System and method for diamond verification and traceability

Abstract

A system for diamond verification and traceability is disclosed, comprising a cloud-based server device and a client device The server device is connected to a database that stores multiple data records for each diamond, identified by a unique ERP ID, across various stages of the supply chain. Each record includes FTIR spectral data, PL spectral data, images, weight, dimensions, color grade, and clarity grade. The server receives client-submitted diamond data and compares it with stored records by generating similarity vectors for each matched data type. These vectors contain feature parameters consisting of peak features, non-peak features, and other non-peak features. A classifier model analyze each vector to produce similarity scores, which are then processed by a probabilistic model to generate a final confidence score. Based on this score, the system generates a diamond verification and traceability report, enabling reliable identification and lineage tracking of diamonds.

Description

SYSTEM AND METHOD FOR DIAMOND VERIFICATION ANDTRACEABILITYTECHNICAL FIELD

[0001] The present disclosure relates generally to a system and method for diamond verification and traceability.BACKGROUND

[0002] In the gemstone industry, tracking the journey of a diamond from extraction at a mine through cutting, polishing, and eventual retail presentation has become increasingly important. Regulatory bodies and consumers alike demand transparent records to ensure ethical sourcing and compliance with legal restrictions, while manufacturers seek objective means to protect high-value natural diamonds from substitution or tampering. Although scientific tools such as Fourier transform infrared spectroscopy can capture atomic-level impurity profiles, integrating such measurements into an end-to-end verification framework remains a technical challenge.

[0003] Various stakeholders in the diamond supply chain require methods that can robustly associate polished stones with their rough precursors, detect unauthorized substitutions, and confirm origin claims. Mines, cutting houses, grading laboratories, and retailers all benefit from streamlined, automated verification processes that reduce reliance on manual record-keeping and expert judgment. In particular, there is a growing need for scalable solutions capable of correlating spectral data, imaging data, dimensional measurements, and other physical attributes to deliver reproducible proof of lineage. Such solutions would enhance supply-chain transparency, support compliance audits, and mitigate financial risks associated with forged certificates or misrepresented products.

[0004] Conventional approaches often depend on trained gemologists examining inclusion patterns, color grading, and cut characteristics, or on isolated spectroscopic assays that provide limited chemical information. Manual inclusion mapping is labor-intensive and may fail on small or inclusion-free stones, while visual assessments are subject to variability in operator expertise and imaging conditions. Single-gate spectroscopic profiles, though useful for confirming certainimpurity signatures, can yield similar results for stones from the same deposit, reducing discriminatory power. Moreover, these techniques generally do not account for the transformations that occur during cutting and polishing — such as shifts in inclusion positions, creation of surface features, or changes in mass and dimensions limiting their efficacy for parent-child matching.

[0005] Reliance on standalone verification strategies can result in false positives when unrelated stones appear spectrally similar or false negatives when expected matches are disrupted by processing effects. Manufacturer-declared provenance offers no independent countermeasure against mislabeling or deceit, and existing systems often overlook multidimensional data such as fluorescence patterns and three-dimensional form. As a result, there remains a need for a more comprehensive, science-based framework that integrates diverse measurement modalities and data-driven analytics to achieve reliable, scalable traceability across the diamond’s lifecycle.

[0006] Therefore, there is an increasing need to provide a reliable way to verify the origin of diamonds, ensuring compliance with government regulations and requirement, and providing a method for parent-child match between rough and cut diamonds to offer greater transparency to consumers. Further, under the current condition of considerable incentive lying with parties involved in manufacturing to switch a high value natural diamond and replacing it with a low value diamond, some science backed objective is need of the hour.SUMMARY

[0007] This summary is provided to introduce concepts of the invention related to a system and method for diamond verification and traceability, as disclosed herein. This summary is neither intended to identify essential features of the invention as per the present invention nor is it intended for use in determining or limiting the scope of the invention as per the present invention.

[0008] In accordance with an embodiment of the present disclosure, a cloud-based server device for diamond verification and traceability is provided. The device comprising: a database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; a processing unit communicatively coupled to the database, the processing unit configured to: (i) receive, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client; (ii) retrieve, from the database, the stored diamond data records corresponding to the ERP ID provided by the client; (iii) for each pair of corresponding data types from the stored diamond data records and the client data, generate a similarity vector, each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other nonpeak features; iv) generate, using a classifier model, a score for each similarity vector; (v) generate, using a probabilistic model, a confidence score from the scores generated by the classifier model; and (vi) generate a diamond verification and traceability report based on the confidence score.

[0009] In accordance with the present embodiment, each data record comprises at least Fourier Transform Infrared (FTIR) spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades of the diamond.

[0010] In accordance with the present embodiment, the client data comprises the ERP ID, a stage identifier, and a plurality of FTIR spectral data, PL spectral data, images, weight values, and associated feature data of the client diamond.

[0011] In accordance with the present embodiment, the peak features comprise at least X-location, prominence, and area; the non-peak features comprise at least band integral, Euclidean distance, cosine similarity, band measurements, and peak counts; and the other non-peak features comprise at least color, fluorescence, weight, dimensional measurement, color grade, and clarity grade.

[0012] In accordance with the present embodiment, each value in the similarity vector is a score between 0 and 1 indicating the degree of similarity for the corresponding feature, where a score close to 1 indicates a match and a score close to 0 indicates no match.

[0013] In accordance with the present embodiment, the confidence score is a value between 0 and 1, where a score close to 1 indicates a match and a score close to 0 indicates a non-match.

[0014] In accordance with the present embodiment, the diamond verification and traceability report comprises at least one of: (a) a determination and a lineage history of the client diamond when the confidence score is equal to or more than a first predetermined threshold (t); (b) INCONCLUSIVE when the confidence score lies between a second predetermined threshold (t low) and the first predetermined threshold (t); and (c) ALERT when the final confidence score is below the second predetermined threshold (t low).

[0015] In accordance with the present embodiment, the processing unit is further configured to store, upon the confidence score being equal to or more than the first predetermined threshold (t), the received client data of the client diamond in the database linked to the ERP ID.

[0016] In accordance with the present embodiment, the stored diamond data record and the client data each comprise a plurality of FTIR spectra data, PL spectra data, and corresponding image data, each captured by rotating the diamond through distinct orientations.

[0017] In accordance with the present embodiment, the processing unit is further configured to enroll a new diamond data record in the database when no record is present for the provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

[0018] In accordance with the present embodiment, the processing unit is further configured to provide a REFER result indicative of rescan the client diamond in the diamond verification and traceability report when the final confidence score is INCONCLUSIVE.

[0019] In accordance with the present embodiment, the corresponding data types comprise pairs of matched categories of diamond-related data from the stored diamond data and the client data, each pair including one data type from the stored data and its corresponding type from the client data, the categories including FTIR spectral data, PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades, each pair being compared to generate the similarity vector comprising the plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non -peak features.

[0020] In accordance with the present embodiment, the processing unit is configured to extract at least one peak feature and at least one non-peak feature from each pair of corresponding data types comprising FTIR spectral data and PL spectral data, and to use the extracted features in the generation of similarity vectors.

[0021] In accordance with another embodiment of the present disclosure, a client device is provided. The client device comprising: a plurality of data acquisition devices, each configured to: (i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond; (ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unitconfigured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to a cloud-based server device; and receiving, from the cloudbased server device, a diamond verification and traceability report for the client diamond.

[0022] In accordance with the present embodiment, the client device further includes a rotation fixture configured to hold the client diamond and rotate the client diamond through a plurality of distinct orientations.

[0023] In accordance with the present embodiment, the processing unit is further configured to preprocess the FTIR and PL data prior to transmitting to the cloudbased server device.

[0024] In accordance with another embodiment of the present disclosure, a method for diamond verification and traceability is provided. The method comprising: (i) receiving, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client; (ii) retrieving, from a database, a stored diamond data records corresponding to the ERP ID provided by the client; (iii) generating, by a processing unit, a similarity vector for each pair of corresponding data types from the stored diamond data and the client data, wherein each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non -peak features; (iv) generating a score for each similarity vector using a classifier model; (v) generating a confidence score from the scores generated by the classifier model using a probabilistic model; and (vi) generating, by the processing unit, a diamond verification and traceability report based on the confidence score generated by the probabilistic model.

[0025] In accordance with the present embodiment, the database configured to store, for each diamond assigned an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; wherein each data record comprises at least FTIR spectral data, photoluminescence (PL) spectral data,images, weight values, dimensional measurements, color grades, and clarity grades of the diamond.

[0026] In accordance with the present embodiment, the client data comprises the ERP ID, a stage identifier, and a plurality of FTIR spectral data, PL spectral data, images, weight values, and associated feature data of the client diamond.

[0027] In accordance with the present embodiment, the peak features comprise at least X-location, prominence, and area; the non-peak features comprise at least band integral, Euclidean distance, cosine similarity, band measurements, and peak counts; and the other non-peak features comprise at least color, fluorescence, weight, dimensional measurement, color grade, and clarity grade.

[0028] In accordance with the present embodiment, the diamond verification and traceability report comprises at least one of: (a) a determination and a lineage history of the client diamond when the final confidence score is equal to or more than a first predetermined threshold (t); (b) INCONCLUSIVE when the final confidence score lies between a second predetermined threshold (t low) and the first predetermined threshold (t); and (c) ALERT when the final confidence score is below the second predetermined threshold (t low).

[0029] In accordance with the present embodiment, the method further comprises: when the confidence score being equal to or more than the first predetermined threshold (t), storing the received client data of the client diamond in the database linked to the ERP ID.

[0030] In accordance with the present embodiment, the stored diamond data record and the client data each comprise a plurality of FTIR spectra, PL spectra, and images, each captured by rotating the diamond through distinct orientations.

[0031] In accordance with the present embodiment, the method further comprises enrolling a new diamond data record in the database when no record is present forthe provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

[0032] In accordance with the present embodiment, the method further comprises providing a REFER result indicative of rescan in the diamond verification and traceability report when the confidence score is INCONCLUSIVE.

[0033] In accordance with the present embodiment, the corresponding data types comprise pairs of matched categories of diamond-related data from the stored diamond data and the client data, each pair including one data type from the stored data and its corresponding type from the client data, the categories including FTIR spectral data, PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades, each pair being compared to generate the similarity vector comprising the plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non -peak features.

[0034] In accordance with the present embodiment, the method further comprises extracting at least one peak feature and at least one non-peak feature from each pair of corresponding data types comprising FTIR spectral data and PL spectral data, and to use the extracted features in the generation of similarity vectors.

[0035] In accordance with yet another embodiment of the present disclosure, a system for diamond verification and traceability is provided. The system comprising a cloud-based server device and a client device. The cloud-based server device, the client device comprising: a database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; a processing unit communicatively coupled to the database, the processing unit configured to: (i) receive, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client; (ii) retrieve, from the database, the stored diamond data records corresponding to the ERP ID provided by the client; (iii) for each pair of corresponding data types from the stored diamond data records and the client data,generate a similarity vector, each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, nonpeak features, and other non-peak features; (iv) generate, using a classifier model, a score for each similarity vector; (v) generate, using a probabilistic model, a confidence score from the scores generated by the classifier model; and (vi) generate a diamond verification and traceability report based on the confidence score; and the client device, the client device comprising: a plurality of data acquisition devices, each configured to: (i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond; (ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to the cloud-based server device; receiving, from the cloudbased server device, a diamond verification and traceability report for the client diamond.BRIEF DESCRIPTION OF ACCOMPANYING DRAWINGS

[0036] The detailed description is described with reference to the accompanying figures. Some implementations of the device(s), in accordance with the present subject matter, are described by way of examples, and with reference to the accompanying figures, in which:

[0037] FIG. 1 illustrates a block diagram of a system for diamond verification and traceability, according to an exemplary implementation of the present disclosure;

[0038] FIG. 2 illustrates a flowchart for a diamond verification and traceability method, according to an exemplary implementation of the present disclosure;

[0039] FIG. 3 illustrates a process flow showing an example data flow for the diamond verification and traceability method, according to an exemplary implementation of the present disclosure;

[0040] FIG. 4A and 4B illustrate example graphs of FTIR spectra for the stored diamond and the client diamond, indicating its most prominent peak, respectively, in accordance with an exemplary implementation of the present disclosure;

[0041] FIG. 5A and 5B illustrate example graphs showing the peak window and baseline for the stored and the client diamonds, respectively, providing a focused view of the spectral region around the most prominent peaks of FIGs. 4A and 4B, in accordance with an exemplary implementation of the present disclosure;

[0042] FIG. 6 illustrates a block diagram for a cloud-based server device, in accordance with an exemplary implementation of the present disclosure; and

[0043] FIG. 7 illustrates a block diagram for a client device, in accordance with an exemplary implementation of the present disclosure.DETAILED DESCRIPTION

[0044] The present disclosure describes a system and method for diamond verification and traceability.

[0045] In the following description, for the purpose of explanation, specific details are set forth in order to provide an understanding of the present disclosure. It will be apparent, however, to one skilled in the art that the present disclosure may be practiced without these details. One skilled in the art will recognize that embodiments of the present disclosure, some of which are described below, may be incorporated into a number of systems.

[0046] However, the systems and methods are not limited to the specific embodiments described herein. Further, structures and devices shown in the figures are illustrative of exemplary embodiments of the presently disclosure and are meant to avoid obscuring of the presently disclosure.

[0047] It should be noted that the description merely illustrates the principles of the present invention. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described herein, embody the principles of the present invention. Furthermore, all examples recited herein are principally intended expressly to be only for explanatory purposes to help the reader in understanding the principles of the invention and the concepts contributed by the inventor to furthering the art and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass equivalents thereof.

[0048] Diamonds are generally found deep within the Earth crust, typically between 120-200 kilometres (75-120 miles) below the surface, under conditions of extreme heat (approximately 900 to 1,300 degrees Celsius) and high pressure (45-60 kilobars). In this environment, carbon crystallizes into diamonds over billions of years, with the carbon necessary for diamond formation coming from the melting of existing rocks in the Earth's upper mantle, which has a rich supply of carbon atoms.

[0049] Diamonds are transported to the Earth’ s surface by magma that rises through volcanic pipes known as kimberlite and lamproite pipes. These eruptions can be violent, leading to the formation of carrot-shaped volcanic structures that serve as the primary source of diamonds. Diamonds are found in two main types of deposits: primary deposits, located in kimberlite and lamproite pipes where diamonds originally form, and secondary deposits, which result from erosion, transporting diamonds away from their primary sources to riverbeds, beaches, and ocean floors. Both types of deposits are critical for diamond mining and extraction.

[0050] Diamonds are renowned for their exceptional hardness, rated a 10 on the Mohs hardness scale, meaning they can only be scratched by other diamonds. They have a specific gravity of 3.52, making them 3.52 times denser than water. Their optical properties grant them high brilliance, with a high refractive index (2.419) that allows them to reflect and refract light beautifully. Additionally, diamonds often contain atomic inclusions and visible defects, such as nitrogen or boron impurities, which can affect their colour and spectral properties. These imperfections provide each diamond with a unique signature, similar to a fingerprint, which can be analysed for identification purposes.

[0051] Today, diamonds are mined in around 25 countries worldwide, with significant deposits found in regions such as Africa, Siberian Russia, and Australia. Although South Africa has historically been associated with diamond mining, diamonds are a global resource found in ancient geological formations that host the kimberlite and lamproite pipes bringing diamonds to the surface.

[0052] It is to be noted here that every diamond contains atomic imperfections, which can be analysed using an FTIR spectrum. These imperfections remain unchanged when the diamond is polished, preserving characteristics from its previous rough stage. Additionally, most diamonds exhibit some level of UV-VIS light absorption and fluorescence. As the diamond undergoes cutting and polishing, its mass may decreases, and similarly, its dimensions may also reduce or stay unchanged during the process.

[0053] In recent years, figuring out where diamonds come from and tracking them has become a major concern in the gemstone industry. This is largely due to government rules and the growing demand from consumers for clear information. For example, it crucial to accurately identify the source of a diamond to comply with government legislation and regulation to meet their requirements, as some countries may not allow the diamond which has origin from some specific country.

[0054] Traditional methods for verifying the origin of diamonds and establishing parent-child relationships between rough and polished stones rely heavily onmanual inspection by expert gemologists. These methods often involve analyzing inclusions or other visible characteristics, which are prone to human error, tampering, and scalability issues. Furthermore, such approaches fail to address scenarios where diamonds lack visible inclusions, suffer accidental damage during processing, or exhibit changes in inclusion patterns due to cutting and polishing. These limitations leave the supply chain vulnerable to forgery, substitution, and misrepresentation, undermining consumer trust and regulatory compliance. Existing solutions, such as those relying solely on FTIR spectral data or singlemodality analysis, are insufficient to counter false claims of lineage, as diamonds from the same mine often share similar spectral characteristics, leading to false positives.

[0055] There is also a pressing need to track diamonds from the mine to the market. The existing methods are unreliable for determining the source and for matching rough diamonds with their cut counterparts. This aspect of diamond traceability is the "parent-child" matching process, which refers to the connection between rough diamonds (the "parent") and their polished counterparts (the "child"). This matching is critical for ensuring that the diamonds sold to consumers are ethically sourced and not from conflict regions. Current techniques struggle with this matching process, making it difficult to establish a clear lineage from rough to cut diamonds.

[0056] Adding to these challenges is the issue of forgery claims by manufacturers. Some companies may misrepresent the origin of their diamonds, falsely claiming that their stones are ethically sourced or from conflict-free regions. This deception not only undermines consumer trust but also complicates regulatory compliance and efforts to promote ethical practices within the industry. Further to this, under the current condition of considerable incentive lying with parties involved in manufacturing to switch a high value natural diamond and replacing it with a low value diamond, some science backed objective is need of the hour.

[0057] Therefore, there is a growing demand for a dependable way to confirm the origins of diamonds, verify the diamond, and matching of parent-child diamond.This would not only help meet regulations but also facilitate the parent-child matching process and reduce the risk of forgery claims.

[0058] The present disclosure provides a system and method for diamond verification and traceability, which addresses these shortcomings by introducing a comprehensive, multi-modal, and multi-stage approach for diamond verification and traceability. Unlike conventional methods, the described system integrates diverse data types including a Fourier Transform Infrared (FTIR) spectral data, a photoluminescence spectral data, UV-VIS absorption profiles, fluorescence imagery, three-dimensional shape models, weight, dimensions, color grade, and clarity grade. By utilizing machine learning models trained on these heterogeneous data sets, the approach performs automated parent-child matching and verification at each stage of the diamond's lifecycle. This facilitates robust validation of a diamond's origin and lineage, even when a rough diamond is cut into multiple polished pieces. The approach further incorporates data acquisition devices and cloud-based processing to enable real-time data collection, secure storage, and seamless access for authorized stakeholders.

[0059] The presently described system enhances the accuracy and reliability of diamond verification and traceability. By employing advance technique to extract and analyze features such as peak features, non-peak features, other non-peak features, fluorescence patterns, and 3D model mismatches, the system ensures that discrepancies are flagged and investigated promptly. This multi-layered approach reduces the risk of forgery, supports regulatory compliance, and provides consumers with improved transparency regarding the ethical sourcing of diamonds.

[0060] Accordingly, the present disclosure provides a system and method for diamond verification and traceability to determine the origin of diamonds, verify the diamond, and establishing connections between cut pieces of the diamond and reduce the risk of forgery claims.

[0061] These and other advantages of the present disclosure would be described in greater detail in conjunction with FIGS. 1-7 in the following description. It shouldbe noted that the description merely illustrates the principles of the present disclosure. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described herein, embody the principles of the present disclosure and are included within its scope. Furthermore, all examples recited herein are intended only to aid the reader in understanding the principles of the present subject matter. Moreover, all statements herein reciting principles, aspects and implementations of the present subject matter, as well as specific examples thereof, are intended to encompass equivalents thereof.

[0062] FIG. 1 illustrates a block diagram of a system 100 for diamond verification and traceability, in accordance with an example implementation of the present disclosure. The system 100 is configured to track and verify diamonds throughout their lifecycle, from mining to retail, by integrating cloud-based data processing, and client-server communication. Although the system 100 is depicted to include few components, a cloud-based server device 102 and a plurality of client devices 108 arranged in a particular arrangement in the present disclosure, it should not be taken to limit the scope of the present disclosure. The system 100 comprises a plurality of client devices 108 and a cloud-based server device 102, which includes at least a processor 104 and a database 106. It is to be noted here that there can n number of client devices 108 communicating with the cloud-based server device 102, therefore the number of client devices 108 is not limited thereto. The abovedescribed components of the system 100 are merely illustrative.

[0063] Various components of the system 100, the server device 102 and the client device 108 are communi cably coupled to each other via a network. The network may include wired or wireless communication protocols. In an embodiment, the network may include, but not limited to a local area network LAN, a wide area network WAN (e.g., the Internet), a mobile network (for e.g., GSM (Global System for Mobile Communication), GPRS (General Packet Radio Service), EDGE (Enhanced Data for Global Evolution), etc.).

[0064] The client device 108 is configured to facilitate diamond verification and traceability by interacting with a cloud-based server device 102 and other components of the system 100. The client device (108) can be deployed at any location worldwide. The client device 108 is configured to collect, process, and transmit diamond-related data. The client device can be employed at various stages of the diamond supply chain, such as mining, cutting, polishing, and retail.

[0065] The client device 108 comprises a user interface, a processor, a memory, and a display. The user interface enable interaction between the client and the client device 108. Through the user interface, users can input data, such as an ERP ID of a client diamond against which the client claims lineage and a stage identifier of the client diamond, and access diamond verification and traceability reports generated by the server device 102.

[0066] The processor, which is responsible for executing various computational tasks. These tasks include collecting diamond data including FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, colour grades, and clarity grades of the diamond. These diamond data are collected via various devices capable of collecting the FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, colour grades, and clarity grades. Each of these data are collected from different orientation by rotating the diamond. The processor is further configured to transmit these diamond data (hereafter referred to as a client diamond data) to the cloud-based server device 102 and receive a diamond verification and traceability report of the client diamond.

[0067] The memory is included in the client device 108 to store data and instructions required for the operation of the client device 108. The memory temporarily holds the diamond-related data collected by the client device 108, such as FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades. This data issubsequently processed by the processor and transmitted to the server device 102 for further analysis.

[0068] The client device 108 is coupled with a plurality of acquisition devices, each capable of collecting data such as FTIR spectroscopy data and PL spectroscopy data, color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of a diamond. These acquisitions device can be remotely connected to the client device or can be embedded with the client device, thereby the scope is not limited to thereto.

[0069] In an exemplary implementation, the client device 108 comprises a plurality of data acquisition devices, each configured to: (i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond; (ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and (iii) a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to a cloud-based server device; and receiving, from the cloud-based server device, a diamond verification and traceability report for the client diamond.

[0070] The cloud-based server device (102) includes at least a processing unit (104) and a database (106). The database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain.

[0071] The ERP ID is a unique identifier assigned to each diamond within the supply chain management system. The ERP ID serves as a reference key that links all data records, measurements, and lifecycle events associated with a particulardiamond across different stages in the diamond supply chain, including mining, cutting, polishing, and retail. This identifier is typically generated and managed by an enterprise resource planning (ERP) system, ensuring that each diamond can be accurately tracked, verified, and traced throughout its journey from origin to final sale. The ERP ID enables seamless integration of data from various sources and supports robust traceability, auditability, and compliance with industry and regulatory standards.

[0072] For each diamond, the database (106) stores multiple data records under a unique ERP ID, capturing essential characteristics of the diamond. These data records include Fourier Transform Infrared (FTIR) spectral data, photoluminescence (PL) spectral data, high-resolution images, weight measurements, dimensional data, color grades, and clarity grades. In an exemplary implementation, each of these data records are recorded from at least eight different orientations of the diamond. This multi-angle data collection ensures a comprehensive and accurate representation of the diamond’s physical and optical characteristics, enabling reliable verification and traceability throughout the diamond supply chain.

[0073] The processing unit (104) is communicatively coupled to the database (106) to store and access data stored in therein. The processing unit 104 is configured to process diamond-related data, perform feature extraction, generate similarity vectors, and use machine learning models for verification and traceability of the client diamond and provide a diamond verification and traceability report. The detailed operations performed by the processing unit 104 will be described in accordance with FIG. 2.

[0074] FIG. 2 shows a flowchart illustrating a method for diamond verification and traceability. The method involves a series of steps that utilize client data, stored diamond records, and advanced computational models to generate a comprehensive verification and traceability report for a diamond. The process ensures the authenticity and lineage of the diamond by leveraging feature extraction, machinelearning, and probabilistic decision-making. The method steps are performed by the processing unit 104 of the cloud-based server device 102 and includes the following steps:

[0075] At Step S201, receiving, from the client device 108, client data related to a client diamond to be verified against an ERP ID provided by a client.

[0076] In this step, the server device 102 receives from the client device 108, client data (also referred to interchangeably as client diamond data) related to a client diamond and an ERP ID against which the client diamond data need to be verified and matched. The client data further comprises a stage identifier, and a plurality of FTIR spectral data, PL spectral data, images, and weight values of the client diamond. In one example, when the client claims that the client diamond is derived from, or is the same as, a specific previously recorded diamond, the client provides the ERP ID corresponding to that specific diamond so that the server device 102 can retrieve the associated stored data records and perform a comparative analysis for verification, match, and lineage determination. For example, when a client claims that a diamond is a polished piece derived from a specific rough diamond, they provide the ERP ID of the specific rough (parent) diamond. Using this ERP ID, the server retrieves the stored data of the parent diamond from the database. It then compares this data with the client’s submitted diamond data to determine if there is a match. Through this comparison, the system can establish a parent-child relationship between the two diamonds, helping verify the diamond’s origin and traceability.

[0077] At step S202, retrieving, from a database, a stored diamond data records corresponding to the ERP ID provided by the client.

[0078] In this step, once the client data is received, the system retrieves stored diamond data records from a database 106 corresponding to the provided ERP ID. The database 106 contains a comprehensive set of data records for each diamond having the ERP ID. These records include FTIR and PL spectral data, images, weight values, dimensional measurements, color grades, clarity grades, and othermetadata. In an exemplary embodiment, the database stores a plurality of FTIR and PL spectral data and images, each captured from distinct orientation by rotating the diamond. This step serves as the foundation for comparing the client diamond data with the existing records in the database. In yet another exemplary implementation, the database stores at least eight data of FTIR and PL spectral data and images, each captured from eight distinct orientation by rotating the diamond.

[0079] In an embodiment, the database may contain data records for each diamond, which are collected at various stages of the diamond supply chain, such as mining, cutting, polishing, and retail. During the mining stage, a rough diamond is obtained from mine. The mining owner sells this rough diamond to a manufacturer. Upon purchase of the rough diamond, a KP certificate, sales invoice, weight is obtained of the rough diamond from the mining owner. Further, the rough diamond is first associated with key data such as its weight and the current stage of processing (whether rough, sawn, makeable, block, or polished). These data are linked to the obtained rough diamond and uploaded to the database link to the respective ERP ID.

[0080] At step S203, generating a similarity vector for each pair of corresponding data types from the stored diamond data and the client data, wherein each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non -peak features.

[0081] In this step, after retrieving the stored diamond data records from the database 106, the processing unit 104 generates a similarity vector for each pair of corresponding data types from the stored diamond data and the client diamond data.

[0082] The corresponding data types comprise pairs of matched categories of diamond-related data from the stored diamond data and the client diamond data. Each pair including one data type from the stored diamond data and its corresponding type from the client diamond data, the categories includes FTIR spectral data, PL spectral data, images, weight values, dimensional measurements,color grades, and clarity grades. For each matched pair of data types, the processor generates a similarity vector. This similarity vector is composed of a plurality of values, with each value representing a specific feature parameter. These parameters includes peak features, non-peak features, and other non-peak features for each matched pair of data type. In an embodiment, the value representing a specific feature parameter may depending on the nature of the data being analyzed.

[0083] In accordance with the present embodiment, the term peak features refers to specific characteristics derived from each matched pair of data type such as FTIR spectral data that assist in identifying and comparing diamonds. These include, but are not limited to, the X-location (position) of the peak, prominence, intensity, and width from spectral data. The term non-peak features includes parameters such as band integral, Euclidean distance, cosine similarity, band measurements, peak counts, peak width at half height, and stability ratios derived from spectral data or the diamond data. The term other non-peak features refers to non-spectral attributes such as color, fluorescence, weight, dimensional measurements, color grade, and clarity grade. Although these are exemplary parameters, the system is not limited thereto and may incorporate additional relevant features as required to enhance the accuracy of diamond verification and traceability.

[0084] In an exemplary embodiment, the operation of generating a similarity vector for each pair of corresponding data type from the stored diamond data and the client diamond data is described in accordance with FIG. 3, FIGs 4A-B, and FIGs 5A-B. FIG. 3 provides a process flowchart for comparing a FTIR data of the client diamond with the stored diamond data, illustrating how similarity vectors are generated and used for verification. This process is exemplified using the FTIR spectral graphs shown in FIG. 4A, FIG. 4B, FIG. 5 A, and FIG. 5B, which represent the spectral data of the stored diamond and the client diamond, respectively.

[0085] For the purpose of simplified explanations, the FTIR spectral data is considered from the other data types of the diamond data. Once the FTIR spectral data is received from the client diamond and retrieved from the database for thestored diamond data, the processing unit 104 preprocess the FTIR spectra. It is to be noted here that the processing unit receives one or more FTIR spectral scan. In one embodiment, the processing unit receives at least eight FTIR spectral scans from the client corresponding to the client diamond and the stored diamond data. These eight FTIR spectral scans are captured from eight distinct orientations and are combinedly referred to as FTIR data. After receiving these FTIR spectral scans, the processing unit performs preprocess and normalization operation on each FTIR scan so that one or more feature parameters can be extracted from each FTIR scan. The processing unit preprocess the FTIR spectral data and PL spectral data prior to feature extraction, the preprocessing comprising: (a) applying a moving-average smoothing filter to reduce high-frequency noise; (b) performing baseline estimation using a polynomial fit over the smoothed signal and subtracting the baseline to remove instrument drift and broad background; (c) applying non-negativity clipping to the resulting data; (d) performing area normalization so that the total value of the spectrum is adjusted to one, enabling comparability across instruments and exposures. This preprocessing operation on the FTIR spectrum ensures that data is comparable.

[0086] It is to be noted that the FTIR spectrum may represent either the entire spectral range or one or more sub-ranges corresponding to known regions associated with diamond impurities (such as boron) and characteristic peak ranges, for example, between 1100 and 1400.

[0087] Refer FIG. 3, at step S301, after acquisition of the FTIR spectra for both the client diamond and the stored diamond. In an example, at least eight FTIR spectra for each of the client diamond and the stored diamond are obtained by the processing unit. Herein, FIG. 4 A shows one FTIR spectrum of the stored diamond, indicating its most prominent peak, while FIG. 4B shows one FTIR spectrum of the client diamond, indicating its most prominent peak. FIG. 5A and FIG. 5B further illustrate the peak window and baseline for the stored and the client diamonds, respectively, providing a focused view of the spectral region around the most prominent peaks of FIGs. 4 A and 4B.

[0088] To generate the similarity vector, the following steps are performed by the processing unit 104:

[0089] Preprocessing: both spectra are preprocessed to ensure comparability. This includes baseline correction, normalization, and alignment of the wavenumber axis so that the area under the peak is computed after baseline correction and both spectra use the same units.

[0090] Feature Extraction: For each FTIR spectrum (stored diamond and client diamond), at least one peak feature, at least one non-peak feature, and at least one other non-peak feature is extracted. Herein below example, the operation for peak features extraction is described. It is noted here that the same operation can be performed for extraction of non-peak features and other non-peak features from the dataset of the client diamond and the stored diamond.

[0091] For each FTIR spectrum of the stored and the client diamond, at least three peak features are extracted from the most prominent peak:• xA, xBX-location (wavenumber, in cm ') of the most prominent peak for the stored (xA) and client (xfi) diamond.• PA’ PB '- Prominence (height above local valley) of the most prominent peak for the stored (pA) and client (pB) diamond.• AA, AB: Area under the most prominent peak (after baseline correction) for the stored (AA) and client (AB) diamond.

[0092] After extracting peak features, for each peak feature, a similarity value is calculated as follows:• X-location similarity, calculated as sx= max• prominence similarity, calculated as sn= 1 -]Pa Pb]and max (PA,PB,£)• area similarity, calculated as sA= 1 -IAa AbImax (AA’AB’£)

[0093] where xA, pA, AAare features from the stored diamond data, xB, pB, ABare features from the client data, and Rxis the instrument or dynamic range; and £ is a small value to avoid division by zero.

[0094] From the above, for each peak feature, a similarity score is obtained. Each similarity score is a value between 0 and 1, where a score close to 1 indicates high similarity and a score close to 0 indicates low similarity. These values are aggregated to generate a single similarity vector into a feature matrix representing inter-sample spectral relationships between two corresponding data type, as in present case is for corresponding FTIR spectrum for the client and stored diamonds.

[0095] Similarity Vector Construction: The similarity vector for the pair of spectra is constructed as:[S%, Sp, S / |]

[0096] This similarity vector quantifies the degree of similarity between the client diamond and the stored diamond for the most prominent peak in their FTIR spectra. It is noted here that, the above example is for peak feature for single pair of FTIR spectra for the stored diamond and the client diamond, a similar technique can be used for generating one or more similarity value for each non-peak features and other non-peak features and then a similarity vector is constructed, which includes values of the peak features, non-peak features and other non-peak features.

[0097] In an example, the similarity vector having peak features for the FTIR spectrum of FIG. 4 A and 4B and a peak window and baseline of FIG. 5 A and 5B is generated as follows:• xA= 2076.05349, xB= 1206.36393 cm1• pA= 1.008093, pB= 0.947307• AA= 292.446924, AB= 170.350872

[0098] The similarity scores are calculated as:869.68956Sr= 1 — 0.845600Sp — 1 £2^ « 0.93971.0080930.5825

[0099] Thus, the similarity vector for each pair of FTIR spectra of the client diamond and the stored diamond is:[0.84183,0.93970,0.58250]

[0100] It is noted here that, the above example describes how the similarity vector is generated for a single pair of the FTIR spectra of the stored diamond and the client diamond. However, in one embodiment, there are total eight distinct FTIR spectra of the stored and the client diamonds. Accordingly, to have best possible match results, the processing unit 106 compare each FTIR spectra of the client diamond with each FTIR spectra of the stores diamond. For example, there are eight FTIR spectra from the stored diamond and eight FTIR spectra from the client diamond, the maximum total similarity vectors can be generated is 64 similarity vectors.

[0101] Similarly, for the processing unit 104 generate similarity vector for all eight pair of FTIR spectra of the client diamond and the stored diamond. Accordingly, after computing for all eight pair of FTIR spectra, 64 similarity vectors are generated by the processing unit 104, see step S302.

[0102] By generating these similarity vectors, the server device can quantify how closely the client diamond data aligns with the stored records. This process enables the device 102 to assess the authenticity of the diamond and determine whether it matches or is derived from the diamond associated with the given ERP ID. The resulting similarity scores form the foundation for further analysis using machine learning models, which ultimately produce a confidence score and generate a comprehensive verification and traceability report.

[0103] At step S303, the processing unit 104 uses a trained classifier model to further analyze each of the 64 similarity vectors and generate a classification score corresponding to each vector. Accordingly, the trained classifier model generates 64 scores, see step S304. The score generated by the model is a value between 0 and 1, where a score close to 1 indicates high similarity and a score close to 0 indicates low similarity between the stored diamond data and the client diamond data.

[0104] In an example, when a similarity vector including a plurality of values, each corresponding to the peak features, non-peak features and the other non-peak features, the trained classifier model compute these values of the similarity vector and generate a single score between 0 and 1, which represents degrees of closeness.

[0105] At step S305, the 64 classification scores generated at the step 304 are then further computed to generate a single final confidence score by employing a probabilistic model or a weighted reducer. Here the score is a value between 0 and 1, where a score close to 1 indicates high similarity and a score close to 0 indicates low similarity between the stored diamond data and the client diamond data.

[0106] At step S306, after obtaining the final confidence score, the processing unit 104 compare the score with predetermined thresholds to generate a diamond verification and traceability report. The diamond verification and traceability report is indicative of whether the client diamond. For example, the processing unit 104 generates any one of the following:(a) a determination and a lineage history of the client diamond when the confidence score is equal to or more than a first predetermined value (t);(b) INCONCLUSIVE when the confidence score lies between a second predetermine value (t low) and t; and(c) ALERT when the final confidence score is below the t low.

[0107] In an embodiment, the first predetermined and the second predetermined thresholds are values between 0 and 1 and can be preset by an administrator of the cloud-based server device 102. In an example, the first predetermined value can be 0.99, representing a high confidence level for confirming that the client diamond and the stored diamond are a definite match. The second predetermined value can be 0.75, representing a moderate confidence level indicating that the client diamond is not related or derived from the stored diamond, such as does not belonging to the same lineage or source.

[0108] It is to be noted that, although the above example in FIG 3. is described with reference to FTIR data, the same technique can be applied to other types of diamond-related data, such as PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades parameters, for determining similarity between the stored diamond data and the client diamond data. In another exemplary embodiment, each datatype is referred to as a gate, that is a gate is denoted as a data modality like FTIR, PL spectrum, images, weight values, dimensional measurements, color grades, and clarity grades. For example, gate 1 corresponds to FTIR data analysis, gate 2 corresponds to PL data analysis, gate 3 corresponds to weight analysis, and the like. Accordingly, in the non-limiting scope, the server device generates a similarity vector for each data type corresponding to each gate. For example, if there are eight FTIR scans from the client diamond and the stored diamond data, the processing unit of the server device will generate 64 similarity vectors only for FTIR data. Where each similarity vector will include plurality of values each corresponding to the feature parameters including peak features, non-peak features, and other non-peak features. Similarly, for every data type multiple of similarity vectors are generated and analyzed to determined the probability of the match between the client diamond and the stored diamond data to establish the match between the parent-and-child diamonds, where the stored diamond data is referred to as parent diamond and the client diamond is referred to as child diamond. It is to be noted that, as the database stores, for each diamonddata, plurality of data records from the diamond supply chain. The diamond data stored at the database can be rough, split, and polished stage diamond.

[0109] Reference is again made to FIG. 2, at step S204, generating a score for each similarity vector using a classifier model.

[0110] In this step, after generating the similarity vectors (S302), the next step is to generate, using a classifier model (S303), a score for each similarity vector (S302). The classifier is trained using a plurality of similarity vectors (S302) generated from pairs of diamond data, with the training process involving generating similarity vectors (S302) from the features parameters extracted from corresponding data types selected from FTIR spectral data, PL spectral data, images, and associated metadata for pairs of diamonds with known relationships. The classifier model (S303) is optimized to generate a probabilistic score between 0 and 1 indicating the likelihood of a match, with the score reflecting a probability of a match between the stored diamond data and the client diamond data. It is to be noted that the detailed operation of the classifier model is explained in step S303 of FIG. 3.

[0111] In an example, the classifier model generates at least one similarity score for each pair of matched data types between the stored diamond data and the client diamond data. For instance, when eight FTIR data records of the client diamond are compared with eight FTIR data records of the stored diamond, the classifier model generates sixty -four (64) similarity scores. Each score is a value between 0 and 1, where a value closer to 1 indicates a higher degree of similarity, and a value closer to 0 indicates a lower degree of similarity between the corresponding data pairs. Accordingly, a total of 64 scores are generated, each representing the probability of a match between corresponding FTIR data pairs.

[0112] At step S205, generating a confidence score from the scores generated by the classifier model using a probabilistic model.

[0113] In this step, once the classifier model generates sixty-four (64) similarity scores. These individual scores are then aggregated to derive a single final confidence score, which represents the overall probability of match between the stored diamond data and the client diamond data by using a probabilistic model or a weighted average method or the like. In an example, the sixty-four (64) similarity scores are aggregated to derive or generate a single final confidence score by employing a Naive Bayes model. In another example, the sixty-four (64) similarity scores are aggregated by employing a weighted average method or the like can be used to generate the final confidence score.

[0114] At step S206, generating, by the processing unit, a diamond verification and traceability report based on the confidence score generated by the probabilistic model.

[0115] In this step, based on the confidence score, the server device generates a diamond verification and traceability report (S206). The report provides a determination of the diamond’s authenticity and lineage. If the confidence score meets or exceeds a first predetermined threshold (t), the report confirms the match between the client diamond and the stored diamond data and provides a lineage history of the client diamond. If the confidence score falls between the first predetermined threshold (t) and a second predetermined threshold (t low), the result is marked as INCONCLUSIVE, prompting further investigation. If the confidence score is below the second predetermined threshold, the report issues an ALERT, indicating a potential mismatch or forgery.

[0116] Further, the processing unit is configured to provide a REFER result in the diamond verification and traceability report when the final confidence score is inconclusive, thereby recommending a rescan or further investigation.

[0117] In an embodiment, the processing unit 104 is further configured to enroll a new diamond data record in the database 106 when no record is present for the provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

[0118] The method step of FIG. 2 ensures a robust and reliable framework for diamond verification and traceability, addressing challenges such as false positives, false negatives, and the need for scalable, automated solutions in the diamond supply chain. The integration of machine learning models and probabilistic decision-making enhances the accuracy and reliability of the verification process, providing stakeholders with a comprehensive and transparent solution for diamond traceability.

[0119] Further, the cloud-based server device generates a diamond verification and traceability report based on the confidence score, which may include a determination and lineage history, an inconclusive result prompting further investigation, or an alert indicating a potential mismatch or forgery. The architecture supports real-time data transfer, automated verification, and scalable analytics, ensuring robust diamond traceability and compliance with regulatory requirements. By integrating cloud computing, machine learning, and multi-modal data acquisition, the disclosed system provides a reliable, objective, and science- backed solution for diamond verification and traceability from mine to market.

[0120] FIGs. 6 and 7 illustrate a simplified block diagram for a cloud-based server device and a client device, respectively, in accordance with an exemplary implementation of the present disclosure.

[0121] FIG. 6 shows a server device 600 configured to facilitate diamond verification and traceability by leveraging advanced computational and communication capabilities. The server device 600 plays a significant role in the system, enabling the processing, storage, and communication of diamond-related data across various stages of the diamond supply chain. The server device 600 comprises a processor 602, which is responsible for executing various computational tasks. These tasks include processing diamond data, extracting features, generating similarity vectors, and executing machine learning models to verify and trace diamonds. The processor 602 plays a central role in ensuring the accuracy and efficiency of the diamond verification and traceability process.

[0122] A user interface 604 is included in the server device 600 to facilitate interaction with users. The user interface 604 allows users to input data, configure system parameters, and access diamond verification and traceability reports. This interface ensures that the system is user-friendly and accessible to various stakeholders in the diamond supply chain.

[0123] The memory 606 is configured to store data and instructions required for the operation of the server device 600. This includes storing the machine learning models, similarity vectors, and other computational data necessary for diamond verification and traceability. The memory 606 also temporarily holds data during processing, ensuring seamless operation of the server device 600.

[0124] A communication module 608 is included in the server device 600 to enable communication with other devices in the system, such as client devices and external databases. The communication module 608 supports various wired and wireless communication protocols, including the Internet, local area networks (LAN), and mobile networks. This module ensures that the server device 600 can receive data from client devices, process the data, and transmit verification and traceability reports back to the respective devices.

[0125] The server device 600 is designed to integrate seamlessly with the overall diamond verification and traceability system, providing robust computational and communication capabilities. By leveraging the processor 602, user interface 604, memory 606, and communication module 608, the server device 600 ensures the efficient and accurate processing of diamond-related data, supporting the system's goal of ensuring diamond authenticity and traceability throughout the supply chain.

[0126] FIG. 7 shows a block diagram of a client device 700 configured to facilitate diamond verification and traceability by interacting with a cloud-based server device 102 and other components of the system 100. The client device 700 is designed to collect, process, and transmit diamond-related data at various stages of the diamond supply chain, such as mining, cutting, polishing, and retail.

[0127] The client device 700 comprises a processor 702, which is responsible for collecting data from various modules capable of collecting data such as a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond; and acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and preparing the data for transmission to the server device 102. The processor 702 is significant in maintaining the accuracy and efficiency of the data collection and processing operations performed by the client device 700.

[0128] A memory 704 is included in the client device 700 to store data and instructions required for the operation of the client device 700. The memory 704 temporarily holds the diamond-related data collected by the client device 700, such as FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades. This data is subsequently processed by the processor 702 and transmitted to the server device 102 for further analysis.

[0129] The user interface 706 is configured to enable interaction between the client and the client device 700. Through the user interface 706, users can input data, such as the ERP ID and stage identifier, and access diamond verification and traceability reports generated by the server device 102. The user interface 706 ensures that the client device 700 is user-friendly and accessible to various stakeholders in the diamond supply chain.

[0130] The communication module 708 is included in the client device 700 to facilitate communication with the server device 102 and other components of the system 100. The communication module 708 supports various wired and wireless communication protocols, such as the Internet, local area networks (LAN), or mobile networks. This module ensures seamless data transfer between the clientdevice 700 and the server device 102, enabling real-time data processing and verification.

[0131] The client device 700 plays a role in the diamond verification and traceability system 100, as the client device 700 enables the collection and transmission of diamond-related data at different stages of the supply chain. By utilizing the processor 702, memory 704, user interface 706, and communication module 708, the client device 700 facilitates efficient and accurate data collection, processing, and communication, supporting the system's objective of maintaining diamond authenticity and traceability throughout the lifecycle of the diamonds.

[0132] In an exemplary implementation, the client device comprising: a plurality of data acquisition devices, each configured to: collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond, acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail. Each acquisition device is communicatively coupled to the client device, the processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to a cloud-based server device and receiving, from the cloudbased server device, a diamond verification and traceability report for the client diamond.

[0133] In an embodiment, the client device can be located worldwide and act as a measurement station for diamond data acquisition and signal conditioning. The client device comprising: (a) a rotation fixture configured to hold a diamond and step the diamond through a plurality of distinct orientations; (b) an FTIR spectroscopy module configured to acquire FTIR spectral scans of the diamond atthe distinct orientations; (c) a photoluminescence (PL) spectroscopy module configured to acquire PL spectral scans of the diamond at the distinct orientations; (d) an imaging module configured to acquire color images and fluorescence images of the diamond; (e) weight and dimensional sensors configured to measure weight values and dimensional measurements of the diamond; (f) a user interface configured to receive at least an ERP ID and a stage identifier selected from mining, sawn, polished, or retail; and (g) a processing unit operably coupled to the modules and sensors and configured to execute an on-device signal-conditioning pipeline on FTIR and PL scans. The on-device signal-conditioning pipeline on FTIR and PL scans comprising: (i) smoothing using a moving-average filter; (ii) baseline estimation using a polynomial fit and baseline subtraction to reduce instrument drift; (iii) non-negativity clipping; and (iv) area normalization to enable comparability of spectra across instruments and exposures. The processing unit further configured to associate the conditioned measurements with the ERP ID and the stage identifier and transmit the associated data to a cloud-based server device for verification and traceability.

[0134] The rotation fixture is configured to acquire at least eight FTIR scans and at least eight PL scans at eight distinct orientations of the diamond. In an embodiment, the fixture includes a precision-controlled rotation mechanism that incrementally rotates the diamond at predefined angular intervals, for example, every 45 degrees, to ensure uniform and comprehensive data capture across all facets of the diamond. Each orientation enables the collection of spectral information corresponding to a unique surface or crystallographic direction, thereby capturing subtle variations in optical or structural characteristics. This multiorientation data acquisition enhances the robustness and accuracy of subsequent feature extraction and similarity computation processes, allowing reliable verification and traceability of the diamond across different stages of the supply chain.

[0135] The FTIR module captures spectra including a sub-range between 1100 cm1and 1400 cm1for peak analysis. In an embodiment, the FTIR module isconfigured to scan either the entire spectral range or specific sub-ranges corresponding to known impurity-related regions, such as those associated with boron, nitrogen, or hydrogen defects in diamonds. The selected sub-range between 1100 cm ' and 1400 cm1is particularly significant as it contains prominent absorption features indicative of diamond lattice impurities and structural characteristics. By focusing on this range, the module enables precise extraction of peak features such as peak position, prominence, intensity, and width, which are critical for generating reliable similarity vectors and ensuring accurate identification, verification, and provenance tracking of the diamond. In an example, the FTIR module and PL module are a spectrometer and are capable of capturing the FTIR and PL spectrum scan.

[0136] The processing unit is operably coupled to the FTIR and PL modules, as well as to the imaging module, weight and dimensional sensors, and is configured to execute an on-device signal-conditioning pipeline that performs preprocessing on the acquired FTIR and PL scans prior to transmission. This preprocessing pipeline refines the raw spectral data to enhance quality, remove noise, and standardize the measurements across instruments and acquisition conditions. The signal-conditioning pipeline comprises multiple stages, including: (i) smoothing, wherein a moving-average filter is applied to minimize high-frequency noise and improve the signal-to-noise ratio without distorting spectral features; (ii) baseline estimation and subtraction, wherein a polynomial fit is used to estimate and remove baseline drift caused by instrument variation or scattering effects; (iii) nonnegativity clipping, which constrains all corrected intensity values to physically meaningful non-negative values; and (iv) area normalization, wherein each spectrum is scaled such that the total spectral area equals one, thereby ensuring consistent amplitude scaling and enabling reliable device-to-device and exposure- to-exposure comparability. The area normalization renders the total spectral area equal to one to provide device-to-device comparability.

[0137] The processing unit is further configured to associate each conditioned FTIR and PL scan with a corresponding ERP ID and stage identifier, maintaining acontinuous digital linkage between the acquired data and its respective diamond throughout the supply chain. Additionally, for each conditioned scan, the processing unit computes a quality metric derived from the baseline-corrected signal. Scans falling below a preset quality threshold are automatically tagged for reacquisition at the station or for REFER handling at the cloud-based server device (102), ensuring that only high -integrity, preprocessed spectral data are utilized for subsequent verification and traceability operations.

[0138] The imaging module is configured to acquire both color images and fluorescence images of the diamond under controlled illumination conditions. In an embodiment, the imaging module captures fluorescence images under ultraviolet (UV) illumination, enabling visualization of characteristic emission patterns and inclusion features that serve as optical fingerprints for diamond identification and provenance analysis. In addition, the module can be a high-definition camera and it acquires high-resolution color images under white or visible light illumination to document the diamond’s visual attributes, including hue, tone, surface reflections, and facet geometry.

[0139] In another embodiment, the imaging module is also configured to capture images in dark-field and other specialized lighting environments such as bright- field, diffused, or oblique illumination. Imaging in these modes enhances the visibility of inclusions, surface textures, polish marks, and internal growth structures that may not be apparent under standard illumination. The module may employ adjustable lighting geometry, ring-light arrays, or directional light sources to optimize contrast and highlight diagnostic features across various diamond orientations.

[0140] The imaging module further incorporates automatic exposure control, spectral calibration, and focus correction mechanisms to ensure consistent color rendition and fluorescence intensity across multiple acquisitions and instruments. Each captured image whether fluorescence, color, or dark-field is digitally tagged with the corresponding ERP ID and stage identifier, maintaining a verifiable linkagebetween the imaging data and the respective diamond’s lifecycle stage within the verification and traceability framework. The tagged images are subsequently transmitted to the cloud-based server device for centralized analysis, feature extraction, and archival within the diamond’s authenticated digital record.

[0141] The processing unit is further configured to time-stamp each conditioned measurement to ensure temporal traceability of data acquisition events. Upon completion of signal conditioning and image capture, the processing unit packages the resulting datasets including the FTIR and PL measurements, fluorescence and color images, weight, and dimensional data into a single structured transmission addressed to the cloud-based server device 102. This integrated transmission ensures that all data elements corresponding to a particular acquisition event are securely and coherently communicated to the server for verification, and analysis. Furthermore, the processing unit embeds the stage identifier with each measurement such that the server can enroll a new rough diamond record when no database record exists for the ERP ID and the stage identifier indicates rough.

[0142] The stage identifier associated with the client device is persisted with the transmitted data, enabling multi-stage lineage tracking of each diamond within the server-side database indexed by the ERP ID. This facilitates the reconstruction of the diamond’s provenance and transformation history such as transitions between rough, split, and polished stages across multiple acquisition points in the supply chain.

[0143] The client device includes a user interface operably coupled to the processing unit, configured to receive from a user at least an ERP ID and a stage identifier, wherein the stage identifier is selected from mining, sawn, polished, or retail. The user interface is further configured to receive and display a diamond verification and traceability report generated by the cloud-based server device. The report provides the user with a confidence score, match status, and lineage summary, thereby enabling authenticated confirmation of the diamond’s identity, verification status, and traceable lifecycle across all registered handling stages.

[0144] In an example, the client device is configured to the verification report and to display one of MATCH, INCONCLUSIVE with REFER, or ALERT states for the client diamond.

[0145] In yet another embodiment, the present disclosure provides A computer-implemented method for diamond verification and traceability performed by a cloud-based server device, the method comprising:

[0146] (i) receiving from a client measurement station, for a client diamond, conditioned FTIR spectra, conditioned PL spectra, images, weight, and dimensional measurements, together with an ERP ID and a stage identifier;

[0147] (ii) retrieving, from a database indexed by the ERP ID, stored diamond data comprising conditioned FTIR and PL spectra, images, weight, and dimensions for a stored diamond;

[0148] (iii) for each pair of corresponding data types between the client diamond and the stored diamond, computing a similarity vector v whose components are unit-interval scores including at least:(A) FTIR peak-location similarity: sx= max{0, 1 - |xs- x_c| / Rx], where xsand x_c are most-prominent-peak wavenumbers ( cm ') of the stored and client spectra, and Rx>0 is a wavenumber tolerance, and Rxis chosen according to an instrument tolerance in cm1;(B) FTIR prominence similarity: sp= 1 - |ps- p_c| / (max(ps, p_c) + a), where psand p c are baseline-relative prominences and e>0, and a is a small positive constant to avoid division by zero;(C) FTIR peak-area similarity: s_A = min(As, A_c) / (max(As, A_c) + a), where Asand A c are areas under the baseline-corrected peak; and(D) one or more non-peak and other-modality similarities selected from band-integral similarity, cosine-similarity mapping,Euclidean-distance-to-score mapping, peak-count similarity, and image / weight / dimension similarity, each normalized into [0,1];

[0149] (iv) for multiple orientations, pairing client and stored FTIR (and PL) scans to produce a plurality of similarity vectors {vkf } and inputting each Vkf to a trained classifier to obtain per-pair probabilities pkf£[0,l];

[0150] (v) aggregating the per-pair probabilities into a final confidence C E [0,1] using a log-odds sum with non-negative reliability weights {wk}: logit(C) = Sk,£ Wk logit(pkf) + b and C = l / (l+eA{-logit(C)}), where k indexes data gates (FTIR, PL, images, weight, dimensions) and £ indexes orientation pairs;

[0151] (vi) comparing C against two predetermined thresholds t and t low to generate a report with outcomes: MATCH and lineage when C > t; INCONCLUSIVE when t low < C < t; and ALERT when C < t low; and

[0152] (vii) when C > t, storing the received client data in the database linked to the ERP ID.

[0153] The classifier is one of a neural network or a gradient-boosted decision tree, and its output probabilities are calibrated using isotonic regression or Platt scaling prior to aggregation. The aggregation of step (v) is implemented as a naive Bayes combiner under independence assumptions across gates and orientations.

[0154] In an example, the predefined threshold can be predefined as per the requirement. In yet another example, the predefined threshold can be t = 0.99 and t low = 0.75. in yet another embodiment, thresholds t and t low are selected by receiver-operating-characteristic (ROC) analysis on a held-out validation set to target predetermined false-match and false-non-match rates.

[0155] In the report, when the outcome is INCONCLUSIVE to prompt rescan of the diamond, the processing unit is configured to generate a REFER instruction.

[0156] The processing unit of the server device is further configured to performs per-device drift reduction by the baseline-estimation and subtraction and thereby reduces cross-instrument variance before transmission.

[0157] After receiving the client diamond data, the server device analyses the FTIR and PL scan data and is configured to reject any FTIR or PL scan whose quality metric indicates insufficient signal quality and excludes corresponding similarity vectors from aggregation. The quality metric is computed from baseline-corrected spectra and includes at least one of: peak-prominence-to-residual-noise ratio, residual variance, or peak-count sanity bounds, each compared to a preset threshold.

[0158] In an exemplary implementation, eight client FTIR scans are paired with eight stored FTIR scans to form sixty-four FTIR similarity vectors, and the same is performed for PL spectra. Further, the image-based similarity includes at least one fluorescence-image similarity score and one color-image similarity score normalized to [0,1],

[0159] The cosine-similarity and Euclidean-distance features used in step (iii)(D) are each monotonically mapped into [0,1] to form unit-interval components of the similarity vector.

[0160] The database stores, for the ERP ID, multi-stage records including rough, split, and polished stages and the verification report includes lineage history when C > t.

[0161] In some embodiments, the signal-conditioning pipeline is executed at the client measurement station prior to transmission to the server. The pipeline comprises: (i) smoothing using a moving-average filter; (ii) baseline estimation by polynomial fit over the smoothed signal and baseline subtraction to reduce instrument drift and broad background; (iii) non-negativity clipping; and (iv) area normalization such that the total spectral area is equal to one to enable comparability across instruments and exposures. Each conditioned scan istime-stamped and associated with an ERP ID and a stage identifier (mining, sawn, polished, or retail) before transmission.

[0162] Variable definitions and units (applies to FTIR / PL features): Unless stated otherwise: wavenumber x is expressed in cmRxdenotes a wavenumber tolerance calibrated per instrument (e.g.,p is peak prominence relative to the local baseline; A is the area under the baseline-corrected peak; a is a small constant (e.g., 10 ) used to avoid division by zero; and all similarity components are normalized to the unit interval [0,1],

[0163] The Quality metric and REFER policy comprises: For each conditioned spectrum, a quality metric is computed from baseline-corrected residuals and peak statistics. Non-limiting examples include (i) peak-prominence-to-residual-noise ratio (PNR), (ii) residual variance after baseline removal, and (iii) peak-count sanity bounds for the selected sub-range (e.g., I I 00-1400 cm1for FTIR). Scans falling below preset thresholds are flagged as low-quality; flagged scans are excluded from server-side aggregation, and the report may issue a REFER instruction prompting reacquisition.

[0164] The orientation pairing and fusion comprises: When eight orientations are captured for the client and eight for the stored diamond, the server forms 8x8 = 64 pairings per modality (FTIR, PL). For each pairing, a similarity vector v is computed (peak and non-peak features), scored by a trained classifier to produce probability p, and the set { piT } is fused by a naive-Bayes log-odds combiner with non-negative reliability weights per gate.

[0165] Threshold selection and calibration comprises: Classifier outputs are probability-calibrated (e.g., Platt scaling or isotonic regression) and fused to a final confidence C using a log-odds sum: logit(C) = Sk,f Wk logit(pkf) + b; C = l / ( l+eA{— logit(C) }). Decision thresholds (t, t_low) are set via ROC analysis on a held-out validation set; in one embodiment t = 0.99 and t low = 0.75.

[0166] In accordance with the various embodiments, the client diamond or the stored diamond data can belong to rough diamond, split diamond, or polished diamond in the diamond supply chain.

[0167] In accordance with the various embodiment, the classifier model is a machine learning model. The classifier model is trained using a plurality of similarity vectors generated from pairs of diamond data, the training comprising:(a) generating similarity vectors from the features parameters extracted from corresponding the datatype selected from FTIR spectral data, PL spectral data, images, and associated metadata for pairs of diamonds with known relationships;(b) labeling each pair as either a positive pair (same diamond or parent-child relationship) or a negative pair (different diamonds); (c) inputting the similarity vectors and labels into a machine learning model selected from a neural network, gradient boosted decision tree, or similar classifier; (d) optimizing the model parameters to output a probabilistic score between 0 and 1 indicating the likelihood of a match; and (e) the model ensure the output score reflects a probability of a match.

[0168] In an embodiment, the classifier model may be implemented using any of a Neural Network (NN), a Gradient Boosted Decision Tree (GBDT), or a Light Gradient Boosting Machine (LGBM). The classifier model is trained using a dataset comprising approximately 150,000 natural diamonds and more than 40 million training pairs of diamond data. In one example, the training includes six classifier sub-models corresponding to different comparison scenarios, namely, rough-to- rough, rough-to-polish, split-to-split, split-to-polish, rough-to-split, and polish-to- polish.

[0169] In an exemplary embodiment, the model training describes herein below: The server device uses a machine learning model to automate the comparison and classification of diamond samples based on multimodal data, including FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, and associated metadata. For example, the model can be any one of a Neural Network(NN), a Gradient Boosted Decision Tree (GBDT), or a Light Gradient Boosting Machine (LGBM).

[0170] Training Dataset: The training process begins with the collection of a comprehensive dataset comprising natural diamonds at various stages, such as rough, split, polished, and jewelry. For each diamond, multiple modalities are captured, including FTIR spectra, PL spectra, images, and weight values. Each scan is tied to a unique stone identifier (ERP ID), and when a rough diamond is split, the resulting child diamonds are assigned new IDs with parent-child relationships recorded in the naming convention.

[0171] To ensure the integrity and representativeness of the training data, the following steps are performed:

[0172] Resampling: All spectral data are resampled to a common wavenumber grid (e.g., 6000 to 500 cm1at 8 cm ' resolution) to standardize input dimensions.

[0173] Splitting: The dataset is divided into training, validation, and test sets, with strict policies to prevent data leakage. All scans of a given stone id (ERP ID) appear in only one of these sets.

[0174] Balancing: Device and stage balancing is maintained across splits to ensure the model generalizes well to different instruments and lifecycle stages.

[0175] Labeling: Each pair of diamond scans is labeled as either a positive pair (same diamond or parent-child relationship) or a negative pair (different diamonds). Additional quality labels may be assigned to indicate usable or poor-quality scans.

[0176] Data Augmentation and Preprocessing: To enhance model robustness and prevent overfitting, data augmentation techniques are applied during training. These include: Preprocessing steps are performed on all input spectra, including smoothing, baseline correction, non-negativity clipping, and area normalization. These steps ensure that the extracted features are comparable across different devices and measurement conditions.

[0177] Feature Extraction and Similarity Vector Generation: For each pair of diamond scans, the specific features are extracted from the spectra and other modalities. These features include peak features (X-location, prominence, area), non-peak features (band integrals, peak counts, width at half height, stability ratios), and other non-peak features (photoluminescence characteristics, color similarity, fluorescence similarity, weight similarity, dimensional measurement similarity, color grade similarity, and clarity grade similarity).

[0178] A similarity vector is generated for each pair, with each value in the vector representing the degree of similarity for a specific feature parameter. The similarity scores are normalized between 0 and 1, where a score close to 1 indicates high similarity and a score close to 0 indicates low similarity.

[0179] For Example: Suppose the training dataset includes 150,000 natural diamonds, each with eight FTIR spectra and eight PL spectra captured at different orientations. For a given diamond pair (which include parent and child diamond), the device generates similarity vectors for all possible combinations of the parent and child diamond spectra (e.g., 8 * 8 = 64 vectors for FTIR alone). For each vector, the following features are calculated:

[0180] Peak features: X-location similarity:

[0181] where XA and XB are the X-locations of the most prominent peak in the parent and child spectra, and Rxis the instrument or dynamic range.

[0182] Prominence similarity:

[0183] where PA and PB are the prominences of the most prominent peak, and £ is a small value to avoid division by zero.

[0184] Area similarity:

[0185] where AA and AB are the areas under the most prominent peak.

[0186] Non-peak and other features: Band integrals, peak counts, photoluminescence characteristics, color similarity, and weight similarity are calculated using domain-specific rules and normalized to the [0, 1] range.

[0187] For a specific pair:• xA= 2076.05, xB= 1206.36 cm ’, Rx= 5600• pA= 1.008, pB= 0.947• AA= 292.45, AB= 170.35

[0188] The similarity scores would be:869.69 sx= 1 — 0.845600„ 0.061 svp= 1 - 0.94 1.0080.58

[0189] The similarity vector for this pair is: [0.84,0.94,0.58]

[0190] The primary model used for classifier model is a neural network or a gradient-boosted decision tree (GBDT), trained to output a probabilistic score indicating the likelihood of a match between two diamond samples. The architecture typically includes: Input layer corresponding to the length of the similarity vector. Multiple hidden layers with batch normalization, dropout, and non-linear activation functions (e.g., ReLU). Output layer with a sigmoid activation to produce a probability score between 0 and 1.

[0191] The model is trained using supervised learning, with the similarity vectors and their corresponding labels (match or non-match) as input. The loss function is optimized to maximize the separation between positive and negative pairs, ensuring the model learns the relative importance and tolerances of each feature.

[0192] After initial training, the model undergoes calibration using techniques such as isotonic regression or Platt scaling. Calibration ensures that the output scores reflect true probabilities and are interpretable for decision-making. The model’s performance is validated on the held-out test set, with metrics such as accuracy, precision, recall, and ROC-AUC used to assess its effectiveness.

[0193] At runtime, the trained model is used to process new client data. For each verification event, the device generates similarity vectors for all possible pairs of scans between the client diamond and the stored reference diamond. The model outputs confidence scores for each similarity vector, which are then aggregated by a naive model to yield a final confidence score. This score is used to generate the diamond verification and traceability report, indicating match, inconclusive, or alert status.

[0194] The cloud-based server device is designed for continuous learning and improvement. As new diamonds are processed and verified, their data and verification outcomes are incorporated into the training dataset, allowing periodic retraining and calibration of the model to maintain high accuracy and adapt to evolving data distributions.

[0195] After training on 150,000 diamonds and 40 million training pairs, the model achieves high accuracy in distinguishing true parent-child relationships from unrelated diamonds. For instance, in field trials, the model may correctly identify a polished diamond as a child of a specific rough diamond, even when both have similar FTIR spectra but differ in other features such as photoluminescence or weight. The model’s calibrated confidence score provides a reliable basis for automated verification, supporting ethical sourcing and supply chain transparency.

[0196] This detailed training process, supported by real-world examples, ensures that the system delivers robust, science-backed verification and classification of diamonds, supporting provenance, parent-child relationship establishment, and traceability throughout the diamond supply chain.

[0197] In an exemplary embodiment according various embodiments of the present disclosure, a system for diamond verification and parent-child match is provided. The system offers a cloud-based platform designed for the automated comparison and classification of diamond samples using multimodal data across the diamond supply chain. At the initial stage, a diamond is assigned an ERP ID, and its data including FTIR spectral data, photoluminescence (PL) spectral data, images, weight, dimensional measurements, color grades, and clarity grades are collected and stored in a centralized database. As the diamond moves through different stages (such as rough, split, polished, and jewelry), new data records are added, each associated with the relevant stage identifier. When a client seeks to verify a diamond, the client device equipped with spectrometers and imaging modules captures updated data from the diamond, including multiple spectra and images, often by rotating the diamond through various orientations to achieve thorough coverage.

[0198] The client device transmits the acquired data, along with the ERP ID and stage identifier, to the cloud-based processing unit. The processing unit retrieves the relevant stored records from the database and preprocesses the incoming and reference spectral data using techniques such as smoothing, baseline correction, non-negativity clipping, and area normalization. Feature extraction is then performed on both FTIR and PL spectra, generating peak features and non-peak features such as peak locations, prominence, area, band calculations, peak counts, width at half height, and stability ratios, and other non-peak features like photoluminescence characteristics, color similarity, fluorescence similarity, weight similarity, and dimensional measurement similarity.

[0199] For each pair of corresponding data types, the server device generates a similarity vector, where each value represents the degree of similarity for a specific feature, between 0 and 1. These similarity vectors are then provided to a classifier model such as a neural network or gradient-boosted decision tree trained on labeled data pairs to generate a probabilistic score for each similarity vector. The scores are further aggregated, for example using a naive model, to yield a final confidence score indicating the likelihood of a match between the client diamond and the reference record.

[0200] Based on the final confidence score, the system generates a diamond verification and traceability report, which may indicate a match and provide a lineage history if the score exceeds a first predetermined threshold, an inconclusive result if the score falls within the first predetermined threshold and a second predetermined threshold, or an alert if the score is below the second predetermined threshold. In the case of a match, the new data is stored in the database along with the updated stage identifier. If the result is inconclusive, the server device may recommend a rescan or further investigation.

[0201] In another embodiment, the above technique assists in establishing parentchild relationships between a rough diamond and the derived polished stones, by integrating multi -gate verification routine that evaluates 3-D morphological model overlap, conservation of total mass, correlation of FTIR peak positions and areas, similarity of UV-VIS absorption bands, and pixel-level similarity of fluorescence images. Each gate produces a probabilistic score, which is aggregated to yield a parent-child likelihood, and the relationship is recorded in the cloud-database if the likelihood exceeds the first predetermined threshold.

[0202] In one aspect accordance with the various embodiments of the present disclosure, a cloud-based server device for diamond verification and traceability is provided. The device comprising: a database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; a processing unit communicatively coupled to the database,the processing unit configured to: (i) receive, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client; (ii) retrieve, from the database, the stored diamond data records corresponding to the ERP ID provided by the client; (iii) for each pair of corresponding data types from the stored diamond data records and the client data, generate a similarity vector, each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other nonpeak features; iv) generate, using a classifier model, a score for each similarity vector; (v) generate, using a probabilistic model, a confidence score from the scores generated by the classifier model; and (vi) generate a diamond verification and traceability report based on the confidence score.

[0203] In accordance with the present aspect, each data record comprises at least Fourier Transform Infrared (FTIR) spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades of the diamond.

[0204] In accordance with the present aspect, the client data comprises the ERP ID, a stage identifier, and a plurality of FTIR spectral data, PL spectral data, images, weight values, and associated feature data of the client diamond.

[0205] In accordance with the present aspect, the peak features comprise at least X-location, prominence, and area; the non-peak features comprise at least band integral, Euclidean distance, cosine similarity, band measurements, peak counts, peak width at half height, and stability ratios; and the other non-peak features comprise at least color, fluorescence, weight, dimensional measurement, color grade, and clarity grade.

[0206] In accordance with the present aspect, each value in the similarity vector is a score between 0 and 1 indicating the degree of similarity for the corresponding feature, where a score close to 1 indicates a match and a score close to 0 indicates no match.

[0207] In accordance with the present aspect, the confidence score is a value between 0 and 1, where a score close to 1 indicates a match and a score close to 0 indicates a non-match.

[0208] In accordance with the present aspect, the diamond verification and traceability report comprises at least one of (a) a determination and a lineage history of the client diamond when the confidence score is equal to or more than a first predetermined threshold (t); (b) INCONCLUSIVE when the confidence score lies between a second predetermined threshold (t low) and the first predetermined threshold (t); and (c) ALERT when the final confidence score is below the second predetermined threshold (t low).

[0209] In accordance with the present aspect, the processing unit is further configured to store, upon the confidence score being equal to or more than the first predetermined threshold (t), the received client data of the client diamond in the database linked to the ERP ID.

[0210] In accordance with the present aspect, the stored diamond data record and the client data each comprise a plurality of FTIR spectra data, PL spectra data, and corresponding image data, each captured by rotating the diamond through distinct orientations.

[0211] In accordance with the present aspect, the processing unit is further configured to enroll a new diamond data record in the database when no record is present for the provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

[0212] In accordance with the present aspect, the processing unit is further configured to provide a REFER result indicative of rescan the client diamond in the diamond verification and traceability report when the final confidence score is INCONCLUSIVE.

[0213] In accordance with the present aspect, the corresponding data types comprise pairs of matched categories of diamond-related data from the stored diamond data and the client data, each pair including one data type from the stored data and its corresponding type from the client data, the categories including FTIR spectral data, PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades, each pair being compared to generate the similarity vector comprising the plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non -peak features.

[0214] In accordance with the present aspect, the processing unit is configured to extract at least one peak feature and at least one non-peak feature from each pair of corresponding data types comprising FTIR spectral data and PL spectral data, and to use the extracted features in the generation of similarity vectors.

[0215] In another aspect of the various embodiments of the present disclosure, a client device is provided. The client device comprising: a plurality of data acquisition devices, each configured to: (i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond; (ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to a cloud-based server device; and receiving, from the cloudbased server device, a diamond verification and traceability report for the client diamond.

[0216] In accordance with the present aspect, the client device further includes a rotation fixture configured to hold the client diamond and rotate the client diamond through a plurality of distinct orientations.

[0217] In accordance with the present aspect, the processing unit is further configured to preprocess the FTIR and PL data prior to transmitting to the cloudbased server device.

[0218] In another aspect of the various embodiments of the present disclosure, a method for diamond verification and traceability is provided. The method comprising: (i) receiving, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client; (ii) retrieving, from a database, a stored diamond data records corresponding to the ERP ID provided by the client; (iii) generating, by a processing unit, a similarity vector for each pair of corresponding data types from the stored diamond data and the client data, wherein each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non-peak features; (iv) generating a score for each similarity vector using a classifier model; (v) generating a confidence score from the scores generated by the classifier model using a probabilistic model; and (vi) generating, by the processing unit, a diamond verification and traceability report based on the confidence score generated by the probabilistic model.

[0219] In accordance with the present aspect, the database configured to store, for each diamond assigned an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; wherein each data record comprises at least FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades of the diamond.

[0220] In accordance with the present aspect, the client data comprises the ERP ID, a stage identifier, and a plurality of FTIR spectral data, PL spectral data, images, weight values, and associated feature data of the client diamond.

[0221] In accordance with the present aspect, the peak features comprise at least X-location, prominence, and area; the non-peak features comprise at least band integral, Euclidean distance, cosine similarity, band measurements, peak counts, peak width at half height, and stability ratios; and the other non-peak features comprise at least color, fluorescence, weight, dimensional measurement, color grade, and clarity grade.

[0222] In accordance with the present aspect, the diamond verification and traceability report comprises at least one of: (a) a determination and a lineage history of the client diamond when the final confidence score is equal to or more than a first predetermined threshold (t); (b) INCONCLUSIVE when the final confidence score lies between a second predetermined threshold (t low) and the first predetermined threshold (t); and (c) ALERT when the final confidence score is below the second predetermined threshold (t low).

[0223] In accordance with the present aspect, the method further comprises: when the confidence score being equal to or more than the first predetermined threshold (t), storing the received client data of the client diamond in the database linked to the ERP ID.

[0224] In accordance with the present aspect, the stored diamond data record and the client data each comprise a plurality of FTIR spectra, PL spectra, and images, each captured by rotating the diamond through distinct orientations.

[0225] In accordance with the present aspect, the method further comprises enrolling a new diamond data record in the database when no record is present for the provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

[0226] In accordance with the present aspect, the method further comprises providing a REFER result indicative of rescan in the diamond verification and traceability report when the confidence score is INCONCLUSIVE.

[0227] In accordance with the present aspect, the corresponding data types comprise pairs of matched categories of diamond-related data from the stored diamond data and the client data, each pair including one data type from the stored data and its corresponding type from the client data, the categories including FTIR spectral data, PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades, each pair being compared to generate the similarity vector comprising the plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non -peak features.

[0228] In accordance with the present aspect, the method further comprises extracting at least one peak feature and at least one non-peak feature from each pair of corresponding data types comprising FTIR spectral data and PL spectral data, and to use the extracted features in the generation of similarity vectors.

[0229] In another aspect of the various embodiments of the present disclosure, a system for diamond verification and traceability is provided. The system comprising a cloud-based server device and a client device. The cloud-based server device, the client device comprising: a database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; a processing unit communicatively coupled to the database, the processing unit configured to: (i) receive, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client; (ii) retrieve, from the database, the stored diamond data records corresponding to the ERP ID provided by the client; (iii) for each pair of corresponding data types from the stored diamond data records and the client data, generate a similarity vector, each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non- peak features, and other non-peak features; (iv) generate, using a classifier model, a score for each similarity vector; (v) generate, using a probabilistic model, a confidence score from the scores generated by the classifier model; and (vi) generate a diamond verification and traceability report based on the confidencescore; and the client device, the client device comprising: a plurality of data acquisition devices, each configured to: (i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond; (ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to the cloud-based server device; receiving, from the cloudbased server device, a diamond verification and traceability report for the client diamond.

[0230] The foregoing description of the invention has been set merely to illustrate the invention and is not intended to be limiting. Since modifications of the disclosed embodiments incorporating the substance of the invention may occur to person skilled in the art, the invention should be construed to include everything within the scope of the invention.

Claims

Claims :

1. A cloud-based server device for diamond verification and traceability, comprising: a database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; a processing unit communicatively coupled to the database, the processing unit configured to:(i) receive, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client;(ii) retrieve, from the database, the stored diamond data records corresponding to the ERP ID provided by the client;(iii) for each pair of corresponding data types from the stored diamond data records and the client data, generate a similarity vector, each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, nonpeak features, and other non-peak features;(iv) generate, using a classifier model, a score for each similarity vector;(v) generate, using a probabilistic model, a confidence score from the scores generated by the classifier model; and(vi) generate a diamond verification and traceability report based on the confidence score.

2. The cloud-based server device as claimed in claim 1, wherein each data record comprises at least Fourier Transform Infrared (FTIR) spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades of the diamond.

3. The cloud-based server device as claimed in claim 1, wherein the client data comprises the ERP ID, a stage identifier, and a plurality of FTIR spectral data,PL spectral data, images, weight values, and associated feature data of the client diamond.

4. The cloud-based server device as claimed in claim 1, wherein the peak features comprise at least X-location, prominence, and area; the non-peak features comprise at least band integral, Euclidean distance, cosine similarity, band measurements, and peak counts; and the other non-peak features comprise at least color, fluorescence, weight, dimensional measurement, color grade, and clarity grade.

5. The cloud-based server device as claimed in claim 1, wherein each value in the similarity vector is a score between 0 and 1 indicating the degree of similarity for the corresponding feature, where a score close to 1 indicates a match and a score close to 0 indicates no match.

6. The cloud-based server device as claimed in claim 1, wherein the confidence score is a value between 0 and 1, where a score close to 1 indicates a match and a score close to 0 indicates a non-match.

7. The cloud-based server device as claimed in claim 1, wherein the diamond verification and traceability report comprises at least one of:(a) a determination and a lineage history of the client diamond when the confidence score is equal to or more than a first predetermined threshold (t);(b) INCONCLUSIVE when the confidence score lies between a second predetermined threshold (t low) and the first predetermined threshold (t); and(c) ALERT when the final confidence score is below the second predetermined threshold (t low).

8. The cloud-based server device as claimed in claim 1, wherein the processing unit is further configured to store, upon the confidence score being equal to or more than the first predetermined threshold (t), the received client data of the client diamond in the database linked to the ERP ID.

9. The cloud-based server device as claimed in claims 2 or 3, wherein the stored diamond data record and the client data each comprise a plurality of FTIRspectra data, PL spectra data, and corresponding image data, each captured by rotating the diamond through distinct orientations.

10. The cloud-based server device as claimed in claim 1, wherein the processing unit is further configured to enroll a new diamond data record in the database when no record is present for the provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

11. The cloud-based server device as claimed in claim 7, wherein the processing unit is further configured to provide a REFER result indicative of rescan the client diamond in the diamond verification and traceability report when the final confidence score is INCONCLUSIVE.

12. The cloud-based server device as claimed in claim 1, wherein the corresponding data types comprise pairs of matched categories of diamond- related data from the stored diamond data and the client data, each pair including one data type from the stored data and its corresponding type from the client data, the categories including FTIR spectral data, PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades, each pair being compared to generate the similarity vector comprising the plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non-peak features.

13. The cloud-based server device as claimed in claim 1, wherein the processing unit is configured to extract at least one peak feature and at least one non-peak feature from each pair of corresponding data types comprising FTIR spectral data and PL spectral data, and to use the extracted features in the generation of similarity vectors.

14. A client device comprising: a plurality of data acquisition devices, each configured to:(i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIR spectroscopy dataand PL spectroscopy data is captured from different angles by rotating the client diamond;(ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to a cloud-based server device; and receiving, from the cloud-based server device, a diamond verification and traceability report for the client diamond.

15. The client device as claimed in claim 14, further comprises a rotation fixture configured to hold the client diamond and rotate the client diamond through a plurality of distinct orientations.

16. The client device as claimed in claim 14, the processing unit is further configured to preprocess the FTIR and PL data prior to transmitting to the cloud-based server device.

17. A method for diamond verification and traceability, the method comprising:(i) receiving, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client;(ii) retrieving, from a database, a stored diamond data records corresponding to the ERP ID provided by the client;(iii) generating, by a processing unit, a similarity vector for each pair of corresponding data types from the stored diamond data and the client data, wherein each similarity vector comprising a plurality of values, each valuecorresponding to a feature parameter selected from peak features, non-peak features, and other non-peak features;(iv) generating a score for each similarity vector using a classifier model;(v) generating a confidence score from the scores generated by the classifier model using a probabilistic model; and(vi) generating, by the processing unit, a diamond verification and traceability report based on the confidence score generated by the probabilistic model.

18. The method as claimed in claim 17, wherein the database configured to store, for each diamond assigned an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; wherein each data record comprises at least FTIR spectral data, photoluminescence (PL) spectral data, images, weight values, dimensional measurements, color grades, and clarity grades of the diamond.

19. The method as claimed in claim 17, wherein the client data comprises the ERP ID, a stage identifier, and a plurality of FTIR spectral data, PL spectral data, images, weight values, and associated feature data of the client diamond.

20. The method as claimed in claim 17, wherein the peak features comprise at least X-location, prominence, and area; the non-peak features comprise at least band integral, Euclidean distance, cosine similarity, band measurements, and peak counts; and the other non-peak features comprise at least color, fluorescence, weight, dimensional measurement, color grade, and clarity grade.

21. The method as claimed in claim 17, wherein the diamond verification and traceability report comprises at least one of:(a) a determination and a lineage history of the client diamond when the final confidence score is equal to or more than a first predetermined threshold (t);(b) INCONCLUSIVE when the final confidence score lies between a second predetermined threshold (t low) and the first predetermined threshold (t); and(c) ALERT when the final confidence score is below the second predetermined threshold (t low).

22. The method as claimed in claim 17, wherein the method further comprises: when the confidence score being equal to or more than the first predetermined threshold (t), storing the received client data of the client diamond in the database linked to the ERP ID.

23. The method as claimed in claims 18 or 19, wherein the stored diamond data record and the client data each comprise a plurality of FTIR spectra, PL spectra, and images, each captured by rotating the diamond through distinct orientations.

24. The method as claimed in claim 17, wherein the method further comprises enrolling a new diamond data record in the database when no record is present for the provided ERP ID and the client diamond is at a rough stage in the diamond supply chain.

25. The method as claimed in claim 21, wherein the method further comprises providing a REFER result indicative of rescan in the diamond verification and traceability report when the confidence score is INCONCLUSIVE.

26. The method as claimed in claim 17, wherein the corresponding data types comprise pairs of matched categories of diamond-related data from the stored diamond data and the client data, each pair including one data type from the stored data and its corresponding type from the client data, the categories including FTIR spectral data, PL spectral data, images, weight values, dimensional measurements, color grades, and clarity grades, each pair being compared to generate the similarity vector comprising the plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non-peak features.

27. The method as claimed in claim 17, wherein the method further comprises extracting at least one peak feature and at least one non-peak feature from each pair of corresponding data types comprising FTIR spectral data and PL spectral data, and to use the extracted features in the generation of similarity vectors.

28. A system for diamond verification and traceability comprising: a cloud-based server device, the device comprising: a database configured to store, for each diamond having an ERP ID, a plurality of data records corresponding to different stages in a diamond supply chain; a processing unit communicatively coupled to the database, the processing unit configured to:(i) receive, from a client device, client data related to a client diamond to be verified against an ERP ID provided by a client;(ii) retrieve, from the database, the stored diamond data records corresponding to the ERP ID provided by the client;(iii) for each pair of corresponding data types from the stored diamond data records and the client data, generate a similarity vector, each similarity vector comprising a plurality of values, each value corresponding to a feature parameter selected from peak features, non-peak features, and other non-peak features;(iv) generate, using a classifier model, a score for each similarity vector;(v) generate, using a probabilistic model, a confidence score from the scores generated by the classifier model; and(vi) generate a diamond verification and traceability report based on the confidence score; and a client device, the client device comprising: a plurality of data acquisition devices, each configured to:(i) collect a plurality of FTIR spectroscopy data and PL spectroscopy data associated with a client diamond, each of the FTIRspectroscopy data and PL spectroscopy data is captured from different angles by rotating the diamond;(ii) acquire one or more color images, fluorescence images, weight values, dimensional measurements, color grades, and clarity grades of the diamond, each captured from different angles by rotating the diamond, and a user interface configured to receive from a client at least: an ERP ID against which the client diamond is to be verified and traced, and a stage identifier indicating a stage of the diamond selected from mining, sawn, polished, or retail; a processing unit configured to: transmit the ERP ID, the stage identifier, and the collected data of the client diamond to the cloud-based server device; receiving, from the cloud-based server device, a diamond verification and traceability report for the client diamond.

Images