between the target image and the reference image; or - The recognition server system (50) scales down the reference image along with the reference image grid (100) to form a scaled-down reference image with the scaled-down reference image grid (100); - The recognition server system (50) aligns the target image grid (110) with the scaled-down reference image grid (100) to provide initial image alignment between the target image and the reference image; - The recognition server system (50) enlarges the scaled-down reference image together with the scaled-down reference image grid (100) to form a first scaled-up reference image having a first scaled-up reference image grid (100); - The recognition server system (50) aligns the target image grid (110) with the first scaled-up reference image grid (100) to provide a first image alignment between the target image and the reference image; and - The recognition server system (50) repeats the scaling up of the reference image along with the reference image grid (100) once or more, and aligns the target image grid (110) with the scaled-up reference image grid (100) to provide a first image alignment of the target image with the reference image.

10. The method according to claim 9, characterized in that, Step d) Aligning the target image with the reference image by the recognition server system (50) further includes: - The recognition server system (50) scales down the reference image along with the reference image grid (100) to form a scaled-down reference image with the scaled-down reference image grid (100); - The recognition server system (50) aligns the target image with the scaled-down reference image by moving the target grid points (113) of the target image grid (110); - The recognition server system (50) scales up the reduced reference image along with the scaled-down reference image grid (100) to form a first scaled-up reference image having a first scaled-up reference image grid (100); and - The recognition server system (50) aligns the target image with the first scaled reference image by moving the target grid points (113) of the target image grid (110); or - The recognition server system (50) scales down the reference image along with the reference image grid (100) to form a scaled-down reference image with the scaled-down reference image grid (100); - The recognition server system (50) aligns the target image with the scaled-down reference image by moving the target grid points (113) of the target image grid (110); - The recognition server system (50) enlarges the scaled-down reference image along with the scaled-down reference image grid (100) to form a first scaled-up reference image having a first scaled-up reference image grid (100). - The recognition server system (50) aligns the target image with the first scaled-up reference image by moving the target grid points (113) of the target image grid (110); and - The recognition server system (50) repeats the reference image along with the reference image grid (100) one or more times and aligns the target image with the sized reference image to provide image alignment between the target image and the reference image.

11. The method according to claim 1, characterized in that: - Step b) includes receiving two or more input images of the object to be identified in the recognition server system (50); as well as - Step c) includes the recognition server system (50) selecting one of the two or more input images as the target image; or - Step b) includes receiving two or more input images of the object to be identified in the recognition server system (50); - Step c) includes the recognition server system (50) comparing the two or more input images with the reference image; as well as - Step c) further includes the identification server system (50) selecting one of the two or more input images as the target image based on the comparison.

12. The method according to claim 1, characterized in that: - Step b) includes receiving two or more input images of the object to be identified in the recognition server system (50); - Step c) includes the recognition server system (50) aligning at least two of the two or more input images relative to the reference image, and the recognition server system (50) generating the target image by combining the at least two of the two or more aligned input images into the target image; or - Step b) includes receiving two or more input images of the object to be identified in the recognition server system (50); Step c) includes the recognition server system (50) aligning at least two of the two or more input images relative to one of the input images, and the recognition server system (50) generating the target image by combining the at least two of the two or more aligned input images into the target image, wherein the alignment of each of the at least two input images with the one input image is performed according to the alignment in step d), each of the at least two input images serving as the target image in step d), and the one input image serving as the reference image.

13. The method according to claim 1 or 12, characterized in that: - Step b) includes receiving two or more input images of the object to be identified in the recognition server system (50); - Step c) includes: The recognition server system (50) aligns at least two of the two or more input images with respect to one of the input images. The recognition server system (50) divides at least two of the two or more input images into two or more sub-parts; The identification server system (50) selects at least two of the two or more sub-parts; and The target image is generated by the recognition server system (50) by combining the selected sub-parts into the target image; or - Step b) includes receiving two or more input images of the object to be identified in the recognition server system (50); - Step c) includes: The recognition server system (50) aligns at least two of the two or more input images with respect to one of the input images. The recognition server system (50) divides at least two of the two or more input images into two or more sub-parts; The recognition server system (50) compares the two or more sub-parts in the input image with the corresponding sub-parts in the reference image; The identification server system (50) selects at least two of the two or more sub-parts based on the comparison; and The target image is generated by the recognition server system (50) by combining the selected sub-parts into the target image.

14. The method according to claim 1, characterized in that, Step d) includes: - The recognition server system (50) extracts an input object image from the one or more input images or from the target image, the target image being composed of the input object image, the input object image conforming to the outline of the object; or - The recognition server system (50) extracts an input object image from the one or more input images or from the target image, and the recognition server system (50) removes the background to form the target image, which is composed of the input object image and conforms to the outline of the object.

15. The method according to claim 1, characterized in that: - Step c) includes the recognition server system (50) aligning the size of the one or more input images with that of the reference image; or - Step c) or step d) includes the identification server system (50) aligning the target image with the reference image size.

16. The method according to claim 1, characterized in that: - Step c) includes the recognition server system (50) aligning the colors of the one or more input images with those of the reference image; or - Step c) or step d) includes the identification server system (50) aligning the target image with the reference image in color.

17. The method according to claim 1, characterized in that, The method further includes, prior to step e): - The recognition server system (50) excludes a predetermined region of the aligned target image from the analysis of the aligned target image relative to the reference image; or - The recognition server system (50) applies a mask on a predetermined region of the aligned target image and a corresponding predetermined region of the reference image to exclude the predetermined region from the analysis of the aligned target image relative to the reference image.

Images