System for identification and sorting of collectible items and related methods

Abstract

An automated evaluation system is used to obtain images of collectible items and analyze different parameters of the images using artificial intelligence (AI), such as edge parameters, corner parameters, and centering parameters. The system can move the collectible items, which can be collectible cards, to a work area using a robotic arm or by manual manipulation to then obtain images of the collectible items using a camera or scanner. The images can be saved to a database and accessible by a control computer to analyze the images with an AI model. The system can provide a first pass grade regarding authenticity and condition of the cards. The grade can be manually verified by a human grader.

Description

Technical Field

[0001] This disclosure generally relates to systems and methods for identifying and classifying collectible items (e.g., baseball cards, game cards, and coins). Specific discussion focuses on automated evaluation systems with machine learning capabilities, which have programmed protocols for imaging collectible items and then identifying characteristics of the imaged items for classification, storage for historical review, and grading. Background Technology

[0002] Collectible items such as sports cards, trading cards, game cards, stamps, and coins are examples of collectible items that hobbyists collect for both tangible and intangible value. While intangible value may be less easy to measure, tangible value can be measured, for example, based on the rarity or age of a particular trading card and the condition of the card.

[0003] Third-party services can be used to classify collectible items based on authenticity and condition. Typically, when a card is sent to a grading service provider for grading, the service provider examines the card, ranks it according to its condition (usually on a numerical scale), seals the card in a tamper-evident holder, and assigns a serial number and other card information for future reference.

[0004] Once a card has been graded by a reputable third-party grader, its authenticity and condition are backed by the grader's reputation, eliminating any subjectivity in self-assessment. This typically increases the card's value. However, the number of human graders is limited, and demand for grading services often leads to backlogs lasting months. Summary of the Invention

[0005] An aspect of the invention includes an automated evaluation system that uses trained artificial intelligence (AI) to analyze various parameters and authenticity of collectible cards. The AI ​​model can also provide a first pass rating for each collectible item or card analyzed. This first pass rating can be used to assist in evaluating the analyzed cards, wherein the final rating of the analyzed cards is input from a human grader.

[0006] In an exemplary embodiment, the automated evaluation system includes: a robotic arm and robotic hand comprising multiple independently movable arm segments; at least one feed hopper including a receiving space for receiving one or more sleeve-shaped collectible items; at least one finishing hopper including a receiving space for receiving one or more sleeve-shaped collectible items; a work platform for supporting sleeve-shaped collectible items positioned adjacent to the robotic arm; a camera spaced apart from the work platform for capturing one or more images of the sleeve-shaped collectible items; and one or more sensors for detecting sleeve-shaped collectible items positioned in the at least one feed hopper and the at least one finishing hopper.

[0007] The automated evaluation system may include a housing and a lamp mounted within the housing for providing backlighting to the camera. In some instances, a front lamp for front illumination may also be provided. The housing may have a frame and / or panel, as well as a working surface for mounting various components.

[0008] In alternative examples, a scanner can be used to capture images of the front and back of a card to be analyzed by an automated evaluation system or card identification and verification system. In yet other examples, the front and back images can be acquired or generated by a user's mobile phone or other handheld device and then transmitted to an automated evaluation system, such as by saving them to the cloud, for analysis.

[0009] The system may include a flipping mechanism for flipping sleeve-shaped collectibles positioned on the work platform. This allows the sleeve-shaped collectibles to be imaged on both sides of the collectible, which can represent any number of collectibles, such as trading cards, game cards, sports cards, coins, and stamps.

[0010] In other instances, the system may include: an imaging system, such as a camera or scanner, for imaging the front and back of a collectable card; and a computer system that runs a card-verification AI model to verify and analyze various features and parameters of the card, wherein the imaged card is manually retrieved, flipped, and transferred without the need for a robotic arm.

[0011] The system may contain two or more independently movable feed hoppers.

[0012] The system may include two or more independently movable finishing hoppers.

[0013] The system may include a control computer containing a database of multiple images captured by a camera. The computer may contain advanced learning software, also known as artificial intelligence, for training and refining processes and decisions within a set of tasks or sub-tasks.

[0014] The database can be stored locally or remotely in the cloud and is accessible by the control computer. Alternatively, the database can be stored on an external hard drive or on a local data disk located with the control computer.

[0015] One or more sensors may be provided to detect the presence and location of hoppers and sleeve-type collectable items. In this example, sensors are provided for each hopper to be used with an automated evaluation system.

[0016] The system can pick up a second sleeve-type collectible item for imaging while the first sleeve-type collectible item is being imaged by the camera.

[0017] This invention relates to, as referenced Figures 3 to 8 One of the methods for displaying and describing.

[0018] Another aspect of the present invention is a method for obtaining an image of a collectable item. The method may include: placing the collectable item into a first plastic sleeve to form a sleeve-type collectable item; placing the sleeve-type collectable item into a receiving space of a hopper; placing a barcode representing an order into a second plastic sleeve to form sleeve-type barcode order information; placing the sleeve-type barcode order information on top of the sleeve-type collectable item; moving the sleeve-type barcode order information to a work platform using a robotic arm; and detecting the information on the barcode using a camera.

[0019] The method may further include using a robotic arm to move the sleeve-shaped collectible item to the work platform and activating the camera to capture one or more images of the sleeve-shaped collectible item.

[0020] Another aspect of the invention includes a method for issuing a first pass rating for a collectable item. The method includes: placing the collectable item into a first plastic sleeve to form a sleeve-type collectable item; placing the sleeve-type collectable item into a receiving space of a hopper; moving the sleeve-type collectable item to a work platform using a robotic arm; activating a camera to obtain one or more images of the sleeve-type collectable item; and issuing a rating for the sleeve-type collectable item based on the captured images of the sleeve-type collectable item.

[0021] The method further includes generating a final grade of the collectible item that is imaged and analyzed by the AI ​​model, the final grade being performed by a human grader.

[0022] An automated evaluation system for identifying and classifying recyclable items includes: at least one feed hopper including a receiving space for receiving one or more sleeve-shaped recyclable items; at least one finishing hopper including a receiving space for receiving one or more sleeve-shaped recyclable items; a work platform for supporting the sleeve-shaped recyclable items; a camera or scanner spaced apart from the work platform for capturing one or more images of the sleeve-shaped recyclable items; and a control computer including at least one hardware processor and a memory storing an artificial intelligence model for analyzing the images of the sleeve-shaped recyclable items acquired by the camera or the scanner.

[0023] The system may further include a robotic arm and a robotic hand comprising multiple independently movable arm segments.

[0024] The system may further include one or more sensors for detecting sleeve-type collectible items located in the at least one feed hopper and the at least one finish hopper. The sleeve-type collectible item may be a sleeve-type collectible card, such as a transaction card.

[0025] The system may further include a housing and a lamp installed within the housing for providing backlighting to the camera.

[0026] The system may further include a flipping mechanism for flipping a sleeve-type collectible item positioned on the work platform.

[0027] The at least one feed hopper may include two or more independently movable hoppers.

[0028] The at least one finishing hopper may include two or more independently movable hoppers.

[0029] The system may further include a database containing multiple images captured by the camera or the scanner.

[0030] The database can be stored remotely or on an external storage drive or optical disc, and can be accessed by the control computer.

[0031] The database can be stored on the memory of the control computer.

[0032] The memory may store instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform the following steps: a) retrieving a first image and a second image from the database, the first image including a front image of a first sleeve-shaped collectible item and the second image including a rear image of the first sleeve-shaped collectible item; b) retrieving from the memory a trained convolutional neural network (CNN) trained on images of collectible items, each of which has at least two collectible item parameters; c) generating and outputting a score for each of the at least two collectible item parameters; and d) generating a first pass rating for the first sleeve-shaped collectible item based on the analysis of the first image and the second image performed by the convolutional neural network.

[0033] The first sleeve-type collectible item may be a first sleeve-type collectible card.

[0034] The system can generate and output scores based on the authenticity of the first sleeve-type collectible card.

[0035] The at least two parameters may include edge parameters and corner parameters.

[0036] The at least two parameters may include color parameters and surface parameters.

[0037] The at least two parameters may include a sharpness parameter and a scratch parameter.

[0038] The at least two parameters may be at least two of the following parameters: edge parameter, corner parameter, color parameter, surface parameter, center parameter, sharpness parameter, printing defect parameter, staining parameter, scratch parameter, peeling parameter, and crease parameter.

[0039] Another aspect of the present invention is a method for obtaining an image of a collectable item, comprising: placing the collectable item into a first plastic sleeve to form a sleeve-type collectable item; placing the sleeve-type collectable item into a receiving space of a hopper; placing a barcode representing an order into a second plastic sleeve to form sleeve-type barcode order information; placing the sleeve-type barcode order information on top of the sleeve-type collectable item; moving the sleeve-type barcode order information to a work platform; and detecting the information on the barcode using a camera.

[0040] The method may further include using a robotic arm to move the sleeve-shaped collectible item to the work platform and activating the camera to capture one or more images of the sleeve-shaped collectible item.

[0041] Another aspect of the invention includes a method for issuing a first pass rating for a collectable item, comprising: placing the collectable item into a first plastic sleeve to form a sleeve-type collectable item; placing the sleeve-type collectable item into a receiving space of a hopper; moving the sleeve-type collectable item to a work platform using a robotic arm; activating a camera to obtain one or more images of the sleeve-type collectable item; and issuing a rating for the sleeve-type collectable item based on the captured images of the sleeve-type collectable item.

[0042] In alternative embodiments of the various processes discussed herein, the collectable item or card may be imaged outside the sleeve or without the sleeve and subsequently placed into the sleeve for further evaluation. Therefore, in the case of sleeved collectable items, it should be understood that a card positioned inside the sleeve may first be imaged without the sleeve. In other instances, the process may continue without ever placing the collectable card or item inside the sleeve. For example, the card may be imaged on the front and back sides and then placed directly into a transparent protective housing with a final grade without first being placed inside the sleeve.

[0043] A method for issuing a first pass rating for a collectible item includes: manually moving the collectible item to a work platform; activating a camera to obtain one or more images of the collectible item; issuing a first pass rating for the collectible item based on analysis of the one or more images of the collectible item using an AI model; issuing a first final rating for the collectible item based on a manual evaluation of the one or more images of the collectible item or based on an evaluation of the collectible item; and wherein the first pass rating may be the same as or different from the first final rating.

[0044] The first final grade may be performed by a first grader, and if the first final grade is different from the first pass grade, then the second final grade of the collectible item may be generated by a second grader based on a manual assessment of one or more images of the collectible item or based on an assessment of the collectible item. Attached Figure Description

[0045] These and other features and advantages of the apparatus, system, and method will become more apparent with reference to the specification, claims, and drawings, wherein:

[0046] Figure 1 and 2 It is a schematic depiction of an automated evaluation system for obtaining images of sleeve-type collectible items and for issuing first pass levels of sleeve-type collectible items.

[0047] Figure 2A This is a schematic diagram illustrating the configuration of an automated evaluation system according to another aspect of the present invention.

[0048] Figure 3 It is a flowchart depicting the process of taking images of incoming collectible items and issuing barcodes and unique identifiers for each set of cylindrical collectible items.

[0049] Figure 4 It is a flowchart depicting the process of obtaining measurements for each collectable item to determine whether the measurements fall within the baseline of the card type in question.

[0050] Figure 5 It is a flowchart depicting the process of teaching a computer to generate the first pass level of collectible items.

[0051] Figure 6 This is an alternative flowchart depicting the process of teaching a computer to generate first pass levels of collectible items.

[0052] Figure 7 It is a flowchart depicting the process of determining whether a collectible item to be evaluated has been previously evaluated by a grading provider.

[0053] Figure 8 It is a flowchart depicting the process used to identify the characteristics of a transaction card.

[0054] Figure 9 It is a schematic depiction of an alternative automated evaluation system for obtaining images of sleeve-type collectible items and for issuing first pass ratings for sleeve-type collectible items.

[0055] Figure 10 It is a schematic diagram depicting a method for evaluating, inspecting and / or certifying sleeve cards. Detailed Implementation

[0056] The detailed description following, taken in conjunction with the accompanying drawings, is intended as a description of a currently preferred embodiment of a robotic system for identifying and classifying collectable articles and their components provided according to aspects of this apparatus, system, and method, and is not intended to represent the only form in which this apparatus, system, and method can be constructed or utilized. The descriptive statements are intended to construct the apparatus, system, and method in conjunction with the illustrated embodiments and to describe the features and steps of using embodiments thereof. However, it should be understood that the same or equivalent functions and structures may be accomplished by different embodiments that are intended to be covered by the spirit and scope of this disclosure. As indicated elsewhere herein, similar element numbers are intended to indicate similar or analogous elements or features.

[0057] Now for reference Figure 1 The block diagram illustrates an embodiment of an automated evaluation system for analyzing, identifying, and / or inspecting collectable articles according to aspects of the invention shown, generally designated as 100. In the illustrated example, the automated evaluation system or card identification and inspection system 100 includes a robotic arm or robot 102 positioned within a housing 104, which may have a frame and panels forming support surfaces for supporting the various components of the automated evaluation system 100. As shown, the robotic arm 102 includes a base 106 and one or more arm segments or sections 108, 110, 112 connected to each other via motorized engagement joints.

[0058] A robotic hand 116, including one or more fingers 118, is provided at the end of a robotic arm 102. The robotic hand 116 is attached to the robotic arm via a movable end engagement 114 configurable to rotate and tilt along an axis. As shown, the hand 116 has a first finger 118a and a second finger 118b, which may include more than two fingers, such as three or four fingers. The fingers 118 may be fixed fingers, movable to move closer and further apart to grip objects, or may be contiguous, for example, fingers with two or more movable segments, thereby enabling them to conform for better gripping of objects. In various embodiments, the hand 116 may utilize one or more suction ports in communication with a suction source. The suction ports may contact a surface, such as a sleeve, and may be lifted or otherwise manipulated by vacuum suction. Other configurations of the robotic arm are also considered.

[0059] At least one feed hopper 122 and at least one finishing hopper 124 are equipped with an automated evaluation system 100, wherein two or more feed hoppers and two or more finishing hoppers are considered. Each hopper 122, 124 is understood as having a structure with storage space for storing or holding sleeve-type collectible items. In the illustrated example, each hopper can be considered as a holding compartment having multiple sides and one or more access openings (e.g., open sides) for holding a stack of collectible items (e.g., a stack of transaction cards to be moved by robotic arm 102 and processed by robotic system 100). For example, and further discussed below, collectible cards to be graded by collectors can each be placed in a plastic sleeve, and if not already in a plastic sleeve, a worker produces sleeve-type collectible items or cards. The sleeve-type cards are then stacked inside feed hopper 122 until filled to the desired height. Optionally, feed hopper 122 is filled with good cards from a discrete number of requesters or customers (e.g., from one to fifty customers). For example, a feed hopper is filled with 68 sleeve cards from customer #344. Next, another feed hopper is filled with 14 sleeve cards from customer #356, 27 sleeve cards from customer #357, and 5 sleeve cards from customer #358. Then, the next feed hopper 122 is loaded with additional sleeve cards to be graded, and so on. The robotic arm 102 is programmed to process each feed hopper 122 one card at a time until completion, and then moves to the next feed hopper to be processed, and so on. Then, each empty feed hopper 122 is filled with additional cards to be processed by the robotic arm 102 and the automated evaluation system 100.

[0060] As further discussed below, the processed sleeve cards are placed inside the finishing hopper 124 after imaging. Once the finishing hopper is filled with processed sleeve cards, the robotic arm then places the next finished sleeve card into the next available finishing hopper 124, and so on. An operator can empty a filled finishing hopper 124 to a selective height for further processing, such as placing the processed sleeve cards inside a more permanent rigid plastic housing, and then marking the rigid plastic housing in which the collectible card is secured with an identifier. The robotic arm 102 can pick up each collectible item to be processed by placing processed items into the finishing hopper 124 through an access opening of the corresponding feed hopper 122 and then through an access opening of the finishing hopper. In some instances, each grading order (which may include a request to grade one or more collectible items) can be positioned in its own hopper such that each feed hopper represents a single order from a single requester. Similarly, each finishing hopper may hold only one or more sleeve-type collectible items from one or more specific orders. In some instances, each feed hopper 122 and each finishing hopper 124 may have more than one order, and each order from a customer or requester may include one or more collectible items or cards.

[0061] As discussed further below, each feed hopper 122 may include multiple cards from one requester or from multiple requesters, where each request from a requester is referred to as an order. For example, order #1 may include ten (10) cards to be graded, order #2 may include fifty-five (55) cards to be graded, and order #3 may include one (1) card to be graded, and so on. Each order may be assigned an order identifier (ID), such as a barcode or a quick response (QR) code. The cards to be graded may be stacked by order number, where each order is separated from the next order by a separator (e.g., a delimiter). The delimiter may be identified by the order ID, enabling the robotic arm 102 and the worker to identify and track the cards to be graded by order ID. In an example, the delimiter may be a barcode or QR code printed on pulp and placed in a sleeve. Alternatively, the delimiter may be a machine-readable identifier, such as a symbol or label, which enables the automated evaluation system to know whether the next order is a new order or an order different from the analyzed orders.

[0062] Multiple sensors 126 may be positioned above hoppers 122, 124 to provide feedback to the robotic arm 102 regarding the position of the hoppers, whether the hoppers contain sleeve cards, and the card stack height in each hopper. For example, a proximity sensor may be used to sense the position of the hoppers. The information provided by the sensors is used by the robotic arm 102 to move sleeve cards between hoppers to handle cards, as further discussed below. Sensors 126 may be mounted on a bracket 128 positioned within the housing 104 or on a bracket forming part of the housing frame.

[0063] The work platform 132 may be provided for placing the sleeve card 134 thereon by the robotic arm 102 after the sleeve card 134 has been removed from one of the feed hoppers 122. The sleeve card 134 is shown schematically and may represent any number of possible collectible items located inside the sleeve, such as sports cards, transaction cards, YUGIOH cards, movie cards (e.g., Star Wars collectible cards), coins, stamps, etc. For the purposes discussed below, the sleeve collectible item may be a sleeve collectible card. The sleeve card 134 has a front and back side or a first side and a second side. The robotic arm 102 may pick up the sleeve card 134 (by way of example) from the first feed hopper 122a and then place the sleeve card 134 onto the work platform 132, wherein the front side or first side of the sleeve card is facing upwards to face the camera 136. The camera can be a high-resolution color digital camera with a fixed-focal-length lens to obtain the desired field of view from a mounting or fixed working distance with reference to a focal plane, such as the surface of the sleeve card when it is placed on the work platform. In some instances, the lens can be a liquid lens containing small cells of light-emitting liquid that change shape when a voltage is applied, allowing for rapid electronic focusing. This changes the focal length of the lens and thus the depth of field and the focal point of the lens, allowing the same camera to capture collectible objects at different heights or distances from the camera lens. In examples, the work platform 132 can have more than one designated location for placing the sleeve card. As shown, the work platform 132 has three designated locations, with the sleeve card 134 placed in the middle designated location. In other instances, there can be a different number of designated locations, such as two or more than three. The different locations can also be at different heights. The sleeve card can be placed in any of the designated locations on the work platform, and more than one sleeve card can be placed on the work platform at a time.

[0064] One or more images of the front side of the sleeve card 134 may be captured by a camera 136 positioned directly above the sleeve card 134. For example, two or more images of the front side of the sleeve card may be captured, such as three, five, or ten images. The images may be enhanced using one or more luminaires 138, 140 to generate different lighting conditions for obtaining a set of images to facilitate the processing of the sleeve card 134, as further discussed below. Alternatively or additionally, the luminaires may be placed below the sleeve card and below the camera to generate background illumination or backlighting during imaging. Background illumination allows the system to better evaluate the edges of the sleeve card 134 from surfaces or areas that would not form part of the card from the captured image. The sleeve card 134 is then flipped by a flipping mechanism or flipping device 144 and placed back on the work platform 132, with the back side of the sleeve card facing upwards, for example, facing the camera 136. One or more images (e.g., two or more images) on the back or second side of the sleeve card 134 can be captured by the camera 136 under different lighting conditions (including under background lighting) produced by the lamps 138, 140.

[0065] In some instances, while the sleeve card 134 is imaged on the front and back sides of the sleeve card by the camera 136, the robotic arm 102 moves to the feed hopper 122 to retrieve a second sleeve card and then places the second sleeve card at a second designated location on the work platform 132. Next, when the first card has been imaged on both sides, the robotic arm 102 can pick up the first card and place it in one of the finishing hoppers 124. During this transfer of the first card to the finishing hopper 124, the first and second sides of the second card can be imaged. While this imaging process of the second card is occurring, the robotic arm 102 can move to the feed hopper 122 to retrieve a third sleeve card and then capture images of the first and second sides of the third sleeve card, and so on. The camera 136 can be moved accordingly relative to the work platform 132 to align the camera lens with multiple designated locations on the work platform to capture images of sleeve cards positioned at different designated locations on the work platform. The process can be repeated with additional cards. In other instances, the work platform 132 may be movable relative to a camera, which may be fixed, allowing multiple sleeve cards to be sequentially placed under the fixed camera for imaging.

[0066] In this example, the flipping mechanism 144 may be a pick-and-place robot that picks up the sleeve card 134, flips the sleeve card, and then places the flipped sleeve card back to the same designated position on the work platform 132 for further imaging of the second side of the sleeve card. The flipping mechanism 144 may utilize fingers, vacuum, or a combination thereof to grip or hold the sleeve card during flipping. The work platform 132 may be structured to cooperate with the flipping mechanism, for example, by incorporating cuts, openings, and different geometries to accommodate the movement and requirements of the flipping mechanism.

[0067] Camera 136 and lamps 138, 140 can be mounted to the same bracket 151 or to different brackets positioned within housing 104 or to one or more brackets forming part of housing 104. In various embodiments, a second or additional and / or other camera can be integrated with an automated evaluation system or card identification and verification system 100. For example, a second camera and a second set of lamps can be positioned below work platform 132 to simultaneously capture one or more images of the second side of the same sleeve card while the upper camera 136 captures one or more images of the first side of the sleeve card. This allows images of both sides of the sleeve card to be captured by different cameras without flipping the card.

[0068] In various embodiments, the robotic system or card identification and inspection system 100 uses image data generated by camera 136 or several cameras to perform a task or set of tasks related to authenticating and evaluating the condition of the sleeve card 134. For example, and further discussed below, the robotic arm 102 may pick up and place the sleeve card from the feed hopper 122 onto the work platform 132, pause to capture one or more images by camera 136, pause to flip the sleeve card using the flipping mechanism 144, pause to capture additional images from a second side, and then pick up and place the sleeve card into the finishing hopper 124. The processor may then compile the images of the sleeve card to determine parameters such as card edges, staining, surface quality, and card size. In examples, artificial intelligence (AI), such as machine learning (ML), is used to analyze various parameters of the images, such as scratches, alignment, print quality, etc., as described below. Figures 3 to 8 Further discussion. Exemplary AI models incorporate convolutional neural networks (CNNs). Preferred CNN models include VGG16 and VGG19 networks, which can be retrained to analyze and classify collectible items based on images, including collectible item defect types (e.g., parameters indicating low scores due to poor quality or high scores due to being original), collectible item classes (e.g., specific types or series), and / or whether the collectible item is authentic. Specifically, CNN models can be used to classify collectible cards based on images, including collectible card defect types, card classes, and / or whether the card is authentic. Other deep neural networks for image processing are also considered, such as LeNet, AlexNet, GoogLeNet / Inception, ResNet, and ZFNet.

[0069] Reference Figure 1 Supplement Figure 2In various embodiments, the robotic arm 102 can perform tasks via different functions of a control system (e.g., a control computer 150). In the illustrated example, the control computer 150 is wirelessly connected to the robotic arm 102, while wired connections are also considered. The robotic arm 102 may include control elements housed in the base 106 or elsewhere, such as wireless or wired communication interfaces (e.g., network interface cards, WiFi transceivers, Bluetooth). TM (or other near-field transceivers, etc.) and I / O ports for diagnostics and troubleshooting. Other components may be housed in base 106 for transmitting instructions to the robotic arm or for controlling the functions of the robotic arm, such as one or more processors or microcontrollers. In the example shown, the model and knowledge base are integrated with control computer 150, which can be used to execute tasks assigned to robotic arm 102. As a task example, protocols are used by robotic arm 102 to determine how to drive various motorized joints to a position to perform a task or sub-task and how to manipulate robotic hand 116 (e.g., by driving motors that then drive arm segments or sections 108, 110, and 112 and hand 116 to their neighbors) to perform tasks involving sleeve cards. The knowledge base may contain protocols that have been trained by machine learning techniques (i.e., artificial intelligence) to enable automated evaluation system 100 and robotic arm 102 to identify and make decisions and steps based on image data generated by camera 136, as further discussed below. In a specific instance, the knowledge base software or AI model operates on a control computer 150, which may be located locally near the robotic arm. In other instances, the AI ​​model may reside on a networked computer, a remote server, or in the cloud and may be accessible via an application or an internet dashboard. The image data files to be analyzed by the AI ​​model may reside on the same computer, different computers, on a network, in the cloud, or on an external hard drive or data disk.

[0070] In some embodiments, if the automated evaluation system or card identification and verification system 100 is unable to perform a task, such as being unable to determine the card category, then the specific sleeve card that could not be performed is marked and stored in the control computer 150 for user intervention, for further user input, or the system may prompt intermediate user intervention, such as providing manual input regarding specific decisions that the control computer and / or knowledge-based protocols have not yet encountered. In various embodiments, the automated evaluation system 100 may continue to learn identification attributes, for example, by updating its knowledge base to reflect information added to the database for retraining and / or through human intervention. For example, the AI ​​model may be retrained or fine-tuned to improve detection accuracy.

[0071] Fine-tuning can involve updating the CNN architecture and retraining it to learn new or different features for different classes or characteristics of the cards. Fine-tuning is a multi-step process involving one or more of the following steps: (1) removing fully connected nodes at the end of the network (i.e., where actual class label prediction is performed); (2) replacing the fully connected nodes with the most recently initialized fully connected nodes; (3) freezing previous or top convolutional layers in the network to ensure that any previously robust features learned by the model are not overwritten or discarded; (4) training only the fully connected layers at a specific learning rate; and (5) unfreezing some or all of the convolutional layers in the network and performing additional training with the same or new dataset at a relatively small learning rate. In practice, fine-tuning is used to improve the accuracy of evaluation authentication and inspection criteria as well as other parameters of the cards to be analyzed, such as references. Figure 5 The listed parameters.

[0072] In embodiments, a task or set of tasks may be assigned to the automated evaluation system or card identification and verification system 100 via a user interface 152 connected to the control computer 150. For example, a technician may provide input via the user interface 152 to override, append, intervene in, or supplement instructions or task resolution sent to the robotic arm 102 in the event of a failure. In some embodiments, the user interface 152 may be used to provide attributes of the sleeve card, such as by designating the card as “real” to leave only the card’s condition or various other factors to be identified.

[0073] In various embodiments, the automated evaluation system 100 and one or more of the control computers 150 may include control logic to determine the attributes of each sleeve card regarding which tasks will be performed (e.g., evaluating only the edge quality and size of the card without evaluating its color). For example, the AI ​​model may be programmed to focus only on specific requested features, rather than all features the AI ​​model has been trained to perform. In one instance, the AI ​​model analyzes the card image and outputs or classifies information about the card image, such as whether the card's border is closed, the color is off, or the signature appears fake, etc., but the final grading or card scoring is performed manually by one or more human graders. In another instance, one or more human graders use information generated by the AI ​​model to provide or assign a grade between 1 and 10 to collectible cards, which includes analyzing the image of the card using the AI ​​model. This allows for a relatively rapid execution of the inspection and certification process for collectible cards using one or more AI models, but the final grade for each card is provided by one or more human graders. In some instances, the AI ​​model is programmed to generate a first pass grade for the collectible card represented by the analyzed image. However, the final grade of a collectible item or card can only be generated, provided, or given by one or more human graders. Less preferably, the automated evaluation system 100 of the present invention, using one or more AI models, can be programmed to generate a final card score or grade, which can be modified or adjusted by one or more human graders. The control logic for various protocols is referenced below. Figures 3 to 8 Further discussion.

[0074] Therefore, if the final grade matches the first passing grade, it can be understood as the unique final grade. If the final grade does not match the first passing grade, then the final grade is the first final grade, and the second final grade is generated by the second grader. The first and second final grades can then form the final grade. In other instances, the final grade may be defined by a third or more final grades, for example, by additionally defining a fourth and fifth final grade.

[0075] Now for reference Figure 2AThis illustration depicts an automated evaluation system 100, according to aspects of the present invention, for automatically inspecting and detecting specific parameters and other aspects of collectable cards using image data and machine learning. A computing device or computer system 150 may receive front and back images of each collectable card to be inspected from a camera 136. In some configurations, the image acquisition system 136a may include more than one camera 136 for acquiring front and back images of each collectable card to be inspected and analyzed. In some embodiments, the computing device 150 may execute at least a portion of a Card Inspection AI Model (CIAI) 162 to inspect and automatically detect characteristics, parameters, and other aspects of the collectable card, as further discussed below. In an example, the computing device 150 has a processor capable of executing a convolutional neural network (CNN) to determine whether the front and back images of the collectable card represent a genuine card and / or contain one of several characteristics or parameters of the card, such as stains, color, edge contours, etc., as referenced below. Figures 3 to 8 Further discussion: Instructions are stored in or on the memory of the computing device. When executed by at least one hardware processor, these instructions cause the at least one hardware processor to operate the CNN to perform several tasks, including accessing data files, analyzing data files, performing analysis of data files, and providing output indicating the characteristics or parameters represented by the data files.

[0076] In an example, the automated evaluation system or card recognition and inspection system 100 includes a software driver and libraries for analyzing image data. An exemplary software driver may be the Python interpreted high-level programming language, operating in conjunction with an NVIDIA CUDA compiler driver that compiles one of several libraries modified in a process known as transfer learning. Libraries may include the cuDNN SDK deep learning GPU-accelerated library, the TensorFlow open-source library for developing and evaluating deep learning models, the Keras open-source library, the NumPy open-source library for working with array and matrix data structures, the matplotlib open-source library for image display and annotation or graphical display, and the OpenCV open-source library for processing card images to identify objects and features. In an exemplary embodiment, a convolutional neural network (CNN) is used as the deep learning model for a vision application to examine and classify multiple features and parameters of card images of collectable cards. When using and training the CNN model, learned patterns at specific locations in the image are used as instances that can be recognized anywhere else in the image. Initial convolutional layers learn small local patterns, such as edges and textures of the card image, while subsequent layers learn larger patterns composed of features learned from the initial layers.

[0077] In some embodiments, computing system 150 may transmit information about image data received from image acquisition system 136a to server 164 via communication network 166, which may execute at least a portion of CIAI model 168. In such embodiments, server 164 may send information back to computing system 150 indicating the output of CIAI model 168, which may be one of several characteristics or parameters of the card collected from at least two images of the card—its front or first surface image and its back or second surface image.

[0078] In some embodiments, computing device 150 and / or server 166 may be any suitable computing device or combination of devices, such as a desktop computer, laptop computer, smartphone, tablet computer, wearable computer, server computer, virtual machine executed by a physical computing device, etc. In some embodiments, CIAI models 162, 168 may use a convolutional neural network (CNN) previously trained as a general image classifier to classify the characteristics or parameters of the collectible card.

[0079] In some embodiments, Figure 1 The image acquisition system 136a is an image source for supplying image data to computer device 150 and / or server computer 164. The image acquisition system 136a may alternatively or additionally include an image scanner 510. Figure 9 The image acquisition system 136a, having one or more cameras 136, is used to scan the front and back images of the collectable card to be analyzed by the system. In some embodiments, the image acquisition system 136a may be housed locally with the computing device 150. For example, the image acquisition system 136a may be integrated with the computing device 150. In other words, the computing device 150 may be configured as part of the device to capture and / or store images from the image acquisition system 136a, for example, to store data files captured by the cameras. In another instance, the image acquisition system 136a may be connected to the computing device 150 via a cable, a direct wireless link, or the like. Alternatively or additionally, in some embodiments, the image acquisition system 136a may be located locally and / or remotely to the computing device 150, and may transmit image data to the computing device 150 and / or the server 164 via a communication network 166. In alternative embodiments, a user or customer may acquire the front and back images of the collectable card and then transmit the acquired images to the computing device 150, a remote server 164, or the cloud for the system 100 to analyze and examine the images.

[0080] In some embodiments, the communication network 166 may be any suitable communication network or combination of communication networks. For example, the communication network 166 may include a Wi-Fi network (which may include one or more wireless routers, one or more switches, etc.), a peer-to-peer network (e.g., a Bluetooth network), a cellular network (e.g., a 4G network conforming to any suitable standard (e.g., CDMA, GSM, LTE, LTE Advanced, WiMAX, etc.), a wired network, etc. In some embodiments, the communication network 166 may be a local area network (LAN), a wide area network (WAN), a public network (e.g., the Internet), a private or semi-private network (e.g., a corporate intranet), any other suitable type of network, or any suitable combination of networks. The communication link may be any suitable communication link or combination of communication links, such as a wired link, a fiber optic link, a Wi-Fi link, a Bluetooth link, a cellular link, etc.

[0081] In some embodiments, the communication system for providing the communication network 166 may include any suitable hardware, firmware, and / or software for transmitting information via the communication network 166 and / or any other suitable communication network. For example, the communication system may include one or more transceivers, one or more communication chips and / or chipsets, etc. In more specific instances, the communication system may include hardware, firmware, and / or software that can be used to establish Wi-Fi connections, Bluetooth connections, cellular connections, Ethernet connections, etc.

[0082] Figure 3 It indicates that it is used for Figure 1 and 2 A schematic diagram of a protocol for an automated evaluation system that images incoming transaction cards so that the captured images can be used or processed in other subsequent processes to evaluate the authenticity, condition, and other parameters of the incoming cards. The protocol may include the use of a printer 160 ( Figure 2 A barcode and a unique identifier are printed and placed on each of the processed cards. Alternatively, a QR code may be used instead of a barcode or as a supplement to a barcode. The imaging protocol or process 170 begins at 172 with the receipt of an order from the customer for grading of cards or other collectable items to be sent to the grading service provider. Each order can range from one card to hundreds of cards to be evaluated. Next, at 174, the operator loads the cards into plastic sleeves to form sleeve-type collectable cards, or simply sleeve cards. In this example, each card to be evaluated is individually sleeved. The sleeve cards are then placed into the feed hopper 122. Figure 1 In the automated evaluation system housing 104, the feed hopper 122 can be positioned within the housing 104. Figure 1The feed hopper is fed from outside the unit until it is full or reaches the desired height or is used for a specified number of orders, such as for one order or for two or more orders. The user then loads additional sleeve cards into the next feed hopper. When the feed hopper is full or reaches the desired height or for an order, the user then places the filled feed hopper into the unit or housing of the automated evaluation system for processing. It is anticipated that a belt or conveyor will transport the filled feed hopper from the filling station or area to the unit for processing by the automated evaluation system.

[0083] In alternative embodiments, the collectable item or card may be imaged outside the sleeve or without the sleeve and then placed into the sleeve for further evaluation. Therefore, in the case of sleeved collectable items, it should be understood that a card positioned inside the sleeve may first be imaged without the sleeve. In other instances, the process may continue without ever placing the collectable card or item inside the sleeve. For example, the card may be imaged on the front and back sides and then placed directly into a transparent protective housing with a final grade without first being placed inside the sleeve.

[0084] At point 176, a barcode, broadly interpreted as encompassing a QR code and indicating order information, is added to the top of the order stack. The barcode can be printed onto a card or pulp and placed inside a sleeve to form a sleeve-type order information barcode card or a sleeve-type order information (OrderInfo) card. Each order information barcode is linked to an OrderID, which contains information about the order (e.g., how many cards are in the order) and information about the requester, as discussed further below.

[0085] At point 178, after the feed hopper is full or after a specified number of sleeve cards have been placed inside, the user places the feed hopper into the unit or housing 104 of the robot system or card identification and inspection system 100. Each feed hopper may have more than one order, with each order identified by an information order barcode or OrderID. In other instances, each hopper is specific to one order. For illustrative purposes, different information order barcodes can be imagined as OrderID1, OrderID2, OrderID3, etc.

[0086] At position 180, robotic arm 102 ( Figure 1 The top sleeve clip is removed from the stack from one of the feed hoppers 122 and the item is placed into the photographic fixture or work platform 132. Figure 1On one of the designated locations. For illustrative purposes, the first sleeve item may be assumed to be an information order barcode or a sleeve order information card. As further discussed below, alternative embodiments include abandoning the robotic arm and manually transferring the sleeve card and sleeve order information card to a photographic fixture or work platform or to a scanner for imaging, and then manually moving the imaged sleeve card and sleeve order information card to a discharge hopper for further processing.

[0087] Next, at position 182, camera 136 ( Figure 1 Decode the barcode to obtain the information stored in the control computer 150 ( Figure 2 The OrderID is stored in the cloud or on a local server. For example, a camera scans a barcode to retrieve the data stored in the barcode.

[0088] Next, at position 184, the OrderID is passed to the Application Programming Interface (API) to retrieve the order data associated with that specific OrderID. The order data may include information such as the requester's name, the number of cards associated with the order, and the tier provider ID number assigned to each card in the order.

[0089] Next, at position 186, the robotic arm 102 picks up the order barcode, such as a sleeve-type order information card, and places it in the finishing hopper 124. Figure 1 Next, at point 188, the robotic arm picks up the next set of sleeve clips from the feed hopper and at point 190 places the sleeve clips into the photographic fixture or work platform 132. Figure 1 Imaging can be performed on the surface.

[0090] At point 192, activate camera 136. Figure 1 The camera captures one or more images of the sleeve card. Lights 138 and 140 can also be illuminated to provide background lighting during imaging. Background lighting may include one or more light sources positioned on the side of the card opposite the camera. At 194, one or more images of the sleeve card are captured without backlighting. The card is then flipped so that a second side of the card can be captured. The same number and image settings of the second side can be captured, including cases with and without backlighting. In other instances, backlighting is not required when capturing images of the second side or the back side.

[0091] At point 196, the protocol queries whether images of both the front and back of the card have been captured. In some instances, only one image of the front and one image of the back are captured by the image acquisition device. In other instances, multiple images of the front of the card and multiple images of the back of the card can be captured under different lighting conditions. If no images have been captured, the card is flipped at point 198 to allow for image capture. If images have been captured, the robotic arm picks up the card from the imaging fixture or work platform at point 200.

[0092] At point 202, the card is moved through printer 160, which prints a barcode specific to the sleeve card that has just been imaged and places the barcode onto the plastic sleeve. In some instances, additional information such as grade, message, date, etc., can be printed on the barcode. Then, at point 204, the sleeve card, which has been imaged and has a specific barcode placed on the sleeve, is placed into the finishing hopper. Thus, each processed card can have a barcode and unique identifier associated with it, attached to the sleeve, allowing each card to be tracked via a barcode and a unique identifier, such as a unique grader service provider ID. In some instances, the unique identifier is automatically (i.e., via automation) printed and placed onto the sleeve card. In other instances, the unique identifier is manually placed onto the sleeve card after the front and back of the sleeve card have been imaged.

[0093] At 206, the protocol queries whether the card just imaged is the last card for a specific OrderID, for example, by detecting whether the next card in the stack has a different sleeve order information card or by counting the total number of cards that have been imaged compared to the number of cards associated with the OrderID. If the card just imaged is not the last card, then return to 188 and repeat. If it is the last card, then at 208, the protocol queries whether the card just imaged is the last card in the feed hopper. If not, then return to 180 and repeat. If yes, then at 210, the operator removes the completed hopper with one or more imaged sleeve cards from the robot workstation. At 212, the robot arm can now move to the next feed hopper and repeat. The image of one or more cards for a specific OrderID is now stored in the system computer and ready to be processed, for example, ready for evaluation using the AI ​​model of the present invention.

[0094] Figure 4 It means to use Figure 1 and 2This is a schematic diagram of protocol 220 for the automated evaluation system to perform card measurement and size verification. The goal of this process is to check the measurements of each card to determine if they fall within the baseline of the card type in question, and to reject or mark cards for further inspection if they do not fall within the baseline. The obtained measurements are compared with existing measurement data from the same type of card stored in a card image database. The new measurement data is then added to the database and stored for future reference. Assuming that a particular order contains more than one card, the process begins at 222 with receiving the order for transaction cards or collectable cards for evaluation, sleeve the cards, and then at 224, take front and back images of each card in the order. At 226, images can be taken with and without backlighting. Steps 222, 224, and 226 can be referenced. Figure 3 The steps described are the same.

[0095] At 228, using known parameters of the camera and lens, the pixel dimensions are converted into physical dimensions. At 230, the precise angles and parallelism of the card edges are also measured. At 232, the measurements are compared with expected values ​​for a specific card product and a specific card (which may be referred to herein as Spec#). At 234, the system queries whether the measurements are within the correct threshold. If so, then at 236, the software displays the measurement data to the card grader during the grading process. In this example, the measurement data is stored in a computer. Then, when the grader grades the imaged card at a later time, the information can be retrieved to assist the grader in grading the card without having to perform physical measurements on the card.

[0096] If the measurement is not within the correct threshold, then at 238 the protocol queries whether the measurement meets the automatic rejection threshold. For example, the edges of a particular card may be trimmed to remove worn edges. However, if over-trimmed, the card size measurement may be far below the acceptable threshold for that particular card type. If so, the card may be rejected because most collectors would not consider an over-trimmed card desirable. Therefore, if the card meets the automatic rejection threshold, then at 240 the card is rejected and marked as such. In this example, the card may be marked as "minimum size requirement" or "evidence of trimming". At 242, the measurement data of the rejected card is then added to the measurement database along with the card's measurement status (e.g., rejected or accepted).

[0097] Rewinding to point 238, if the measurement does not meet the automatic rejection threshold, a size warning is added to the card grader interface at point 244 during the grading process. Therefore, during the grading process, the grader can see warnings associated with the card and determine at point 246 whether the card has the appropriate size for grading. If not, the card is rejected at point 240. If yes, the grader presents the card's grade at point 248, such as a number between 1 and 10, and the process continues. At point 242, the measurement data of the card now graded by the grader is added to the database.

[0098] At point 250, the machine learning model (e.g., a CNN model) for the specific card product is retrained and the thresholds are adjusted. For example, if the card is a "1952 Topps baseball" card, the thresholds for that particular type of card are adjusted to fine-tune the acceptable and unacceptable parameters. For example, the AI ​​model can be retrained or fine-tuned to improve the detection of acceptable card edges. Fine-tuning may involve updating the CNN architecture and retraining it to learn new or different features of different classes or characteristics of cards, such as different thresholds for acceptable card edges. Fine-tuning is a multi-step process and involves one or more of the following steps: (1) removing fully connected nodes at the end of the network (i.e., where actual class label prediction is performed); (2) replacing the fully connected nodes with the most recently initialized fully connected nodes; (3) freezing previous or top convolutional layers in the network to ensure that any previously robust features learned by the model are not overwritten or discarded; (4) training only the fully connected layers at a specific learning rate; and (5) unfreezing some or all of the convolutional layers in the network and performing additional training with the same or new dataset at a relatively small learning rate. In this example, fine-tuning is used to improve the accuracy of assessments of certification and inspection conditions, as well as other parameters of the card being analyzed. Therefore, when the same specific card is encountered again in the future, the threshold for that particular card can be better determined.

[0099] At position 254, the protocol queries whether there are more images to process. If yes, it returns to position 226. If no, the process completes at position 256.

[0100] Figure 5 This is a schematic diagram of protocol 264 used to teach a computer to generate a first pass rating for a transaction card (which may be reviewed by one or more human graders at a later stage). The process attempts to have the computer present an initial opinion, which is then confirmed or rejected by one or more human graders. This can significantly reduce the time human graders spend evaluating each card and grading them, and accelerate the grading process for a large number of cards to be graded. The first pass rating can be considered implied, but the final grading decision for the collectible card to be graded depends entirely on one or more human graders.

[0101] Assuming a particular order contains more than one card, the process begins at point 266 where the transaction card order is received and evaluated, and then at point 268, front and back images of each card in the order are taken. Steps 266 and 268 can be referenced. Figure 3 The steps for receiving and processing card orders are described identically. Next, an evaluation can be performed on the imaged sleeve card and a score can be provided based on several parameters, such as the measurement at step 270, which may involve reference... Figure 4 The steps described.

[0102] In addition to measurements, the protocol can evaluate other card parameters and factors, including edges, corners, one or more colors, surface, realism, prominence, centering, focus or sharpness, typographical defects, staining, scratches, peeling, creases, and optionally other factors at 272 to 300 points. A score can be assigned to each parameter or characteristic being examined. At step 302, the score for each factor or characteristic is evaluated by a machine learning model that outputs a rating or score on a grading scale (e.g., 1 to 10). In one instance, the first pass score for the card can be an average score, rounded to the nearest integer, or obtained from multiple scores of the different factors and characteristics being examined. In another instance, one or more of the different factors and characteristics can be weighted more than the others. For example, if the realism score is 6, then its score generated from the AI ​​model could be 1.1 multiplied by 6 or 6.6. The outputs of various individual factors and characteristics from the AI ​​model can be adjusted with weighting factors of 1.1 to 1.3 or higher (preferably, no more than 1.5). Preferably, authenticity, printing defects, and centering are scaled proportionally. Scaled and non-scaled values ​​are averaged and rounded to the nearest integer. Next, at step 304, the card is evaluated by a human grader. Preferably, the human grader ignores the grade assigned to the card by the computer.

[0103] At step 306, the protocol queries whether the human grade matches the computer's first pass grade. If yes, the grade is finalized at 308. If not, the card is graded by different human graders at 310. The human graders may discuss and publish a final score or grade that differs from the first pass grade or score generated by the computer using machine learning. Less preferably, a single human grader may publish a final score that differs from the computer-generated first pass score. In some instances, a second human grader may sample the final scores of those analyzed cards that received scores from human graders, where the final scores differ from the computer-generated first pass score. Then, at 314, the score of each of the evaluated factors is added to the database and associated with the correct image and the final grade.

[0104] At point 316, the machine learning model for each factor is retrained to include the new data, and at point 318, the machine learning model for evaluating scores and determining ranks is retrained to take the new data into account. Fine-tuning and retraining can be performed as discussed elsewhere in this paper.

[0105] At 320, the protocol queries whether there are more images to process. If yes, it returns to 270. If no, the process completes at 322.

[0106] Figure 6 This is a schematic diagram of protocol 330 used to teach a computer to generate a first pass level for a transaction card (which may be reviewed by one or more human graders at a later stage). Figure 6 The process can be Figure 5 Alternative implementation schemes for the process may be compatible with Figure 5 The process is used in parallel. The process attempts to elicit an initial opinion from the computer, which is then confirmed or rejected by one or more human graders. This significantly reduces the time human graders spend on each card and speeds up the grading process for a large number of cards.

[0107] Assuming a particular order contains more than one card, the process begins at point 332 with receiving and evaluating the transaction card order, and at point 334 with taking front and back images of each card in the order. Steps 332 and 334 can be referenced. Figure 3 The steps described are the same. Based on the barcode from OrderID, card type data from the database is retrieved at point 336 (if available). Next, at point 338, a protocol query is performed to determine if sufficient data exists for this card type. If yes, at point 340, the card images are evaluated against images trained from cards of the same type. If not, then at point 342, the card images from the cards to be analyzed are evaluated against a machine learning model trained from cards of the same group. Following this, after evaluating the card images against any cards of the same type or group, at point 346, the machine learning model outputs a first pass rating.

[0108] Next, at point 348, a human grader evaluates the card and assigns a grade or score. Preferably, the human grader ignores the grade assigned to the card by the computer. At point 350, the protocol queries whether the human grade matches the first pass grade generated by the computer. If yes, the grade is finalized at point 352. If no, the card is evaluated by a second human grader at point 354. The human graders may discuss and issue a final score or grade that differs from the first pass grade or score generated by the computer using machine learning. Then, at point 356, the card image and the final grade are added to the database.

[0109] At position 358, the machine learning model for the card type is retrained, and at position 360, the machine learning model for the deck is retrained. Fine-tuning and retraining can be performed as discussed elsewhere in this document.

[0110] At position 362, the protocol queries whether there are more images to process. If yes, it returns to position 336. If no, the process completes at position 364.

[0111] Figure 7 This is a schematic diagram of protocol 370 used to determine whether a card has been evaluated and whether the data and / or image linked to the card are in the grading database. The process identifies when the card was previously evaluated and compares previous and current state images to look for changes that indicate tampering. In this example, incoming image data is compared with data on a card of the same Spec#. In a specific instance, the card to be evaluated is the 1952 Topps Mickey Mantle #311 card. The incoming card is compared with the image on the file for the 1952 Topps Mickey Mantle #311 card. New images and data are added to the database and will be used for subsequent searches of similar cards.

[0112] Assuming a particular order contains more than one card, process 370 begins at 372 with receiving and evaluating the transaction card order, and at 374 with taking front and back images of each card in the order. Steps 372 and 374 can be referenced. Figure 3 The steps described are the same. At point 376, a computer vision algorithm is used to analyze the card image to generate a feature point vector. At point 378, the digital fingerprint generated from the feature factors is compared with the digital fingerprint stored in the database.

[0113] At 380, the protocol queries whether there is a match between the generated feature point vector and the digital fingerprint stored in the database. If so, at 382, ​​the process moves to retrieve the image associated with the previously stored fingerprint. Next, at 384, the images are compared using an algorithm, and at 386, measurements derived from the previous and current images are compared. Then, at 388, a composite image is generated and presented to an expert or grader to emphasize the card's variations. Next, at 390, the protocol queries whether there are additional images to process, and if so, returns to 376; otherwise, the process terminates at 394.

[0114] Simultaneously, if no match is found between the generated feature point vector and the digital fingerprint stored in the database, the newly acquired digital fingerprint is added to the database at step 392. The process then moves to step 390 and repeats. If the analyzed card has not been previously classified, it can be classified as discussed elsewhere in this document.

[0115] Figure 8 This is a schematic diagram of Protocol 400 used to determine whether a process can correctly identify the defined characteristics of a transaction card. These characteristics include individual data elements such as year, manufacturer, group, Card#, athlete, etc. For example, a 1996 Skybox E-X2000 Kobe Bryant #30 has been previously evaluated and given internal numbers, such as Spec#. If the card to be analyzed is of the same type, then ideally the process outputs a Spec# that matches the Spec# in the database for the same card type. The new information for the new card is then stored in the database for future use.

[0116] Assuming a particular order contains more than one card, process 400 begins at 402 with receiving and evaluating the transaction card order, and at 404 with taking front and back images of each card in the order. Steps 402 and 404 can be referenced. Figure 3 The steps described are the same. At point 406, a computer vision algorithm is used to analyze the card image to generate a feature point vector. At point 408, the digital fingerprint generated from the feature point vector is compared with the digital fingerprint stored in the database.

[0117] Next, at 410, the process queries whether a match is found between the generated feature point vector and the digital fingerprint stored in the database. If yes, then at 412, the process queries whether the confidence factor is higher than a threshold. If yes, then at 414, the process assigns an internal identifier (e.g., Spec#) to the card. Next, the process queries whether more images need to be processed. If yes, then the process moves to 406, and if no, it moves to 418, thus indicating that the process is complete.

[0118] Moving back to point 410, if no match is found between the generated feature point vector and the digital fingerprint stored in the database, the card is manually identified by a professional grader at point 420. Next, the fingerprint is added to the database at point 422, and the process is repeated at point 414.

[0119] Rewind to point 412, where the procedure checks if the confidence factor is above a threshold. If the answer is no, then at point 424, a possible match is presented to the grader for verification and / or correction. Next, at point 426, the confidence factor is adjusted, and at point 414, an internal identifier or Spec# is assigned to the card. The process is then repeated as previously described. If the analyzed card has not been previously graded, it can be graded as discussed elsewhere in this document.

[0120] In alternative embodiments of the various processes discussed herein, the collectable item or card may be imaged outside the sleeve or without the sleeve and subsequently placed into the sleeve for further evaluation. Therefore, in the case of sleeved collectable items, it should be understood that a card positioned inside the sleeve may first be imaged without the sleeve. In other instances, the process may continue without ever placing the collectable card or item inside the sleeve. For example, the card may be imaged on the front and back sides and then placed directly into a transparent protective housing with a final grade without first being placed inside the sleeve.

[0121] Now for reference Figure 9 The block diagram illustrates an alternative embodiment of an automated evaluation system for identifying and / or verifying collectable articles according to another aspect of the invention shown, generally designated as 500. In the illustrated example, the automated evaluation system or card identification and verification system 500 is similar to... Figure 1 and 2 System 100 has several exceptions. Most notably, this automated assessment system 500 does not have a robotic arm. Instead, various transfer functions, such as moving the sleeve card from the feed hopper to the work platform for imaging, flipping the sleeve card for further imaging, and moving the imaged card to the completion hopper, are operated manually by a hand 600, for example, by a technician or human grader.

[0122] As shown, the automated evaluation system 500 includes a housing 502, which may be a compartmentalized space comprising multiple walls or panels defining a workspace 504 in which multiple automated evaluation system components are located, or located adjacent to, or near their locations. External light interference should be minimized by providing appropriate coverage and shading. At least one feed hopper 122 and at least one finishing hopper 124 are equipped with the automated evaluation system 500, wherein two or more feed hoppers and two or more finishing hoppers are considered. The hoppers are similar to those discussed elsewhere herein for temporarily holding sleeve cards to be inspected, certified, and / or evaluated.

[0123] The work platform 132 can be provided for a worker's hand 600 to place a sleeve card 134 thereon. The sleeve card 134 has a front and back side, or a first side and a second side. A technician or grader can pick up the sleeve card 134 (for example) from the first feed hopper 122a and then place the sleeve card 134 onto the work platform 132, with the front side or first side of the sleeve card facing upwards to face the camera system 136a including the camera 136, similar to Figure 1 The content.

[0124] One or more images of the front side of the sleeve card 134 can be captured by a camera 136 positioned directly above the sleeve card 134. For example, two or more images of the front side of the sleeve card can be captured, such as three, five, or ten images. The images can be enhanced using one or more lamps 138, 140 to generate different lighting conditions for obtaining a set of images to facilitate the processing of the sleeve card 134. Alternatively or additionally, a lamp 139 can be placed below the sleeve card and below the camera 136 to generate background illumination or backlighting during imaging. Thereafter, the sleeve card 134 can be manually flipped over by a technician's hand 600 and placed back on the work platform 132 with the back side of the sleeve card facing upwards, for example, facing the camera 136. One or more images (e.g., two or more images) of the back or second side of the sleeve card 134 can be captured by the camera 136 under different lighting conditions generated by the lamps 138, 140 (including under background illumination 139). The work platform 132 may include markings or boundaries to guide technicians in placing the sleeve card 134. Alternatively, the sleeve card 134 may be imaged using a scanner 510.

[0125] Camera 136 and lamps 138, 140 can be mounted to the same bracket 151 or to different brackets positioned within housing 502 or to one or more brackets forming part of housing 502. In various embodiments, a second or additional and / or other camera can be integrated with an automated evaluation system or card identification and verification system 500. For example, a second camera and a second set of lamps can be positioned below the work platform to simultaneously capture one or more images of the second side of the same sleeve card while the upper camera 136 captures one or more images of the first side of the sleeve card. This allows images of both sides of the sleeve card to be captured by different cameras without flipping the card.

[0126] In various embodiments, the automated evaluation system or card identification and verification system 500 uses image data generated by camera 136 or several cameras to perform a task or set of tasks related to authenticating and evaluating the condition of the sleeve card 134. The processor on the control computer 150 can then compile the image of the sleeve card to determine parameters (e.g., card edges, staining, surface quality, and card size) as a task, similar to... Figure 1 and 2 System 100.

[0127] Figure 10This is a schematic diagram depicting a method 1000 for evaluating, inspecting, and / or certifying sleeve cards according to aspects of the present invention. In an example, a customer or order is placed to evaluate, inspect, certify, and / or classify one or more cards, wherein only one collectible card 550 is schematically shown. As discussed above, the card 550 is then placed inside a plastic sleeve 552 to create a sleeve card 134. Next, an order identifier 554 is placed outside the sleeve 552.

[0128] Next, the sleeve-type card 134 is used Figure 1 and 2 Automated evaluation system 100 or Figure 9 The automated evaluation system 500 imaging involves manual human intervention 600 to perform various flipping and movement steps, as previously discussed. Subsequently, the image-collectible card 550 can be evaluated, inspected, certified, and / or provided with a first pass rating, as previously referenced. Figures 3 to 8 Discussion. In the case where the first pass grade is generated by an AI model, a grader 602, who ignores the first pass grade and can provide manual labor for flipping and moving cards, can provide a human-generated grade. If the human-generated grade is the same as the first pass grade, then the grader's grade in the final grade "#" matches the first pass grade. If the human-generated grade is different from the first pass grade, then one or more additional graders are consulted, and the final grade "#" is given to the collectible card by two or more graders. The final grade "#" may be the same as or different from the first pass grade. In any case, the final grade "#" preferably has a human component or human input and is not strictly a machine-generated final grade.

[0129] Therefore, if a final grade matches the first passing grade, it can be considered the unique final grade. If the final grade does not match the first passing grade, then the final grade is the first final grade, and the second final grade is generated by the second grader. The first and second final grades then form the final grade. In other instances, the final grade is defined by a third or more final grades, for example, by additionally defining a fourth and fifth final grade. Additional final grades can be performed by different human graders.

[0130] Next, the collectable card 500 is placed inside a card holder 570, which typically consists of two transparent plastic housings irreversibly locked together to provide a secure housing that is extremely difficult or impossible to separate. A tag 558 may also be added inside the card holder 570. The tag 558 may contain information about the collectable card, an identification number, a barcode, and / or a QR code, which can be used to access a database to verify the identification or certificate number, the name of the tiered entity, the final grade "#", and other information. In some examples, a Near Field Communication (NFC) chip may be included in the card holder 570. A smartphone can be used to access the information stored in the NFC chip, which may include a link to a webpage providing information about the collectable card 500.

[0131] In alternative embodiments of the various processes discussed herein, the collectable item or card may be imaged outside the sleeve or without the sleeve and subsequently placed into the sleeve for further evaluation. Therefore, in the case of sleeved collectable items, it should be understood that a card positioned inside the sleeve may first be imaged without the sleeve. In other instances, the process may continue without ever placing the collectable card or item inside the sleeve. For example, the card may be imaged on the front and back sides and then placed directly into a transparent protective housing with a final grade without first being placed inside the sleeve.

[0132] Methods for creating and using robotic systems for identifying and classifying collectable items and their components are within the scope of this invention.

[0133] Although limited embodiments of robotic systems for identifying and classifying collectable articles and their components have been clearly described and illustrated herein, many modifications and variations will be apparent to those skilled in the art. Therefore, it should be understood that robotic systems for identifying and classifying collectable articles and their components, constructed based on the principles of the disclosed devices, systems, and methods, may be embodied differently from those explicitly described herein. This disclosure is also defined in the appended claims.

Claims

1. An automated evaluation system for collectable items, comprising: At least one feed hopper includes a receiving space for receiving collectable items, said collectable items including a first collectable item and a second collectable item, each collectable item including a front side and a back side; At least one finishing hopper, which includes a receiving space for receiving one or more collectable items; A belt or conveyor for transporting the collectable items to a work platform located within a housing, the housing including a frame and a panel, wherein the work platform includes a surface for supporting at least one of the collectable items; A barcode reader is used to read machine-readable codes associated with the collectable items to query order information from multiple requests, each request relating to one or more of the collectable items; A first camera is used to capture an image of the front side of the collectable item, and a second camera is used to capture an image of the back side of the collectable item; A photographic clamp for supporting the collectible when imaging it; A control computer includes at least one hardware processor and a memory, wherein the memory stores an artificial intelligence (AI) model for analyzing images of the front and back sides of the collectable item acquired by the first camera and the second camera. A first pass rating is generated for the first collectible item by the AI ​​model running on the control computer based on analysis of images of the front and back sides of the first collectible item; and The final grade of the first collectible item is based on a manual evaluation of the image of the first collectible item. The first pass rating is based on reviewing the image of the first collectable item with respect to at least two of the following parameters: edge parameter, corner parameter, color parameter, surface parameter, center parameter, sharpness parameter, puncture parameter, staining parameter, scratch parameter, peeling parameter, crease parameter, and authenticity parameter.

2. The system of claim 1, wherein the collectable item is a sleeve-type collectable item.

3. The system of claim 1, further comprising a robotic arm having an end effector for removing the collectable items one by one and placing the collectable items onto the photographic fixture.

4. The system of claim 3, further comprising one or more sensors for detecting sleeve-type collectible items positioned in the at least one feed hopper and the at least one finishing hopper.

5. The system of claim 3, further comprising a lamp mounted within the housing for providing backlighting to the first camera.

6. The system of claim 3, further comprising a flipping mechanism for flipping a sleeve-type collectible item positioned on the work platform.

7. The system of claim 1, wherein the first collectable item is positioned within a two-piece rigid plastic housing.

8. The system of claim 1, wherein the first pass level is a score from 1 to 10.

9. The system of claim 2, further comprising a database having a plurality of images captured by the first camera and the second camera.

10. The system of claim 1, wherein the final grade is a score from 1 to 10.

11. The system of claim 10, wherein the first pass level and the final level are different.

12. The system of claim 1, wherein the AI ​​model is programmed to reject collectable items because one or more of the parameters are not met.

13. The system of claim 1, further comprising a separator having a machine-readable identifier to identify a second order request that is different from the first order request.

14. An automated evaluation system for collectable items, comprising: A feed hopper includes a receiving space of a size and shape adapted to receive collectable items, the collectable items including a first collectable item and a second collectable item, each collectable item including a front side and a back side; A belt or conveyor for transporting the collectable items to a work platform located within a housing, the housing including a frame and a panel, wherein the work platform includes a surface defining a plane for supporting at least one of the collectable items; A barcode reader for reading machine-readable codes associated with the collectable items to query order information from multiple requests, each request relating to one or more of the collectable items; A first camera is used to capture an image of the front side of the collectable item, and a second camera is used to capture an image of the back side of the collectable item; A photographic clamp for supporting the collectible item when imaging it, wherein when the collectible item is located on the photographic clamp, a first camera is arranged relative to the photographic clamp to acquire one or more images of each of the front sides of the collectible item, and a second camera is arranged relative to the photographic clamp to acquire one or more images of each of the back sides of the collectible item. The hopper is completed, which includes a receiving space that is adapted in size and shape to receive the collectable item after it has been imaged; A control computer, comprising at least one hardware processor and a memory, wherein the memory stores an artificial intelligence (AI) model for analyzing images of the front and back sides of the collectable item that have already been imaged; as well as A robotic arm having an end effector for removing the collectable item and placing the collectable item onto the photographic fixture to image the collectable item; The AI ​​model is programmed to accept or reject one or more of the collectable items based on analysis of the images acquired by the first and second cameras.

15. The system of claim 14, wherein the AI ​​model accepts or rejects the first collectable item based on examining one or more images of the first collectable item with respect to at least two of the following parameters: edge parameter, corner parameter, color parameter, surface parameter, center parameter, sharpness parameter, puncture parameter, staining parameter, scratch parameter, peeling parameter, crease parameter, and authenticity parameter.

16. The system of claim 14, wherein the AI ​​model is programmed to issue a first pass rating for the first collectible item based on analysis of the images of the front and back sides of the first collectible item.

17. The system of claim 14, wherein the image data file analyzed by the AI ​​model resides on the control computer, a different computer, a network, the cloud, or an external hard drive.

18. The system of claim 14, wherein each of the collectable items is located inside a transparent plastic sleeve.

19. The system of claim 16, wherein the first pass level is a combination of levels based on a plurality of parameters.

20. The system of claim 19, wherein at least one of the levels of the parameters is adjusted using a weighting factor of 1.1 to 1.

3.

21. A method for acquiring images of collectable items using an automated evaluation system, comprising: Collectible items are placed into a feed hopper, the feed hopper including a receiving space of a size and shape adapted to receive the collectible items, the collectible items including a first collectible item and a second collectible item, each collectible item including a front side and a back side; The collectable items are transported to a work platform located within a housing, which includes a frame and a panel, wherein the work platform includes a surface that defines a plane for supporting at least one of the collectable items. A barcode reader is used to read machine-readable codes to retrieve order information from multiple requests, each request involving one or more of the collectable items; A photographic clamp is used to support the collectible when imaging it. Use a camera to obtain one or more images of the front side of the collectable item; The collectable item is moved to the receiving space of the finishing hopper, the size and shape of which are adapted to receive the collectable item after it has been imaged; An operation control computer, the control computer including at least one hardware processor and a memory, the memory storing an artificial intelligence (AI) model for analyzing images of the collectable items that have been imaged, wherein the AI ​​model is programmed to accept or reject one or more of the collectable items that have been imaged based on the analysis of the images. A robotic arm with an end effector is used to move the collectible item onto the photographic fixture; as well as The one or more collectible items are rejected based on the analysis performed by the AI ​​model on the images of the one or more collectible items.

22. The method of claim 21, wherein the AI ​​model accepts or rejects the first collectable item based on examining the image of the first collectable item with respect to at least two of the following parameters: edge parameter, corner parameter, color parameter, surface parameter, center parameter, sharpness parameter, puncture parameter, staining parameter, scratch parameter, peeling parameter, crease parameter, and realism parameter.

23. The method of claim 21, wherein each collectable item is placed inside a protective cover before being imaged by the camera and a second camera spaced apart from the camera.

24. The method of claim 23, wherein each collectable item is coupled to a tag containing information about the collectable item.

Images