Clothing characteristics evaluation method
By using a clothing feature evaluation system that combines visible light and infrared spectral information with a deep learning model, the reuse value of clothing can be quantified, solving the problem of high costs associated with manual evaluation and achieving efficient clothing reuse and environmentally friendly classification.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- KOBE UNIV
- Filing Date
- 2024-10-30
- Publication Date
- 2026-05-15
AI Technical Summary
In existing technologies, clothing feature assessment mainly relies on human vision and intuition, resulting in high labor and time costs, and failing to effectively assess the reuse value of clothing.
By analyzing the color changes and material composition of clothing, utilizing visible light and infrared spectral information, and combining deep learning models, the reuse value of clothing is quantified, and its reuse path is determined based on the evaluation results.
It enables quantitative assessment of clothing characteristics, reduces labor and time costs, improves reuse efficiency, and ensures that clothing is recycled according to material, thus reducing environmental burden.
Smart Images
Figure 2026079072000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for evaluating clothing characteristics for evaluating the characteristics of clothing.
Background Art
[0002] Patent Document 1 discloses a waste sorting system for efficiently sorting waste. Patent Document 2 discloses a spectrometer system for analyzing the materials of recycled products. Patent Document 3 discloses a method for measuring the fiber composition of fiber samples.
Prior Art Documents
Patent Documents
[0003]
Patent Document 1
Patent Document 2
Patent Document 3
Summary of the Invention
Problems to be Solved by the Invention
[0004] The first evaluation of the characteristics of clothing is the evaluation of the reuse value. The evaluation of the reuse value is currently performed based on human visual information and human sensibility. None of the systems in Patent Documents 及3 perform the evaluation of the reuse value. In the conventional method for evaluating clothing characteristics, a great deal of human cost and time cost are incurred.
[0005]
Means for Solving the Problems
[0006] It should be noted that in the original text, "特許文献1乃至3" was translated as "Patent Documents 1 to 3" in a more general sense in the above translation. If a more literal translation is needed, it could be "Patent Documents 1 through 3". Also, the text "特許文献1乃至3のシステムはいずれもリユース価値の評価を行っていない。" was translated as "None of the systems in Patent Documents 1 to 3 perform the evaluation of the reuse value." Here, "1 to 3" can also be adjusted according to the more literal translation requirements as above.The present invention provides a method for evaluating the characteristics of clothing, which involves analyzing the color change of clothing based on visible spectrum information, evaluating the reuse value of the clothing based on the color change, and, if it is determined in the evaluation of the reuse value that the clothing is not reusable, further evaluating the material of the clothing based on infrared spectrum information.
[0007] The present invention's method for evaluating clothing characteristics analyzes color changes caused by stains, wear and tear, fraying, etc., to assess its reusability. For example, the method determines that clothing is reusable if the similarity exceeds a predetermined threshold. Conversely, it determines that clothing is not reusable if the similarity falls below a predetermined threshold. As a result, the present invention's method for evaluating clothing characteristics quantifies the evaluation of reusability, which was previously done manually and time-consuming, thereby reducing human and time costs. Furthermore, by evaluating the material of clothing determined to be unreusable based on infrared spectral information, the method enables the unreusable clothing to be sent to an appropriate recycling process according to its material. [Effects of the Invention]
[0008] According to this invention, the reuse value and the characteristics of clothing can be quantitatively evaluated, thereby reducing human and time costs. [Brief explanation of the drawing]
[0009] [Figure 1] This is a block diagram showing the configuration of the clothing characteristics evaluation system 1. [Figure 2] This is a flowchart showing the operation of the clothing characteristics evaluation system 1. [Figure 3] This figure shows the spectral information of stain-free and stained areas in the same material. [Figure 4] This figure shows the spectral information of PET (Poly Ethylene Terephthalate), cotton, and a mixture of these two materials. [Figure 5]This figure shows the results of imaging the spectral intensity at a wavelength of 1480 nm. [Figure 6] This is a block diagram showing the configuration of the clothing characteristics evaluation system 1A related to a modified example. [Figure 7] This is a time-series temperature fluctuation diagram. [Figure 8] This figure shows the results of the analysis of the phase information of the frequency components. [Figure 9] This diagram shows the relationship between the thickness and type of interlining and the phase difference. [Modes for carrying out the invention]
[0010] Embodiments of the present invention will be described below. Figure 1 is a block diagram showing the configuration of the clothing characteristics evaluation system 1. Figure 2 is a flowchart showing the operation of the clothing characteristics evaluation system 1.
[0011] The clothing characteristics evaluation system 1 comprises a light source 11, a hyperspectral camera 12, and a computer 13. The light source 11 irradiates the clothing 100 with light including the visible spectrum. The reflected light from the light source 11 on the clothing 100 is input to the hyperspectral camera 12.
[0012] The hyperspectral camera 12 includes a spectrometer and an image sensor. The hyperspectral camera 12 allows reflected light from the clothing 100 to pass through, for example, a slit, in only one line in a first direction (y direction). The hyperspectral camera 12 spectrally analyzes the light beam of the one line that passed through the slit using a spectrometer. The hyperspectral camera 12 acquires the intensity of each spectrally analyzed wavelength using a pixel array of the image sensor in a second direction (x direction). The hyperspectral camera 12 performs scanning, for example, by oscillating the built-in mirror of the spectrometer. As a result, the hyperspectral camera 12 acquires the intensity of each wavelength as spectral information (λ) in a predetermined range including the visible range (e.g., 400 nm to 1000 nm). In other words, the hyperspectral camera 12 acquires a three-dimensional data cube consisting of two-dimensional position information (x, y) and spectral information (λ).
[0013] The computer 13 consists of an information processing device such as a personal computer or a smartphone having at least one processor. The computer 13 evaluates the characteristics of the clothing 100 using the three-dimensional data cube acquired by the hyperspectral camera 12.
[0014] First, the computer 13 analyzes the color change of the clothing based on the spectral information in the visible range (S11). The color change is caused by stains such as spots, sagging, fraying, etc.
[0015] FIG. 3 is a diagram showing the spectral information of a stain-free part and a stained part in the same material as an example of the analysis of color change. The horizontal axis in FIG. 3 is the wavelength, and the vertical axis is the intensity. As shown in FIG. 3, when stains such as spots, sagging, fraying, etc. occur on the clothing and cause a color change, a significant change occurs in the spectral information. For example, the analysis of color change is performed using the data cube when the clothing not used in the clothing characteristic evaluation system 1 as a reference spectrum. The computer 13, for example, obtains the similarity between the spectral information of the clothing 100 and the reference spectrum. The computer 13 may obtain the similarity between each pixel and a certain reference pixel as the reference spectrum among the data cubes obtained by measuring the clothing 100 with the clothing characteristic evaluation system 1. The similarity is obtained by SAM (Spectral Angle Mapper), which is the angle between spectra, or the Euclidean distance, etc. Alternatively, the similarity may be obtained based on histogram analysis.
[0016] Next, the computer 13 evaluates the reuse value of the clothing 100 based on the color change (S12). The reuse value is determined based on the analysis result (e.g., similarity) of the color change obtained in the process of S11. The computer 13, for example, determines that it is reusable when the similarity exceeds a predetermined threshold. The clothing 100 determined to be reusable is sent for reuse. Also, the computer 13 determines that it is not reusable when the similarity is below a predetermined threshold.
[0017] Note that the computer 13 may construct a trained model trained with a DNN (Deep Neural Network). As a training stage, the computer 13 that constructs the trained model acquires a large number of datasets showing the spectral information of clothing determined to be reusable by human evaluation and the spectral information of clothing determined to be non-reusable by human evaluation. The computer uses the acquired large number of datasets to train a predetermined model using a predetermined algorithm so as to output the reuse value for the input spectral information. As a usage stage, the computer 13 inputs the spectral information of clothing and outputs the reuse value. Since the spectral information of clothing and the reuse value have a certain correlation as described above, the trained model can output the corresponding reuse value when the spectral information of clothing is input.
[0018] Note that any algorithm may be used for training the model. For example, any machine learning algorithm such as CNN (Convolutional Neural Network) or RNN (Recurrent Neural Network) can be used as the algorithm.
[0019] Next, when the computer 13 determines that the clothing 100 is non-reusable in the evaluation of the reuse value in S12, the computer 13 further evaluates the material of the clothing 100 based on the spectral information in the infrared region (S13).
[0020] The spectral information in the infrared region may be acquired by the hyperspectral camera 12 described above, or may be acquired by a near-infrared camera different from the hyperspectral camera 12.
[0021] Clothing materials have unique absorption spectra (fingerprint spectra) according to their molecular structures. The unique absorption spectra appear in the infrared wavelength region. Therefore, the computer 13 can identify the material substance based on the absorption spectrum.
[0022] Figure 4 shows the spectral information of PET (Poly Ethylene Terephthalate), cotton, and a mixture of these two materials. As shown in Figure 4, there is no spectral difference between the materials in the short wavelength range shorter than 1400 nm. In the long wavelength range above 1400 nm, spectral differences are observed depending on the material.
[0023] Figure 5 shows the results of imaging the spectral intensity at a wavelength of 1480 nm. Each material in the spectral image shown in Figure 5 has a different color in the visible range. However, as shown in Figure 5, the higher the degree of PET inclusion, the higher the spectral intensity and the higher the brightness when imaged, and it can be seen that this does not depend on the "color" in the visible range.
[0024] Furthermore, the PET / cotton 50 / 50 material shown in Figure 5 has a check pattern created by 100% PET yarn and 100% cotton yarn. In spectral intensity, the brightness of the 100% PET yarn portion is high, and the brightness of the 100% cotton yarn portion is low, and the check pattern in the visible range is directly reflected in the spectral intensity image. Therefore, composite states such as interwoven or blended materials, and the presence or absence of post-processing, can be detected as changes in brightness.
[0025] Computer 13 registers the spectral intensity of each material in the long wavelength range of 1400 nm or more (e.g., 1480 nm) within the infrared region as a database and compares it with the detected spectral intensity to evaluate the material of clothing 100, identifying the composition ratio of PET and cotton, the composite state such as weaving or blending, and whether or not post-processing has been performed. Computer 13 selects the material with the spectral intensity closest to the detected spectral intensity using methods such as cross-correlation functions or pattern matching. Alternatively, computer 13 may construct a trained model trained with a DNN. As a training stage, the computer constructing the trained model acquires a large number of datasets showing spectral intensity, material composition ratio, composite state such as weaving or blending, and whether or not post-processing has been performed. Using the acquired large number of datasets, computer 13 trains a predetermined model using a predetermined algorithm to output information on the material composition ratio, composite state such as weaving or blending, and whether or not post-processing has been performed in relation to the input spectral information. Computer 13, in its operational phase, takes the spectral intensity of the garment as input and outputs information on the material composition ratio, composite state such as weaving or blending, and whether or not post-processing has been performed. As described above, there is a certain correlation between the spectral intensity of the garment and the material composition ratio, composite state such as weaving or blending, and whether or not post-processing has been performed. Therefore, a trained model can output the corresponding material composition ratio, composite state such as weaving or blending, and whether or not post-processing has been performed, simply by inputting the spectral intensity of the garment.
[0026] As a result, the computer 13 can identify the material of the garment regardless of the color of the fabric, and can also identify the mixing form and mixing ratio. In this way, the garment characteristic evaluation method of this embodiment allows for the quantitative evaluation of reuse value by a computer, which was previously done based on human visual information and human sensibilities. Furthermore, the garment characteristic evaluation method of this embodiment can further identify the material if it is determined that the garment is not reusable. As a result, the garment characteristic evaluation method of this embodiment can significantly reduce human and time costs. In addition, the garment characteristic evaluation method of this embodiment can ensure that appropriate garments are put back into reuse. Therefore, the garment characteristic evaluation method of this embodiment can also reduce the burden on the natural environment by reducing waste. Furthermore, the garment characteristic evaluation method of this embodiment can output information indicating the composition ratio of the material of garments determined to be unreusable, the composite state such as weaving or blending, and whether or not post-processing has been performed, so that the garments can be sent to the appropriate recycling process. For example, if the garment characteristic evaluation method of this embodiment determines that the material of the garment is cotton, it can be sent for material recycling such as recycled wool and respun yarn, and if the material of the garment is determined to be PET, it can be sent for chemical recycling through depolymerization.
[0027] (modified version) Figure 6 is a block diagram showing the configuration of a modified garment characteristic evaluation system 1A. The garment characteristic evaluation system 1A includes an infrared sensor 21, a light source 22, a power supply 23, and a computer 13. The infrared sensor 21 and the power supply 23 are connected to the computer 13.
[0028] Computer 13 further evaluates the materials contained within the clothing based on their thermal response characteristics.
[0029] Computer 13 controls the power supply 23 to pulse-likely turn the light source 22 on and off. The light source 22 is, for example, a xenon flash lamp, which emits light in the near-infrared region (e.g., 750nm to 1500nm) with a wavelength shorter than the measurement wavelength of the infrared sensor 21. As a result, the reflected light from the light source 22 is less likely to affect the measurement of the infrared sensor 21. In addition, since xenon flash lamps do not have residual heat from a heating tube, the influence of reflected light from the light source 22 during measurement can be reduced.
[0030] The infrared sensor 21 is an infrared thermograph equipped with an InSb (Indium antimonide) sensor. The InSb sensor is a quantum type that is used after cooling the infrared sensor to liquid nitrogen temperature, and it has a fast response speed and high sensitivity.
[0031] In Figure 6, as an experimental example, a surface material 70 made of 100% PET with a thickness of 2 mm is placed on the top surface of an aluminum plate 90. Two layers of 100% PET core material 80 are placed on top of the surface material 70, and another layer of surface material 70 is placed on top of that, and the whole thing is covered with sapphire glass 50. The sapphire glass 50 is placed there to ensure that the surface material 70 and the core material 80 are in close contact. Since the sapphire glass 50 transmits infrared rays, it does not affect the measurement of the infrared sensor 21.
[0032] Figure 7 shows a time-series temperature fluctuation. In Figure 7, "Outer fabric only" represents the temperature fluctuation of the area where the upper and lower outer fabrics 60 are in contact, "Outer fabric (air layer)" represents the temperature fluctuation of the area where the upper and lower outer fabrics 60 are not in contact, and "Interlining" represents the temperature fluctuation of the area where the interlining 80 is placed between the upper and lower outer fabrics 70.
[0033] As shown in Figure 7, the areas with and without the interlining 80 exhibit different thermal response characteristics, specifically different time-series temperature decrease trends. Compared to areas with only the outer fabric 70, areas with the interlining 80 experience heat transfer in the thickness direction. This is because the heat capacity of the interlining 80 is larger than that of the outer fabric 70, resulting in faster heat transfer in the thickness direction.
[0034] Thus, the temperature distribution and temperature fluctuations measured on the surface of the fabric differ depending on the presence or absence of the interlining 80. In evaluating the material of the garment, the computer 13 further detects whether or not the garment contains interlining based on its thermal response characteristics and displays the detection result on the display. This allows users to easily determine whether or not the garment contains interlining and send the garment to the appropriate recycling process. For example, if the garment characteristic evaluation method of the modified example determines that there is interlining, the garment can be sent to the dismantling recycling process, and if it determines that there is no interlining, the garment can be sent to the melting recycling process.
[0035] Computer 13 may construct a pre-trained model trained with a DNN. As part of the training phase, Computer 13 acquires numerous datasets showing time-series temperature fluctuations and the presence or absence of interlining. Using these acquired datasets, Computer 13 trains a predetermined model using a predetermined algorithm to output information indicating the presence or absence of interlining in response to input time-series temperature fluctuations. As part of the usage phase, Computer 13 takes time-series temperature fluctuations as input and outputs information indicating the presence or absence of interlining. Since there is a certain correlation between the time-series temperature fluctuations of clothing and the information indicating the presence or absence of interlining, the pre-trained model can output information indicating the presence or absence of interlining when given time-series temperature fluctuations of clothing as input.
[0036] Furthermore, the computer 13 can also quantify the thermal response characteristics by converting the time-series temperature changes after heating the clothing into frequency components obtained through Fourier analysis. In Fourier analysis, a Fourier series expansion is performed as shown in Equation 1 below, decomposing the time-series temperature change T(t) into the cosine and sine components of each frequency component.
[0037]
number
[0038]
number
[0039] Figure 9 shows the relationship between the thickness and type of interlining and the phase difference. As shown in Figure 9, the greater the thickness of the interlining, the greater the phase difference between the outer fabric and the interlining. Furthermore, even with the same interlining thickness, different phase differences are obtained for PET and cotton materials. Thus, different temperature changes appear depending on the thickness and type of interlining. Since computer 13 can quantitatively evaluate the characteristics of temperature changes by using phase analysis, it evaluates the type and thickness of the material embedded in the garment. Alternatively, computer 13 may construct a trained model trained with a DNN. In the training stage, computer 13 acquires a large number of datasets showing phase difference information and the type and thickness of the embedded material. Using the acquired large number of datasets, the computer trains a predetermined model using a predetermined algorithm to output information showing the type and thickness of the embedded material in relation to the input phase difference information. In the usage stage, computer 13 takes phase difference information as input and outputs information showing the type and thickness of the embedded material. As described above, there is a certain correlation between phase difference information and information indicating the type and thickness of the contained material. Therefore, a trained model can output information indicating the type and thickness of the contained material when given phase difference information as input.
[0040] Furthermore, computer 13 may evaluate the material contained in the clothing based on the thermal response characteristics described in the modified example, without performing evaluation of the reuse value based on color change or evaluation of the material based on spectral information in the infrared region. In other words, the present invention may have the following technical concept.
[0041] (1) Heat the clothing, The thermal response characteristics to the heating are measured by a sensor. The materials contained in the clothing are evaluated based on their thermal response characteristics. Methods for evaluating clothing characteristics. As described above, the temperature distribution and temperature fluctuations measured on the surface of the fabric differ depending on the presence or absence of interlining. The computer detects whether or not interlining is embedded in the garment based on its thermal response characteristics and displays the detection result on a display. This allows users to easily determine whether or not interlining is embedded in the garment and send it to the appropriate recycling process. For example, if the garment characteristic evaluation method of the modified example determines that interlining is present, it can be sent to the dismantling recycling process; if it determines that there is no interlining, it can be sent to the melting recycling process.
[0042] In (1) above, the thermal response characteristics may be determined based on the frequency components obtained by Fourier analysis of the time-series temperature change after heating the clothing. Computers can also quantify thermal response characteristics by converting the time-series temperature changes after heating clothing into frequency components obtained through Fourier analysis.
[0043] In (2) above, the Fourier analysis may be performed based on the phase information of the frequency components. Because the outer fabric portion, the outer fabric (air layer) portion, and the portion with interlining each exhibit different phases, users can easily determine which part of the garment contains interlining or not.
[0044] In (3) above, the type and thickness of the material contained within the garment may be evaluated based on the phase difference between the phase information of the outer fabric and the phase information of the interlining. The greater the thickness of the interlining, the greater the phase difference between the outer fabric and the interlining. Furthermore, even with the same interlining thickness, different phase differences can be obtained between PET and cotton materials.
[0045] The description of these embodiments should be considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the claims, rather than by the embodiments described above. Furthermore, the scope of the invention is intended to include all modifications within the meaning and scope equivalent to the claims. [Explanation of Symbols]
[0046] 1,1A: Clothing characteristics evaluation system 11:Light source 12: Hyperspectral camera 13: Computer 21: Infrared sensor 22:Light source 23: Power supply 50: Sapphire glass 60, 70: Outer fabric 80: Interlining 90: Flat plate 100: Clothing
Claims
1. Analyzes color changes in clothing based on spectral information in the visible range. Based on the aforementioned color change, the reuse value of the clothing is evaluated. If, in the evaluation of the reuse value, it is determined that the clothing is not reusable, the material of the clothing is further evaluated based on spectral information in the infrared region. Methods for evaluating clothing characteristics.
2. In the material evaluation of the aforementioned garment, the material contained within the garment is further evaluated based on its thermal response characteristics. The method for evaluating clothing characteristics according to claim 1.
3. The thermal response characteristics are determined based on the frequency components obtained by Fourier analysis of the time-series temperature change after heating the garment. The method for evaluating clothing characteristics according to claim 2.
4. The Fourier analysis described above is performed based on the phase information of the frequency components. The method for evaluating clothing characteristics according to claim 3.
5. The type and thickness of materials contained within the garment are evaluated based on the phase difference between the phase information of the outer fabric and the phase information of the interlining. The method for evaluating clothing characteristics according to claim 4.
6. The aforementioned evaluation of reuse value is performed based on similarity determination with the reference spectrum. A method for evaluating clothing characteristics according to any one of claims 1 to 5.
7. The aforementioned similarity determination is performed based on histogram analysis. The method for evaluating clothing characteristics according to claim 6.