Gray hair identification device, gray hair identification method, and gray hair identification program

Abstract

To provide a gray hair identification device, a gray hair identification method, and a gray hair identification program which can accurately identify whether or not hair is gray hair.SOLUTION: The gray hair identification device identifies gray hair and comprises: a feature quantity acquisition unit which acquires feature quantities relating to hair of an identification object person; and a determination unit which determines whether or not his or her hair is gray hair on the basis of a plurality of feature quantities acquired by the feature quantity acquisition unit. As the feature quantities, at least two of information of a color of the hair and information of thickness of the hair are set.SELECTED DRAWING: Figure 9

Description

Technical Field

[0001] The present invention relates to a white hair discrimination device, a white hair discrimination method, and a white hair discrimination program for discriminating whether hair is white or not.

Background Art

[0002] Since white hair has a great impact on the visual impression, many products that promote restoring white hair to black hair or dyeing white hair are provided. However, the evaluation based on the visual impression is affected not only by the variation due to the ability of the evaluator and environmental factors but also by, for example, the illusion due to using products related to white hair countermeasures. Therefore, a method for objectively and highly accurately evaluating and discriminating white hair in hair is desired.

[0003] On the other hand, a conventionally known evaluation method described in Patent Document 1 for evaluating the amount of white hair of an evaluation target person according to an index related to the redness of the scalp or redness is known.

[0004] According to the above document, a hair treatment agent suitable for the state of hair can be selected based on the amount of white hair evaluated from information related to the color of the scalp, the magnitude of variation in hair thickness, the fine hair rate, etc. However, this is an evaluation method for evaluating the state of hair based on the color of the scalp and does not actually identify white hair in hair. Therefore, there is a problem that the state of white hair of the target person may not be accurately grasped.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] An object of the present invention is to provide a white hair discrimination device, a white hair discrimination method, and a white hair discrimination program that can accurately discriminate whether hair is white or not. [Means for solving the problem]

[0007] To solve the above problems, the present invention firstly provides a gray hair identification device for identifying gray hair, comprising a determination unit that determines whether or not the hair is gray based on a plurality of feature quantities relating to the hair of the person to be identified, wherein at least two of the feature quantities are set as information on the color of the hair and information on the thickness of the hair.

[0008] Secondly, it is characterized by having a feature acquisition unit that acquires the aforementioned feature quantities.

[0009] Thirdly, the feature acquisition unit is characterized by comprising: an image acquisition unit that acquires information about an image taken such that it includes at least a portion of at least one hair from the scalp of the person to be identified; and a feature extraction unit that extracts each of a predetermined number of features from the image acquired by the image acquisition unit.

[0010] Fourth, the determination unit is characterized by determining whether or not the hair is gray based on a pre-trained model that has been trained in advance to determine whether or not the hair is gray based on a plurality of features.

[0011] Fifth, the determination unit is characterized by determining whether the hair in the captured image acquired by the captured image acquisition unit is gray, based on a trained model that has been trained in advance to determine whether the hair is gray or not based on a plurality of features.

[0012] Sixth, the pre-trained model is characterized by the use of a model that has been pre-trained using supervised learning.

[0013] Seventh, the determination unit is characterized in that it limits the hair to be determined to those with a thickness of 20 μm or more.

[0014] Eighth, the aforementioned feature quantity is characterized by including information about the brightness of the hair.

[0015] Ninth, the captured image includes at least a portion of the scalp from which hair grows, and the feature quantities also include information on the color of the scalp that can be extracted from the captured image.

[0016] Tenth, the feature acquisition unit includes a related information acquisition unit that acquires information on the age of the person to be identified, and is characterized in that the information on the age of the person to be identified is also set as a feature.

[0017] Eleventh, Executed by a computer A method for identifying gray hair, comprising a step of performing a determination process to determine whether or not the hair is gray based on multiple feature quantities related to the hair of the person to be identified, characterized in that at least two of the feature quantities are set as information on the color of the hair and information on the thickness of the hair.

[0018] Twelfth, the present invention is characterized by having a feature acquisition step for acquiring the aforementioned feature quantities.

[0019] Thirteenth, the feature acquisition step includes a step of performing an image acquisition process to acquire information about an image taken such that the image includes at least a portion of at least one hair from the scalp of the person to be identified, and a step of performing a feature extraction process to extract at least two features from the image: information about the hair color and information about the hair thickness.

[0020] The 14th feature is that the computer is made to perform a determination process to determine whether or not the hair is gray based on multiple features related to the hair of the person to be identified, and at least two of these features are set as information on hair color and information on hair thickness.

[0021] Fifteenth, it is characterized by executing a feature acquisition process to obtain the aforementioned feature quantities.

[0022] Sixteenth, the feature amount acquisition process includes a photographed image acquisition process for acquiring information on a photographed image that is photographed so as to include at least a part of at least one hair, and the photographed image acquisition process of the photographed image acquired Information and a feature amount extraction process for extracting at least two of the information on the color of the hair and the information on the thickness of the hair as feature amounts.

Advantages of the Invention

[0023] The gray hair discrimination device can discriminate at least one hair from an evaluation site image acquired from an evaluation target person and extract appropriate feature amounts, so that it can accurately discriminate whether each hair shown in the evaluation site image is gray hair or not. Therefore, the state of the gray hair of the evaluation target person can be evaluated more objectively with higher accuracy.

Brief Description of the Drawings

[0024] [Figure 1] It is a block diagram showing an example of the configuration of the gray hair discrimination device according to the present invention. [Figure 2] It is a flowchart showing an example of the information processing flow in the gray hair discrimination device according to the present invention. [Figure 3] (A) is a diagram showing an example of an evaluation site image, and (B) is a diagram showing an example of performing hair discrimination on the evaluation site image. [Figure 4] It is a graph showing the RGB (Red) data of all hairs and the determination result by an evaluator. [Figure 5] It is a graph showing the RGB (Green) data of all hairs and the determination result by an evaluator. [Figure 6] It is a graph showing the RGB (Blue) data of all hairs and the determination result by an evaluator. [Figure 7] It is a graph showing the thickness data of all hairs and the determination result by an evaluator. [Figure 8] It is a scatter diagram showing the thickness and luminance of all hairs. [Figure 9]This table shows the parameters used in machine learning and the accuracy evaluation. [Modes for carrying out the invention]

[0025] The following describes an example of a gray hair identification device with reference to the drawings. Figure 1 is a block diagram showing an example of the configuration of a gray hair identification device according to the present invention. As shown in Figure 1, the gray hair identification device 10 of the present invention comprises an image acquisition unit 11, a hair identification unit 12, a feature acquisition unit (feature acquisition means) including a feature extraction unit 13 and a related information acquisition unit 14, a determination unit 15, an output unit 16, and a storage unit 17.

[0026] The gray hair recognition device 10 may be an information processing device designed as a dedicated machine, but it should be implementable using a general-purpose computer or server. In other words, the gray hair recognition device 10 is equipped with a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), memory, storage such as a hard disk drive or SSD, and a communication device for connecting to a communication network, which are typically found in general-purpose computers and server devices, and these are connected via a bus. The processing in each part of the gray hair recognition device 10 is realized by reading a program for executing the processing in each part from memory and executing it in the CPU or GPU, which functions as a control unit (control circuit). In other words, the execution of the program enables the processor (processing circuit) to execute the processing of each device.

[0027] Furthermore, while this description focuses on a standalone device, it is not limited to this configuration. It may also be integrated into a server device accessible via a communication network, or the server device may be configured to perform inference based on a pre-trained model. In addition, configurations that connect to other devices via a communication network are also possible in areas not specifically mentioned.

[0028] The image acquisition unit 11 has the function of acquiring images of the scalp of the person being evaluated (hereinafter referred to as evaluation area images) (see Figure 2). The image acquisition unit 11 acquires evaluation area images by taking pictures of the scalp of the person being evaluated using a digital camera or the like, but any means that can take pictures of the scalp of the person being evaluated is acceptable.

[0029] To explain the specific image acquisition method, the hair in the area of ​​the subject's scalp to be evaluated (1 square centimeter in the illustrated example) is cut short, and the area to be evaluated is photographed at close range with a digital camera. This obtains an image of the evaluation area (captured image) showing the scalp and hair. The length of the hair after cutting is not particularly limited as long as the feature extraction unit can extract features, but is preferably about 0.5 to 20 mm, more preferably about 1 to 10 mm, even more preferably about 1 to 5 mm, and most preferably about 1 to 3 mm, with about 1 mm being the most preferable.

[0030] The hair identification unit 12 is configured to analyze the evaluation area image, identify each individual hair within the evaluation area image, and assign an identification number to each hair identified within the evaluation area image (see Figure 3(B)). The system may also be configured to acquire positional information for each identified hair within the evaluation area image.

[0031] The feature extraction unit 13 is configured to analyze the evaluation area image and acquire data on the RGB of the hair, the brightness of the hair, the thickness of the hair, and the RGB of the scalp for each hair identified within the evaluation area image. The data acquired by the feature extraction unit 13 can be used as parameter values ​​(prediction parameters, features, explanatory variables) for determining whether or not a hair is gray.

[0032] The features extracted here are as follows: RGB (Red, Green, Blue) of hair is the RGB value for each individual hair (0-255 each); brightness of hair is the value obtained by converting the RGB of hair to grayscale (0-255); hair thickness is the diameter of a single hair; and RGB (Red, Green, Blue) of scalp is the RGB value (0-255) of the scalp within the shooting range where the evaluation area image was taken.

[0033] The feature acquisition unit described above (image acquisition unit 11, hair identification unit 12, feature extraction unit 13) is not limited to the above means as long as it can acquire the aforementioned features related to hair, and may also be configured to include an input unit for inputting features acquired by other methods.

[0034] The related information acquisition unit 14 is configured to acquire data such as the age information and scalp care status of the person being evaluated by having them answer questions or input information themselves. Each piece of data acquired by the related information acquisition unit 14 can be used as a parameter value for determining gray hair.

[0035] The age information of the subjects evaluated, obtained here, was the age of the subject at the time the evaluation area image was taken, and was classified into one of the following age groups: 20s, 30s, 40s, 50s, or 60s, and used as a parameter value.

[0036] The determination unit 15 receives hair information identified by the hair identification unit 12, and sets the RGB of the hair, the brightness of the hair, the thickness of the hair, the RGB of the scalp, and the age of the person being evaluated, extracted by the feature extraction unit 13, as prediction parameters (features, explanatory variables). Based on this, it is configured to execute a gray hair determination process that classifies (determines) whether each hair in the evaluation area image is gray or not (whether it is gray or black) based on a pre-trained learning model (supervised learning model).

[0037] In this case, the determination unit 15 sets at least the RGB values ​​of the hair (and the brightness of the hair) and the thickness of the hair as prediction parameters, so that even if there are individual differences in the color and shade of hair other than gray hair (black hair) among evaluators, gray hair can be determined (classified) with high accuracy. Furthermore, by limiting the thickness of the hair used for gray hair determination to hair that is somewhat thick (20 μm, 40 μm), the accuracy of gray hair determination can be further improved. More details will be described later.

[0038] Incidentally, the gray hair identification device 10 of the present invention uses one of the following as a classification-type supervised learning model (machine learning algorithm): Logistic regression, decision tree, random forest, K-nearest neighbors, SVM (support vector machine), or XGBoost (gradient boosting). However, it is not limited to these, as long as it is a model that can classify hair as gray or not (gray or black hair).

[0039] The output unit 16 is a monitor that displays the gray hair determination results made by the determination unit 15, or a printer that can print on paper, etc., and is configured to output the gray hair determination results, specifically the total number of hairs in the evaluation area image, the number of gray hairs (determined as gray hairs) in the image, the number of non-gray hairs, information on the thickness of the gray hairs, and information on the distribution of gray hairs in the image.

[0040] The memory unit 17 is a storage device such as a hard disk drive or SSD, and has the function of storing information necessary for processing each part of the gray hair recognition device 10, and also storing various types of information generated during the processing of each part.

[0041] Next, the information processing flow (process) in the gray hair identification device will be explained based on Figure 2. Figure 2 is a flowchart showing an example of the information processing flow in the gray hair identification device according to the present invention. First, as shown in Figure 2, a feature acquisition process (captured image acquisition process S101 → hair identification process S102 → feature extraction process S103) is performed to acquire the feature quantities of the hair of the person to be identified. Then, a determination process (S104) is performed to determine gray hair based on the feature quantities, and then an output process (S105) is performed to output the determination result.

[0042] To explain in more detail, the gray hair identification device 10 is started by performing an image acquisition process in which the image acquisition unit 11 acquires an image of the evaluation area by taking a photograph of the scalp of the person to be evaluated for gray hair determination (step S101).

[0043] Next, the gray hair identification device 10 analyzes the evaluation area image acquired in step S101 using the hair identification unit 12 to identify each individual hair in the evaluation area image and performs a hair identification process in which it assigns an identification number to each hair in the evaluation area image for identification purposes (step S102).

[0044] Next, the gray hair identification device 10 analyzes each hair in the evaluation area image acquired in steps S101 and S102 using the feature extraction unit 13 to acquire data on the RGB of the hair, the brightness of the hair, the thickness of the hair, and the RGB of the scalp for each hair in the evaluation area image, and acquires data such as age information about the person being evaluated, which is input by the related information acquisition unit 14, thereby executing a feature extraction process to acquire prediction parameters (features, explanatory variables) for gray hair determination (step S103).

[0045] Next, the determination unit 15 of the gray hair identification device 10 sets necessary data as prediction parameters from the data extracted in step 103, such as the RGB of the hair, the brightness of the hair, the thickness of the hair, the RGB of the scalp, and age information, and performs a determination process to classify whether each hair in the evaluation area image is gray or not (gray or black) based on a machine learning (supervised learning) model that has been trained in advance (step S104).

[0046] Next, the output unit 16 of the gray hair identification device 10 outputs gray hair information based on the gray hair determination result output in step S104 (step S105), and the process ends.

[0047] According to the above configuration, the gray hair identification device 10 can identify each individual hair in the evaluation area image, extract information that serves as a prediction parameter for each hair, and determine whether each hair is gray or not. As a result, it is possible to output various information related to gray hair with high accuracy, such as the total number of hairs in the evaluation area image, the number of gray hairs among the total number of hairs, the distribution of gray hairs in the image, and the thickness of gray hairs, thereby enabling a more objective evaluation of the gray hair condition of the evaluator. Below, a specific example of a gray hair determination method using the gray hair identification device 10 described above will be explained. [Examples]

[0048] Next, specific examples of gray hair detection using the gray hair identification device will be described based on Figures 3 to 9. Figure 3(A) is an example of an image of an evaluation area, Figure 3(B) is an example of hair identification performed on an evaluation area image, Figures 4 to 6 are graphs showing the RGB data of all hair and the judgment results by the evaluator, Figure 7 is a graph showing the thickness data of all hair and the judgment results by the evaluator, and Figure 8 is a scatter plot showing the thickness and brightness of all hair.

[0049] The gray hair detection (gray hair identification) process using the gray hair identification device 10 described above was performed on a total of 65 evaluation sites.

[0050] <Acquisition of images of the evaluated area> An evaluation target was created on a portion of the scalp of the subject by cutting the hair to approximately 1 mm in length over an area of ​​1 square centimeter. An image of the evaluation area was obtained by taking a close-up photograph of this evaluation target with a digital camera (image acquisition unit 11). In this embodiment, 65 evaluation area images were obtained from the scalps of 65 subjects.

[0051] <Getting parameters> First, the hair identification unit 12 analyzes the evaluation area image to identify the hairs visible in the evaluation area image, and assigns an identification number to each identified hair (see Figure 2(B)). In this embodiment, a total of 14,245 hairs were identified from 65 evaluation area images.

[0052] Next, by analyzing the evaluation area image using the feature extraction unit 13, the following parameters were obtained for each of the 14,245 hairs in the evaluation area image identified by the hair identification unit 12: RGB of the hair (see Figures 3 to 5), hair thickness (see Figure 6), hair brightness (see Figure 7), and RGB of the scalp.

[0053] Furthermore, by inputting data about the subjects whose evaluation area images were obtained via the related information acquisition unit 14, the age information of the subjects was acquired. This allowed for the classification of the subjects into age groups (20s to 60s) based on their age information at the time of hair photography.

[0054] <Evaluator's Judgment> Next, in order to verify the accuracy of supervised learning and machine learning-based judgment, the evaluators first determined whether each hair in the 65 evaluation area images obtained was white or black, creating correct data (label data). As a result of the evaluators' judgment, out of a total of 14,245 hairs identified in all 65 evaluation cases, 13,728 (96.4%) were black and 517 (3.6%) were white.

[0055] In this process, the evaluation was conducted by two evaluators, and only when the two evaluators' results matched was the evaluation result recorded. If the two evaluators' results did not match, the evaluation was conducted again by the two evaluators, and this process was repeated until the results matched. In addition, to ensure that the evaluation criteria were consistent, the evaluators were trained to distinguish between gray hair and black hair using sample images of gray hair and black hair before performing the actual evaluation work.

[0056] Based on the information obtained above, Figures 4 to 6 show histograms where the horizontal axis represents the RGB (Red, Green, Blue) of the hair extracted by the feature extraction unit 13, and the vertical axis represents gray hair and black hair as determined by the evaluator. Figure 7 shows a histogram where the horizontal axis represents hair thickness, and the vertical axis represents gray hair and black hair as determined by the evaluator. Figure 8 shows a scatter plot where the horizontal axis represents hair thickness, the vertical axis represents hair brightness, and each plotted data point is color-coded according to whether it is gray or black hair as determined by the evaluator.

[0057] <Decision based on machine learning (supervised learning model)> First, in order to perform gray hair detection using the supervised learning (trained) model by the judgment unit 15, 70% of the 14,245 hairs identified from all 65 cases to be evaluated were used as training data (labels, target variable) for supervised learning, and the remaining 30% were divided as test data. At this time, the data was divided so that the proportion of gray hair in the training data and the proportion of gray hair in the test data were approximately equal.

[0058] The determination unit 15 sets the data obtained in the above process regarding the RGB of the hair, the brightness of the hair, the thickness of the hair, the RGB of the scalp, and the age of the person being evaluated as prediction parameters to be used for gray hair determination by the trained model, and performs supervised learning using the training data. In this embodiment, a total of six models were used as algorithms (learning models) for gray hair determination (classification): Logistic regression, decision tree, random forest, K-nearest neighbors, SVM (support vector machine), and XGBoost (gradient boosting), and supervised learning was performed for each.

[0059] Subsequently, the determination unit 15 performed gray hair determination on each hair constituting the test data, classifying each hair as either gray or black (black hair) according to the determination results of the supervised learning model, for each combination of parameters and algorithm described above. The accuracy of each determination result was verified using the method described later.

[0060] Specifically, the determination unit 15 performed gray hair determination (Example 1) with three prediction parameters: RGB of the hair, brightness of the hair, and thickness of the hair. In addition, it performed gray hair determination (Examples 2 to 4) that included either or both of the RGB of the scalp and the age of the person being evaluated as prediction parameters. Furthermore, in addition to the above conditions, gray hair determination was performed targeting only hair with a thickness (diameter) of 20 μm or more (Examples 5 to 8), and gray hair determination was performed targeting only hair with a thickness (diameter) of 40 μm or more (Examples 9 to 12), and these were compared.

[0061] <Verification of the accuracy of the judgment results> The judgment unit 15 performed gray hair detection by changing the combination of prediction parameters (a total of 12 patterns, including combinations with the range of the test data) and the learning model (a total of 6 types). For each judgment result, the accuracy rate of gray hair detection was calculated, assuming that the judgment result identified as gray hair by the evaluator was correct. The accuracy rate of gray hair detection was calculated as (number of correctly identified gray hairs) / (number of hairs identified as gray by the evaluator in the test data) (see Figure 9).

[0062] Figure 9 is a table showing the parameters used in machine learning and the accuracy evaluation. In Example 1, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, and the thickness of the hair. In this case, the accuracy rate for gray hair was 84.74% on average across all models tested, and 87.01% for the most accurate model (see Figure 9).

[0063] In Example 2, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, the thickness of the hair, and the RGB values ​​of the scalp. In this case, the accuracy rate for gray hair was 82.03% on average across all models tested, and 85.71% for the most accurate model (see Figure 9).

[0064] In Example 3, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, the thickness of the hair, and the age of the person being evaluated. In this case, the accuracy rate for gray hair was 81.60% on average across all models tested, and 89.61% for the most accurate model (see Figure 9).

[0065] In Example 4, the prediction parameters were set to the RGB of the hair, the brightness of the hair, the thickness of the hair, the RGB of the scalp, and the age of the person being evaluated. In this case, the accuracy rate for gray hair was 81.71% on average across all models tested, and 87.66% for the most accurate model (see Figure 9).

[0066] In Example 5, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, and the thickness of the hair. Only hair with a thickness (diameter) of 20 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 92.28% on average across all models tested, and 95.79% for the most accurate model (see Figure 9).

[0067] In Example 6, the prediction parameters were set to the RGB of the hair, the brightness of the hair, the thickness of the hair, and the RGB of the scalp. Only hair with a thickness (diameter) of 20 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 89.82% on average across all models tested, and 93.68% for the most accurate model (see Figure 9).

[0068] In Example 7, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, the thickness of the hair, and the age of the person being evaluated. Only hair with a thickness (diameter) of 20 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 91.40% on average across all models tested, and 96.84% for the most accurate model (see Figure 9).

[0069] In Example 8, the prediction parameters were set to hair RGB, hair brightness, hair thickness, scalp RGB, and the age of the person being evaluated. Only hair with a thickness (diameter) of 20 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 90.00% on average across all models tested, and 93.68% for the most accurate model (see Figure 9).

[0070] In Example 9, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, and the thickness of the hair. Only hair with a thickness (diameter) of 40 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 93.68% on average across all models tested, and 96.55% for the most accurate model (see Figure 9).

[0071] In Example 10, the prediction parameters were set to the RGB of the hair, the brightness of the hair, the thickness of the hair, and the RGB of the scalp. Only hair with a thickness (diameter) of 40 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 91.95% on average across all models tested, and 94.83% for the most accurate model (see Figure 9).

[0072] In Example 11, the prediction parameters were set to the RGB values ​​of the hair, the brightness of the hair, the thickness of the hair, and the age of the person being evaluated. Only hair with a thickness (diameter) of 40 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 92.82% on average across all models tested, and 98.28% for the most accurate model (see Figure 9).

[0073] In Example 12, the prediction parameters were set to hair RGB, hair brightness, hair thickness, scalp RGB, and the age of the person being evaluated. Only hair with a thickness (diameter) of 40 μm or more was included in the accuracy evaluation. In this case, the accuracy rate for gray hair was 92.53% on average across all models tested, and 94.83% for the most accurate model (see Figure 9).

[0074] Incidentally, as Comparative Example 1 for comparison with each example shown in Figure 9, we also output the results when only the RGB values ​​of the hair and the brightness of the hair were set as prediction parameters. In this case, the accuracy rate for gray hair was 78.25% on average across all models tested, and 81.82% for the most accurate model (see Figure 9).

[0075] Furthermore, as a comparative example 2, we also output the results when only the RGB data of the scalp was set as the prediction parameter, but in this case, it was not possible to determine gray hair, and the calculation was impossible (see Figure 9).

[0076] Based on the above, it has been shown that the gray hair identification device 10 can determine (evaluate) whether hair is gray or not with a higher accuracy compared to the comparative examples, regardless of which learning model is used, by setting at least the RGB (and brightness) of the hair and the thickness of the hair as prediction parameters (see Examples 1 to 12).

[0077] Furthermore, it was confirmed that the gray hair identification device 10 improves the accuracy of gray hair identification by selecting only hairs with a thickness of 20 μm or more, compared to cases where the hair thickness is not limited (see Examples 5 to 8). In addition, it was confirmed that the accuracy of gray hair identification improves by selecting only hairs with a thickness of 40 μm or more, compared to cases where the hair thickness is not limited or when the hair thickness is limited to 20 μm or more (see Examples 9 to 12).

[0078] According to the gray hair identification device 10 described above, the gray hair condition of the person being evaluated can be objectively evaluated with a certain level of accuracy, without being affected by variations in the evaluator's abilities and environmental factors, as well as biases due to the person's gray hair countermeasures. [Explanation of symbols]

[0079] 10. Gray hair identification device 11 Image acquisition unit 13 Feature Extraction Unit 14 Related Information Acquisition Department 15 Judgment section

Claims

1. A gray hair identification device, It includes a determination unit that determines whether or not the hair is gray based on multiple features related to the hair of the person to be identified. As the aforementioned features, at least two pieces of information were set: information about hair color and information about hair thickness. A gray hair identification device characterized by the following features.

2. The system includes a feature acquisition unit that acquires the aforementioned feature quantities. The gray hair identification device according to claim 1.

3. The feature acquisition unit, An image acquisition unit acquires information about an image taken such that it includes at least a portion of at least one hair from the scalp of the person to be identified. It comprises a feature extraction unit that extracts each of a predetermined number of feature quantities from the captured image acquired by the aforementioned image acquisition unit. The gray hair identification device according to claim 2.

4. The determination unit determines whether or not the hair is gray based on a pre-trained model that has been trained in advance to determine whether or not the hair is gray based on a plurality of the aforementioned features. The gray hair identification device according to claim 1.

5. The determination unit determines whether the hair in the captured image acquired by the captured image acquisition unit is gray, based on a pre-trained model that has been trained in advance to determine whether or not the hair is gray based on a plurality of features. The gray hair identification device according to claim 3.

6. The aforementioned trained model is one that has been pre-trained using supervised learning. The gray hair identification device according to claim 4.

7. The determination unit limits the hair to be determined to those with a thickness of 20 μm or more. A gray hair identification device according to any one of claims 1 to 6.

8. The aforementioned feature quantity includes information about the brightness of the hair. A gray hair identification device according to any one of claims 1 to 6.

9. The aforementioned captured image includes at least a portion of the scalp from which hair grows. As part of the aforementioned features, information on scalp color that can be extracted from the captured image was also set. A gray hair identification device according to claim 3 or 5.

10. The feature acquisition unit includes a related information acquisition unit that acquires information on the age of the person to be identified. As part of the aforementioned features, information on the age of the person to be identified was also set. A gray hair identification device according to any one of claims 2, 3, or 5.

11. A method for identifying gray hairs performed by a computer, The process includes a step of determining whether or not the hair is gray based on multiple features related to the hair of the person to be identified. As the aforementioned features, at least two pieces of information were set: information about hair color and information about hair thickness. A method for identifying gray hair, characterized by the following features.

12. The process includes a feature acquisition step for acquiring the aforementioned features. The method for identifying gray hair according to claim 11.

13. The feature acquisition process described above is: The process involves obtaining information about a captured image, which is an image taken so as to include at least a portion of at least one hair from the scalp of the person to be identified, and The process includes a feature extraction operation that extracts at least two features from the captured image: information on hair color and information on hair thickness. The method for identifying gray hair according to claim 12.

14. On the computer, The system performs a determination process to determine whether the hair is gray or not based on multiple features related to the hair of the person being identified. As the aforementioned features, at least two pieces of information were set: information about hair color and information about hair thickness. A gray hair identification program characterized by the following features.

15. The feature acquisition process is executed to obtain the aforementioned features. The gray hair identification program according to claim 14.

16. The aforementioned feature acquisition process is as follows: Image acquisition process to obtain information about the captured image, which is an image taken so as to include at least a portion of at least one hair, This process includes a feature extraction process that extracts at least two features from the information of the captured image obtained by the aforementioned image acquisition process: information on hair color and information on hair thickness. The gray hair identification program according to claim 15.

Images