Apparatus and method for predicting contact angle through extraction of droplet boundary characteristics from low-level images based on learning model

Abstract

A method for predicting a contact angle includes the step of a data processing unit generating training data comprising a droplet image vector, which is a low-resolution droplet image, and a target vector representing a contact angle calculated in correspondence with said droplet image vector, and the step of a model generation unit generating a training model that predicts the contact angle of a droplet through weighted operations on said droplet image vector using said training data.

Description

Technology Field

[0001] The present invention relates to a technology for predicting a contact angle, and more specifically, to an apparatus and a method for predicting a contact angle through the extraction of droplet boundary characteristics from a low-level image based on a learning model. Background Technology

[0002] When a liquid droplet is placed on a solid surface in air, the angle formed between the solid surface and the triplicate point—that is, the point of contact between the three phases (solid, liquid (droplet), and gas (air))—is defined as the angle on the side containing the liquid as the contact angle of the droplet with respect to the solid. Prior art literature

[0003] Korean Published Patent No. 2020-0063981 (Registered June 5, 2020) The problem to be solved

[0004] The objective of the present invention is to provide an apparatus and a method for predicting a contact angle through the extraction of droplet boundary characteristics from a low-level image based on a learning model. means of solving the problem

[0005] A method for predicting a contact angle according to a preferred embodiment of the present invention for achieving the above-described purpose comprises the steps of: a data processing unit generating training data including a droplet image vector, which is a low-level droplet image with reduced resolution, and a target vector representing a contact angle calculated in correspondence with said droplet image vector; and a model generating unit generating a training model that predicts the contact angle of a droplet through weighting operations from said droplet image vector using said training data.

[0006] The step of generating the above training data includes the step of the data processing unit generating a droplet image, the step of the data processing unit calculating the contact angle of the droplet image, and the step of the data processing unit reducing the resolution of the droplet image to generate a droplet image vector, which is a low-level droplet image.

[0007] The step of generating the droplet image vector comprises the step of the data processing unit horizontally cutting an ellipse based on the triple point of the droplet to generate a droplet image having the overall shape of the droplet having a ratio of the major axis to the minor axis of a predetermined range, and the step of the data processing unit extracting a part of the droplet image having the overall shape to generate a droplet image including at least the triple point and the highest point.

[0008] The ratio of the major axis to the minor axis is characterized as being 1.00 to 0.80.

[0009] The step of calculating the contact angle of the above droplet image is performed by the data processing unit using a mathematical formula

[0010]

[0011] A contact angle corresponding to the droplet image is calculated according to the above, wherein θ is the contact angle of the droplet image, a is the major axis of the ellipse forming the basis of the droplet image, b is the minor axis of the ellipse forming the basis of the droplet image, x1 and y1 are the center coordinates of the ellipse forming the basis of the droplet image, and x and y are the coordinates of the triple point of the droplet image.

[0012] The step of generating a droplet image vector, which is a low-level droplet image, is characterized in that the data processing unit generates a low-level droplet image having a first size and a second resolution lower than the first resolution by resizing the droplet image having a first size and a first resolution corresponding to the first size.

[0013] The step of generating the above learning model includes the step of the model generating unit inputting the droplet image vector into the learning model; the step of the learning model performing a plurality of operations on the droplet image vector to which weights that have not yet been learned between a plurality of layers are applied to derive a prediction vector representing the predicted contact angle of the droplet image corresponding to the droplet image vector; the step of the model generating unit deriving a loss representing the difference between the target vector and the prediction vector through a loss function; and the step of the model generating unit performing optimization to update the weights of the learning model so that the loss is minimized.

[0014] A method for predicting a contact angle according to a preferred embodiment of the present invention for achieving the above-described purpose comprises: a step in which a preprocessing unit preprocesses an input droplet image to generate a droplet image vector when a droplet image with an unknown contact angle is input; a step in which an inference unit inputs the droplet image vector to a learning model trained using learning data including a learning droplet image vector, which is a low-level droplet image with reduced resolution, and a target vector representing a contact angle calculated in correspondence with the learning droplet image vector; a step in which the learning model performs a plurality of operations in which weights learned between a plurality of layers are applied to the droplet image vector to derive a prediction vector that predicts the contact angle of the droplet in the droplet image; and a step in which an inference unit outputs the contact angle of the prediction vector.

[0015] The step of generating the droplet image vector includes the step of the preprocessing unit sequentially performing gray scale, Gaussian blur, and binarization on the input droplet image to remove external noise, and the step of the preprocessing unit performing image-to-image operations in a flood fill manner on the droplet image from which external noise has been removed to remove internal noise.

[0016] The above training data is characterized by including a target vector, which is a contact angle calculated from a droplet image generated based on an ellipse, and a droplet image vector, which is a low-level droplet image generated by reducing the resolution of the droplet image generated based on the ellipse.

[0017] An apparatus for predicting a contact angle according to a preferred embodiment of the present invention for achieving the purpose described above comprises: a data processing unit that generates training data including a droplet image vector, which is a low-resolution droplet image, and a target vector representing a contact angle calculated in correspondence with said droplet image vector; and a model generation unit that generates a training model that predicts the contact angle of a droplet through weighting operations from said droplet image vector using said training data.

[0018] The above data processing unit is characterized by generating a droplet image, calculating the contact angle of the generated droplet image, and then reducing the resolution of the generated droplet image to generate a droplet image vector, which is a low-level droplet image.

[0019] The above data processing unit is characterized by horizontally cutting an ellipse based on the triple point of the droplet to generate a droplet image having the overall shape of the droplet having a ratio of the major axis to the minor axis within a predetermined range, and extracting a part of the droplet image having the overall shape to generate a droplet image including at least the triple point and the highest point.

[0020] The ratio of the major axis to the minor axis is characterized as being 1.00 to 0.80.

[0021] The above data processing unit is a mathematical formula

[0022]

[0023] A contact angle corresponding to the droplet image is calculated according to the above, wherein θ is the contact angle of the droplet image, a is the major axis of the ellipse forming the basis of the droplet image, b is the minor axis of the ellipse forming the basis of the droplet image, x1 and y1 are the center coordinates of the ellipse forming the basis of the droplet image, and x and y are the coordinates of the triple point of the droplet image.

[0024] The data processing unit is characterized by generating a low-level droplet image having a first size and a second resolution lower than the first resolution through resizing of the droplet image having a first size and a first resolution corresponding to the first size.

[0025] The above model generation unit inputs the droplet image vector into a learning model, and when the learning model performs multiple operations on the droplet image vector to which weights that have not yet been learned between multiple layers are applied to derive a prediction vector representing the predicted contact angle of the droplet image corresponding to the droplet image vector, it derives a loss representing the difference between the target vector and the prediction vector through a loss function, and performs optimization by updating the weights of the learning model so that the loss is minimized.

[0026] An apparatus for predicting a contact angle according to a preferred embodiment of the present invention for achieving the purpose described above comprises: a preprocessing unit that, when a droplet image with an unknown contact angle is input, preprocesses the input droplet image to generate a droplet image vector; and an inference unit that inputs the droplet image vector to a learning model trained using learning data including a learning droplet image vector, which is a low-resolution droplet image, and a target vector representing a contact angle calculated in correspondence with the learning droplet image vector, and when the learning model performs a plurality of operations in which weights learned between a plurality of layers are applied to the droplet image vector to derive a prediction vector predicting the contact angle of the droplet in the droplet image, outputs the contact angle of the prediction vector.

[0027] The above preprocessing unit is characterized by sequentially performing grayscale, Gaussian blur, and binarization on the input droplet image to remove external noise, and performing image-to-image operations using a floodfill method on the droplet image from which external noise has been removed to remove internal noise.

[0028] The above training data is characterized by including a target vector, which is a contact angle calculated from a droplet image generated based on an ellipse, and a droplet image vector, which is a low-level droplet image generated by reducing the resolution of the droplet image generated based on the ellipse. Effects of the invention

[0029] According to the present invention, even in the case of low-level images, droplet boundary characteristics can be extracted through a learning model to accurately predict the contact angle. Brief explanation of the drawing

[0030] FIG. 1 is a diagram illustrating the configuration of a device for predicting a contact angle through the extraction of droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention. FIG. 2 is a diagram illustrating a learning model for predicting a contact angle through the extraction of droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention. FIG. 3 is a flowchart illustrating a method for generating training data for a learning model according to an embodiment of the present invention. FIG. 4 is an illustrative diagram for explaining a method of generating training data for a learning model according to an embodiment of the present invention. FIG. 6 is a flowchart illustrating a method for predicting a contact angle through the extraction of droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention. FIG. 7 is an illustrative diagram for explaining a method for capturing a droplet image according to an embodiment of the present invention. FIG. 8 is an illustrative diagram explaining a method for predicting a contact angle through the extraction of droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention. FIG. 9 is a drawing showing a computing device according to an embodiment of the present invention. Specific details for implementing the invention

[0031] Before the detailed description of the present invention, it should be understood that the embodiments described in this specification and the configurations illustrated in the drawings are merely the most preferred embodiments of the present invention and do not represent all of the technical ideas of the present invention, and that various equivalents and modifications that can replace them may exist at the time of filing this application.

[0032] Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that identical components in the accompanying drawings are indicated by the same reference numerals whenever possible. Furthermore, detailed descriptions of known functions and configurations that could obscure the essence of the invention will be omitted. For the same reason, some components in the accompanying drawings may be exaggerated, omitted, or schematically depicted, and the size of each component does not entirely reflect its actual size.

[0033] In addition, the terms and words used in the specification and claims described below should not be interpreted as being limited to their ordinary or dictionary meanings, but should be interpreted in a meaning and concept consistent with the technical spirit of the invention, based on the principle that the inventor can appropriately define the concept of the terms to best describe his invention.

[0034] First, we will describe an apparatus for predicting a contact angle by extracting droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention. FIG. 1 is a diagram illustrating the configuration of an apparatus for predicting a contact angle by extracting droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention. FIG. 2 is a diagram illustrating a learning model for predicting a contact angle by extracting droplet boundary characteristics of a low-level image based on a learning model according to an embodiment of the present invention.

[0035] Referring to FIG. 1, a prediction device (10) according to an embodiment of the present invention includes a data processing unit (100), a model generation unit (200), a preprocessing unit (300), and an inference unit (400).

[0036] The data processing unit (100) is intended to generate training data according to an embodiment of the present invention. The training data includes a droplet image vector, which is a low-level droplet image, and a target vector, which is a contact angle obtained from a high-level droplet image corresponding to the droplet image vector. Here, the high-level droplet image refers to an image having a resolution greater than or equal to the size of the image, and the low-level droplet image refers to an image having a resolution less than or equal to the size of the image. The data processing unit (100) generates a high-level droplet image, calculates a target vector, which is a contact angle from the high-level droplet image, and then reduces the resolution of the droplet image to generate a droplet image vector, which is a low-level droplet image. Here, the droplet image vector may be a binary representation of the pixel values ​​of the droplet image.

[0037] The model generation unit (200) is intended to generate a learning model (LM) through learning using training data (Deep Learning). The model generation unit (200) generates a learning model (LM) having learned weights by training a prototype of a learning model (LM) having unlearned weights using training data (Deep Learning). According to an embodiment of the present invention, the learning model (LM) that has completed learning receives a droplet image vector, which is a droplet image, as input, and can predict the contact angle of the droplet image to produce a prediction vector, which is the predicted contact angle. The learning model (LM) may be a Convolutional Neural Network (CNN). The learning model (LM) may include an input layer (IL), at least one pair of alternating convolution layers (CL) and pooling layers (PL), at least one fully-connected layer (FL), and an output layer (OL). An example of such a learning model (ML) is a convolutional neural network, as shown in Fig. 2.

[0038] As described, the learning model (LM) comprises multiple layers, namely, an input layer (IL), a convolution layer (CL), a pooling layer (PL), a fully connected layer (FL), and an output layer (OL) in sequence. Each of the multiple layers receives a value obtained by applying weights to the operation result of the previous layer, performs an operation on the weighted input value using a predetermined activation function, and transmits the operation result to the next layer. The operation result of the final layer is then output as the output value of the learning model (LM), i.e., a prediction vector. This prediction vector represents the predicted contact angle of the droplet.

[0039] The input layer (IL) serves as a buffer for receiving a droplet image vector, which is a droplet image, and is formed according to the specifications of the droplet image so that each of the multiple pixels of the input droplet image vector maintains position information (e.g., pixel coordinates).

[0040] The convolution layer (CL) is composed of at least one first feature map (FM 1: Feature Map). When a droplet image vector is input to the input layer (IL), the convolution layer (CL) performs a convolution operation using a first filter (or kernel) on the input data of the input layer (IL), applies weights, and performs an operation using an activation function to derive a plurality of first feature maps (FM1).

[0041] The pooling layer (PL) consists of at least one second feature map (FM 2: Feature Map 2). The pooling layer (PL) performs a pooling (or sub-sampling) operation using a second filter (or kernel) on at least one first feature map (FM1) of the convolution layer (CL) to derive at least one second feature map (FM2).

[0042] The fully connected layer (FL) consists of multiple operation nodes (f1 to fn). The multiple operation nodes (f1 to fn) of the fully connected layer (FL) produce multiple operation values ​​through operation by an activation function on at least one second feature map (FM2) of the pooling layer (PL). Optionally, a flattening layer (FLL) may be interposed between the pooling layer (PL) and the fully connected layer (FL) to modify the form of the value of the previous layer to match the form of the value of the next layer and input it to the next layer. That is, the flattening layer (FLL) performs the role of modifying the value of the second feature map (FM2), which is in matrix form of the pooling layer (PL), to match the operation node of the fully connected layer (FL) and inputting it. For example, if the second feature map of the pooling layer (PL) is in the form of (rows × columns × channels), the flattening layer (FLL) can convert it into the form of the operation node of the fully connected layer (FL), which is (columns × channels), and input it to that operation node.

[0043] The output layer (OL) includes a single output node (O). Each of the multiple operation nodes (f1 to fn) of the completion connection layer (FL) is connected to the output node (O) of the output layer (OL) via a channel having a weight (W). In other words, multiple operation values ​​of the multiple operation nodes (f1 to fn) are input to the output node (O) with weights applied. Accordingly, the output node (O) of the output layer (OL) calculates a prediction vector through an operation using an activation function on the multiple operation values ​​of the completion connection layer (FL) to which weights are applied. This prediction vector represents the contact angle of the droplet in the input droplet image vector.

[0044] Examples of activation functions used for computation in the aforementioned convolution layer (CL), pooling layer (PL), complete connection layer (FL), and output layer (OL) include sigmoid, hyperbolic tangent (tanh), exponential linear unit (ELU), rectified linear unit (ReLU), leaky reLU, Maxout, minout, and softmax. The convolution layer (CL), complete connection layer (FL), and output layer (OL) perform computations by selecting one of these activation functions.

[0045] Referring again to FIG. 1, the preprocessing unit (300) is intended to generate a droplet image vector through preprocessing of the captured droplet image. The preprocessing unit (300) removes external noise by sequentially performing grayscale, Gaussian blur, and binarization on the input droplet image. Then, the preprocessing unit (300) removes internal noise by performing image-to-image operations on the droplet image from which external noise has been removed using a floodfill method. Subsequently, the droplet image from which internal noise has been removed can be vertically cut, and the image from which the cut left portion is extracted can be generated as a droplet image vector.

[0046] The inference unit (400) is intended to derive a prediction vector that predicts the contact angle of the droplet in the droplet image by analyzing the droplet image vector using a learned learning model (LM) that has completed learning.

[0047] Next, a method for generating training data for a learning model according to an embodiment of the present invention will be described. FIG. 3 is a flowchart illustrating a method for generating training data for a learning model according to an embodiment of the present invention. FIG. 4 is an example diagram illustrating a method for generating training data for a learning model according to an embodiment of the present invention.

[0048] Referring to FIG. 3, the data processing unit (100) generates a high-level droplet image in step S110. Here, a high-level droplet image refers to an image having a resolution greater than or equal to the size of the image. For example, if the size of the image is 300×300 pixels, it refers to an image with a resolution greater than or equal to 300×300 pixels.

[0049] At this time, the data processing unit (100) horizontally cuts the ellipse based on the triple point of the droplet to generate a high-level droplet image having the overall shape of the droplet with a ratio of the major axis to the minor axis of a predetermined range. In this droplet image, the inside of the droplet boundary is 0 (black) and the outside is 1 (white). Then, the data processing unit (100) extracts a part of the droplet image having the overall shape to generate a droplet image including at least the triple point and the highest point.

[0050] For example, as shown in FIG. 4, an ellipse (e) having a ratio of a major axis (a) to a minor axis (b) within a predetermined range is provided. At this time, the ratio of the major axis to the minor axis is preferably 1.00 to 0.80. Then, a triple point Q(x, y) of the droplet is set, and a droplet image (fdi) having the overall shape of the droplet having a ratio of a major axis to a minor axis within a predetermined range is generated by cutting horizontally based on the set triple point Q(x, y).

[0051] Next, the data processing unit (100) calculates a target vector, which is a contact angle, from a high-level droplet image in step S120. At this time, the data processing unit (100) can calculate the contact angle according to the following mathematical formula 1.

[0052]

[0053] Here, θ represents the contact angle of the droplet image, i.e., the target vector. Referring to FIG. 4, a represents the major axis of the ellipse forming the basis of the droplet image, and b represents the minor axis of the ellipse forming the basis of the droplet image. Also, x1 and y1 are the center coordinates P(x1, y1) of the ellipse forming the basis of the droplet image, and x and y represent the coordinates Q(x, y) of the triple point of the droplet image.

[0054] Next, the data processing unit (100) reduces the resolution of the droplet image in step S130 to generate a droplet image vector, which is a low-level droplet image. A low-level droplet image refers to an image having a resolution less than the size of the image. For example, if the size of the image is 300×300 pixels, it refers to an image with a resolution less than 300×300 pixels.

[0055] At this time, the data processing unit (100) can generate a low-level droplet image having a first size and a second resolution lower than the first resolution by resizing a high-level droplet image having a first size and a first resolution corresponding to the first size.

[0056] For example, assume that the first size of the image is 300×300 pixels and the first resolution is 300×300 pixels. Here, while maintaining the first size of the image at 300×300 pixels, a low-level droplet image can be generated by reducing the resolution from the first resolution to the second resolution of 50×50 pixels through resizing.

[0057] Next, a method for generating a learning model using learning data according to an embodiment of the present invention will be described. FIG. 5 is a flowchart illustrating a method for generating a learning model using learning data according to an embodiment of the present invention.

[0058] Referring to FIG. 5, the model generation unit (200) can receive training data generated as described with reference to FIG. 3 and FIG. 4 in step S210. When the training data is received, the model generation unit (200) inputs the droplet image vector into the training model (LM) in step S220.

[0059] Then, in step S230, the learning model (LM) performs multiple operations on the droplet image vector to which weights that have not yet been learned between multiple layers are applied, thereby deriving a prediction vector representing the predicted contact angle of the droplet in the droplet image corresponding to the droplet image vector.

[0060] Accordingly, the model generation unit (200) can calculate a loss representing the difference between the prediction vector and the target vector through a loss function in step S240. That is, the model generation unit (100) can calculate a loss representing the difference between the predicted contact angle of the prediction vector and the calculated contact angle of the target vector through a loss function.

[0061] Next, the model generation unit (200) performs optimization to update the weights of the learning model (LM) so that the loss derived through the loss function in step S250 is minimized.

[0062] The aforementioned steps S220 through S250 are performed repeatedly using multiple different learning data, and the weights of the learning model (LM) are repeatedly updated according to these repetitions. This repetition is repeated until a predetermined condition is satisfied. According to one embodiment, the predetermined condition may be that the loss converges and is less than a predetermined value. Accordingly, the model generation unit (100) determines in step S260 whether the loss calculated previously (S240) converges and is less than or equal to a preset target value, and if the loss is less than or equal to the preset target value, the learning of the detection model (DM) is completed in step S270.

[0063] Next, a method for predicting a contact angle by extracting droplet boundary features from a low-level image based on a learning model according to an embodiment of the present invention will be described. FIG. 6 is a flowchart illustrating a method for predicting a contact angle by extracting droplet boundary features from a low-level image based on a learning model according to an embodiment of the present invention. FIG. 7 is an illustrative diagram illustrating a method for capturing a droplet image according to an embodiment of the present invention. FIG. 8 is an illustrative diagram illustrating a method for predicting a contact angle by extracting droplet boundary features from a low-level image based on a learning model according to an embodiment of the present invention.

[0064] Referring to FIG. 6, the preprocessing unit (300) may receive a droplet image with an unknown contact angle at step S310. This droplet image is captured by a camera (73) while illuminating it with light through a light source (72) when a droplet (D) is dropped onto a substrate (S) through a needle (71), as shown in FIG. 8. In this way, the droplet image generated through the capture is in a state where the contact angle is unknown.

[0065] When a droplet image is input, the preprocessing unit (300) preprocesses the input droplet image in step S320 to generate a droplet image vector. At this time, as illustrated in FIG. 8, the preprocessing unit (300) sequentially performs grayscale, Gaussian blur, and binarization on the input droplet image to remove external noise. Then, the preprocessing unit (300) removes internal noise by performing image-to-image operations on the droplet image from which external noise has been removed using a floodfill method. Subsequently, the droplet image from which internal noise has been removed is vertically cut, and the image from which the cut left portion is extracted is generated as a droplet image vector.

[0066] When a droplet image vector is generated, the inference unit (400) inputs the droplet image vector into a learning model (LM) that was previously learned in the manner described in Fig. 5 at step S330.

[0067] Then, at step S340, the learning model (LM) performs multiple operations in which learned weights between multiple layers are applied to the droplet image vector to derive a prediction vector that predicts the contact angle of the droplet corresponding to the droplet image vector. Then, at step S350, the inference unit (400) outputs the contact angle of the prediction vector.

[0069] FIG. 9 is a drawing showing a computing device according to an embodiment of the present invention. The computing device (TN100) of FIG. 9 may be a device described in the present specification, for example, a prediction device (10).

[0070] In the embodiment of FIG. 9, the computing device (TN100) may include at least one processor (TN110), a transceiver (TN120), and a memory (TN130). Additionally, the computing device (TN100) may further include a storage device (TN140), an input interface device (TN150), an output interface device (TN160), etc. The components included in the computing device (TN100) may be connected by a bus (TN170) to communicate with each other.

[0071] The processor (TN110) can execute a program command stored in at least one of the memory (TN130) and the storage device (TN140). The processor (TN110) may refer to a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed. The processor (TN110) may be configured to implement the procedures, functions, and methods described in relation to embodiments of the present invention. The processor (TN110) can control each component of the computing device (TN100).

[0072] Each of the memory (TN130) and the storage device (TN140) can store various information related to the operation of the processor (TN110). Each of the memory (TN130) and the storage device (TN140) may be composed of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory (TN130) may be composed of at least one of read-only memory (ROM) and random access memory (RAM).

[0073] The transmitting and receiving device (TN120) can transmit or receive wired or wireless signals. The transmitting and receiving device (TN120) can be connected to a network to perform communication.

[0074] In particular, the data processing unit (100), model generation unit (200), preprocessing unit (300), and inference unit (400) according to an embodiment of the present invention may be implemented in the form of a program readable through a computing device and stored in memory (TN130), then executed on a processor (TN110), or implemented as a lower module of the processor (TN110).

[0075] Meanwhile, the method according to the embodiment of the present invention described above may be implemented in the form of a program readable through various computer means and recorded on a computer-readable recording medium. Here, the recording medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the recording medium may be those specifically designed and configured for the present invention, or those known and available to those skilled in the art of computer software. For example, the recording medium includes magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions may include machine language wires, such as those generated by a compiler, as well as high-level language wires that can be executed by a computer using an interpreter, etc. Such hardware devices may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

[0076] Although the present invention has been described above using several preferred embodiments, these embodiments are illustrative and not limiting. As such, those skilled in the art will understand that various changes and modifications can be made in accordance with the doctrine of equivalents without departing from the spirit of the invention and the scope of rights set forth in the appended claims. Explanation of the symbols

[0077] 10: Prediction device 100: Data processing unit 200: Model Creation Section 300: Preprocessing section 400: Inference section

Claims

Claim 1 A method for predicting a contact angle comprises: a step in which a data processing unit generates training data including a droplet image vector, which is a low-level droplet image with reduced resolution, and a target vector representing a contact angle calculated corresponding to the droplet image vector; and a step in which a model generation unit generates a training model that predicts the contact angle of a droplet through weighted operations on the droplet image vector using the training data; wherein the step of generating training data comprises: a step in which the data processing unit generates a droplet image; a step in which the data processing unit calculates the contact angle of the droplet image; and a step in which the data processing unit generates a droplet image vector, which is a low-level droplet image, by reducing the resolution of the droplet image; and wherein the step of calculating the contact angle of the droplet image comprises a mathematical formula A method for predicting a contact angle, characterized by calculating a contact angle corresponding to the droplet image according to the above, wherein θ is the contact angle of the droplet image, a is the major axis of the ellipse forming the basis of the droplet image, b is the minor axis of the ellipse forming the basis of the droplet image, x1 and y1 are the center coordinates of the ellipse forming the basis of the droplet image, and x and y are the coordinates of the triple point of the droplet image. Claim 2 delete Claim 3 A method for predicting a contact angle according to claim 1, wherein the step of generating the droplet image vector comprises: a step in which the data processing unit horizontally cuts an ellipse based on the triple point of the droplet to generate a droplet image having the overall shape of the droplet having a ratio of the major axis to the minor axis of a predetermined range; and a step in which the data processing unit extracts a part of the droplet image having the overall shape to generate a droplet image including at least the triple point and the highest point. Claim 4 A method for predicting a contact angle according to claim 3, characterized in that the ratio of the major axis to the minor axis is 1.00 to 0.

80. Claim 5 delete Claim 6 A method for predicting a contact angle according to claim 1, wherein the step of generating a droplet image vector, which is a low-level droplet image, is characterized in that the data processing unit generates a low-level droplet image having a first size and a second resolution lower than the first resolution through resizing of the droplet image having a first size and a first resolution corresponding to the first size. Claim 7 A method for predicting a contact angle according to claim 1, wherein the step of generating the learning model comprises: a step in which the model generating unit inputs the droplet image vector into the learning model; a step in which the learning model performs a plurality of operations on the droplet image vector to which weights that have not yet been learned between a plurality of layers are applied to derive a prediction vector representing the predicted contact angle of the droplet image corresponding to the droplet image vector; a step in which the model generating unit derives a loss representing the difference between the target vector and the prediction vector through a loss function; and a step in which the model generating unit performs optimization to update the weights of the learning model so that the loss is minimized. Claim 8 A method for predicting a contact angle according to claim 1, further comprising: a step in which a preprocessing unit preprocesses an input droplet image to generate a droplet image vector when a droplet image with an unknown contact angle is input; a step in which an inference unit inputs the droplet image vector to a learning model; a step in which the learning model performs a plurality of operations in which weights learned between a plurality of layers are applied to the droplet image vector to derive a prediction vector that predicts the contact angle of the droplet in the droplet image; and a step in which the inference unit outputs the contact angle of the prediction vector. Claim 9 A method for predicting a contact angle according to claim 8, wherein the step of generating the droplet image vector comprises: a step in which the preprocessing unit sequentially performs gray scale, Gaussian blur, and binarization on the input droplet image to remove external noise; and a step in which the preprocessing unit performs inter-image operations on the droplet image from which external noise has been removed using a flood fill method to remove internal noise. Claim 10 delete Claim 11 An apparatus for predicting a contact angle comprises: a data processing unit that generates training data including a droplet image vector, which is a low-level droplet image with reduced resolution, and a target vector representing a contact angle calculated corresponding to the droplet image vector; and a model generation unit that generates a training model for predicting the contact angle of a droplet through weighted operations on the droplet image vector using the training data; wherein the data processing unit generates a droplet image, calculates the contact angle of the generated droplet image, and then generates a droplet image vector, which is a low-level droplet image, by reducing the resolution of the generated droplet image, and the data processing unit [uses] a mathematical formula An apparatus for predicting a contact angle, characterized by calculating a contact angle corresponding to the droplet image according to the above, wherein θ is the contact angle of the droplet image, a is the major axis of the ellipse forming the basis of the droplet image, b is the minor axis of the ellipse forming the basis of the droplet image, x1 and y1 are the center coordinates of the ellipse forming the basis of the droplet image, and x and y are the coordinates of the triple point of the droplet image. Claim 12 delete Claim 13 An apparatus for predicting a contact angle according to claim 11, wherein the data processing unit generates a droplet image having the overall shape of the droplet having a ratio of the major axis to the minor axis of a predetermined range by horizontally cutting an ellipse based on the triple point of the droplet, and generates a droplet image including at least the triple point and the highest point by extracting a part of the droplet image having the overall shape. Claim 14 An apparatus for predicting a contact angle according to claim 13, characterized in that the ratio of the major axis to the minor axis is 1.00 to 0.

80. Claim 15 delete Claim 16 An apparatus for predicting a contact angle according to claim 11, wherein the data processing unit generates a low-level droplet image having a first size and a second resolution lower than the first resolution by resizing the droplet image having a first size and a first resolution corresponding to the first size. Claim 17 An apparatus for predicting a contact angle according to claim 11, wherein the model generation unit inputs the droplet image vector into a learning model, and when the learning model performs multiple operations with weights applied to the droplet image vector that have not yet completed learning between multiple layers to derive a prediction vector representing the predicted contact angle of the droplet image corresponding to the droplet image vector, derives a loss representing the difference between the target vector and the prediction vector through a loss function, and performs optimization to update the weights of the learning model so that the loss is minimized. Claim 18 An apparatus for predicting a contact angle according to claim 11, further comprising: a preprocessing unit that preprocesses an input droplet image to generate a droplet image vector when a droplet image with an unknown contact angle is input; and an inference unit that inputs the droplet image vector to a learning model, and when the learning model performs a plurality of operations in which weights learned between a plurality of layers are applied to the droplet image vector to derive a prediction vector predicting the contact angle of the droplet in the droplet image, the inference unit that outputs the contact angle of the prediction vector. Claim 19 An apparatus for predicting a contact angle according to claim 18, wherein the preprocessing unit sequentially performs grayscale, Gaussian blur, and binarization on the input droplet image to remove external noise, and performs image-to-image operations on the droplet image from which external noise has been removed using a floodfill method to remove internal noise. Claim 20 delete

Images