Detecting elastane from textile samples

Abstract

According to an example aspect of the present disclosure, there is provided a method comprising determining at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane, determining an average spectrum of each of the at least one first hyperspectral image, determining a mathematical classification model based on the average spectrum of each of the at least one first hyperspectral image, determining at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane, determining at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane, evaluating the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples and detecting using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

Description

DETECTING ELASTANE FROM TEXTILE SAMPLESFIELD

[0001] Embodiments of the present disclosure relate in general to detecting elastane from textile samples.BACKGROUND

[0002] Elastane is a synthetic polymer, which increases the elasticity of yams and fabrics. Only small concentrations (<10%) of elastane may be needed to change the properties of yams. Elastane may be blended with natural or synthetic fibers before yarn spinning or elastane fibers may be wrapped with other fibers. Elastane may thus be found on the surface of a fabric or in the core of a yam surrounded by other fibers. This, together with the low concentration, makes it difficult to detect elastane in a fabric. There is therefore a need to provide improvements for detecting elastane in textile samples.SUMMARY

[0003] According to some aspects, there is provided the subject-matter of the independent claims. Some embodiments are defined in the dependent claims.

[0004] According to a first aspect of the present disclosure, there is provided an apparatus comprising at least one processor; and at least one memory including computer program code; the at least one memory and the computer program code being configured to, with the at least one processor, cause the apparatus at least to determine at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane, determine an average spectrum of each of the at least one first hyperspectral image, determine a mathematical classification model based on the average spectrum of each of the at least one first hyperspectral image, determine at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane, determine at leastone third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane, evaluate the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples and detect using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

[0005] According to a second aspect of the present disclosure, there is provided a system, comprising a hyperspectral camera and the apparatus of the first aspect of the present disclosure, wherein the hyperspectral is configured to take the hyperspectral images, and transmit the hyperspectral images to the apparatus.

[0006] According to a third aspect of the present disclosure, there is provided a method, comprising determining at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane, determining an average spectrum of each of the at least one first hyperspectral image, determining a mathematical classification model based on the average spectrum of each of the at least one first hyperspectral image, determining at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane, determining at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane, evaluating the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples and detecting using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

[0007] According to a fourth aspect of the present disclosure, there is provided a non- transitory computer readable medium having stored thereon a set of computer readable instructions that, when executed by at least one processor, cause an apparatus to at least perform the method. According to a fifth aspect of the present disclosure, there is provided a computer program comprising instructions which, when the program is executed by an apparatus, cause the apparatus to perform the method.BRIEF DESCRIPTION OF THE DRAWINGS

[0008] FIG. 1 illustrates operation of a hyperspectral imaging system in accordance with at least some embodiments of the present disclosure;

[0009] FIG. 2 illustrates a first hyperspectral image processing workflow in accordance with at least some embodiments of the present disclosure;

[0010] FIG. 3 illustrates a second hyperspectral image processing workflow in accordance with at least some embodiments of the present disclosure;

[0011] FIG. 4a illustrates the results of a SIMCA model in accordance with at least some embodiments of the present disclosure;

[0012] FIG. 4b illustrates the results of a EDA model in accordance with at least some embodiments of the present disclosure;

[0013] FIG. 5 illustrates an example apparatus capable of supporting at least some embodiments of the present disclosure;

[0014] FIG. 6 illustrates a flow graph of a method in accordance with at least some embodiments of the present disclosure.EMBODIMENTS

[0015] Embodiments of the present disclosure provide improvements for detecting elastane in textile samples. Detection of elastane is important at least for textile sorting and recycling. Elastane is a widely used synthetic polymer and only small amounts (<10%) may be required to improve the elasticity of yarns and fabrics. However, elastane may be blended with other fibers or embedded inside the core of the yarn making it difficult to detect elastane in a fabric. Embodiments of the present disclosure enable detecting elastane, for example, in cotton by separating a mixture of cotton and elastane from pure cotton. For example, hyperspectral imaging, such as Near-Infrared, NIR, imaging, may be used. The image spectra may then be mathematically coupled with cotton and elastane or pure cotton using, for example, a classification model based on a training set with known sample compositions.

[0016] Embodiments of the present disclosure provide a quick and non-invasive way to detect elastane in cotton, which can be fully automatized. In some example embodiments, an average spectrum may be determined based on NIR images to decrease sampling uncertainties. Average spectra may enable distinguishing detailed changes in the spectra and to reliably estimate model complexity. A class modelling approach or a discrimination modelling approach may be then exploited. A class model may be different from a discriminant model as only samples from the modelled class may be used for training the class model and samples from at least two classes are required for training a discriminant model. A class modelling approach can provide advantages in modelling new classes as all previous training samples are not required for training a new class model. On the other hand, a discriminant model can provide advantages in separating the properties of the classes as the features of at least two classes are used for determining class separation.

[0017] FIG. 1 illustrates operation of a hyperspectral imaging system in accordance with at least some embodiments of the present disclosure. Hyperspectral imaging system 100 may comprise hyperspectral camera 110, object 120 and apparatus 130. Hyperspectral camera 110 may be configured to take hyperspectral images of object 120. Object 120 may comprise one fabric at a time. Said fabric may comprise textile samples.

[0018] Apparatus 130 may be a computer configured to detect whether object 120 comprises elastane. Apparatus 130 may comprise for example at least one processor, and at least one memory including computer program code, the at least one memory and the computer program code being configured to, with the at least one processor, cause the apparatus at least to detect elastane from object 120 from at least one image taken by hyperspectral camera 110.

[0019] In case of hyperspectral imaging, every image may comprise information of the near infrared region, like 780 nm - 2600 nm, preferably 1000 - 2500 nm. Near infrared hyperspectral imaging may be used to analyse the chemical composition of materials, as the NIR light and different materials interact characteristically. The individual wavelength images may be stacked to form a hyperspectral data cube. Each hyperspectral data cube may contain spatial coordinates and a spectrum in each spatial pixel. Hyperspectral imaging may thus combine digital imaging and spectroscopy.

[0020] The measured spectra may be used to identify materials based on the spectral fingerprints. Different materials may interact differently with electromagnetic radiation andthus have unique spectrum. Measured spectra may be used to identify different areas or, for example, to identify material compounds in the sample.

[0021] Hyperspectral camera 110 may comprise at least an objective, a spectrograph, and a grayscale camera. Hyperspectral camera 110 may also comprises at least one source of illumination, i.e., at least one light source 112. Angles of at least one light source 112 and at least one detector 114 need to be set optimally to reduce these effects and to obtain reliable imaging results. Diffuse illumination, detection and correct image preprocessing methods may be used to avoid these spectral and optical artefacts. Hyperspectral imaging may provide significant benefits for textile applications because elastane may be detected efficiently, even if elastane would be located inside the core of the yarn, thereby enabling efficient sorting and recycling.

[0022] When light signal 122, produced by light source 112, interacts with material of object 120, absorbed light may be approximated to be the difference between light 124 reflected from a reference material and the sample. In some example embodiments, photons in the incident light signal may not be measured. Instead, sample reflectance may be estimated based on a signal from a highly reflective white reference, possibly with a dark current background.

[0023] Imaging applications may comprise light source 112, object 120 that reflects the incident light and detector 114 that measures reflected light 124. Light source 112 may be directed to transmit light signal 122 to object 120 in a desired angle and reflected light 124, along with the reflection angle, may be measured using detector 114. Reflectance may be the fraction of light signal 122 that is reflected compared with a reference material.

[0024] FIG. 2 illustrates a first hyperspectral image processing workflow in accordance with at least some embodiments of the present disclosure. At step 210, imaging of hyperspectral images, such as NIR images, may be performed by hyperspectral camera 110. For instance, the images may be scanned in a line-scanning mode with a defined number of spatial pixels in one line and spectral wavelength bands for each pixel. Spectral bands may cover the wavelength region of 1000-2500 nm with a given spectral sampling width and spectral resolution. Samples may be illuminated for example with polychromatic light produced by quartz halogen lamps. Each sample may be imaged separately.

[0025] The field of view of hyperspectral camera 110 may be set to cover the textile samples. During imaging the samples may be moved under detector 114, a constant distance away from the objective of detector 114. The speed of a moving table, i.e., the location of object 120, may be set based on the frame rate and integration time of hyperspectral camera 110.

[0026] Hyperspectral camera 110 may transmit at least one first signal and receive a reflected version of the at least one first signal to take at least one first hyperspectral image. Hyperspectral camera 110 may take at least one second, third and target hyperspectral images in the same way. Apparatus 130 may cause taking of the hyperspectral images, e.g., by transmitting a control signal to hyperspectral camera 110, to command hyperspectral camera 110 to take the images. That is, apparatus 130 may cause transmission of the at least one first signal by transmitting a first control signal to hyperspectral camera 110. Apparatus 130 may then and receive, responsive to the first control signal, a reflected version of the at least one first signal. Apparatus 130 may cause taking of the at least one second, third and target hyperspectral images in the same way.

[0027] Loading of images to apparatus 130 may be performed as well. That is, hyperspectral camera 110, or any other suitable apparatus, may transmit said images to apparatus 130, possibly via a communication network. Said hyperspectral images may be raw hyperspectral images, measured in detector signal intensity counts. Hyperspectral camera 110 may take at least one first, second, third and target hyperspectral images, and transmit the at least one first, second, third and target hyperspectral images to apparatus 130. For example, hyperspectral camera 110 may be a near- infrared camera and samples comprising cotton and elastane may be imaged within 780 - 2600 nm, preferably 1000 - 2500 nm.

[0028] Apparatus 130 may then determine at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane. In some example embodiments, said first text samples may comprise cotton in addition to elastane.

[0029] At step 220, apparatus 130 may convert the determined raw signal intensity counts into reflectance based on the measured white reference and dark current intensities. Apparatus 130 may then remove the hyperspectral image background, which does not belong to the sample, and determine an average spectrum of each of the at least one firsthyperspectral image. In some example embodiments, apparatus 130 may then preprocess the at least one average spectrum with appropriate methods to, for example, reduce measurement uncertainties. These appropriate methods may comprise but are not limited to, for example, Multiplicative Scatter Correction, MSC, Standard Normal Variate, SNV, and Savitzy-Golay smoothing and differentiation.

[0030] At steps 230 and 240, apparatus 130 may determine the mathematical classification model based on the average spectrum of each of the at least one first textile sample image. In the example of FIG. 2, apparatus 130 may determine that the mathematical classification model is a class model, e.g., based on user input. That is, the user may decide that the mathematical classification model is a class model, and the class model may be determined based on, e.g., Soft Independent Modelling Class Analogy, SIMCA. Apparatus 130 may then determine a multivariate description of said multiple first textile samples of the cotton and elastane class by using, for example, a cross-validated Principal Component Analysis, PCA, model. The PCA model may describe the spectral features of the cotton and elastane class, and the model loadings may equal the eigenvectors of a covariance matrix determined from the class spectra. The PCA model may be used project the average spectra of the cotton and elastane samples to lower dimensional sub-space, where the number of dimensions may equal the number of extracted Principal Components, PCs. This number may be considerably smaller than the number of original spectral variables and may refer to a dimensionality reduction. The number of extracted PCs may govern PCA model complexity, which may be evaluated using model cross-validation. Latent variables are a more general concept and PCs are one example of such latent variables. Even though PCs are used as an example in various example embodiments of the present disclosure, it is noted that the embodiments may be applied similarly for any other suitable latent variables.

[0031] Model cross-validation may, for example, divide the training set objects into internal training and test sets, and be used to determine True Positive Rate, TPR, as the share of correctly classified internal test set objects. A suitable number of PCs may then be selected to maximize TPR or to set TPR close to a predefined value. TPR may resemble a confidence level.

[0032] Apparatus 130 may then, for example, determine at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprisingelastane. Apparatus 130 may also determine at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane. Together the second and third textile samples may constitute an external test set.

[0033] Apparatus 130 may then evaluate the performance of the mathematical classification model based at least on the average spectra of said hyperspectral images of second and third textile samples, which may constitute the test set. Model performance may, for example, be evaluated by determining the predicted class memberships.

[0034] In some example embodiments, apparatus 130 may evaluate the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples and their known class assignments. That is, the mathematical classification model may be used to predict the class assignments of the external test set objects, which may then be compared with their known class memberships to evaluate the reliability of the model predictions.

[0035] In some example embodiments, apparatus 130 may then determine the class model based on one class and use latent variable weights of the class model to project an average spectrum of the at least one target hyperspectral image to a lower dimensional subspace, the lower dimensional sub-space describing properties of said one class, wherein a dimensionality of the lower dimensional sub-space equals a number of required latent variables.

[0036] At step 250, apparatus 130 may assign predicted test set classes based on two distances determined based on the mathematical classification model. First, the classification model may be used to determine the squared Euclidean distance of a test set spectrum to the projection of this spectrum to the lower dimensional class sub-space, which may be described using the PCA model. The projection itself may be determined by multiplying a preprocessed spectrum with class-specific weights, which may equal the loading vectors of the PCs. Second, the classification model may be used to determine the squared Mahalanobis distance of the projection to the origin of the class subspace. Together these distances may thus describe the distance of the test set object to the model and its distance to the center of the subspace within the model. These distances may then be combined to determine, for example, a combined index, and the combined index may be used to decide whether the test set object belongs to the modelled class based on a chosen confidence level. Apparatus 130may hence determine a distance of the average spectrum to the projection of the average spectrum to the lower dimensional sub-space and a distance of the projection to a centre of the lower dimensional sub-space, wherein the centre of the lower dimensional sub-space is a centre of said one class in the lower dimensional sub-space.

[0037] The predicted class memberships may then be compared with known class assignments to determine TPR and True Negative Rate, TNR, as the share of correctly classified external test set objects. Apparatus 130 may thus evaluate, based at least on the distances, whether the at least one target textile sample belongs to said one class. For example, apparatus 130 may evaluate the performance of the classification model based on the external test set using the determined test set TPR and TNR. Apparatus 130 may, for example, compare the determined test set TPR and TNR to some pre-set requirements. After positive evaluation, apparatus 130 may then determine the average spectrum of at least one hyperspectral image of at least one target textile sample and evaluate whether the at least one target textile sample comprises elastane. Positive evaluation may refer to determining that the performance mathematical classification model is satisfactory, e.g., TPR and TNR above a threshold. Apparatus 130 may for example determine that the at least one hyperspectral target textile sample comprises elastane when the at least one target textile sample belongs to said one class, wherein said one class may be a class of textile samples comprising elastane.

[0038] FIG. 3 illustrates a second hyperspectral image processing workflow in accordance with at least some embodiments of the present disclosure. Steps 310 - 330 may be the same as steps 210 - 230 in FIG. 2. At step 340, apparatus 130 may determine a mathematical classification model based on the average spectrum of each of the at least two hyperspectral images of first textile samples where the first textile samples comprise mixtures of textile fibers with and without elastane. In the example of FIG. 3, apparatus 130 may determine that the mathematical classification model is a discriminant model, e.g., based on user input. That is, the user may decide that the mathematical classification model is a discriminant model, and the discriminant model may comprise but is not limited to, for example, Linear Discriminant Analysis, LDA, or Quadratic Discriminant Analysis, QDA. Apparatus 130 may determine the discriminant model based on at least a first class of textile samples comprising elastane and a second class without elastane.

[0039] Apparatus 130 may determine a multivariate description of said multiple first samples of the textile mixtures with and without elastane by using, for example, a crossvalidated PCA model. The PCA model may describe the spectral features of the classes with and without elastane and the model loadings may equal the eigenvectors of a covariance matrix determined from the combined class spectra.

[0040] PCA model complexity may equal the number of extracted PCs which may be evaluated using model cross-validation. Model cross-validation may, for example, divide the training set objects into internal training and test sets and determine TPR and TNR as the share of correctly classified internal test set objects. A suitable number of PCs may then be selected to maximize TPR and TNR or set TPR and TNR close to some pre-set requirements.

[0041] At step 350, apparatus 130 may then evaluate the performance of the mathematical classification model based on the average spectra of said hyperspectral images of second and third textile samples, which may constitute the test set. Model performance may, for example, be evaluated by determining the predicted class memberships. At step 350, apparatus 130 may assign predicted test set classes based a distance determined based on the mathematical classification model. First, the classification model may be used to project the test set spectrum to the model subspace, which may be described using the conjoint PCA model. Apparatus 130 may further use latent variable weights of the discriminant model to project an average spectrum of the at least one hyperspectral target image to a lower dimensional sub-space describing class memberships, wherein a dimensionality of the lower dimensional sub-space equals a number of required latent variables. The projection itself may be determined by multiplying a preprocessed spectrum with the model weights, which may equal the loading vectors of the PCs.

[0042] Second, the classification model may be used to determine the Mahalanobis distances of the projections to the class centroids and the distances may be corrected with the determinants of the covariance matrices of each class. These distances may thus describe the distance of the test set object to the centre of each class within the subspace of the model. The predicted class memberships may then be assigned by minimizing the Mahalanobis distances to the class centroids and may then be compared with known class assignments to determine TPR and True Negative Rate, TNR, as the share of correctly classified external test set objects. Apparatus 130 may thus determine a distance of the projection to classcentroids in the lower dimensional sub-space, wherein a centroid is the centre of one class (e.g., first or second class) in the lower dimensional sub-space.

[0043] Apparatus 130 may then, after positive evaluation, determine the average spectrum of at least one hyperspectral image of at least one target textile sample and evaluate whether the at least one target textile sample comprises elastane. Positive evaluation may refer to determining that the performance mathematical classification model is satisfactory. Apparatus 130 may thus evaluate, based at least on the distance, whether the at least one target textile sample belongs to said first class or said second class. For example, apparatus 130 may determine that the at least one target textile sample comprises elastane when the at least one target textile sample belongs to said first class.

[0044] Alternatively, apparatus 130 may at step 340 determine regression coefficients of the discriminant model based on the first textile samples where the first textile samples comprise mixtures of textile fibers with and without elastane. In such a case, the user may decide that the mathematical classification model is a discriminant model, and the discriminant model may include but is not limited to, for example, Partial Least Squares Discriminant Analysis, PLS-DA. Apparatus 130 may then determine a vector of coefficients by regressing the class memberships of the first textile samples onto the average image spectra by using, for example, Partial Least Squares, PLS, regression. Apparatus 130 may thus use regression coefficients of the discriminant model to estimate a class membership of an average spectrum of the at least one hyperspectral target image.

[0045] The PLS model may comprise the spectral features which best separate the classes with and without elastane, and the PLS model complexity may equal the number of extracted latent variables which may be evaluated using model cross-validation. A suitable number of latent variables may then be selected to maximize TPR and TNR or set TPR and TNR close to some pre-set requirements.

[0046] At step 350, apparatus 130 may then evaluate the performance of the mathematical classification model based on the average spectra of said hyperspectral images of second and third textile samples, which may constitute the test set. At step 350, apparatus 130 may assign predicted test set classes based on the predicted class memberships based on the mathematical classification model. These predicted class memberships may be determined by multiplying a preprocessed spectrum with the regression coefficients of the PLS model. The predicted class memberships may then be assigned by minimizing theEuclidean distances of the predicted values to the original class notations of the first textile samples or, for example, using some suitable decision threshold determined during cross- validation and may then be compared with known class assignments to determine TPR and True Negative Rate, TNR, as the share of correctly classified external test set objects. Apparatus 130 may thus determine a distance of the estimated class membership to class notations of the first and second class.

[0047] Apparatus 130 may then, after positive evaluation, determine the average spectrum of at least one hyperspectral image of at least one target textile sample and evaluate whether the at least one target textile sample comprises elastane. Positive evaluation may refer to determining that the performance mathematical classification model is satisfactory. Apparatus 130 may thus evaluate, based at least on the distance, whether the at least one target textile sample belongs to said first class or said second class. For example, apparatus 130 may determine that the at least one target textile sample comprises elastane when the at least one target textile sample belongs to said first class.

[0048] In some example embodiments, apparatus 130 may hence evaluate the mathematical model, the class model or the discriminant model, by comparing distances of the average spectrum of said hyperspectral images of second and third textile samples to a confidence level, like a certain TRP threshold value.

[0049] FIG. 4a illustrates the results of a SIMCA model in accordance with at least some embodiments of the present disclosure. The horizontal axis in FIG. 4a may denote the distance of the test set objects to the origin of the lower dimensional sub-space within the model and the vertical axis may denote the distance of the objects to the model. The colour of the symbols may denote the known class memberships of the test set objects and the dotted line may describe the combined class threshold used to assign predicted memberships based on the class model.

[0050] FIG. 4b illustrates the results of a EDA model in accordance with at least some embodiments of the present disclosure. The objects in FIG. 4b may denote the location of the projected average spectra in the lower dimensional model subspace described by the PCs shown on the axes. The colour of the symbols may denote the known class memberships of the test set objects and the dotted line may describe the class boundary used to assign predicted class memberships based on the discriminant model. The axes in FIG. 4b may describe the location of the projected objects on the PCs.

[0051] FIG. 5 illustrates an example apparatus capable of supporting at least some embodiments of the present disclosure. Illustrated is apparatus 500, which may comprise or be apparatus 130 of FIG. 1.

[0052] Comprised in apparatus 500 may be processing unit, i.e., processing element, 510, which may further comprise, for example, a single- or multi-core processor wherein a single-core processor comprises one processing core and a multi-core processor comprises more than one processing core. Processing unit 510 may comprise, in general, a control device. Processing unit 510 may comprise one or more processors. Processing unit 510 may be a control device. Processing unit 510 may comprise at least one Application-Specific Integrated Circuit, ASIC. Processing unit 510 may comprise at least one Field- Programmable Gate Array, FPGA. Processing unit 510 may be means for performing method steps in apparatus 500. Processing unit 510 may be configured, at least in part by computer instructions, to perform actions.

[0053] Apparatus 500 may comprise memory 520. Memory 520 may comprise Random- Access Memory, RAM, and / or permanent memory. Memory 520 may comprise at least one RAM chip. Memory 520 may comprise solid-state, magnetic, optical and / or holographic memory, for example. Memory 520 may be at least in part accessible to processing unit 510. Memory 520 may be at least in part comprised in processing unit 510. Memory 520 may be means for storing information, such as a phase and amplitude of a reflected signal. Memory 520 may comprise computer instructions that processing unit 510 is configured to execute. When computer instructions configured to cause processing unit 510 to perform certain actions are stored in memory 520, and apparatus 500 overall is configured to run under the direction of processing unit 510 using computer instructions from memory 520, processing unit 510 and / or its at least one processing core may be configured to perform said certain actions. Memory 520 may be at least in part comprised in processing unit 510. Memory 520 may be at least in part external to apparatus 500 but accessible to apparatus 500.

[0054] Apparatus 500 may comprise a transmitter 530. The transmitter 530 may be used to transmit via a communication network, for example, to transmit to hyperspectral camera 110 or to another apparatus. Apparatus 500 may also comprise a receiver 540. The receiver 540 may be used to receive via a communication network, for example, to receive from hyperspectral camera 110 or from another apparatus.

[0055] Apparatus 500 may also comprise a user interface, UI, 550. UI 550 may comprise at least a display or a touchscreen. A user may be able to operate apparatus 500 via UI 550. Also, UI 550 may be used for displaying information to the user. For example, UI 550 may be used for displaying an indication that the at least one target textile sample comprises elastane.

[0056] Processing unit 510 may be furnished with a transmitter arranged to output information from processing unit 510, via electrical leads internal to apparatus 500, to other devices comprised in apparatus 500. Such a transmitter may comprise a serial bus transmitter arranged to, for example, output information via at least one electrical lead to memory 520 for storage therein. Alternatively to a serial bus, the transmitter may comprise a parallel bus transmitter. Likewise processing unit 510 may comprise a receiver arranged to receive information in processing unit 510, via electrical leads internal to apparatus 500, from other devices comprised in apparatus 500. Such a receiver may comprise a serial bus receiver arranged to, for example, receive information via at least one electrical lead from receiver 540 for processing in processing unit 510. Alternatively to a serial bus, the receiver may comprise a parallel bus receiver.

[0057] Processing unit 510, memory 520, transmitter 530, receiver 540 and / or UI 550 may be interconnected by electrical leads internal to apparatus 500 in a multitude of different ways. For example, each of the aforementioned devices may be separately connected to a master bus internal to apparatus 500, to allow for the devices to exchange information. However, as the skilled person will appreciate, this is only one example and depending on the embodiment various ways of interconnecting at least two of the aforementioned devices may be selected without departing from the scope of the present disclosure.

[0058] FIG. 6 illustrates a flow graph of a method in accordance with at least some embodiments of the present disclosure. The phases of the illustrated method may be performed by apparatus 130. The method may be a computer-implement method. The method, and apparatus 130, may be for detecting elastane from textile samples.

[0059] The method may comprise, at step 610, determining at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane. The method may also comprise, at step 620, determining an average spectrum of each of the at least one first hyperspectral image and at step 630, determining a mathematicalclassification model based on the average spectrum of each of the at least one first hyperspectral image. The method may further comprise, at step 640, determining at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane and, at step 650, determining at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane. Finally, the method may comprise, at step 660, evaluating the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples and detecting, at step 670, using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

[0060] The embodiments disclosed provide a technical solution to a technical problem. One technical problem being solved is how to detect elastane for example from cotton. In practice, this is problematic because only small amounts of elastane are required for increasing the elasticity of yams and elastane may be embedded inside the core of the yam making it difficult to detect. Detection of elastane is important at least for textile sorting and recycling.

[0061] The embodiments herein overcome these limitations by providing a way to detect elastane, for example from cotton, by an apparatus. In this manner, detection of elastane can be accomplished in a more accurate and robust fashion. This results in several advantages. First, elastane can be detected quickly. Second, elastane may be detected in a non-invasive way. Third, detection of elastane may be fully automized. Fourth, processing of target samples may be fully automized. Other technical improvements may also flow from these embodiments, and other technical problems may be solved.

[0062] For example, apparatus 130 may be configured to control hyperspectral camera 110 to take images and transmit said images to apparatus 130. Apparatus 130 may then process the received images according to embodiments of the present disclosure, to detect elastane. Depending on whether elastane is detected by apparatus 130, apparatus 130 may further transmit an indication about whether a target sample comprises elastane, for example to another apparatus via a communication network. Alternatively, apparatus 130 may display the indication on UI 550.

[0063] If elastane is not detected, the indication may indicate that elastane is not detected and hence cause sorting of the target sample accordingly. Apparatus 130 may, in a fully automized system, cause another apparatus to sort the target sample to a container, which is not to be recycled, unless the target sample is processed by removing elastane. On the other hand, if elastane is detected, the indication may indicate that elastane is detected and hence cause sorting of the target sample accordingly. Apparatus 130 may, in a fully automized system, cause another apparatus to sort the target sample to a container, which is to be recycled, without further processing the target sample. Apparatus 130 may hence control another apparatus, such as a sorting apparatus for recycling, depending on whether the at least one target textile sample is detected to comprise elastane.

[0064] Alternative ways for exploiting the indication may be possible as well. For example, apparatus 130 may transmit the indication to cause removal of elastane from the target sample, possibly by another apparatus or a user, when elastane is detected. Hence, the indication may be transmitted to cause processing of the target sample by another apparatus, to make the target sample recyclable.

[0065] It is to be understood that the embodiments of the disclosure disclosed are not limited to the particular structures, process steps, or materials disclosed herein, but are extended to equivalents thereof as would be recognized by those ordinarily skilled in the relevant arts. It should also be understood that terminology employed herein is used for the purpose of describing particular embodiments only and is not intended to be limiting.

[0066] Reference throughout this specification to one embodiment or an embodiment means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Where reference is made to a numerical value using a term such as, for example, about or substantially, the exact numerical value is also disclosed.

[0067] As used herein, a plurality of items, structural elements, compositional elements, and / or materials may be presented in a common list for convenience. However, these lists should be construed as though each member of the list is individually identified as a separate and unique member. Thus, no individual member of such list should be construed as a de facto equivalent of any other member of the same list solely based on theirpresentation in a common group without indications to the contrary. In addition, various embodiments and example of the present disclosure may be referred to herein along with alternatives for the various components thereof. It is understood that such embodiments, examples, and alternatives are not to be construed as de facto equivalents of one another, but are to be considered as separate and autonomous representations of the present disclosure.

[0068] In an exemplary embodiment, an apparatus, such as, for example, apparatus 130, may comprise means for carrying out the embodiments described above and any combination thereof.

[0069] In an exemplary embodiment, a computer program may be configured to cause a method in accordance with the embodiments described above and any combination thereof. In an exemplary embodiment, a computer program product, embodied on a non-transitory computer readable medium, may be configured to control a processing unit to perform a process comprising the embodiments described above and any combination thereof.

[0070] In an exemplary embodiment, an apparatus, such as, for example, apparatus 130, may comprise at least one processing unit, and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, with the at least one processing unit, cause the apparatus at least to perform the embodiments described above and any combination thereof.

[0071] Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the preceding description, numerous specific details are provided, such as examples of lengths, widths, shapes, etc., to provide a thorough understanding of embodiments of the disclosure. One skilled in the relevant art will recognize, however, that the disclosure can be practiced without one or more of the specific details, or with other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the disclosure.

[0072] While the forgoing examples are illustrative of the principles of the present disclosure in one or more particular applications, it will be apparent to those of ordinary skill in the art that numerous modifications in form, usage and details of implementation can be made without the exercise of inventive faculty, and without departing from the principlesand concepts of the disclosure. Accordingly, it is not intended that the disclosure be limited, except as by the claims set forth below.

[0073] The verbs “to comprise” and “to include” are used in this document as open limitations that neither exclude nor require the existence of also un-recited features. The features recited in depending claims are mutually freely combinable unless otherwise explicitly stated. Furthermore, it is to be understood that the use of "a" or "an", that is, a singular form, throughout this document does not exclude a plurality.INDUSTRIAL APPLICABILITY

[0074] At least some embodiments of the present disclosure find industrial application in recycling and reusing waste textiles.ACRONYMS LISTASIC Application-Specific Integrated CircuitFPGA Field-Programmable Gate ArrayLDA Linear Discriminant AnalysisMSC Multiplicative Scatter CorrectionNIR Near-InfraredPC Principal ComponentPCA Principal Component AnalysisPLS Partial Least SquaresPLS-DA PLS Discriminant AnalysisQDA Quadratic Discriminant AnalysisRAM Random- Access MemorySIMCA Soft Independent Modelling by Class AnalogySNV Standard Normal VariateTNR True Negative RateTRP True Positive RateUI User InterfaceREFERENCE SIGNS LIST

Claims

CLAIMS:

1. An apparatus comprising at least one processor; and at least one memory including computer program code; the at least one memory and the computer program code being configured to, with the at least one processor, cause the apparatus at least to:- determine at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane;- determine an average spectrum of each of the at least one first hyperspectral image;- determine a mathematical classification model based on the average spectrum of each of the at least one first hyperspectral image;- determine at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane;- determine at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane;- evaluate the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples; and- detect using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

2. An apparatus according to claim 1, wherein the mathematical classification model is a class model, and the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- determine the class model based on one class and use latent variable weights of the class model to project an average spectrum of the at least one target hyperspectral image to a lower dimensional sub-space, the lower dimensional sub-space describing properties of said one class, wherein a dimensionality of the lower dimensional subspace equals a number of required latent variables;- determine a distance of the average spectrum to the projection of the average spectrum to the lower dimensional sub-space and a distance of the projection to acentre of the lower dimensional sub-space, wherein the centre of the lower dimensional sub-space is a centre of said one class in the lower dimensional subspace; and- evaluate, based at least on the distances, whether the at least one target textile sample belongs to said one class.

3. An apparatus according to claim 2, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- determine that the at least one hyperspectral target textile sample comprises elastane when the at least one target textile sample belongs to said one class.

4. An apparatus according to claim 2 or claim 3, wherein said one class is a class of textile samples comprising elastane.

5. An apparatus according to claim 1, wherein the mathematical classification model is a discriminant model, and the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- determine the discriminant model based on at least a first class of textile samples comprising elastane and a second class without elastane;- use latent variable weights of the discriminant model to project an average spectrum of the at least one hyperspectral target image to a lower dimensional sub-space describing class memberships, wherein a dimensionality of the lower dimensional sub-space equals a number of required latent variables;- determine a distance of the projection to class centroids in the lower dimensional sub-space, wherein a centroid is the centre of one class in the lower dimensional subspace; and- evaluate, based at least on the distance, whether the at least one target textile sample belongs to said first class or said second class.

6. An apparatus according to claim 1, wherein the mathematical classification model is a discriminant model, and the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- determine the discriminant model based on at least a first class of textile samples comprising elastane and a second class without elastane;- use regression coefficients of the discriminant model to estimate a class membership of an average spectrum of the at least one hyperspectral target image;- determine a distance of the estimated class membership to class notations of the first and second class; and- evaluate, based at least on the distance, whether the at least one target sample belongs to said first class or said second class.

7. An apparatus according to claim 5 or claim 6, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- determine that the at least one target textile sample comprises elastane when the at least one target textile sample belongs to said first class.

8. An apparatus according to any of the preceding claims, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- evaluate the mathematical model by comparing distances of the average spectrum of said hyperspectral images of second and third textile samples to a confidence level.

9. An apparatus according to any of the preceding claims, wherein the at least one first, second, third and target hyperspectral images are near infrared images, such as with wavelengths from 780 nm to 2600 nm, preferably 1000 - 2500 nm.

10. An apparatus according to any of the preceding claims, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- receive, from a hyperspectral camera, the at least one first, second, third and target hyperspectral images.

11. An apparatus according to any of the preceding claims, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to:- transmit an indication about whether the at least one target sample comprises elastane.

12. An apparatus according to any of the preceding claims, wherein the at least one first, second, third and target textile samples comprise cotton.

13. A method comprising:- determining at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane;- determining an average spectrum of each of the at least one first hyperspectral image;- determining a mathematical classification model based on the average spectrum of each of the at least one first hyperspectral image;- determining at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane;- determining at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane;- evaluating the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples; and- detecting using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

14. A computer program comprising instructions which, when the program is executed by an apparatus, cause the apparatus to perform:- determining at least one first hyperspectral image from multiple first textile samples, wherein said multiple first textile samples comprise a first mixture of textile fibers, the first mixture further comprising elastane;- determining an average spectrum of each of the at least one first hyperspectral image;- determining a mathematical classification model based on the average spectrum of each of the at least one first hyperspectral image;- determining at least one second hyperspectral image from multiple second textile samples, wherein said second textile samples comprise a second mixture of textile fibers, the second mixture further comprising elastane;- determining at least one third hyperspectral image from multiple third textile samples, wherein said third textile samples comprise textile fibers without elastane;- evaluating the mathematical classification model based at least on the average spectrum of said hyperspectral images of second and third textile samples; and - detecting using the mathematical classification model, after positive evaluation, from an average spectrum of at least one hyperspectral target image of at least one target textile sample whether the at least one target textile sample comprises elastane.

15. A system, comprising: - a hyperspectral camera and the apparatus according to any of claims 1 to 12, wherein the hyperspectral is configured to take the hyperspectral images, and transmit the hyperspectral images to the apparatus.

Images