Hair damage assessment system and hair damage assessment method

Abstract

One embodiment of the present disclosure discloses a hair damage assessment method comprising the steps of: acquiring an image including hair from an image acquisition device; identifying a hair portion in the image; randomly extracting a predetermined number of patches from the identified hair portion; determining whether the extracted patches are damaged on the basis of a pre-trained model; outputting a total damage score of the image on the basis of the number of patches determined as damaged patches among the randomly extracted patches; and visualizing a damaged portion on the image on the basis of the determination result.

Description

Hair Damage Assessment System and Hair Damage Assessment Method

[0001] The present disclosure relates to a hair damage assessment system and a hair damage assessment method, and more specifically, to a hair damage assessment system and a hair damage assessment method that evaluates the degree of hair damage based on a pre-trained model and visualizes the damaged parts of the hair.

[0002] Social interest in hair health is increasing. To treat damaged hair or maintain and manage its health, it is first necessary to determine the degree of damage. Traditionally, the degree of hair damage was assessed based on individual subjective criteria. Since evaluation standards inevitably vary depending on the evaluator, it was difficult to ensure the reliability of existing hair damage assessments. Furthermore, even when the degree of damage was assessed, it was difficult to intuitively identify the damaged areas at a glance.

[0003] The aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot be considered as publicly known technology disclosed to the general public prior to the filing of the present invention.

[0004] The present disclosure aims to solve the aforementioned problems by providing a hair damage assessment system and a hair damage assessment method that evaluate the degree of hair damage and visualize the damaged parts of the hair based on a pre-trained model. However, these problems are exemplary, and the problems to be solved by the present invention are not limited thereto.

[0005] One embodiment of the present disclosure discloses a method for evaluating hair damage, comprising the steps of: acquiring an image containing hair from an image acquisition device; identifying a hair portion in the image; randomly extracting a predetermined number of patches from the identified hair portion; determining whether the extracted patches are damaged based on a pre-trained model; outputting a total damage score of the image based on the number of patches determined to be damaged among the randomly extracted patches; and visualizing the damaged portion on the image based on the determination result.

[0006] The above image acquisition device may include a scanning electron microscope.

[0007] The above image acquisition device may include an optical microscope.

[0008] In the above determination step, the pre-trained model can output a patch damage value ranging from 0 to 1 depending on the degree of damage for each patch.

[0009] In the above determination step, if the patch damage value output corresponding to the patch exceeds a first limit value, the patch is determined to be the damaged patch, and if the patch damage value is less than the first limit value, the patch is determined to be the non-damaged patch.

[0010] In the above-mentioned damaged portion visualization step, for a patch identified as a damaged patch among the randomly extracted patches, a first pixel value can be added to an existing pixel value of a predetermined color.

[0011] In the above-mentioned damaged portion visualization step, a second pixel value is added to an existing pixel value of a predetermined color regarding the patch, and the second pixel value may be a value obtained by multiplying the output patch damage value by a predetermined weight.

[0012] In the above-mentioned damaged portion visualization step, a patch damage value may be assigned to the center point of the patch, and a third pixel value may be assigned to all points of the image based on density estimation.

[0013] In the above damage score output step, the total damage score of the image may be proportional to the number of patches identified as damaged among the randomly extracted patches.

[0014] Another embodiment of the present disclosure discloses a hair damage evaluation system comprising an image acquisition device and a server that acquires an image containing hair from the image acquisition device, identifies a hair portion in the image, randomly extracts a predetermined number of patches from the identified hair portion, determines whether the extracted patches are damaged based on a pre-trained model, outputs a total damage score of the image based on the number of patches determined to be damaged among the patches, and displays a damaged portion on the image based on the determination result.

[0015] The above image acquisition device may include a scanning electron microscope.

[0016] The above image acquisition device may include an optical microscope.

[0017] The above-mentioned pre-trained model can output a patch damage value ranging from 0 to 1 depending on the degree of damage for each patch.

[0018] The server can determine the patch as a damaged patch if the patch damage value output in response to the patch exceeds a first limit value, and determine the patch as an undamaged patch if the patch damage value is less than the first limit value.

[0019] The above server can add a first pixel value to an existing pixel value of a predetermined color for a patch identified as a damaged patch among the randomly extracted patches.

[0020] The server adds a second pixel value to an existing pixel value of a predetermined color with respect to the patch, and the second pixel value may be a value obtained by multiplying the output patch damage value by a predetermined weight.

[0021] The server can assign a patch damage value to the center point of the patch and assign a third pixel value to all points of the image based on density estimation.

[0022] The total damage score of the above image may be proportional to the number of patches identified as damaged among the randomly selected patches.

[0023] Other aspects, features, and advantages other than those described above will become clear from the following specific details, claims, and drawings for implementing the invention.

[0024] The hair damage evaluation system and hair damage evaluation method according to the embodiments of the present disclosure can improve the reliability of the evaluation results by outputting the total damage score of an image containing hair based on a pre-trained model according to objective criteria.

[0025] The hair damage assessment system and hair damage assessment method according to the embodiments of the present disclosure can intuitively and quickly identify damaged areas by visualizing damaged areas on an image containing hair.

[0026] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from this specification and the accompanying drawings.

[0027] FIG. 1 is a schematic diagram showing the configuration of a hair damage evaluation system according to one embodiment of the present invention.

[0028] FIG. 2 is a schematic diagram showing the configuration of a server according to one embodiment of the present invention.

[0029] FIG. 3 is a diagram illustrating an exemplary structure of a model learned by a server according to an embodiment of the present invention.

[0030] FIGS. 4 and 5 are drawings for explaining the process of a server learning a model using a plurality of sample hair images according to an embodiment of the present invention.

[0031] FIG. 6 is a diagram illustrating the process of a server determining whether a patch is damaged according to an embodiment of the present invention.

[0032] FIG. 7 is a diagram illustrating the process of visualizing a damaged part according to an embodiment of the present invention.

[0033] Figure 8 is a photograph showing the result of visualizing the damaged area along Figure 7.

[0034] FIG. 9 is a diagram illustrating the process of visualizing a damaged part according to another embodiment of the present invention.

[0035] Figures 10 and 11 are photographs showing the results of visualizing the damaged area along Figure 9.

[0036] FIG. 12 is a diagram illustrating the process of visualizing a damaged part according to another embodiment of the present invention.

[0037] Figure 13 is a photograph showing the result obtained by performing density estimation using a Gaussian kernel during the damaged part visualization process according to another embodiment of the present invention.

[0038] Figure 14 is a photograph showing the result obtained by performing density estimation using a cosine kernel during the damaged part visualization process according to another embodiment of the present invention.

[0039] FIG. 15 is a diagram illustrating an example of an initial screen displayed on a user terminal according to an embodiment of the present invention.

[0040] FIG. 16 is a diagram illustrating an example of a result screen displayed on a user terminal according to an embodiment of the present invention.

[0041] FIG. 17 is a flowchart illustrating a method for evaluating hair damage according to one embodiment of the present invention.

[0042] One embodiment of the present disclosure discloses a method for evaluating hair damage, comprising the steps of: acquiring an image containing hair from an image acquisition device; identifying a hair portion in the image; randomly extracting a predetermined number of patches from the identified hair portion; determining whether the extracted patches are damaged based on a pre-trained model; outputting a total damage score of the image based on the number of patches determined to be damaged among the randomly extracted patches; and visualizing the damaged portion on the image based on the determination result.

[0043] The terms used in this invention are used merely to describe specific embodiments and are not intended to limit the scope of other embodiments. A singular expression may include a plural expression unless the context clearly indicates otherwise. Terms used herein, including technical or scientific terms, may have the same meaning as generally understood by those skilled in the art as described in this invention. Terms used in this invention that are defined in a general dictionary may be interpreted as having the same or similar meaning as they have in the context of the relevant technology, and are not interpreted in an ideal or overly formal sense unless explicitly defined in this invention. In some cases, even terms defined in this invention may not be interpreted to exclude embodiments of this invention.

[0044] Below, various embodiments are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present invention. However, since the technical concept of the present invention can be modified and implemented in various forms, it is not limited to the embodiments described in this specification. In describing the embodiments disclosed in this specification, if it is determined that specifically describing related known technologies may obscure the essence of the technical concept of the present invention, such specific description of known technologies is omitted. Identical or similar components are assigned the same reference numerals, and redundant descriptions thereof are omitted.

[0045] In this embodiment, the term "~part" refers to a component that performs a specific function performed by software or hardware such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). However, "~part" is not limited to being performed by software or hardware. "~part" may exist in the form of data stored on an addressable storage medium, or it may be implemented by instructions and configured so that one or more processors execute a specific function.

[0046] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or instruct the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave in order to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media. Software may be read into main memory from other computer-readable media, such as data storage devices, or from other devices via communication interfaces. Software instructions stored in main memory may cause the processor to perform processes or steps, which will be described in detail below. Alternatively, processes consistent with the principles of the present invention may be executed using a fixed wiring circuit instead of or in combination with software instructions. Accordingly, embodiments consistent with the principles of the present invention are not limited to any specific combination of hardware circuits and software.

[0047] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as “comprising” or “having” are intended to indicate the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. Terms such as “first,” “second,” etc., may be used to describe various components, but the components should not be limited by these terms. These terms are used solely for the purpose of distinguishing one component from another.

[0048] The 'learning model' referred to in the present invention may include any form of algorithm or methodology used to learn or understand specific patterns or structures from data. That is, the learning model may include not only machine learning models such as regression models, decision trees, random forests, support vector machines, K-nearest neighbors, Naive Bayes, and clustering algorithms, but also deep learning models such as neural networks, convolutional neural networks, recurrent neural networks, Transformer-based neural networks, GANs (Generative Adversarial Networks), and autoencoders. The 'learning model' may indicate a set of learned parameters or weight values ​​used to predict or classify an output for a specific input, and this model may be trained through methods such as supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Furthermore, it may include various learning methods and structures, such as not only a single model but also ensemble models, multi-modal models, and models through transfer learning. Such a learning model may be pre-trained on a computer device separate from the computer device that predicts the output for the input and used on another computer device.

[0049] A hair damage evaluation system and a hair damage evaluation method according to an embodiment of the present invention will be described below with reference to FIGS. 1 to 17.

[0050] FIG. 1 is a schematic diagram showing the configuration of a hair damage evaluation system according to one embodiment of the present invention.

[0051] A hair damage evaluation system can evaluate the degree of damage to a user's hair using an image containing the user's hair and visualize the damaged parts of the hair. The hair damage evaluation system identifies hair parts in an image containing hair, randomly extracts patches from the hair parts, determines whether the extracted patches are damaged based on a pre-trained model, outputs a total damage score based on the number of patches determined to be damaged, and can visualize the damaged parts on the image. As shown in FIG. 1, the hair damage evaluation system may include a server (100), an image acquisition device (200), a user terminal (300), and a network (400).

[0052] The server (100) can evaluate the degree of damage to an image containing the user's hair. The server (100) can receive an image containing hair from an image acquisition device (200) or a user terminal (300). The server (100) can identify the hair portion in the image containing the hair. The server (100) can randomly extract a predetermined number of patches from the hair portion and determine whether the extracted patches are damaged based on a pre-trained model. The server (100) can output a total damage score of the image based on the number of patches determined to be damaged among the randomly extracted patches. The server (100) can visualize the damaged portion on the image based on the determination result. The server (100) can transmit the total damage score of the image and the visualized image to the user terminal (300).

[0053] FIG. 2 is a schematic diagram showing the configuration of a server (100) according to one embodiment of the present invention.

[0054] Referring to FIG. 2, the server (100) may include a communication unit (110), a first processor (120), a memory (130), and a second processor (140). Additionally, although not shown in the drawing, the server (100) may further include an input / output unit, a program storage unit, etc.

[0055] The communication unit (110) may be a device including hardware and software necessary for the server (100) to transmit and receive signals, such as control signals or data signals, through a wired or wireless connection with another network device, such as an image acquisition device (200) or a user terminal (300).

[0056] The first processor (120) may be a device that controls a series of processes for evaluating the degree of damage to an image containing hair received from an image acquisition device (200) or a user terminal (300).

[0057] In this case, a processor may refer to a data processing device embedded in hardware that has a physically structured circuit to perform functions expressed by code or instructions included in a program, for example. Examples of such data processing devices embedded in hardware may include microprocessors, central processing units (CPUs), processor cores, multiprocessors, application-specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs), but the scope of the present invention is not limited thereto.

[0058] The memory (130) performs the function of temporarily or permanently storing data processed by the server (100). The memory (130) may include magnetic storage media or flash storage media, but the scope of the present invention is not limited thereto. The memory (130) may temporarily and / or permanently store images received from an image acquisition device (200) or a user terminal (300).

[0059] The second processor (140) may refer to a device that performs calculations under the control of the aforementioned first processor (120). In this case, the second processor (140) may be a device having a higher computational capability than the aforementioned first processor (120). The second processor (140) may be composed of a GPU (Graphics Processing Unit). However, this is exemplary and the concept of the present invention is not limited thereto. The second processor (140) may be plural or singular.

[0060] The second processor (140) may provide resources used for computation by a model or artificial neural network implemented by the server (100). The second processor (140) may provide resources used for computation by a model or artificial neural network that evaluates the degree of hair damage from an image containing hair. However, this is exemplary and the scope of the invention is not limited thereto, and a detailed description of the artificial neural network will be provided later.

[0061] The image acquisition device (200) may be a device that acquires an image containing the user's hair and transmits it to a server (100) or a user terminal (300). In one embodiment, the image acquisition device (200) may include a scanning electron microscope (SEM). A scanning electron microscope can obtain an image by intensively scanning an electron beam. In another embodiment, the image acquisition device (200) may include an optical microscope.

[0062] The user terminal (300) can receive and store an image containing hair from the image acquisition device (200). The user terminal (300) may be a device that receives an image containing hair with a total damage score and a visualized damaged part from the server (100) and provides it to the user. In one embodiment, the user terminal (300) may be implemented as a general-purpose information processing device such as a mobile phone, a tablet PC, and a PC. However, this is exemplary and the scope of the invention is not limited thereto.

[0063] The network (400) may refer to a communication network that mediates the transmission and reception of data between each component of the hair damage evaluation system. In one embodiment, the network (400) may include wired networks such as LANs (Local Area Networks), WANs (Wide Area Networks), MANs (Metropolitan Area Networks), and ISDNs (Integrated Service Digital Networks), or wireless networks such as wireless LANs, CDMA, Bluetooth, and satellite communication, but the scope of the present invention is not limited thereto.

[0064] FIG. 3 is a diagram illustrating an exemplary structure of a model learned by a server (100) according to one embodiment of the present invention.

[0065] The model may be an artificial neural network based on a Convolutional Neural Network (CNN) model as shown in FIG. 3. In this case, the CNN model may be a layered model used to extract features from input data by alternately performing multiple computational layers (Convolutional Layer, Pooling Layer). In the following, the terms "model" and "artificial neural network" may be used interchangeably and may have the same meaning.

[0066] The first processor (120) can build or train an artificial neural network model by processing training data according to a supervised learning technique. A detailed description of how the first processor (120) trains the model will be provided later.

[0067] The first processor (120) can train an artificial neural network by repeatedly performing the process of updating the weight values ​​of each layer and / or each node so that the output value generated by inputting one of the input data into the artificial neural network approaches the value labeled in the corresponding training data using a plurality of training data.

[0068] At this time, the first processor (120) can update the weight values ​​(or coefficients) of each layer and / or each node according to the back propagation algorithm.

[0069] The first processor (120) can generate a convolution layer for extracting feature values ​​of input data and a pooling layer for combining the extracted feature values ​​to form a feature map.

[0070] Additionally, the first processor (120) can combine the generated feature maps to create a fully connected layer that is prepared to determine the probability that the input data corresponds to each of the multiple items.

[0071] The first processor (120) can produce an output layer that includes an output corresponding to the input data.

[0072] In the example illustrated in FIG. 3, the input data is divided into 5x7 blocks, a 5x3 unit block is used for the creation of the convolution layer, and a 1x4 or 1x2 unit block is used for the creation of the pooling layer; however, this is exemplary and the concept of the present invention is not limited thereto. Accordingly, the type of input data and / or the size of each block can be configured in various ways.

[0073] Meanwhile, such an artificial neural network may be stored in the aforementioned memory (130) in the form of the type of the artificial neural network model, the coefficients of at least one node constituting the artificial neural network, the weight values ​​of the nodes, and the coefficients of a function defining the relationship between multiple layers constituting the artificial neural network. Of course, the structure of the artificial neural network may also be stored in the memory (130) in the form of source code and / or a program.

[0074] The types and / or structures of the models described in FIG. 3 are exemplary and the scope of the invention is not limited thereto.

[0075] FIGS. 4 and FIGS. 5 are drawings for explaining the process of a server (100) according to an embodiment of the present invention learning a model using a sample hair image.

[0076] The server (100) can train a model to evaluate the degree of hair damage of a user. The model may be an artificial neural network trained to output a patch damage value of a patch based on the input of a patch randomly extracted from a hair portion identified in an image containing hair. Such an artificial neural network may be trained in a supervised learning manner based on multiple training data.

[0077] The size of the patch is not specifically limited. In one embodiment, the patch may have a size of 224 pixels in width and height. The patch damage value is a value between 0 and 1 and may be a value assigned according to the degree of damage to the patch.

[0078] To this end, the server (100) may acquire sample hair images for multiple training data. With reference to FIGS. 4 and FIGS. 5, the sample hair images may include damaged hair images (S1, S2, S5, S6) and non-damaged hair images (S3, S4, S7, S8).

[0079] In one embodiment, a sample hair image may be made using human or dog hair. In one embodiment, damaged hair may be produced using a basic solution. This can reproduce cases where hair is damaged during a perming or dyeing process. Multiple damaged hairs may be produced by varying the concentration and contact time of the basic solution. In another embodiment, damaged hair may be produced by directly bleaching or heat treating the hair. However, this is merely an example, and the method of obtaining damaged hair is not limited thereto.

[0080] Damaged hair images and non-damaged hair images can be obtained using an image acquisition device (200) for damaged hair and non-damaged hair. In one embodiment, damaged hair images and non-damaged hair images can be obtained using a scanning electron microscope or an optical microscope for damaged hair and non-damaged hair.

[0081] The server (100) can train the model by dividing it according to the type of image acquisition device (200). Referring to FIG. 4, the server (100) can train the model using sample hair images (S1, S2, S3, S4) obtained using a scanning electron microscope. Referring to FIG. 5, the server (100) can train the model using sample hair images (S5, S6, S7, S8) obtained using an optical microscope.

[0082] The server (100) can obtain a plurality of training data from a sample hair image. The training data may include patches extracted from the sample hair image and patch damage values ​​of the patches.

[0083] The server (100) can train a model using a first processor (120) and / or a second processor (140). The server (100) can input a patch extracted from a sample hair image into the model and update at least one coefficient and / or weight value that constitutes the model so that the output value is close to the patch damage value of the patch. The server (100) can perform training in a direction that reduces the error rate by repeating the above-described process using multiple training data.

[0084] The pre-trained model mentioned in this specification may refer to a model that has been pre-trained according to the aforementioned process.

[0085] The pre-trained model may be a neural network that learns the correlation between a patch extracted from a sample hair image and the patch damage value of said patch based on the aforementioned multiple training data. Accordingly, the pre-trained model can output the patch damage value of said patch as input a patch randomly extracted from a hair portion identified in an image containing hair. In other words, the server (100) inputs a patch randomly extracted from a hair portion identified in an image containing hair into the pre-trained model and obtains the patch damage value of said patch as the output.

[0086] The following describes the process by which a server (100) evaluates the degree of hair damage based on a pre-trained model.

[0087] The server (100) can acquire an image containing hair from an image acquisition device (200). The image acquisition device (200) can photograph hair to be evaluated for damage. The image containing hair obtained by the image acquisition device (200) can be transmitted and stored in a user terminal (300). The server (100) can also acquire an image containing hair from a user terminal (300).

[0088] The server (100) can identify hair portions in an image containing hair. The method by which the server (100) identifies hair portions is not specifically limited. In one embodiment, the server (100) can identify hair portions using a boundary detection filter. The boundary detection filter may include the basic derivative filter of OpenCV (Open Source Computer Vision Library), Roberts Cross Filter, Prewitt Filter, Sobel Filter, Scharr Filter, or Laplacian Filter.

[0089] The server (100) can randomly extract a predetermined number of patches from an identified hair portion. The size of the predetermined number is not specifically limited. In one embodiment, the server (100) can randomly extract 100 to 200 patches from an identified hair portion.

[0090] FIG. 6 is a diagram illustrating the process of a server (100) according to an embodiment of the present invention determining whether a patch is damaged.

[0091] The server (100) can determine whether the extracted patch is damaged based on a pre-trained model. As described above, the pre-trained model can output a patch damage value between 0 and 1 depending on the degree of damage for each patch.

[0092] Referring to FIG. 6, the server (100) can determine that a patch is a damaged patch if the patch damage value output in response to the patch exceeds a first limit value, and determine that a patch is a non-damaged patch if the patch damage value is less than the first limit value. The first limit value is a value between 0 and 1 and can be predetermined according to the evaluation criteria for the degree of damage. The first limit value is not specifically limited. For example, when the first limit value is 0.5, if the patch damage value is 0.76, the patch can be determined as a damaged patch, and if the patch damage value is 0.23, the patch can be determined as a non-damaged patch.

[0093] The server (100) can output a total damage score of the image based on the number of patches identified as damaged among the randomly extracted patches. In one embodiment, the total damage score of the image may be proportional to the number of patches identified as damaged among the randomly extracted patches.

[0094] For example, the damage score can be a value between 0 and 100. If 83 out of 100 randomly selected patches are identified as damaged, the total damage score of the image may be 83. If 17 out of 100 randomly selected patches are identified as damaged, the total damage score of the image may be 17. In this case, a lower total damage score indicates healthier hair, while a higher total damage score indicates more severe hair damage.

[0095] The server (100) can visualize the damaged parts on the image based on the determination result.

[0096] FIG. 7 is a diagram illustrating the process of visualizing a damaged part according to an embodiment of the present invention. FIG. 8 is a photograph showing the result of visualizing the damaged part according to FIG. 7.

[0097] In one embodiment, the server (100) may change the pixel value of a patch depending on whether the patch is determined to be a damaged patch. Referring to FIG. 7, the server (100) may add a first pixel value to an existing pixel value of a predetermined color for a patch determined to be a damaged patch among randomly selected patches. The server (100) may maintain an existing pixel value of a predetermined color for a patch determined to be a non-damaged patch among randomly selected patches.

[0098] The predetermined color is not specifically limited. In one embodiment, the predetermined color may be red. The pixel value is a value representing the color assigned to the patch. In one embodiment, the pixel value may include the R value among RGB. The size of the first pixel value is not specifically limited. In one embodiment, the first pixel value may be a value between 10 and 20.

[0099] For example, when the first pixel value is 10, in the case of a damaged patch among the randomly selected patches, the pixel value representing the red color of the patch can be changed to a value that is 10 higher than the existing pixel value. In the case of a non-damaged patch among the randomly selected patches, the pixel value representing the red color of the patch can be maintained at the existing pixel value.

[0100] An example of the result visualized through the above process is shown in Fig. 8. Fig. 8(a) shows the case where the total damage score of the image is 100 points, and Fig. 8(b) shows the case where the total damage score of the image is 84 points. That is, since there are more patches identified as damaged in the case of Fig. 8(a), it can be seen that the damaged area in the visualized image is wider than in Fig. 8(b). Since the damaged area in the hair is marked in red, the location of the damaged area can be intuitively identified.

[0101] FIG. 9 is a diagram illustrating a process for visualizing a damaged part according to another embodiment of the present invention. FIG. 10 and FIG. 11 are photographs showing the results of visualizing the damaged part according to FIG. 9.

[0102] In another embodiment, the server (100) can change the pixel value of the patch using the patch damage value output by a pre-trained model. Referring to FIG. 9, the server (100) can add a second pixel value to the existing pixel value of a predetermined color for the patch. The second pixel value may be a value obtained by multiplying the output patch damage value by a predetermined weight.

[0103] The predetermined color is not specifically limited. In one embodiment, referring to FIG. 9, the predetermined color may be red. The pixel value is a value representing the color assigned to the patch. In one embodiment, the pixel value may include the R value among RGB. The weight may be any constant multiplied by the patch damage value. The size of the weight is not specifically limited. In one embodiment, the first pixel value may be a value between 10 and 30.

[0104] For example, when the weight is 20, if the patch damage value of a randomly selected patch is 0.82, the second pixel value may be 0.83 × 20 = 16.6. In this case, the pixel value representing red of the corresponding patch may be changed to a value that adds 16.6 to the existing pixel value. Meanwhile, if the patch damage value of a randomly selected patch is 0.37, the second pixel value may be 0.37 × 20 = 7.4. In this case, the pixel value representing red of the corresponding patch may be changed to a value that adds 7.4 to the existing pixel value.

[0105] In other words, the second pixel value added to the existing pixel value can vary depending on the difference in patch damage values. As the patch damage value increases, the second pixel value becomes larger, which can make the predetermined color darker. Conversely, as the patch damage value decreases, the second pixel value becomes smaller, which can make the predetermined color lighter. Therefore, by using patch damage values ​​in visualization, it is possible to intuitively distinguish between severely damaged areas and non-damaged areas.

[0106] Examples of the results visualized through the above process are shown in FIGS. 10 and 11. FIG. 10(a) represents the case where the total damage score of the image obtained from the scanning electron microscope is 100 points, and FIG. 10(b) represents the case where the total damage score of the image obtained from the scanning electron microscope is 84 points. That is, FIG. 10(a) has a greater number of patches identified as damaged than FIG. 10(b). Since a patch damage value must exceed a first threshold value to be identified as a damaged patch, it can be expected that the patch damage values ​​of the patches extracted in FIG. 10(a) will generally be greater than the patch damage values ​​of the patches extracted in FIG. 10(b). Furthermore, even within a single image, the red color becomes darker in areas with more severe damage, so the location of severely damaged areas can be intuitively identified through the visualized image.

[0107] FIG. 11 (a) shows the case where the total damage score of an image obtained from an optical microscope is 100 points, and FIG. 11 (b) shows the case where the total damage score of an image obtained from a scanning electron microscope is 0 points. FIG. 12 is a diagram illustrating the process of visualizing a damaged area according to another embodiment of the present invention. FIG. 13 is a photograph showing the result obtained by performing density estimation using a Gaussian kernel during the process of visualizing a damaged area according to another embodiment of the present invention. FIG. 14 is a photograph showing the result obtained by performing density estimation using a cosine kernel during the process of visualizing a damaged area according to another embodiment of the present invention.

[0108] In another embodiment, the server (100) can change the pixel values ​​of the patch by incorporating density estimation into the patch damage value. In other words, the server (100) can assign a patch damage value to the center point of the patch and assign a third pixel value to all points of the image based on density estimation.

[0109] The third pixel value may be a value proportional to the damage density assigned to all points in the image when considering all patch damage values ​​assigned to the center points of all randomly selected patches. In an image containing hair, integrating the damage density at all points may yield 1.

[0110] The server (100) can assign a third pixel value to every point in the image. That is, the third pixel value can be assigned a different value to each point within a single patch. Additionally, the third pixel value can be assigned to every point in the image in addition to the randomly selected patch. In one embodiment, each point may mean a single pixel.

[0111] For example, referring to FIG. 12, randomly extracted patches may include a first patch (P1), a second patch (P2), a third patch (P3), and a fourth patch (P4). When the first patch (P1) is input into a pre-trained model, a first patch damage value (dv1) may be output. When the second patch (P2) is input into a pre-trained model, a second patch damage value (dv2) may be output. When the third patch (P3) is input into a pre-trained model, a third patch damage value (dv3) may be output. When the fourth patch (P4) is input into a pre-trained model, a fourth patch damage value (dv4) may be output. The first patch damage value (dv1), the second patch damage value (dv2), the third patch damage value (dv3), and the fourth patch damage value (dv4) can be assigned to the center points of the first patch (P1), the second patch (P2), the third patch (P3), and the fourth patch (P4), respectively, as shown in FIG. 10. The server (100) can assign a third pixel value to all points of the image by taking into account the size and location of the first patch damage value (dv1), the second patch damage value (dv2), the third patch damage value (dv3), and the fourth patch damage value (dv4).

[0112] In one embodiment, the density estimation may include kernel density estimation (KDE). Kernel density estimation may include, but is not limited to, a Gaussian kernel, an Epanechnikov kernel, or a cosine kernel. Additionally, in the density estimation, the bandwidth may be determined according to Silverman's rule of thumb.

[0113] Referring to FIGS. 13 and 14, areas with larger patch damage values ​​are expressed in darker colors, allowing for intuitive identification of severely damaged areas. Additionally, since pixel values ​​are not assigned per patch but rather different third pixel values ​​are assigned to each point within a single patch, the boundaries of the damaged areas can be displayed smoothly. Furthermore, since the relative magnitude of patch damage values ​​according to location can be compared within a single image, the degree of damage according to location within a single image can be intuitively identified regardless of the total damage score of the image.

[0114] FIG. 15 is a diagram illustrating an example of an initial screen displayed on a user terminal (300) according to an embodiment of the present invention. FIG. 16 is a diagram illustrating an example of a result screen displayed on a user terminal (300) according to an embodiment of the present invention.

[0115] An example of an initial screen displayed on a user terminal (300) is illustrated in FIG. 15. However, FIG. 15 is merely an example, and the initial screen displayed on the user terminal (300) is not limited to this.

[0116] The user can use the user terminal (300) to obtain an image containing hair or transmit the obtained image to the server (100).

[0117] In one embodiment, when a user places hair on an image acquisition device (200) and presses the "shoot" button on the display screen shown in FIG. 15, the image acquisition device (200), which is connected to the user terminal (300) by a network (400), can acquire an image containing the hair. The image acquisition device (200) can transmit the acquired image to a server (100) and / or the user terminal (300).

[0118] Meanwhile, the user can press the "Find File" button on the initial screen shown in FIG. 15 to transmit an image containing hair stored in the user terminal (300) to the server (100). When the image containing hair is transmitted to the server (100), the server (100) can display the image on the initial screen shown in FIG. 15.

[0119] After an image containing hair is transmitted to the server (100), if the user presses the “Start Evaluation” button on the initial screen shown in FIG. 15, the server (100) can evaluate the degree of hair damage based on a pre-trained model. The evaluation result can be displayed on the result screen of the user terminal (300).

[0120] An example of a result screen displayed on a user terminal (300) is illustrated in FIG. 16. However, FIG. 16 is merely an example, and the result screen displayed on the user terminal (300) is not limited thereto.

[0121] In one embodiment, a result screen displayed on a user terminal (300) may display a total damage score, a damage status, and a visualized image. The damage status may be displayed as bad, dangerous, average, good, etc., according to a predetermined range regarding the total damage score. For example, as shown in Table 1 below, if the total damage score is 0 or more and less than 10, it may be displayed as 'good'; if it is 10 or more and less than 60, it may be displayed as 'average'; if it is 60 or more and less than 90, it may be displayed as 'dangerous'; and if it is 90 or more and less than 100, it may be displayed as 'bad'. However, the above range is merely an example and is not limited thereto.

[0122] Total damage score (x) Damage Condition Good Average Dangerous Poor

[0123] As shown in Fig. 16, for example, if the total damage score is 92 points as a result of evaluating the degree of hair damage, it may be output as "Hair damage: 92%", and since the total damage score is between 90 and 100, it may be output as "Damage condition: Poor". As another example, if the total damage score is 16 points as a result of evaluating the degree of hair damage, it may be output as "Hair damage: 16%", and since the total damage score is between 10 and 60, it may be output as "Damage condition: Average".

[0124] In one embodiment, referring to FIG. 16, a graph that allows the total damage score to be intuitively checked may be displayed on the result screen of the user terminal (300). Although a bar graph is shown in FIG. 16, this is merely an example and the shape or form of the graph is not limited thereto.

[0125] In another embodiment, if the total damage score is lower than a predetermined value, the phrase "Your hair is healthy!" may be displayed on the display screen of the user terminal (300) along with the total damage score (not shown). If the total damage score is higher than a predetermined value, the phrase "Your hair is damaged and requires care" may be displayed on the display screen of the user terminal (300) along with the total damage score (not shown). However, the above phrases are exemplary and are not limited thereto.

[0126] FIG. 17 is a flowchart illustrating a method for evaluating hair damage according to one embodiment of the present invention.

[0127] A hair damage evaluation method may be a method for evaluating the degree of damage to a user's hair and visualizing the damaged parts of the hair using the aforementioned hair damage evaluation system. Referring to FIG. 17, the hair damage evaluation method may include the steps of: acquiring an image containing hair from an image acquisition device (200) (S10); identifying a hair portion in the image (S20); randomly extracting a predetermined number of patches from the identified hair portion (S30); determining whether the extracted patches are damaged based on a pre-trained model (S40); outputting a total damage score of the image based on the number of patches determined to be damaged among the randomly extracted patches (S50); and visualizing the damaged parts on the image based on the determination result (S60).

[0128] The server (100) can acquire an image containing hair from the image acquisition device (200) (S10). The image acquisition device (200) can photograph hair to be evaluated for the degree of damage. In one embodiment, the image acquisition device (200) may include a scanning electron microscope (SEM). In another embodiment, the image acquisition device (200) may include an optical microscope.

[0129] An image containing hair obtained by an image acquisition device (200) can be transmitted to and stored in a user terminal (300). The server (100) may also acquire an image containing hair from the user terminal (300).

[0130] The server (100) can identify hair portions in an image containing hair (S20). The method by which the server (100) identifies hair portions is not specifically limited. In one embodiment, the server (100) can identify hair portions using a boundary detection filter. The boundary detection filter may include the basic derivative filter of OpenCV (Open Source Computer Vision Library), Roberts Cross Filter, Prewitt Filter, Sobel Filter, Scharr Filter, or Laplacian Filter.

[0131] The server (100) can randomly extract a predetermined number of patches from the identified hair portion (S30). The size of the predetermined number is not specifically limited. In one embodiment, the server (100) can randomly extract 100 to 200 patches from the identified hair portion.

[0132] The server (100) can determine whether the extracted patch is damaged based on a pre-trained model (S40). The training process of the pre-trained model is the same or similar as described in FIGS. 4 and FIG. 5, so a detailed description is omitted. As described above, the pre-trained model can output a patch damage value, which is a value between 0 and 1, depending on the degree of damage for each patch.

[0133] Referring to FIG. 6, the server (100) can determine that a patch is a damaged patch if the patch damage value output in response to the patch exceeds a first limit value, and determine that a patch is a non-damaged patch if the patch damage value is less than the first limit value. The first limit value is a value between 0 and 1 and can be predetermined according to the evaluation criteria for the degree of damage. The first limit value is not specifically limited. For example, when the first limit value is 0.5, if the patch damage value is 0.76, the patch can be determined as a damaged patch, and if the patch damage value is 0.23, the patch can be determined as a non-damaged patch.

[0134] The server (100) can output a total damage score of the image based on the number of patches identified as damaged among the randomly extracted patches (S50). In one embodiment, the total damage score of the image may be proportional to the number of patches identified as damaged among the randomly extracted patches. In this case, a lower total damage score indicates that the hair is healthy, while a higher total damage score indicates that the degree of hair damage is severe.

[0135] The server (100) can visualize damaged parts on an image based on the determination result (S60). As an example of visualization, the server (100) can add a first pixel value to an existing pixel value of a predetermined color for a patch identified as a damaged patch among randomly extracted patches. Since the description of this embodiment is the same or similar as that described in FIGS. 7 and FIGS. 8, a detailed description is omitted.

[0136] In another embodiment of visualization, the server (100) may add a second pixel value to an existing pixel value of a predetermined color with respect to the patch. The second pixel value may be a value obtained by multiplying the output patch damage value by a predetermined weight. Since the description of this embodiment is the same or similar as that described in FIGS. 9 to 11, a detailed description is omitted.

[0137] In another embodiment of visualization, the server (100) may assign the patch damage value to the center point of the patch and assign a third pixel value to all points of the image based on density estimation. The third pixel value may be a value proportional to the damage density assigned to all points of the image when considering all patch damage values ​​assigned to the center points of all randomly selected patches. Since the description of this embodiment is the same or similar as that described in FIGS. 12 to 14, a detailed description is omitted.

[0138] When the damage score output step (S50) and the damage part visualization step (S60) are completed, the total damage score of the image and the visualized image can be displayed on a result screen displayed on a user terminal (300). An example of a result screen displayed on a user terminal (300) is illustrated in FIG. 16. However, FIG. 16 is merely an example, and the result screen displayed on a user terminal (300) is not limited thereto.

[0139] Although the present invention has been described with reference to the embodiments illustrated in the drawings, this is merely illustrative. Those skilled in the art will fully understand that various modifications and equivalent alternative embodiments are possible from the embodiments. Accordingly, the true technical scope of protection of the present invention should be determined based on the appended claims.

[0140] The specific technical details described in the embodiments are merely examples and do not limit the technical scope of the embodiments. To make the description of the invention concise and clear, descriptions of general prior art and configurations may be omitted. Furthermore, the connections of lines or connecting members between components depicted in the drawings are illustrative of functional connections and / or physical or circuit connections, and may be replaced or represented by various additional functional, physical, or circuit connections in actual devices. Additionally, unless specifically stated with terms such as "essential" or "importantly," a component may not be strictly necessary for the application of the present invention.

[0141] The term "the above" or similar designations in the description of the invention and claims may refer to both singular and plural forms unless specifically limited otherwise. Furthermore, where a range is described in the embodiments, it is considered to include the invention with respect to individual values ​​within said range (unless otherwise stated), and is equivalent to describing each individual value constituting said range in the description of the invention. Additionally, regarding the steps constituting the method according to the embodiments, the steps may be performed in a suitable order unless explicitly stated or otherwise stated. The embodiments are not necessarily limited by the order in which the steps are described. The use of all examples or exemplary terms (e.g., etc.) in the embodiments is merely for the purpose of describing the embodiments in detail, and the scope of the embodiments is not limited by said examples or exemplary terms unless otherwise limited by the claims. Furthermore, a person skilled in the art will understand that various modifications, combinations, and changes may be made according to design conditions and factors within the scope of the claims or equivalents to which they are added.

[0142] The present invention can be used in the industry regarding a method for evaluating hair damage.

Claims

1. A step of acquiring an image containing hair from an image acquisition device; A step of identifying the hair portion in the above image; A step of randomly extracting a predetermined number of patches from the identified hair portion; A step of determining whether the extracted patch is damaged based on a pre-trained model; A step of outputting a total damage score of the image based on the number of patches identified as damaged among the randomly extracted patches; and A method for evaluating hair damage, comprising the step of visualizing a damaged portion on the image based on the determination result.

2. In Paragraph 1, A method for evaluating hair damage, wherein the image acquisition device described above includes a scanning electron microscope.

3. In Paragraph 1, A method for evaluating hair damage, wherein the image acquisition device described above includes an optical microscope.

4. In Paragraph 1, In the above determination step, A hair damage evaluation method in which the above-mentioned pre-trained model outputs a patch damage value ranging from 0 to 1 depending on the degree of damage for each patch.

5. In Paragraph 4, In the above determination step, A hair damage evaluation method that determines the patch as a damaged patch when the patch damage value output in response to the patch exceeds a first limit value, and determines the patch as a non-damaged patch when the patch damage value is less than the first limit value.

6. In Paragraph 5, In the above damaged area visualization step, A hair damage assessment method that adds a first pixel value to an existing pixel value of a predetermined color for a patch identified as a damaged patch among the randomly extracted patches.

7. In Paragraph 4, In the above damaged area visualization step, With respect to the above patch, a second pixel value is added to an existing pixel value of a predetermined color, and The above second pixel value A hair damage evaluation method, wherein the value obtained by multiplying the output patch damage value by a predetermined weight.

8. In Paragraph 4, In the above damaged area visualization step, A hair damage evaluation method that assigns a patch damage value to the center point of the patch and assigns a third pixel value to all points of the image based on density estimation.

9. In Paragraph 1, In the above damage score output step, A hair damage evaluation method in which the total damage score of the above image is proportional to the number of patches identified as damaged among the randomly selected patches.

10. Image acquisition device; and A hair damage evaluation system comprising: a server that acquires an image containing hair from the image acquisition device, identifies a hair portion in the image, randomly extracts a predetermined number of patches from the identified hair portion, determines whether the extracted patches are damaged based on a pre-trained model, outputs a total damage score of the image based on the number of patches determined to be damaged among the patches, and displays a damaged portion on the image based on the determination result.

11. In Paragraph 10, The above image acquisition device is a hair damage evaluation system comprising a scanning electron microscope.

12. In Paragraph 10, The above image acquisition device is a hair damage evaluation system comprising an optical microscope.

13. In Paragraph 10, A hair damage evaluation system in which the above-mentioned pre-trained model outputs a patch damage value ranging from 0 to 1 depending on the degree of damage for each patch.

14. In Paragraph 13, A hair damage evaluation system in which the server determines the patch as a damaged patch if the patch damage value output in response to the patch exceeds a first limit value, and determines the patch as a non-damaged patch if the patch damage value is less than the first limit value.

15. In Paragraph 14, The above server is A hair damage evaluation system that adds a first pixel value to an existing pixel value of a predetermined color for a patch identified as a damaged patch among the randomly extracted patches.

16. In Paragraph 13, The above server is With respect to the above patch, a second pixel value is added to an existing pixel value of a predetermined color, and The above second pixel value A hair damage evaluation system, which is a value obtained by multiplying the output patch damage value by a predetermined weight.

17. In Paragraph 13, The above server is A hair damage evaluation system that assigns a patch damage value to the center point of the patch and assigns a third pixel value to all points of the image based on density estimation.

18. In Paragraph 10, A hair damage evaluation system in which the total damage score of the above image is proportional to the number of patches identified as damaged among the randomly selected patches.

Images