KR102539486B1 - method and apparatus for determining 3-dimension model data based on sintering process in 3-dimension printing system - Google Patents

Abstract

A 3-dimension (3D) printing system learns a 3D printer that controls to print artificial teeth composed of a plurality of slices, and an artificial neural network for data necessary for a printing operation of the 3D printer, and the learned artificial neural network And a server for determining 3D model data from a target image of a tooth or initial 3D model data of the tooth using a 3D printer, wherein the 3D printer prints the artificial tooth according to the 3D model data determined by the server, The artificial neural network may determine 3D model data in which items to be corrected are pre-reflected in consideration of deformation due to sintering applied to the artificial tooth.

Description

Method and apparatus for determining 3-dimension model data based on sintering process in 3-dimension printing system in 3D printing system

The present invention relates to a method and apparatus for determining 3D model data considering a firing process in a 3-dimension (3D) printing system.

A 3-dimension (3D) printer is a system for forming a molding by laminating molding materials based on three-dimensional data created by a computer design program. FDM (Fused Deposition Modeling) method, which uses a 3D printing method to pull out a solid plastic material that melts in heat like a skein and melt it little by little and stack it up In the SLA method, functional polymers or metal powders are used instead of photocurable resins, and SLS (Selective Laser Sintering) method uses the principle of scanning laser beams to solidify and mold. There is a DLP (Digital Light Processing) method using the principle of partially curing by using

In the case of the DLP method, after applying a photocurable resin to a water tank, when a plate for separating the cured resin comes into contact with the resin, an operation of irradiating light is performed to cure the resin in a desired shape. When a molded article is created by irradiation of light, the plate is raised, and the molded article is separated from the uncured resin. At this time, in order to facilitate separation of the water tank and the molding, a releasing film is inserted between the resin and the water tank.

A Learning-Based Framework for Error Compensation in 3D Printing. Zhen Shen et al. 5, IEEE TRANSACTIONS ON CYBERNETICS, VOL. 49, no. 11, NOVEMBER 2019 <10.1109/TCYB.2019.2898553>
Shape Deviation Generator? A Convolution Framework for Learning and Predicting 3-D Printing Shape Accuracy. Qiang Huang et al. 3, IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 17, no. 3, JULY 2020 <10.1109/TASE.2019.2959211>

In the present specification, a method and apparatus for effectively determining 3D model data considering a firing process in a 3-dimension (3D) printing system are proposed.

A 3-dimension (3-dimension) printing system according to an embodiment for solving the above problems includes a 3D printer that controls to print an artificial tooth composed of a plurality of slices, and data necessary for a printing operation of the 3D printer. and a server for learning an artificial neural network for learning and determining 3D model data from a target image of a tooth or initial 3D model data of the tooth using the learned artificial neural network, wherein the 3D printer determines The artificial tooth is printed according to 3D model data, and the artificial neural network may determine 3D model data in which items to be corrected are pre-reflected in consideration of deformation due to sintering applied to the artificial tooth.

The target image may include an image of a tooth at a position other than the tooth to be replaced by the artificial tooth.

The server may determine the 3D model data by modifying the initial 3D model data by reflecting the items to be corrected in advance.

The server obtains learning data including a tooth image, extracts at least one learning factor selected from the tooth position, the photographing angle of the tooth, the color of the tooth, and the age of the image provider from the tooth image, and extracts the extracted learning factor. The artificial neural network can be learned using

The server may acquire training data including a tooth image, add a coordinate axis graphic to the tooth image, and learn the artificial neural network using the tooth image to which the coordinate axis graphic is added.

The server acquires learning data including a tooth image or initial 3D model data, adds a marking indicating a portion deformed by plasticity to the tooth image or the initial 3D model data, and The artificial neural network may be learned using the added tooth image or the initial 3D model data.

The server may receive learning data from a central server accessible through an external communication network, and learn the artificial neural network using the learning data.

The server may receive parameter information of the learned artificial neural network from a central server accessible through an external communication network, and perform transfer learning of the artificial neural network using the parameter information.

In addition, a method for 3D printing an artificial tooth in a 3D (3-dimension) printing system according to an embodiment for solving the above problems includes learning an artificial neural network for data necessary for a printing operation of a 3D printer, Determining 3D model data from a target image of a tooth or initial 3D model data of the tooth using the learned artificial neural network, and printing an artificial tooth composed of a plurality of slices according to the 3D model data Including, the artificial neural network, considering the deformation due to sintering (sintering) applied to the artificial tooth can determine the 3D model data to be corrected in advance.

In addition, the computer-readable recording medium according to an embodiment for solving the above problems is a computer on which a computer program for performing a method of molding an object using the 3D printing system according to the above-described embodiment is recorded. It may be a recording medium that can be read by

According to the technical details described in this specification, the 3D printing system may determine a slice color set capable of expressing a color similar to that of a target tooth by using an artificial intelligence algorithm.

1 is a diagram illustrating a 3-dimension (3D) printing system according to an embodiment.
2A to 2D are diagrams illustrating a structure of a 3D printer in a 3D printing system according to an embodiment.
3A to 3C are diagrams illustrating an example of a structure of a 3D printer including a plurality of water tanks disposed on a rectangular plate moving linearly in a 3D printing system according to an embodiment.
4 is a diagram illustrating an example of a plurality of water tanks disposed on a rotating circular plate in a 3D printing system according to an embodiment.
5 is a view showing another example of a plurality of water tanks disposed on a rotating rectangular plate in a 3D printing system according to an embodiment.
6 is a diagram showing the structure of an artificial neural network applicable to a 3D printing system according to an embodiment.
7 is a diagram illustrating an example of a shape change by a firing process of a 3D printed artificial tooth in a 3D printing system according to an embodiment.
8 is a diagram illustrating a procedure for performing 3D printing in a 3D printing system according to an embodiment.
9 is a diagram showing the structure of local 3D printing systems interoperable in a 3D printing system according to an embodiment.

The following merely illustrates the principle of the invention. Therefore, those skilled in the art can invent various devices that embody the principles of the invention and fall within the concept and scope of the invention, even though not explicitly described or shown herein. In addition, it should be understood that all conditional terms and embodiments listed in this specification are, in principle, expressly intended only for the purpose of making the concept of the invention understood, and are not limited to such specifically listed embodiments and conditions. .

The above objects, features and advantages will become more apparent through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art to which the invention belongs will be able to easily implement the technical idea of the invention. .

In the specification and claims, terms such as "first", "second", "third" and "fourth", if any, are used to distinguish between similar elements, but not necessarily in a particular sequence. Or used to describe the order of occurrence. It will be understood that the terminology so used is compatible with the embodiments of the invention described herein, under appropriate circumstances, such that, for example, they may operate in sequences other than those shown or described herein. Likewise, where a method is described herein as comprising a series of steps, the order of those steps presented herein is not necessarily the order in which those steps may be performed, and any recited steps may be omitted and/or here Any other step not described may be added to the method.

In addition, terms such as "left", "right", "front", "rear", "top", "bottom", "above", "below" in the specification and claims are used for explanation, It is not necessarily intended to describe an invariant relative position. It will be understood that the terminology so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein may, for example, operate in directions other than those shown or described herein. As used herein, the term "connected" is defined as being directly or indirectly connected in an electrical or non-electrical manner. Objects described herein as "adjacent" to each other may be in physical contact with each other, in close proximity to each other, or in the same general scope or area as is appropriate for the context in which the phrase is used. The presence of the phrase “in one embodiment” herein refers to the same embodiment, although not necessarily.

In addition, in the specification and claims, 'connected', 'connected', 'engaged', 'fastened', 'coupled', 'coupled', etc., refer to various variations of these expressions directly with other components. It is used in the meaning of being connected or indirectly connected through other components.

In addition, the suffixes "module" and "unit" for the components used in this specification are given or used together in consideration of ease of writing the specification, and do not have meanings or roles that are distinguished from each other by themselves.

Also, terms used in this specification are for describing embodiments and are not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase. As used herein, 'comprise' and/or 'comprising' means that a stated component, step, operation, and/or element is the presence of one or more other components, steps, operations, and/or elements. or do not rule out additions.

In addition, in describing the invention, if it is determined that a detailed description of a known technology related to the invention may unnecessarily obscure the subject matter of the invention, the detailed description will be omitted. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating a 3-dimension (3D) printing system according to an embodiment.

Referring to FIG. 1 , the 3D printing system includes a 3D printer 110 and a learning server 120 .

The 3D printer 110 is a device that performs 3D printing based on set 3D model data. The 3D printer 110 prints a 3D shape of a desired shape through a process of curing the mixed solution by irradiating light. The 3D printer 110 may use pre-stored 3D model data or 3D model data input from the outside. For example, the 3D printer 110 may perform 3D printing based on 3D model data provided from the learning server 120 .

The learning server 120 determines at least part of the 3D model data used in the 3D printer 110 . According to an embodiment, the learning server 120 generates at least a portion of 3D model data using an AI (Artificial Intelligence) based algorithm. To this end, the learning server 120 may acquire training data and perform machine learning or deep learning. In addition, the learning server 120 may generate at least part of the 3D model data by performing inference using a learned algorithm (eg, an artificial neural network). Then, the learning server 120 provides the generated 3D model data to the 3D printer 110 . To this end, the learning server 120 may include at least one processor, a communication unit, and a storage unit, and may further include an input unit and an output unit for interaction with a user.

According to one embodiment, the 3D printer 110 and the learning server 120 may be installed in the same space and form a direct connection over a wired link or a wireless link. Through the established connection, the 3D printer 110 and the learning server 120 may transmit and receive necessary data. Alternatively, according to another embodiment, the 3D printer 110 and the learning server 120 may be connected through an external network (eg, Internet network). In this case, the learning server 120 may be implemented in the form of a cloud server.

2A to 2D are diagrams illustrating a structure of a 3D printer in a 3D printing system according to an embodiment.

Referring to FIGS. 2A to 2D , the 3D printer according to an embodiment includes a printing bed 210 supporting a molded object cured by irradiation with light emitted from the light irradiation unit 230, a lower portion of the printing bed and the light irradiation unit 230. ) is located in the printing space specified above and includes at least one tank for accommodating a mixture of powder and binder, and is located at the bottom of the tank and irradiates controlled light to harden the mixture. It may include a light irradiation unit 230 that generates a surface-divided layer of the molding and a control unit (not shown) that electronically controls at least one of the components described above.

The light irradiation unit 230 may include a light source 231 that emits curable light and a reflector 232 that reflects the curable light. Alternatively, the light irradiation unit 230 may be formed of a light emission diode (LED) panel provided in close contact with the lower surface of the water tank without a separate reflector. The lower surface 221 of the water tank may be made of a light-transmitting film. Accordingly, the curable light irradiated from the light irradiation unit 230 may reach the liquid mixture contained in the water tank. Accordingly, the light irradiated from the light irradiation unit 230 may pass through the lower surface of the water tank to cure the liquid mixture contained in the water tank. Since the light transmittance of the liquid mixture is low, the molded product produced by curing the liquid mixture according to the light irradiated from the light irradiation unit 230 forms a surface-divided layer of the molded product.

In one embodiment, the cured molding may be created on the lower surface of the printing bed 210 . In order to mold the surface-divided layer of the molding on the lower surface of the printing bed 210, the printing bed 210 may move up and down. For example, similar to FIG. 2B, when the printing bed 210 is spaced apart from the lower surface of the water tank at a predetermined distance, the mixed solution is positioned by the thickness according to the predetermined distance between the printing bed 210 and the lower surface of the water tank will do When the curable light emitted from the light irradiation unit 230 is irradiated to the lower surface of the water tank, the mixture is cured, and accordingly, a surface-divided layer of the molding is generated on the lower surface of the printing bed 210 .

In the same way, after the surface-divided layer of the molding produced on the lower surface of the printing bed 210 is positioned at a predetermined distance from the lower surface of the water tank, the lower surface of the water tank may be irradiated with curing light. Accordingly, on the lower surface of the printing bed 210, a surface-divided layer of the molding may be continuously generated. By repeating this, a molding may be created on the lower surface of the printing bed 210 .

Meanwhile, when the printing bed 210 is moved up and down in the tank, the mixed solution may be non-uniformly present between the lower surface of the printing bed 210 and the lower surface of the water tank due to the viscosity of the mixed solution. The water tank according to an embodiment may further include an agitator 222 as shown in FIG. 2C so that the mixed solution is uniformly present. The stirring unit according to an embodiment may include a blade 222a made of a plate body. The blade may make a linear reciprocating motion on the surface of the liquid mixture in the water tank to make the surface of the liquid mixture in the water tank even.

The agitator 222 may be connected to the actuator 224 to perform linear reciprocating motion. In one embodiment, the actuator 224 may be a motor that provides rotational kinetic energy. Rotational force of the motor may be provided to the stirring unit 222 through the power transmission unit 223 . For example, the power transmission unit 223 may include a ball screw, and the stirring unit 222 may include a screw tube.

On the other hand, as the mixed solution is stirred in the tank as described above, one mixed solution is accommodated in one tank. Accordingly, it is necessary to change the mixed solution in order to form each layer of the molded product with different components. For example, if you want to mold each layer of the molded product in a different color, the mixed solution in the water bath must be replaced each time before molding each layer.

The water tank unit 220 of the 3D printer according to an embodiment may include a plurality of water tanks. Previously, an example in which a 3D printer is configured in which a single water tank is located in a printing space, which is a space between the printing bed 210 and the light irradiation unit 230, has been described with reference to FIGS. 2A and 2D. However, as will be described later, the water tank unit 220 including a plurality of water tanks may be adopted in a 3D printer.

3A to 3C are diagrams illustrating an example of a structure of a 3D printer including a plurality of water tanks disposed on a rectangular plate moving linearly in a 3D printing system according to an embodiment.

3A to 3C , the water tank unit 220 according to an embodiment may include a plurality of water tanks, for example, a first water tank 321, a second water tank 322, and a third water tank ( 323) may be included. In this example, an example including three tanks is described, but the number of tanks is not limited. In addition, the components of the mixed solution contained in the water tank may be different from each other.

Each bath may be accommodated on a plate 320 . An area of the plate 320 where the water tank is located may be made of a light-transmitting material or may be substantially perforated so that the curable light is irradiated to the lower surface of the water tank.

As in the examples of FIGS. 3A to 3C , the plate may have a rectangular shape, and a portion of the plate 320 may be positioned in the printing space 310 of the 3D printer. 3A and 3B conceptually illustrate an example in which the second water tank 322 is located in a printing space. As described above, the printing space may be a space between the printing bed 210 and the light irradiation unit 230 .

In one embodiment, the plate 330 can move left and right. In order to move the plate left and right, the water tank unit 220 may further include a water tank moving unit composed of a gear, belt, wheel, or the like. The water tank moving means may change the position of the water tank by moving the plate 330 under the control of the controller.

As the plate 320 is moved, the water tank located in the printing space may be changed. Accordingly, the 3D printer according to an embodiment may change the water bath used to mold each layer whenever forming each layer of the molded product. For example, when the plate shown in FIGS. 3A and 3B moves to the right, the second water tank 322 moves out of the printing space 310, and the first water tank 321 moves to the printing space 310 as shown in FIG. 3C. ) can be located. Accordingly, the printing bed 210 may move up and down inside the first water tank 321 , and the light irradiation unit 230 may irradiate curable light to the lower surface of the first water tank 321 .

Meanwhile, the stirring unit 222 described above may receive power from the actuator 224 only when the water tank is located in the printing space. In one embodiment, the stirring unit 222 and the power transmission unit 223 are provided for each tank, but only one actuator 224 may be provided near the printing space 310 . For example, when the water tank is located in the printing space 310, the power transmission unit 223 and the actuator 224 are mechanically coupled, and when the water tank is not located in the printing space 310, the power transmission unit 223 and the actuator 224 may be mechanically separated. For example, the power transmission unit 223 and the actuator 224 may be connected through a detachable connection unit such as a gear pair and a clutch.

Another embodiment of the plate will be described below. 4 is a view showing a plate 420 having a curved shape and a plurality of water tanks provided thereon. The water tank unit 220 according to an embodiment may include a circular plate 420 and four water tanks provided thereon. As described above, the number of tanks is not limited, and the components of the mixed solution contained in the tank may be different.

The circular plate 420 may include an actuator 430 for rotating the circular plate 420 . Accordingly, the water tank 421 located in the printing space 410 may be changed by rotating the circular plate 420 . Accordingly, the 3D printer according to an embodiment may change the water bath used to mold each layer by rotating the circular plate 420 whenever each layer of the molded object is molded.

Meanwhile, the plate may be a square plate 520 having a square shape as shown in FIG. 5 . 5 is a view showing a plate 520 having a square shape and a plurality of water tanks provided thereon. As described above, the number of tanks is not limited, and the components of the mixed solution contained in the tank may be different. As the square plate 520 is also rotated by the actuator 530, the water tank 521 located in the printing space 510 may be changed.

As described with reference to FIGS. 3A to 5 , the water tank unit 220 may include a plurality of water tanks. In order to perform a molding operation using a plurality of tubs, the tub unit 220 is rotated or moved so that the tubs are aligned with the printing bed 210 . However, according to another embodiment, the water tank unit 220 is fixed, and the printing bed 210 may be moved to align with the corresponding water tank. To this end, an actuator for moving the printing bed 210 is further included. In this case, according to an embodiment, the light irradiation unit 230 may move together with the printing bed 210 . Alternatively, according to another embodiment, the light irradiator 230 may include a rotatable reflector, and may irradiate light to the water tank where the molding operation is performed by adjusting the posture (eg, angle, height, etc.) of the reflector. Alternatively, according to another embodiment, a plurality of light irradiation units for a plurality of water tanks may be disposed.

The 3D printer described with reference to FIGS. 2 to 5 has a structure for Digital Light Processing (DLP). The DLP method is a method using the principle of partially curing by irradiating light to the bottom of a storage tank in which a photocurable resin is stored. However, according to various embodiments, in addition to the DLP method, a Fused Deposition Modeling (FDM) method in which a solid plastic material that melts in heat is pulled out like a skein and melted little by little and stacked, a principle in which laser light is injected into a photocurable resin and the scanned part is cured According to other methods, such as SLA (Stereo Lithography Apparatus) method using SLA method and SLS (Selective Laser Sintering) method using the principle of using functional polymer or metal powder instead of photocurable resin and forming by scanning laser beam to solidify A 3D printer may be used for various embodiments of the present invention.

The aforementioned 3D printer 110 interacts with the learning server 120 . Specifically, the learning server 120 learns an AI algorithm based on the learning data, and provides data for 3D printing generated using the learned AI algorithm to a 3D printer. Here, AI means that machines such as computers perform thinking, learning, and analysis possible with human intelligence. Recently, the technology that applies such AI to the 3D printing industry is increasing. To implement AI, Artificial Neural Networks (ANNs) are widely used. An example of an artificial neural network applicable to the present invention is shown in FIG. 6 below.

6 is a diagram showing the structure of an artificial neural network applicable to a 3D printing system according to an embodiment. Referring to FIG. 6 , an artificial neural network includes an input layer 610, at least one hidden layer 620, and an output layer 630. Each of the layers 610, 620, and 630 is composed of a plurality of nodes, and each node is connected to an output of at least one node belonging to the previous layer. Each node adds a bias to the inner product of each output value of the nodes in the previous layer and the corresponding connection weight, and then generates a non-linear activation function The output value multiplied by is delivered to at least one neuron in the next layer.

Artificial neural network models used in various embodiments of the present invention include a fully convolutional neural network, a convolutional neural network (CNN), a recurrent neural network (RNN), and a restricted Boltzmann machine. It may include at least one of a Boltzmann machine (RBM) and a deep belief neural network (DBN), but is not limited thereto. Alternatively, machine learning methods other than deep learning may also be included. Alternatively, a hybrid model combining deep learning and machine learning may also be included. For example, a deep learning-based model may be applied to extract features of an image, and a machine learning-based model may be applied when the image is classified or recognized based on the extracted features. The machine learning-based model may include a support vector machine (SVM), AdaBoost, and the like, but is not limited thereto.

When 3D printing is performed using the 3D printer having the above-described structure, the mixed solution is cured according to light irradiation, and one slice is created through the process of irradiation and curing. By bringing the formed slices into contact with the liquid mixture again and curing, new slices are produced layer by layer. When the process of light irradiation and curing is repeatedly performed, an artificial tooth is formed as a plurality of slices are stacked.

The 3D printed artificial tooth then goes through a sintering process. Firing refers to a process of applying sufficient temperature and pressure to make particles having a large specific surface area, such as powder, into a more dense mass. Through the firing process, the density, porosity, pore size and size distribution of the material can be adjusted, and finally, desired material properties can be realized. Physical properties may be improved through the firing process, but as shown in FIG. 7 , it is common that a change in shape occurs.

7 is a diagram illustrating an example of a shape change by a firing process of a 3D printed artificial tooth in a 3D printing system according to an embodiment. Referring to FIG. 7 , a firing process is performed on the 3D printed artificial tooth 702 . The fired artificial teeth 704 may include various modifications. For example, compared to the artificial tooth 702 before firing, the fired artificial tooth 704 may have a shorter length in at least one of the x-axis, y-axis, or z-axis. In addition, partial shape change, that is, shrinkage, distortion, etc., may occur for a part rather than the whole.

It is not easy to correct the shape change by the firing process as shown in the example of FIG. 7 ex post facto. Therefore, according to an embodiment of the present invention, the artificial tooth may be 3D printed in a shape in which the correction items are pre-applied in consideration of the distortion by reflecting the change due to the firing process in advance. To this end, 3D model data designed to reflect the correction in consideration of distortion is used, and a procedure for generating the 3D model data and performing 3D printing is shown in FIG. 8 below.

8 is a diagram illustrating a procedure for performing 3D printing in a 3D printing system according to an embodiment. That is, FIG. 8 shows 3D printing data in which items to be corrected are pre-reflected in consideration of distortion due to firing from a current pre-firing artificial tooth image using a neural network model showing a correlation between an artificial tooth image before firing and an artificial tooth image after firing. This is an example of a procedure for generating and performing 3D printing based thereon. 8 illustrates the operation and signal exchange of the 3D printer 110 and the learning server 120.

Referring to FIG. 8 , in step S801, the learning server 120 acquires learning data. The training data is an input data set for generating a 3D mapping model in step S803. For example, the input data set may include an artificial tooth image before firing and an artificial tooth image after firing. Here, the image may have a 2D or 3D image format. For example, as the 2D image, a pre-firing image of the 3D output product of the artificial tooth and a post-firing image of the 3D output product of the artificial tooth may be used. Alternatively, as the 3D image, a pre-firing image of the 3D output product of the artificial tooth and a post-firing image of the 3D output product of the artificial tooth may be used. Here, as the 2D or 3D image, a 2D image or a 3D image of modeling data for generating a 3D output product of the artificial tooth may be used instead of an image of the 3D output product of the artificial tooth before firing. Here, the image before firing may be a 2D or 3D image of 3D modeling data or 3D output in which distortions generated by firing are pre-reflected.

Additionally, the input data set may further contain information about firing conditions. The information on the firing conditions may include at least one of temperature, time, material component ratio of the artificial tooth to be fired, and a manufacturing method of the artificial tooth to be fired. Furthermore, the input data set may further include additional information (eg, material constituting the artificial tooth, method used for printing, type of 3D printer, model name of the 3D printer, etc.).

In step S803, the learning server 120 generates a 3D mapping model. The learning server 120 creates an artificial tooth 3D mapping model using the input data set. Here, the 3D mapping model is an artificial neural network model based on an AI algorithm, and is an algorithm that infers 3D model data for 3D printing by using a target tooth image or the like as an input. In other words, the learning server 120 performs learning (eg, machine learning, deep learning) using the learning data. In this case, the pre-sintering artificial tooth image and the post-sintering artificial tooth image may be pre-processed in the same direction and the same length. For example, the target tooth image and the slice color set are expressed in the same coordinate system, and corresponding points in each coordinate system may represent parts corresponding to each other. According to an embodiment, the artificial neural network model to be trained may include a convolutional neural network (CNN) model.

In step S805, the learning server 120 receives target image and firing condition information. For example, the target image may include an image related to a tooth to be replaced with an artificial tooth. Specifically, the target image is the tooth to be replaced with the artificial tooth to be 3D printed (e.g., the second left molar when the artificial tooth is to be fabricated) or another tooth of the same person (e.g., the artificial tooth of the left second molar). If you want to make it as a tooth, include an image of a tooth other than the second small molar on the left. Here, the other tooth is a tooth that is symmetrical to the target tooth (e.g., the right second molar when the left second molar is to be fabricated as an artificial tooth) or an adjacent tooth (e.g., the left second molar when the left second molar is to be fabricated as an artificial tooth) or the left first tooth at least one of a premolar or a left first molar). Although not shown in FIG. 8 , additional information required for an inference operation for generating 3D model data may be further received in addition to the tooth image.

In step S807, the learning server 120 generates 3D model data. The learning server 120 infers 3D model data from a tooth image using a 3D mapping model. For example, the learning server 120 may determine an artificial tooth image before firing from a tooth image using a 3D mapping model. As described above, the artificial tooth image before firing may be a 2D image or a 3D image of the artificial tooth 3D output before firing, or a 2D image or 3D image of modeling data for generating an artificial tooth 3D output. When the image of the artificial tooth before firing is a 2D image or a 3D image of the 3D output of the artificial tooth before firing, 3D modeling data corresponding thereto may be inferred as a result value.

Specifically, the learning server 120 may obtain 3D model data by converting a tooth image into input data according to the structure of an artificial neural network and inputting the input data into the artificial neural network. In other words, the learning server 120 may obtain corresponding 3D model data as a result by inputting the target tooth image to the artificial tooth 3D mapping model. Here, the 3D model data is a blueprint for 3D printing and may include a modeling image before firing. For example, the modeling image is an artificial tooth image before shape change by plasticity, and may have a 2D or 3D format. By using the 3D mapping model, 3D model data can be obtained considering various deformations caused by plasticity, for example, length reduction of a specific axis (eg, the z axis of FIG. 7), partial distortion, discoloration, and the like.

In step S809, the learning server 120 transmits 3D model data to the 3D printer 110. In step S811, the 3D printer 110 performs 3D printing. At this time, the 3D printer 110 performs 3D printing based on the 3D model data provided from the learning server 120 .

In the procedure described with reference to FIG. 8 , the learning server 120 generates a 3D mapping model using learning data. For learning, the learning server 120 may extract learning factors from the tooth image included in the learning data. For example, the learning factors may include at least one of a position of a tooth expressed in an image, a photographing angle of a tooth, a color of a tooth, and an age of an image provider. If the learning factor to be extracted is provided as additional information, the extraction operation may be omitted.

In the procedure described with reference to FIG. 8 , the learning server 120 generates a 3D mapping model using the acquired learning data. In this case, additionally, images included in the learning data (eg, images before firing and images after firing) may be preprocessed for effective learning. According to an embodiment, a coordinate axis graphic may be added to each image to facilitate aligning and contrasting an image before firing and an image after firing in the same coordinate system. In this case, similar pre-processing may be performed on the target image even when 3D model data is generated. That is, prior to performing step S807, a coordinate axis graphic may be added to the target image. The pre-processing operation may be performed by the learning server 120 or in a separate device. A separate AI algorithm for the pre-processing operation may be used.

In addition, according to another embodiment, a marking indicating a portion where deformation exists may be added by comparing an image before firing and an image after firing. For example, marking may be performed by painting a portion deformed by firing in a pre-firing image in a predefined color. At this time, the system may define a level according to the degree of deformation, and different color values may be allocated according to the level. At this time, the color may be painted with a certain transparency. The pre-processing operation may be performed by the learning server 120 or in a separate device. A separate AI algorithm for the pre-processing operation may be used.

In the procedure described with reference to FIG. 8 , the learning server 120 converts a pre-firing image in which distortion due to firing is pre-reflected (eg, a 2D image or 3D image of a 3D output, or a 2D image or 3D image of a 3D model for 3D output). ) and a 3D mapping model with an image (e.g., 2D image or 3D image) after firing the artificial tooth 3D output according thereto, and to generate an input tooth image using the generated 3D mapping model and the input tooth image. It has been described that 3D model data can be created.

On the other hand, in one embodiment, the learning server 120 is a pre-firing image in which distortion due to firing is not reflected (eg, a 2D image or 3D image of a 3D output, or a 2D image or 3D image of a 3D model for 3D output) A 3D mapping model may be generated by inputting a post-firing image (eg, a 2D image or a 3D image) of the 3D output of the artificial tooth and the resultant artificial tooth as a learning input pair of the neural network model. In addition, 3D model data for generating an input tooth image may be generated using the 3D mapping model and the input tooth image. Since such a learning input pair does not reflect the distortion due to firing in advance, it may include an input pair for firing results determined to be unqualified after firing. However, since such an input pair also includes a feature expressed in a plastic object during firing (eg, distortion due to firing), it can be used as an effective input pair for training a 3D mapping model composed of a neural network. Therefore, the 3D mapping model generation step (S803) of FIG. 8 is performed using only the learning input pair in which distortion is pre-reflected, or is performed using only the learning input pair in which distortion is not reflected, or the distortion is pre-reflected in the learning input pair and This may be performed using a learning input pair generated by mixing learning input pairs in which distortion is not reflected at a predetermined ratio. In this case, the learning input pair may be labeled with information for distinguishing a learning input pair in which distortion is pre-reflected and a learning input pair in which distortion is not reflected.

In the procedure described with reference to FIG. 8 , the learning server 120 infers 3D model data from the received tooth image. Here, the 3D model data expresses the shape of an artificial tooth in which deformation by plasticity is reflected in advance. This may be determined by primarily generating 3D model data that does not consider deformation due to plasticity and modifying the primarily generated initial 3D model data.

Therefore, according to another embodiment, 3D model data that does not consider plasticity other than the tooth image, ie, initial 3D model data, may be used as input data for reasoning of the learning server 120 . That is, the target tooth image may be replaced with initial 3D model data. In this case, the learning server 120 may modify the distortion in which the 3D output object before firing, which is generated according to the initial 3D model data, is deformed according to firing, by reflecting it in advance on the initial 3D model data. That is, the 3D mapping model may be a neural network model for generating 3D model data considering plasticity by modifying initial 3D model data.

In order to create such a neural network model, when learning a 3D mapping model, 3D model data designed without considering plasticity may be used as learning data, and additionally, an image of a corresponding tooth may be further used. According to this embodiment, existing 3D model data can be utilized in the practice of the present invention without newly obtaining a tooth image by modifying already generated 3D model data without considering a predetermined value.

In the procedure described with reference to FIG. 8 , the learning server 120 provides 3D model data to the 3D printer 110 so that the 3D printer 110 prints artificial teeth based on the 3D model data. At this time, additionally, before the learning server 120 provides the 3D model data to the 3D printer 110, a verification procedure for the 3D model data inferred using the 3D mapping model may be performed. For example, the verification procedure may be performed based on an expected result of the printed artificial tooth estimated based on the determined 3D model data.

According to an embodiment, the learning server 120 may perform verification based on image comparison. To this end, the learning server 120 generates an expected result image of the artificial tooth based on the 3D model data. Then, the learning server 120 compares the target image and the expected result image received in step S805. A matter to be confirmed through comparison is whether or not the printed artificial tooth is smaller than the artificial tooth expressed in the target image. In general, artificial teeth are reduced by a firing process. Therefore, it is preferable that the printed artificial tooth is not smaller than the result expressed in the target image.

For this comparison, the learning server 120 extracts evaluation indicators from each of the expected result image and the target image. For example, the indicator may include a distance between specific points, an area of a specific region, and the like. When a 3D image is used, a distance between at least one predefined point for comparison and a tooth center may be measured as an index. When a 2D image is used, at least one of the area, width or height of at least one predefined region on the tooth surface can be measured as an index. If the index measured in the expected result image is greater than the index measured in the target image, the learning server 120 determines that the inferred 3D model data is appropriate, and transmits the 3D model data to the 3D printer 110 .

In the description of the verification procedure described above, it is assumed that a set of 3D model data is inferred using a 3D mapping model. However, depending on the structure of the 3D mapping model, multiple sets of 3D model data can be inferred. In this case, the above-described verification operation may be utilized as an operation of selecting one of a plurality of sets. That is, by excluding sets that do not satisfy the verification criterion (eg, the index measured in the expected result image is greater than the index measured in the target image), an appropriate set may be selected.

Furthermore, a plurality of expected result images may be derived by using the 3D model data determined using the 3D mapping model as an input of the 3D mapping model again. In this case, the satisfaction may be determined based on the number of images satisfying the satisfaction condition among the total number of expected result images, and the satisfaction index may be expressed in a percentage form. In this case, one suitable set described above may be determined as a set having the highest satisfaction percentage.

In the procedure described with reference to FIG. 8 , the learning server 120 performs learning using learning data including images, predetermined conditions, and the like. According to another embodiment, the learning server 120 may perform additional learning using information generated after firing the artificial teeth printed through the procedure of FIG. 8 . In order to obtain information generated after firing, a separate device (hereinafter referred to as 'feedback device') may be further used.

The feedback device acquires an image of the fired artificial tooth and transmits it to the learning server 120 . For example, the feedback device is a device that can be connected to a communication network, and can acquire an image (hereinafter referred to as a 'result image') of an artificial tooth fired by a user's manipulation (eg, shooting, offline input, etc.) of the feedback device. can After receiving the resulting image, the learning server 120 compares the received target image with the resulting image in step S805. As a result of the comparison, if the error between the target image and the resulting image is greater than the threshold value, the learning server 120 performs additional learning using the 3D model data and result image generated to print the artificial tooth. Here, the error between the target image and the result image may be evaluated using the index used in the verification procedure described above.

9 is a diagram illustrating a structure of local 3D printing systems interoperable in a 3D printing system according to an embodiment. 9 illustrates a case in which a plurality of 3D printing systems exist.

Referring to FIG. 9 , a plurality of local 3D printing systems 910a, 910b, and 910c may be connected to a central server 920 through a communication network. Each of the plurality of local 3D printing systems 910a, 910b, and 910c corresponds to the 3D printing system of FIG. 1 and includes at least one learning server and at least one 3D printer.

According to an embodiment, the central server 920 may provide learning data for use in learning AI algorithms. That is, when a new local 3D printing system is installed, the central server 920 may help the new local 3D printing system reach a usable state quickly by providing learning data to the new local 3D printing system. Learning data stored in the central server 920 may be provided by a service provider. Alternatively, the central server 920 may store data uploaded from the local 3D printing systems 910a, 910b, and 910c as learning data.

According to an embodiment, the central server 920 may support transfer learning. Transfer learning is a learning method that performs learning using parameters of AI algorithms that have already been learned. The central server 920 receives parameters (eg, weights) of the artificial neural network learned from the local 3D printing systems 910a, 910b, and 910c, and transmits the received parameters to the local 3D printing system to perform learning. can do.

According to another embodiment, instead of the learning server being included in each of the 3D printing systems 910a, 910b, and 910c, the central server 920 is included in each of the 3D printing systems 910a, 910b, and 910c. The function of the learning server for the printer can be performed in an integrated manner. In this case, an artificial neural network for a 3D mapping model may be shared among a plurality of local 3D printing systems.

The 3D printing system according to one embodiment described above may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the medium may be specially designed and configured according to embodiments or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.

Each of the drawings referred to in the description of the previous embodiment is only one embodiment shown for convenience of explanation, and items, contents and images of information displayed on each screen may be modified and displayed in various forms.

Although the present invention has been described with reference to an embodiment shown in the drawings, this is only exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the attached claims.

Claims (10)

In a 3-dimension (3D) printing system,
A 3D printer that controls to print an artificial tooth composed of a plurality of slices; and
A server for learning an artificial neural network for data necessary for a printing operation of the 3D printer and determining 3D model data from a target image of a tooth or initial 3D model data of the tooth using the learned artificial neural network and
The 3D printer prints the artificial tooth according to the 3D model data determined by the server,
The artificial neural network determines 3D model data in which items to be corrected are pre-reflected in consideration of deformation due to sintering applied to the artificial tooth,
The plurality of slices include a first slice and a second slice,
The first slice is created by contact with the first mixed solution and curing of the first mixed solution according to light irradiation,
The second slice is created by contact with the second mixed solution and curing of the second mixed solution according to the irradiation of light,
The firing is a 3D printing system including an operation of applying a designated temperature and a designated pressure to the printed tooth.
According to claim 1,
The target image includes an image of a tooth at a position other than the tooth to be replaced by the artificial tooth.
According to claim 1,
The 3D printing system further comprises a server that acquires the target image and determines 3D model data in which the items to be corrected are pre-reflected using the artificial neural network.
According to claim 1,
The server determines the 3D model data by modifying the initial 3D model data by pre-reflecting the items to be corrected.
According to claim 1,
The server acquires learning data including a tooth image, adds a coordinate axis graphic to the tooth image, and uses the tooth image to which the coordinate axis graphic is added to learn the artificial neural network.
According to claim 1,
The server acquires learning data including a tooth image or initial 3D model data, adds a marking indicating a portion deformed by plasticity to the tooth image or the initial 3D model data, and A 3D printing system for learning the artificial neural network using the added tooth image or the initial 3D model data.
According to claim 1,
The server receives learning data from a central server accessible through an external communication network, and uses the learning data to learn the artificial neural network.
According to claim 1,
The server receives parameter information of the learned artificial neural network from a central server accessible through an external communication network, and transfer-learns the artificial neural network using the parameter information.
In the method of 3D printing an artificial tooth in a 3-dimension (3D) printing system,
Learning an artificial neural network for data necessary for a printing operation of a 3D printer;
determining 3D model data from a target image of a tooth or initial 3D model data of the tooth using the learned artificial neural network; and
Printing an artificial tooth composed of a plurality of slices according to the 3D model data,
The artificial neural network determines 3D model data in which items to be corrected are pre-reflected in consideration of deformation due to sintering applied to the artificial tooth,
The plurality of slices include a first slice and a second slice,
The first slice is created by contact with the first mixed solution and curing of the first mixed solution according to light irradiation,
The second slice is created by contact with the second mixed solution and curing of the second mixed solution according to the irradiation of light,
The firing method comprises an operation of applying a specified temperature and a specified pressure to the printed tooth.
A computer-readable recording medium on which a computer program for performing the method of claim 9 is recorded.

Images