Method and device for determining a location associated with a gemstone - Patents.com
A data model-based approach for gemstone feature detection addresses the inefficiencies in existing systems by accurately locating facets and orientations, improving the precision and efficiency of laser marking and grading processes.
Patent Information
- Application Number
- JP2022541969
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-01-08
- Filing Date
- 2020-12-21
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2040-12-21
AI Technical Summary
Existing gemstone processing systems face challenges in accurately determining the location of specific facets and features due to varying lighting conditions and complex geometries, leading to inefficiencies in laser marking and grading processes.
A computer-implemented method using a trained data model, such as an artificial neural network (ANN), to detect and locate gemstone features like vertices, facets, and orientations by analyzing images, allowing for precise gemstone marking and grading with reduced image processing complexity.
The method enables accurate and efficient determination of gemstone locations and orientations, enhancing the speed and reliability of laser marking and grading processes by minimizing the need for extensive image processing and complex illumination setups.
Smart Images

Figure 0007586609000001 
Figure 0007586609000002 
Figure 0007586609000003
Abstract
Description
[Technical field]
[0001] The present disclosure relates to methods, devices and systems for determining locations associated with gemstones, and in particular, but not exclusively, to techniques for training data models to detect characteristics associated with at least one type of gemstone. [Background technology]
[0002] Measuring the size and shape of gemstones is a common procedure used by a range of industrial processing machines in the gemstone trade. For example, laser surface marking a code number onto a stone requires the location of a specific facet of the gemstone to align the mark. See, for example, the GEMscribe Laser Marking System and Method, accessed from the following web page: http: / / www.ogisystems.com / genscribegem-diamond-laser-inscription.html
[0003] In existing systems such as these, a laser is focused on the surface of a stone (e.g., a diamond) and a small-scale pattern is marked on the surface with the laser beam. To locate the appropriate area on the stone for marking, a user of the system must typically manually specify the location of the mark by placing the stone at the correct location / angle or by adjusting the location / orientation of the mark to match the stone's facets. See, for example, U.S. Patent No. 5,399,993 for existing methods. In some cases, there may be some form of image processing software to assist in locating the stone. This processing software may use live video or a large series of individual images from a camera observing the stone.
[0004] Similarly, grading and sizing of gemstones also requires accurate measurement of the gemstone's geometry. This may involve taking multiple images of the stone and calculating the stone's 3D geometry by locating edges, vertices or facets, such as the method and system for imaging cut stones discussed in US Pat. No. 5,399,633. Other examples include: https: / / sarine.com / 4c-grading / or https: / / sarine.com / products / diamension-axiom / These are examples of imaging techniques for grading diamonds, with reference to artificial intelligence for determining the color and clarity of diamonds, and automated methods for determining the shape of a diamond (otherwise called cut, e.g. brilliant cut, rose cut, Ceylon cut, checkerboard cut, etc.). Such systems and associated software use laser scanning or complex lighting configurations to attempt to map the surface of the diamond.
[0005] In US Patent No. 5,369,433, entitled Transparent Crystal Growth Process Edge and Volume Real-Time Monitoring Method and System, a system is proposed that uses an artificial neural network (ANN) to measure the growth of transparent crystals by manually measuring edges from a set of different images at different angle ranges, and generates a training data set that allows prediction of the three-dimensional (3D) structure of the growing crystal to estimate the volume for growth monitoring.
[0006] In other technical fields unrelated to gemstone applications, the use of orientation detection using neural networks is also widely used.For example, autonomous vehicle guidance software and human pose estimation all use some form of object orientation detection by extracting information from images after localization or segmentation routines are applied to the images.See, for example, U.S. Patent No. 5,399,433 entitled Simultaneous Object Detection and Rigid Transform Estimation Using Neural Networks, in which the orientation of traffic signs is estimated using neural networks.
[0007] For applications of laser marking or escribing gemstones, the use of software to locate gemstones in 2D / 3D space is complicated by the different appearances of gemstones under illumination. For example, there are many different types of stones of different sizes and cuts, which all scatter light in complex patterns due to the nature / purpose of the cut. This poses a problem for many traditional image processing software routines, such as edge detection, thresholding or Hough lines transform. The Sarine DiaMeenting system / software mentioned above uses laser scanning or complex lighting configurations to map the surface of the diamond. This increases the complexity and reduces the applicability of the process to simpler optical systems. For example, the Sarine DiaMeention system is an external system used to map the stone before passing it to the marking / cutting system. In another example, the Sarine DiaScribe software for laser marking uses a strong backlight to generate a "shadow image" of the stone. The stone is then rotated around its center and its shape is mapped by using a vision routine to measure the edges of the shadows (edge detection).The data from all the images is then seen to be combined into a 3D map of the stone.
[0008] Therefore, methods and systems that can accommodate more varied lighting conditions and accommodate fewer (or only one) images for laser marking applications are desirable because this can make the process faster and more reliable. Summary of the Invention [Problem to be solved by the invention]
[0009] To address the shortcomings of existing systems discussed above, the inventors of the present disclosure provide a new and improved method and associated devices and / or systems for automatically determining locations relative to a surface, such as an edge or facet of a gemstone. Specifically, but by no means limited to, the techniques proposed in this disclosure enable the determination of locations on a gemstone that are suitable for marking or inscribing the gemstone. [Means for solving the problem]
[0010] The present disclosure relates to a computer-implemented method, device and system for determining a location associated with a gemstone. The method includes training a data model to detect features associated with a configuration of interest on a surface of a gemstone. Training includes providing a plurality of training images of a gemstone, each training image being associated with a label for a configuration of interest for a given region of the gemstone type and an output indicating a specific gemstone type for the label. The training image may be an image including multiple configurations. In this case, the training data also includes a level and coordinates within a larger image that define the region or regions that include the configuration. An input image of a gemstone whose location is to be determined is then provided to the trained data model, which detects at least one feature in the input image that corresponds to the configuration of interest in the plurality of training images, determines that the feature in the input image is associated with a label in the trained data model that corresponds to the configuration of interest, identifies the gemstone in the input image as being of the specific gemstone type associated with the label, and finally provides an output representation of each configuration of interest for each region for the specific gemstone type identified. Then, based on the output representation, a location on a corresponding region of the gemstone associated with the input image can be determined.
[0011] The present disclosure also relates to a method, device and system for marking a gemstone using a laser marking system, including one or more processors associated with a memory or data storage device, a camera and a laser source or pointer. The method includes obtaining an image of the gemstone whose location is to be determined, as captured by a camera, and providing this image as an input to a data model. A suitable location for marking the gemstone can be determined based on the method implemented by the data model discussed above. Then, marking can be performed at the determined location on the surface of the gemstone, as either the gemstone or the laser pointer is moved or adjusted until a match is reached with the output representation from the data model. [Brief description of the drawings]
[0012] [Figure 1] FIG. 1 is a schematic diagram of an example of a laser marking system for gemstones for one application related to the present disclosure. [Diagram 2] 1 is a flow chart outlining the identification of a location of a feature within a gemstone according to a first embodiment of the present disclosure; [Figure 3a] 1 shows an example of an annotated image of a gemstone apex according to a first embodiment. [Figure 3b] 1 shows an example of an annotated image of a gemstone apex according to a first embodiment. [Figure 4] 1 shows one possible two-dimensional geometrical method for determining the centre of a facet according to a first embodiment. [Diagram 5] FIG. 1 is a schematic diagram of a process for training an artificial neural network according to a first embodiment; [Figure 6] FIG. 1 is a schematic diagram of a possible Artificial Neural Network (ANN) architecture based on Faster-RCNN (Faster-Region Convolutional Neural Network) for the first embodiment. [Figure 7]FIG. 2 is a flow diagram illustrating a predictive process using an ANN to locate a particular configuration within a gemstone according to a first embodiment. [Figure 8] FIG. 11 is a flow diagram illustrating a predictive process using an ANN to segment a particular facet in a gemstone according to a second embodiment of the present disclosure. [Figure 9] 1 illustrates a computing device for implementing one or more methods according to aspects and embodiments of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0013] The present disclosure relates to a computer-implemented method for determining a location associated with a gemstone.
[0014] The method includes training a data model to detect characteristics associated with features of interest on the surface of the gemstone. In some embodiments, the data model is implemented as an artificial neural network (ANN). In most cases, the ANN is a convolutional neural network (CNN), but is not limited to such. Examples of such networks may be, but are not limited to, Region(R) CNN, Fast-RCNN, Faster-RCNN, YOLO, YOLOv3, and other derivatives, including custom implementations designed with a similar structure and written in a suitable computer programming framework, such as TensorFlow.
[0015] Training the data model includes providing a plurality of training images of gemstones. In some embodiments, a large number of images are provided, typically more than 1000. In some embodiments, the plurality of training images can be obtained from a database of known gemstone types, or by crowdsourcing such images of gemstones, or by generating synthetic or artificial images using an image generation program for the plurality of known gemstone types. Each training image relates to a gemstone type among the plurality of types of gemstones. In some embodiments, the gemstone type relates to the cut, i.e., shape, of the gemstone. Examples of gemstone types include cuts such as Step, Brilliant, Princess, Ceylon, Rose, Cabochon, Barion, Checkerboard, Eight, and Old Mine. In some cases, shapes such as round, heart, etc. may also be used to specify the type, i.e., round brilliant cut gemstone. In some embodiments, the multiple images can be provided as a database, or as a single data file of an appropriate format, or as individual images. In some embodiments, the multiple training images may be subjected to image processing operations before being provided to the data model, including rotating, resizing, shifting, and / or modifying the brightness or contrast of the images. This may be advantageous to ensure that any differences in the dimensions or orientation or other properties of the training images do not pose inhibitions or introduce inaccuracies when applied to train the data model or when used to make decisions or predictions or recognitions based on the training images.
[0016] Then, for a given training image in said plurality of training images, the following information is provided to train the data model: A given training image in the plurality of training images may relate to each given image in the plurality of training images, or in some embodiments may relate to one or some or all of the training images, i.e., each of the training images. The training images and information may in some embodiments be provided in the form of a data file or data structure or database that includes annotations or descriptions related to the given training image, i.e., training inputs including images of a given region with labels or annotations about structures of interest or gemstone types associated with the given region; and Training output that identifies specific gemstone types associated with configurations of interest associated with the labels.
[0017] The provision of the above training inputs and outputs may in some embodiments be repeated for one or some or all of the images in the plurality of training images. In most cases, each training image will have at least one training input and one output.
[0018] In some embodiments of the training input, the features of interest may include multiple features of interest that have a common feature, i.e., the same classification or type or type of features. Advantageously, the combination of features of interest increases the accuracy of any decision or prediction made based on the features. For example, a gemstone of a given type may have multiple surfaces, each surface having a feature or feature that is specific or unique to the gemstone type. For example, a brilliant cut gemstone type has an area, i.e., a main surface or table facet, that has a different configuration from the features of interest of the same surface, i.e., the table facet of a gemstone of a checkerboard cut or princess cut type. In some embodiments, multiple features of interest for different surfaces or facets may be possible for a single training image. Thus, for example, the table facet of a princess cut gemstone type has a different configuration from different surfaces of the same gemstone type.
[0019] In other embodiments, the feature of interest may be or may include an area of interest on a given facet or surface of a gemstone type.
[0020] In some embodiments related to the training input, the given region is associated with a facet of the gemstone type associated with the given training image being considered. In some embodiments, the given region is a particular surface or facet of the gemstone type, such as a table facet or a girdle facet or a crown facet, but the disclosure is not so limited and can be applied to any surface, subsurface or facet of the gemstone. The table facet is the large horizontal facet on the top that acts as a window to the interior of the gemstone. The girdle facet is the widest part of the gemstone. The crown facet is also usually around the top and is referred to as such for certain cuts. Other facets can also be considered, such as the pavilion, which is the bottom of the faceted gemstone. The disclosure discusses the table facet below for ease of reference, but the disclosure is not so limited.
[0021] In a first aspect of the present disclosure, the configuration of interest for the training input is a vertex or vertices associated with a given facet for a given gemstone type.
[0022] Thus, for example, the feature of interest may be one or more vertices of the table facet (or other facets) (however, all vertices of the same facet should be considered together as one type or classification of features of interest. Other types of vertices are considered as different features of interest). Thus, the label or annotation in the training input indicates that the feature of interest is a vertex of the table facet. Other vertices relating to different facets, such as girdles, of the same gemstone type may also be possible. In this case, therefore, another label for the same training image is included in the training input, which indicates that the feature of interest in this case is a vertex of a girdle facet. To enable a more accurate and error-free detection of said vertices belonging to a particular facet by the data model, it is advantageous to train the data model based on different vertices of different facets of the gemstone.
[0023] In a first embodiment, the training output includes an indication or result or annotation that specifies the gemstone type of a given training image that corresponds to the label in the training input. Thus, in the above example, assume that the label for the feature of interest, which is a vertex, indicates that it is a vertex of a table facet, and consider that the gemstone type of the training image in question is brilliant cut. In this case, the training output indicates that the label (for that vertex) is related to a brilliant cut gemstone type. Thus, from both the training input and the training output for the training image, the data model can be trained to identify that such a feature is related to a vertex of a table facet of a brilliant cut gemstone when such a feature is encountered or detected.
[0024] Advantageously, the data model can be trained in a similar manner for multiple different configurations of interest for the same and different types of gemstones. Thus, for the first aspect, the data model is trained to identify and classify vertices as belonging to a particular facet of a particular gemstone type, i.e., to classify vertices as belonging to the table facet of a princess cut gemstone.
[0025] In some embodiments, when a plurality of training images are provided with a label identifying the vertices belonging to the feature of interest, i.e., table facet, and a training output indicating that such feature of interest is related to or classified as a brilliant-cut gemstone type, after a certain number of training images with the feature of interest for a given gemstone type, the data model can automatically infer and predict that such gemstone is a brilliant-cut gemstone (without needing this to be confirmed or described by the training output or any other label).In this case, in some embodiments, if the number of vertices is, for example, 8, the data model trained to identify the table vertices of brilliant-cut gemstones can be configured to generate a polygonal contour or boundary connecting the identified vertices for such a particular gemstone type, i.e., brilliant-cut.
[0026] In a second aspect of the present disclosure, the feature of interest is a portion of a given facet among a plurality of facets of a given gemstone type. In some embodiments, this may be a segment or region associated with the given facet, such as a portion or segment of the table facet of the gemstone type in the training image in question. In some embodiments, the portion is the entire given facet, i.e., the portion is, for example, the entire table facet.
[0027] In a second embodiment, the labels or annotations provided during training input for a given training image include a segmented mask indicating portions within a given facet that correspond to configurations of interest of the gemstone type in the given training image.
[0028] Thus, if the portion is the entire table facet, the segmented mask indicates a segment or portion of the mask that corresponds to the table facet for the training image. The label also specifies that the feature of interest, i.e., the portion, relates to a table facet. In some embodiments, the segmented mask is a further image representing the given training image, and the portion shown in the segmented mask has a first pixel intensity that is different from the pixel intensity of the remainder of the segmented mask. Thus, for example, if the feature of interest is a portion or the entire table facet, according to the second aspect, the label or annotation in the training input includes the segmented mask as a black and white image for the given training image, where the portion corresponding only to the table facet is black, i.e., has a lower pixel intensity, while the remainder of the mask is white, i.e., has a high intensity. The label also indicates that such portion in the segmented mark is a table facet.
[0029] As in the first embodiment, the training output includes an indication or result or annotation that specifies the gemstone type of a given training image that corresponds to the label with the segmented mask. Thus, in the above example, let us assume that the label for the feature of interest, which is a part or segment, indicates that the segmented mask has a high intensity part corresponding to the table facet, and consider that the gemstone type of the training image in question is brilliant cut. In this case, the training output indicates that the label is related to a brilliant cut gemstone type. Thus, from both the training input and the training output for the training image, the data model can be trained to identify that when a part or segment of an image with the same or similar characteristics as the segment or part of interest is encountered, it is related to the whole or part of the table facet of a brilliant cut gemstone.
[0030] Advantageously, the data model can be trained in a similar manner for multiple different configurations of interest, i.e., parts, for the same and different types of gemstones. Thus, for the second embodiment, the data model is trained to identify and classify parts or segments of an image as belonging to a particular facet of a particular gemstone type, i.e., to classify a segment as belonging to the table facet of a princess cut gemstone.
[0031] In some embodiments, the training method or data associated with the first and second inputs can be combined. Thus, the training inputs can be based on a first feature of interest, i.e., one or more vertices for a given region or facet as in the first aspect, and a second feature of interest, i.e., a portion of a given facet of a particular gemstone type. The features of interest can be provided at the same time as the data model or at a different time. For example, in some embodiments, the second feature of interest can be used only after the data model has been trained based on the first feature of interest. Advantageously, having a data model trained based on multiple different types of features improves the ability of the data model to easily and accurately detect and classify images having one or more or both types of features in a scalable, efficient, less complicated and accurate manner.
[0032] In a third aspect of the present disclosure, the configuration of interest is an orientation parameter associated with a given facet for a given gemstone type. In some embodiments, the orientation parameter is a rotation angle associated with a given facet, or a distance from a given point on a given facet. In the third aspect, the label in the training input includes the value of the orientation parameter. For the third aspect, rather than being trained to classify one or more input configurations, as in the first and second aspects, the data model in the third aspect is trained using linear regression techniques to identify or extract an orientation parameter, such as the rotation angle (e.g., 0-360 degrees) of a gemstone image, or a position in millimeters from a given point, i.e., to extract and minimize error until a suitable output is obtained.
[0033] After the data model is trained, whether according to the first, second or third aspect and related embodiments described above, the present disclosure includes providing an input image of a given gemstone to the trained data model. The gemstone is in this case a gemstone whose location is to be determined. In some embodiments, the location may be required for marking a facet of the gemstone. In some implementations, the input image is a recorded or live image of the given gemstone captured by a camera. In some embodiments, the camera may be associated with a laser marking device.
[0034] The trained data model then implements the following steps after receiving an input image:
[0035] (a) detecting at least one feature in an input image that corresponds to a structure of interest in a plurality of training images;
[0036] In a first aspect, where the feature of interest is a vertex or vertices of a gemstone-type table facet in the training images, in this step one or more features that match or correspond to the feature of interest are detected as vertices of the table facet.
[0037] In a second embodiment, where the features of interest are portions or segments of a facet, the steps include detecting portions of the input image that correspond to portions in a segmented mask of gemstone-type training images that have first or different pixel intensities, i.e., are different colors.
[0038] In a third aspect, this step includes identifying orientation parameters associated with input images that have low or minimal error or difference values when compared to values contained in the labels of corresponding orientation parameters in the training input.
[0039] (b) determining that at least one feature in the input image is associated with a label in the training input that corresponds to a configuration of interest;
[0040] Thus, for all three aspects, the label in the trained data model that corresponds to or matches the detected feature is determined as the label or annotation for the detected feature.
[0041] (c) identifying a given gemstone in the input image as being of a particular gemstone type associated with the label;
[0042] Thus, for all three aspects, once the labels are determined, the training output associated with the labels is identified. As mentioned above, this reveals the overall classification result associated with the gemstone being a particular type of gemstone, i.e., whether the input image is a princess cut gemstone or a brilliant cut gemstone, etc.
[0043] (d) providing an output including a representation of each feature of interest in each region associated with the particular gemstone type identified.
[0044] In this step, once a particular type has been identified, an exemplary representation of the configuration of interest for the particular type is provided by the data model.
[0045] In a first embodiment, the output representation comprises a polygonal outline or boundary or bounding box based on or including a boundary indicating the outline of a polygon connecting all the vertices of a given facet of the input image that corresponds to the particular gemstone type being identified, i.e., the table facet of a princess cut gemstone.
[0046] For the second aspect, the output representation is a representation of a segment or portion of each facet of the input image that is detected as having a first intensity different from the remainder of the image. Thus, for example, the output representation shows a segment that corresponds to a table facet of a particular gemstone type, i.e., a princess cut table facet.
[0047] In an embodiment where both the first and second aspects are combined during the training phase, a first output representation is obtained based on the features in the input image of the gemstone that correspond to the vertices of the first configuration, i.e., the table facets of the brilliant-cut gemstone. As mentioned above, this is the polygonal outline of the vertices or the bounding box of the vertices. This first output representation can then be used to generate one or more segmentation masks for the specific gemstone type. The generated segmentation masks can then be provided as feedback to be included in further or next or future training inputs to the data model for use in detecting the second configuration of interest, i.e., segments or parts. Advantageously, further training the same data model for a specific gemstone type with feedback obtained from one configuration of interest further improves the accuracy, speed and reliability of the disclosed method for classifying the features of the input image for that specific gemstone type.
[0048] For the third embodiment, the output representation is a representation of the values of the orientation parameters of the input image associated with each region, i.e., facet or surface, of the identified specific gemstone type with low or minimal error or difference. Thus, unlike the first and second embodiments, in the third embodiment, the data model does not predict or provide a bounding box or segment representation, but directly outputs predicted values of the orientation parameters, e.g., rotation angles based on extracted rotation angles learned during training.
[0049] Next, once an output representation from the data model has been obtained, whether using the first, second or third aspect, the method of the present disclosure includes determining a location on the corresponding area of the gemstone that is associated with the input image.
[0050] Advantageously, the output from the data model is based on a representation for a particular gemstone type of the input image, so that this output can be used to accurately determine the location on the surface of the gemstone. Thus, if the type is a brilliant cut, the output representation can be used to easily and accurately locate the center of the table facet of this type of cut. This location can be used for marking the gemstone.
[0051] For example, for the first aspect, the polygonal outline can be overlapped or superimposed on the gemstone or the image of the gemstone to identify the actual location of the table facet in the actual gemstone. In some embodiments related to the first aspect, for two or more vertices of the table facet of a particular gemstone type, the intersection of the perpendicular bisectors of the connecting lines of such vertices can also be used to more precisely delineate the center of the table facet of the particular gemstone type. Thus, the accuracy of determining the location on the table facet is further increased. The output representation of the second aspect can also be used to identify the portion of interest in the segmented mask. In a third aspect, the location on the corresponding facet or region of the gemstone associated with the input image indicates the orientation of the gemstone.
[0052] The present disclosure also relates to a computing device or system including one or more processors and a memory for storing a computer program, the one or more processors configured to execute the computer program to implement the method steps of any of the aspects and embodiments discussed above.
[0053] In some embodiments, the computing device or system also includes one or more processors for implementing the data model, the one or more processors configured to implement an artificial neural network (ANN) for implementing the steps performed by the data model described above.
[0054] As is evident from the above detailed discussion of the data model, the data model, i.e. an artificial neural network (particularly, but not limited to, a convolutional neural network architecture), is trained to recognize and locate structures or features of interest on the surface of a gemstone. Advantageously, with various training data, which may be in the form of labeled images, it is possible to generate an ANN that is capable of identifying these features even in the presence of significant visual noise, contamination and lighting changes. The ANN also identifies a particular type or classification (class) of gemstone cut by recognizing certain features that are characteristic or unique to these cuts.
[0055] Advantageously, the present disclosure allows for the identification and location of specific facets and vertices from a small number (or even a single) of images. The location of these facets can be used to determine the shape or cut of the gemstone and their associated locations. Advantageously, the use of a trained ANN can reduce the complexity associated with the imaging system and increase the speed and reliability of image processing routines in certain matters such as those described in the present method.
[0056] For example, US Patent No. 6,239,993, cited above in the Background section, proposes a process that appears to be limited to constructing a set of images from a range of angles and then predicting the volume of the crystal. In contrast, the present disclosure proposes an improvement over this and other existing image classification techniques, as it allows for the classification of features of interest, such as vertices, to be constructed into a logical data model for comparison with known cuts (or geometric shapes) of gemstones. This advantageously allows for a more powerful set of geometric arguments and classifications to be applied to known vertices, based on prior knowledge of the approximate cut / shape.
[0057] Moreover, the disclosed method is a significant improvement over conventional image processing routines used in other implementations of location-based applications on gemstones. Conventional methods require specific customized solutions that take into account, for example, image illumination, and typically require extensive processing of the image to extract meaningful information. For example, to find the edges of the stone, a black / white thresholding routine is typically required to be applied to a properly illuminated image, followed by an edge matching routine. Darker or lighter areas can then be segmented as belonging to the stone or the background. Such methods only work if the illumination does not change significantly, the reflectance / scattering of the stone is predictable, and the stone edges are free of impurities and blemishes. While the disclosed method described herein is an advancement in many ways (as described above) over these more structured and complex methods, the steps implemented to train the data model according to the above aspects and embodiments are more robust to such changes.
[0058] The application of the disclosed method to laser marking gemstones according to the present disclosure will now be discussed. The method is implemented using a laser marking system that includes a laser source or pointer, a camera, and one or more processors for managing the operation of the system.
[0059] The method includes the step of acquiring by or from a camera an image of the gemstone whose location is to be determined.
[0060] The captured image is then provided as an input image to a trained data model associated with one or more processors of the laser marking system. In some embodiments, the data model is provided within the laser marking system, while in other embodiments, the data model may be on a remote server or location or may be associated with a cloud platform that is communicatively coupled to the laser marking system via a communications network.
[0061] Next, the method includes determining a location on the surface of the gemstone for marking. This step includes the method of determining a location associated with the gemstone as discussed above with respect to the first, second and third aspects of the present disclosure. In some embodiments, in relation to the first aspect of the present disclosure, the output representation from the data model is a polygonal outline or boundary associated with a number of vertices of a given facet of the identified specific gemstone type. In this case, the method includes a step of fitting or superimposing the outline or boundary of said polygon onto the corresponding facet of the gemstone on which the laser marking is to be performed.
[0062] The method then includes moving the gemstone so that it matches the output representation from the data model, and / or moving the laser source so that it matches the output representation.Laser marking is then performed with a laser beam from a laser source or pointer focused at a location on the gemstone that corresponds to the determined location associated with the output representation.In some embodiments, related to the first aspect of the present disclosure, a center of a given facet based on the intersection of multiple vertices is determined, and the gemstone or the laser source or pointer is moved so that the laser beam is focused on the determined center.
[0063] In some embodiments, the location on the gemstone for marking is also used to apply optical aberration corrections to the laser beam from the laser source in order to focus the laser beam into a small volume within the gemstone to increase marking accuracy.
[0064] The present application also relates to a laser marking system including a base or mount for receiving a gemstone, a camera for obtaining an image of the gemstone, a laser source for providing a laser beam for marking a surface or sub-surface associated with the gemstone, and one or more processors configured to implement the above-mentioned methods using aspects and embodiments for determining a location associated with a gemstone of the present disclosure to enable a more accurate, scalable, less complex, reliable, effective and efficient method for laser marking a gemstone.
[0065] Certain specific components and embodiments of the disclosed methods are now described, by way of example only, with reference to the accompanying drawings, in which like reference numerals refer to like features and in which:
[0066] FIG. 1 is a laser marking system for implementing a possible application of the method of the present disclosure discussed above, i.e., for laser marking on or under the surface of facets of cut and polished gemstones (e.g., diamonds). FIG. 1 shows a laser marking system 100. The system 100 includes a camera 102 for capturing an image of a gemstone 104 mounted on a mount 106. The mount 106 is movable in some embodiments to adjust the position of the gemstone 104 it holds. A laser source or pointer 108 is provided for use in marking. The laser source 108 is also movable to be aligned with the surface of the gemstone to be marked. Computational resources 110 associated with memory or image storage are also shown as part of the laser marking system.
[0067] In order to laser mark a gemstone, the location where the marking is to be made on the gemstone typically needs to be determined to a reasonable degree of accuracy, e.g., + / - 20 microns. The high level schematic of FIG. 1 shows a laser marking system (e.g., Opsydia DM5000 Diamond Marking System) in which a camera 102 is connected to a computational resource 110 for storing in a digital format the image captured by the camera 102. When using the image from this camera 102, the exact location, shape and orientation of the diamond must be determined in order to make the laser marking at the correct location on the diamond's facets. While it is possible for a user of the system to manually specify this location, this is slower and more prone to human error when compared to an automated location and / or alignment process for marking.
[0068] Furthermore, if the laser marking occurs below the surface of the gemstone's facets, in some embodiments it may be necessary to determine the location of the facets and the orientation of the gemstone in order to apply optical aberration corrections to the laser beam in order to focus the laser beam into a small volume inside the stone. In this case, it may be necessary, and in some cases even important, to determine the orientation of the facets relative to the laser beam (not shown in FIG. 1) in order to determine the correct correction to apply.
[0069] Therefore, for laser marking applications such as discussed in Figure 1, as well as other applications related to the location of structures on gemstones, the present disclosure proposes a technique for accurately determining the location relative to the gemstone using a data model such as an artificial neural network (ANN) architecture, which can be, but is not limited to, a convolutional neural network (CNN) or a network with at least several convolutional layers.
[0070] FIG. 2 is a flow chart outlining the disclosed method for determining a location associated with a gemstone.
[0071] Step 202 illustrates training data models to detect features associated with structures of interest associated with gemstones.
[0072] As mentioned above, the data model is an ANN, and in some cases, a convolutional neural network (CNN).
[0073] ANNs (including CNNs) are computational models inspired by biological neural networks and are used to approximate functions that are not publicly known. ANNs can be hardware (neurons are represented by physical components) or software-based (computer models) and can use a variety of topologies and learning algorithms. ANNs can be configured to approximate and derive functions without prior knowledge of the task to be performed; instead, they evolve their own set of relevant characteristics from the learning material they process. Convolutional Neural Networks (CNNs) use the mathematical operation of convolution in at least one of their layers and are widely used for image mapping and classification applications.
[0074] In some instances, an ANN typically has three interconnected layers. The first layer may consist of input neurons. These input neurons feed data to a second layer, called a hidden layer, which performs a function and then sends output neurons to a third layer. As for the number of neurons in the input layer, this may be based on training data or reference data provided to train the ANN.
[0075] The second or hidden layer in a neural network implements one or more functions. There may be multiple hidden layers in an ANN. For example, the function or functions may each compute a linear transformation of the previous layer, or may compute a logical function. For example, considering that an input vector can be represented as x, the hidden layer acts as h, and the output acts as y, the ANN may be understood as implementing a function f using the second or hidden layer that maps from x to h, and another function g that maps from h to y. Thus, the activation of the hidden layer is f(x) and the output of the network is g(f(x)).
[0076] To train the data model to detect features related to structures of interest on the surface of a gemstone, the following information is provided to the data model: - a plurality of training images of gemstones, each training image relating to a gemstone type among a plurality of types of gemstones. For a given training image among the plurality of training images, - training inputs including labels for configurations of interest associated with given regions of gemstone types within given training images; and - Training output to identify specific gemstone types associated with configurations of interest for the label.
[0077] In one example related to the first aspect, the features of interest are vertices with a label indicating that the features are related to a gemstone type table facet. The training output in this example may be an indication that the gemstone type is princess cut. This therefore indicates that the vertices in a given training image are classified as vertices of a table facet of a princess cut gemstone. Sufficient instances of such a classification of vertices allow a bounding box to be established or generated based on the location of the vertices.
[0078] Thus, in this step, the data model is trained to detect vertices and apply a classification to each of these vertices, e.g., "This is a vertex that is recognized as one of the vertices from the table facet of a round brilliant cut gemstone" or "This is one of the vertices from a trillion cut grid." A plurality of all such classifications or a collection of all such classifications provides an overall classification of the stone as the training output, e.g., "This is a round brilliant cut gemstone."
[0079] Thus, in a first embodiment, the data model is trained to classify each vertex found in the training images into a different classification or type or set of vertices based on its location and the cut of the gemstone. An example of one of these sets could be "vertices around the table facet of a round brilliant cut gemstone." If the data model has generated about eight of these similarly classified points, the data model is trained to determine that the gemstone is round brilliant cut and that the vertices are all primarily table facets. These eight (or fewer) points can then be used to construct or generate a polygonal contour that defines the table facet for the round brilliant cut gemstone type, as seen in FIG. 3a.
[0080] Some gemstone types have two different kinds of vertices around their table facets, as seen in FIG. 3b. For example, princess cuts have four "sharp vertices 1" and four "wide vertices 2" as shown in FIG. 3(b). In this case, during training, the data model outputs regions or configurations that it considers to be "wide vertices of princess cut gemstone table facets" or "sharp vertices of princess cut gemstone table facets." With enough training images of either subset (sharp vertices or wide vertices), a model of the princess cut gemstone table facet can be formed.
[0081] The training images, training input and training output have already been described in some more detail above in the discussion of the methods of the present disclosure.
[0082] A large number of training images of features of interest or classifications of features are required, for example, over 1000. In some embodiments, a computer data file is constructed by human input (also known as annotation) that includes the defined locations of features of interest and associated labels for all training images. For example, for the first aspect described above where the features of interest are vertices of a given facet, FIG. 3 shows a schematic example of a gemstone having a particular type or types of vertices annotated with labels to define the type or classification of the vertices. FIG. 3a shows vertices associated with an octagonal symmetrically cut gemstone, while FIG. 3b shows two classifications of vertices associated with princess cut gemstones.
[0083] FIG. 5 illustrates a data structure used to train an ANN. In this example, a directory 502 of training image files, a directory 504 of annotation data files or labels, and a data file 506 linking annotations to images are shown, but the data structure is not limited to these and others are possible. In some examples, the ANN architecture 508 may be one that can define bounding boxes or polygonal contours or regions for structures of interest in the training images and can also assign pre-trained classifications to those structures using annotations or labels in the training input. Examples of such networks may include, but are not limited to, RCNN, Fast-RCNN, Faster-RCNN, YOLO, YOLOv3, and other derivatives including custom implementations designed with similar structures and written in a suitable programming framework 510. Execution of ANN routines is typically completed by running the ANN 508 on a suitable computational resource. The computations can be performed on a computer's central processing unit (CPU) or graphics processing unit (GPU) or dedicated tensor processing unit (TPU), or a combination of any of the above. The location of the data files and any code / structure / weights for the ANN can be stored on the computational resource or can be accessed via the internet, for example via a cloud storage platform.
[0084] FIG. 6 shows an example of a schematic implementation of using the Faster-RCNN architecture for a data model such that a region proposal network (RPN) is part of the data model. In this implementation, an input of multiple training images is shown as 602. For the training step in step 202 of FIG. 2, the training input and the training output as described above are also provided. The ANN includes one or more convolutional layers 604 that can perform feature extraction from the input images 602. In some examples, the feature map 606 is also considered before the training images are analyzed in the region proposal network 608, which may perform processing related to the labels and / or training output. Region or feature proposals 610 may then be made to train the data model. In some cases, a region of interest pooling layer 612 may be provided. The output obtained during training may include a classification of features of interest and bounding boxes related to the type gemstones in the training images. An example of bounding boxes for vertices of different classifications can be seen in FIG. 3.
[0085] In step 204, an input image of a gemstone whose surface-related locations are to be detected is sent to the trained data model. In some embodiments, this input image may be a live image of the gemstone or a dynamic image such as a recorded still image or video. In some embodiments, the input image is provided from a camera 102 from a laser marking system 108 in FIG. 1, where locations for marking the gemstone 108 are detected by the data model discussed herein.
[0086] Step 206 depicts detecting at least one feature in the input image obtained in step 204 that corresponds to a feature of interest in the data model.
[0087] Thus, considering the first aspect, when the data model is trained based on the vertices of a given surface, such as the table facet of a particular type of gemstone, it can detect that the input image has such a vertex if such a characteristic, i.e., a table facet vertex, corresponding to the type or classification of the vertex is found in the image. This step then further includes classifying the detected vertex by determining that at least one characteristic detected in the input image is associated with a label in the training input corresponding to the configuration of interest. Thus, in this step, the trained data model has determined that the detected vertex corresponds to the table facet of the gemstone type.
[0088] In step 208, the data model then identifies a given gemstone in the input image as being of a particular gemstone type. Considering the first embodiment as described above, if there are a sufficient number of features corresponding to the feature of interest that indicate it is a vertex of a table facet, the trained data model can determine that the vertex is associated with a particular type of gemstone, i.e., a particular cut, based on the indication in the training output corresponding to the label assigned to the same feature of interest during training. Thus, if the characteristic detected in the image is determined to belong to the gemstone type-table facet of a round brilliant cut gemstone, in this step, the type of gemstone in the input image whose location is to be predicted can be identified as a round brilliant cut gemstone.
[0089] In some embodiments, steps 206 and 208 can be combined or implemented as a single step, and the trained data model can immediately identify that vertices in a given input image relate to particular types of table facets.
[0090] As mentioned above, the data model is an ANN or CNN that is trained (at step 202) by deep learning techniques to locate and classify specific features in recorded or live video images of the gemstone. These features are typically the vertices of the gemstone's facets, as in the first embodiment, but in another implementation they can be entire specific facets, as discussed in connection with the second embodiment in which a segmented mask is used. An appropriately structured ANN, as discussed in connection with Figures 5 and 6, can classify regions of those input images as belonging to a defined classification after appropriate training. Thus, the ANN is able to predict the cut of the gemstone based on an amalgam of the detected or extracted features of interest. As an example, a "round brilliant" diamond cut has a distinctive set of vertices that differs from other diamond cuts such as checkerboard or princess.
[0091] Advantageously, aspects and embodiments of the present disclosure are an advancement of the existing systems discussed in the Background section for determining the shape of gemstones using ANN methods, because the present disclosure includes a classification step for determining and classifying the detected features based on one or more individual features of interest of the images, each of which is classified during training. The collection of such classifications of individual features of interest allows for the classification or identification of the particular cut of the gemstone, and thus an improved set of geometric / symmetry arguments can be applied to determine the geometric shape.
[0092] In some embodiments, as part of the classification step for the input image in steps 206 and 208, conventional image processing routines such as those discussed in the Background section may also be used to extract extra information from the input image or from regions of the input image identified as containing features of interest. For example, a Hough lines transformation may be applied to pixels in regions previously identified as containing vertices to extract linear features. Once an image region of interest (free of other visual clutter) has been obtained, this additional processing may be used to more accurately identify the location of the vertices. This has also been discussed above with respect to an embodiment in which the first and second aspects may be combined, such that once the boundaries associated with the vertices of the first aspect have been obtained, this may be used to generate a segmented mask that is used as a training input to the data model.
[0093] Step 210 shows providing an output representation of the configuration of interest for the particular gemstone type identified in step 208. As described above, once a particular number, e.g., eight, of the vertices of the table facet of a brilliant-cut gemstone have been detected and classified as being of said particular gemstone type, a bounding box or contour of the table facet of the particular gemstone type can be generated. Thus, the output of this step for the first embodiment is a polygonal contour for the table facet of the particular gemstone. For the second embodiment, this representation may be a representation of a portion of the input image of the gemstone that corresponds to the region of interest representing the table facet of the gemstone, and may be shown with different pixel intensities or colors.
[0094] In step 212, a location on the corresponding area of the gemstone associated with the input image can be determined based on the output representation in step 210. In an embodiment related to the first aspect, this may include overlaying or aligning a polygonal contour based on the detected vertices associated with the input image of the gemstone or the gemstone itself. For the laser marking system in FIG. 1, the gemstone 104 or the laser source 108 can be moved or rotated so that the table facets match, for example, the polygonal contour.
[0095] In some embodiments, once the cut of a gemstone has been classified for laser marking applications, the location of a subset or subsets of features of interest within the image can be used to calculate the exact location and orientation of the gemstone based on simple geometric considerations to identify where marking should occur.
[0096] An example of this application is the automatic matching of the polygonal contour of a facet from an ANN (such as the table facet of a gemstone) with the corresponding table facet on which laser marking is to be performed.
[0097] FIG. 4 shows another example of using the same type of set of detected vertices to calculate the facet center. This is by constructing a set of lines between these vertices and calculating the perpendicular bisectors of these lines. A facet contour 402 as provided for the output representation from step 210 is shown for a vertex 408 that is detected to be, for example, the wider vertex of a table facet of a princess cut gemstone. The connection between the vertices 408 is shown as a solid connector 404. The perpendicular bisector of the connector 404 is then shown as a dashed line 406. In this example, the intersection of these perpendicular bisectors 406 is, in this case, the center of the table facet of a princess cut gemstone. A similar set of geometric rules can be constructed to analyze the geometry of other facets. In some embodiments, with a complete or near-complete set of vertices, the deviation of the gemstone from perfect symmetry can be calculated for a given facet.
[0098] As mentioned above, the features of interest considered for the training and prediction or classification phases can be the apex of a gemstone's facet in the first embodiment, a portion of a facet in the second embodiment, or an orientation parameter in the third embodiment. It will be understood that although the steps of Fig. 2 have been described in relation to the first embodiment, the present disclosure is not limited thereto.
[0099] In relation to the first embodiment, FIG. 7 shows a schematic process flow for using an ANN trained by the techniques discussed above in relation to FIG. 2 to predict the location of a particular feature on a gemstone. A new image, or input image 702 provided from a camera (e.g., camera 106 in FIG. 1), is obtained. This image can be either a live image or a recorded image. In some embodiments, the new image 702 may undergo image processing such as resizing, reorienting, brightness adjustment, etc. to provide a processed image 704 before it is provided as an input image to a trained ANN 706, which is trained using the techniques discussed above for the first embodiment. Next, the output 708 of the ANN 706 is shown, where the vertices of the gemstone in the input image are identified in the output representation 708 as the vertices of table facet 1 of the particular gemstone type. The output representation in this case includes the identification or location of the feature of interest (dashed block), which has been recognized as table facet vertex 1 for the image 702. This is recognized by the ANN as it matches the characteristics of the table apex for a round brilliant cut gemstone as seen in the training input and associated labels and annotations of Figure 3a. Therefore, it can be inferred that the gemstone in Figure 7 is a round brilliant cut gemstone.
[0100] Figure 8 also shows a process flow for using a trained ANN, similar to Figure 7, but for the technique discussed in the second aspect to predict the location of a particular facet on a gemstone. In this case, the feature of interest is a portion of a facet, which may be the entire facet.
[0101] Steps 802, 804, and 806 correspond to steps 702, 704, and 706, respectively, except that ANN 806 is trained to identify segments or portions of facets that are features of interest, identified in this case as or by having a different pixel intensity relative to the rest of the image. In this case, ANN 806 would be trained to detect segments or portions of table facets of round brilliant gemstones. The output 808 of ANN 806 is then a representation that includes the identification or location of features of interest (segmented images with high intensity pixels) that have been recognized as being segments of the table facet of round brilliant gemstones for image 802.
[0102] FIG. 9 illustrates a block diagram of one exemplary implementation of a computing device 900 that can be used to implement the steps shown in FIG. 8 and described throughout the detailed description. The computing device is associated with executable instructions to cause the computing device to perform any one or more of the methodologies discussed herein. The computing device 900 may operate in the capacity of a data model or one or more computational resources for implementing a data model for performing the methods of the present disclosure. In alternative implementations, the computing device 900 may be connected (networked) to other machines (machines) in a local area network (LAN), an intranet, an extranet, or the Internet. The computing device may operate in the capacity of a server or a client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The computing device may be a personal computer (PC), a tablet computer, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a server, a network router, a switch or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by the machine. Further, although only a single computing device is shown, the term "computing device" is also understood to include any collection of machines (e.g., computers) that individually or collectively execute a set (or sets) of instructions to perform any one or more of the methodologies discussed herein.
[0103] The exemplary computing device 900 includes a processing device 902, a main memory 904 (e.g., a dynamic random access memory such as read only memory (ROM), flash memory, synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM), etc.), a static memory 906 (e.g., flash memory, static random access memory (SRAM), etc.), and a secondary memory (e.g., a data storage device 918), which communicate with each other via a bus 930.
[0104] The processing device 902 represents one or more general-purpose processors, such as a microprocessor, a central processing unit, or the like. More specifically, the processing device 902 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. The processing device 902 may also be one or more special-purpose processing devices, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, or the like. The processing device 902 is configured to execute processing logic (instructions 922) for performing the operations and steps discussed herein.
[0105] The computing device 900 may further include a network interface device 908. The computing device 900 may also include a video display unit 910 (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)), an alphanumeric input device 912 (e.g., a keyboard or touch screen), a cursor control device 914 (e.g., a mouse or touch screen), and an audio device 916 (e.g., a speaker).
[0106] The data storage device 918 may include one or more machine-readable storage media (or more specifically, one or more non-transitory computer-readable storage media) 928 on which one or more sets of instructions 922 are stored that embody any one or more of the methodologies or functions described herein. The instructions 922 may reside, completely or at least partially, within the main memory 904 and / or within the processing device 902 during execution by the computer system 900, the main memory 904 and the processing device 902 also constituting computer-readable storage media.
[0107] The various methods described above may be implemented by a computer program. The computer program may include computer code configured to instruct a computer to perform one or more functions of the various methods described above. The computer program and / or code for performing such methods may be provided to an apparatus such as a computer on one or more computer readable media, or more generally on a computer program product. The computer readable media may be transitory or non-transitory. The one or more computer readable media may be, for example, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, or a propagation medium for data transmission, for example, for downloading code via the Internet. Alternatively, the one or more computer readable media may take the form of one or more physical computer readable media, such as semiconductor or solid state memory, magnetic tape, removable computer diskettes, random access memory (RAM), read only memory (ROM), rigid magnetic disks, and optical disks such as CD-ROM, CD-R / W, or DVD.
[0108] In implementation, the modules, components, and other structures described herein may be implemented as separate components or may be integrated in the functionality of a hardware component such as an ASIC, FPGA, DSP, or similar device.
[0109] A "hardware component" is a tangible (e.g., non-transient) physical component (e.g., a set of one or more processors) that can perform specific operations and may be configured or arranged in a specific physical manner. A hardware component may include dedicated circuitry or logic that is permanently configured to perform specific operations. A hardware component may be or include a special purpose processor, such as a field programmable gate array (FPGA) or an ASIC. A hardware component may include programmable logic or circuitry that is temporarily configured by software to perform specific operations.
[0110] Thus, the phrase "hardware component" should be understood to encompass a tangible entity that may be physically constructed, permanently configured (e.g., hardwired), or temporarily configured (e.g., programmed) to operate in a particular manner or to perform particular operations described herein.
[0111] In addition, the modules and components may be implemented as firmware or functional circuitry within hardware devices. Further, the modules and components may be implemented in any combination of hardware devices and software components, or solely in software (e.g., code stored or otherwise embodied in a machine-readable medium or transmission medium).
[0112] Unless otherwise noted, and as will be apparent from the discussion that follows, discussions utilizing terms such as "providing," "calculating," "computing," "identifying," "detecting," "establishing," "training," "determining," "storing," "generating," "ascertaining," "obtaining," or equivalent terms throughout the description are understood to refer to operations and processes of a computer system or similar electronic computing device that manipulates and converts data represented as physical (electronic) quantities within the computer system's registers and memory into other data that is similarly represented as physical quantities within the computer system's memory or registers or other such information storage, transmission or display device.
[0113] It should be understood that the above description is intended to be illustrative and not restrictive. Many other implementations will become apparent to those skilled in the art after reading and understanding the above description. Although the present disclosure has been described with reference to certain exemplary implementations, it will be recognized that the present disclosure is not limited to the described implementations, but can be practiced with modification and alteration within the scope of the appended claims. Accordingly, the specification and drawings should be regarded in an illustrative, and not a restrictive, sense. The scope of the present disclosure should therefore be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. [Prior art documents] [Patent documents]
[0114] [Patent Document 1] European Patent Application Publication No. 0054840A2 [Patent Document 2] U.S. Patent No. 9,519,961B2 [Patent Document 3] China Patent Application Publication No. 110033486A [Patent Document 4] U.S. Patent No. 10,262,218B2
Claims
1. 1. A computer-implemented method for determining a location associated with a gemstone, comprising: training the data model to determine features associated with structures of interest on the surface of the gemstone; providing a plurality of training images of gemstones, each training image relating to a gemstone type among a plurality of types of gemstones; For a given training image in the plurality of training images, providing training input including labels for features of interest associated with given regions of said gemstone type in said given training images; and providing a training output that identifies a particular gemstone type associated with the feature of interest on the label; A training step; providing an input image of a given gemstone to the trained data model; The method further comprises: detecting at least one feature in the input image that corresponds to a structure of interest in the plurality of training images; determining that the at least one feature in the input image is associated with a label in the training input that corresponds to the structure of interest; identifying the given gemstone in the input image as being of the particular gemstone type associated with the label; and providing an output including a representation of said respective features of interest of said respective regions with respect to said identified particular gemstone type; [0023] Further comprising implementing Based on the output representation, the method comprises: determining a location on a corresponding area of the gemstone associated with the input image; method.
2. The method of claim 1 , wherein the feature of interest is a vertex or a number of vertices associated with a given facet for a given gemstone type.
3. 3. The method of claim 2, wherein the output representation comprises a polygonal outline or boundary based on the vertices of the given facet of the input image that corresponds to the particular gemstone type identified.
4. 4. The method of claim 2 or 3, further comprising: for two or more features of the input image that correspond to vertices of the given facet of the particular gemstone type, connecting the vertices and obtaining a perpendicular bisector of such connections, the intersection of the perpendicular bisectors indicating the center of the given facet for the particular gemstone type.
5. The method of claim 1 , wherein the feature of interest is a portion of a given facet among a plurality of facets of a given gemstone type.
6. The method of claim 5 , wherein the portion comprises an entire given facet.
7. 7. The method of claim 5 or 6, wherein the step of providing a label for a given training image comprises providing a segmented mask for the given training image, the segmented mask indicating the portion in the given facet that corresponds to the configuration of interest of the gemstone type in the given training image.
8. 8. The method of claim 7, wherein the segmented mask is a further image representing the given training image in which the portion shown in the segmented mask has a first pixel intensity that is different from a pixel intensity of a remainder of the segmented mask.
9. 9. The method of claim 8, wherein the step of detecting at least one feature corresponding to a configuration of interest comprises detecting a portion of the input image that corresponds to the portion having the first pixel intensity in a segmented mask of a given training image that is a given gemstone type.
10. 10. The method of claim 9, wherein the output representation is a representation of the portion of the input image having the first intensity, the portion within the facet that corresponds to the respective given facet of the identified particular gemstone type.
11. The method according to any one of claims 5 to 10, including the method according to any one of claims 2 to 4, In the step of providing training input, the structures of interest include a first structure having one or more vertices for a given area and a second structure having a portion of the given area, the given area being a facet of a particular gemstone type; obtaining a first output representation of a first feature in the input image of a gemstone that corresponds to a vertex of the first configuration; Based on the first output representation, the method includes generating one or more segmentation masks for the particular gemstone type, the one or more segmentation masks being provided as feedback to be included in the training input to the data model for the second configuration. method.
12. The method of claim 1 , wherein the feature of interest is an orientation parameter associated with a given facet for a given gemstone type, and the label includes a value for the orientation parameter.
13. The method of claim 12 , wherein the orientation parameter is a rotation angle associated with the given facet or a distance from a predetermined point on the given facet.
14. 14. The method of claim 12 or 13, wherein the step of detecting at least one feature in the input image corresponding to a configuration of interest comprises identifying an orientation parameter associated with the input image that has a low or minimal error or difference in value when compared to the value included in the label of the corresponding orientation parameter in the training input.
15. 15. The method of claim 14, wherein the output representation is a representation of the values of the orientation parameters of the input images associated with the respective regions of the identified particular gemstone type having the low or minimal error or difference, and the determined locations on the corresponding regions of the gemstone associated with the input images indicate the orientation of the gemstone.
16. A computing device or system comprising one or more processors and a memory storing a computer program, the one or more processors being configured to execute the computer program to implement the steps of the method according to any one of claims 1 to 15.
17. 20. A computing device or system according to claim 16, comprising one or more processors for implementing a data model, the one or more processors being configured to implement an artificial neural network.
18. 1. A method for marking a gemstone, comprising: The method is implemented by one or more processors associated with a laser marking system, the laser marking system including a camera and a laser source or pointer, the method comprising: acquiring from or by a camera an image of the gemstone whose location is to be determined; providing said acquired images as input images to a data model to be trained; - determining a location for marking according to the method of any one of claims 1 to 15; moving the gemstone so that it matches an output representation from the data model, and / or moving the laser source so that it matches the output representation; and performing laser marking using a laser beam from said laser source or pointer focused at said location on said gemstone corresponding to said determined location associated with said output representation. method.
19. 20. The method of claim 18, wherein the output representation from the data model is a polygonal outline or boundary associated with vertices of a given facet of the particular gemstone type identified, the method comprising the step of fitting or superimposing the polygonal outline or boundary onto the corresponding facet of the gemstone on which the laser marking is to be performed.
20. 20. The method of claim 19, comprising determining a center of the given facet based on an intersection of a number of vertices, and moving the gemstone or the laser source or pointer such that the laser beam is focused on the determined center.
21. 21. The method according to any one of claims 18 to 20, wherein the determined locations on the gemstone for marking are used to apply optical aberration corrections to the laser beam in order to focus the laser beam into a small volume within the gemstone.
22. a base or mount for receiving the gemstone; a camera for acquiring an image of said gemstone; a laser source providing a laser beam for creating a surface or subsurface marking associated with said gemstone; and one or more processors configured to implement the method according to any one of claims 18 to 20. Laser marking system.
Citation Information
Patent Citations
Diamond 4C standard automatic grading method and device, and storage device
CN107909103A
Deep learning-based jewelry picture classifying system and method
CN108122009A
Transparent crystal growth process edge and volume real-time monitoring method and system
CN110033486A
Method and apparatus for providing an identifying indicium to a diamond
EP0054840A2
Measurement of parameters of cut rough gemstones
JP2016528489A
