EP3887591A1

EP3887591A1 - Textile identification apparatus and method for identifying a textile type

Info

Publication number: EP3887591A1
Application number: EP19808794.2A
Authority: EP
Inventors: Monika Hohmann; Christina KRICK CALDERON; Philipp Neumaier
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2018-11-27
Filing date: 2019-11-22
Publication date: 2021-10-06
Also published as: WO2020109170A1; CN113348278B; DE102018220370A1; CN113348278A

Abstract

A textile identification apparatus (1) comprises an IR scanner (2) and a data processing device (3) for identifying a textile type of a textile for measurement on the basis of an IR measurement spectrum recorded from the textile for measurement by means of the IR scanner (2), wherein the data processing device (3) is configured to classify the recorded IR measurement spectrum, or an IR measurement spectrum calculated therefrom, on the basis of comparisons with IR reference spectra that correspond to different reference textile types and to adopt as the textile type that reference textile type whose IR reference spectrum exhibits the best correspondence with the IR measurement spectrum. A method (S1-S7) serves to identify a textile type, wherein at least one IR measurement spectrum of a textile for measurement is recorded (S1); the recorded IR measurement spectrum, or an IR measurement spectrum calculated therefrom, is classified by comparison with IR reference spectra that correspond to different reference textile types (S6); and that reference textile type whose IR reference spectrum exhibits the best correspondence with the IR measurement spectrum is adopted as the textile type of the textile for measurement (S6). The invention is particularly advantageously applicable to the identification of a textile and possibly stains present thereon, especially in a domestic setting.

Description

Textile recognition device and method for recognizing a

Type of textile

The invention relates to a textile detection device, comprising an infrared (IR) scanner and a data processing device, the data processing device being set up to recognize a textile type of the measuring textile on the basis of an IR measurement spectrum recorded by a measuring textile by means of the IR scanner . The invention also relates to a method for recognizing a type of textile, in which at least one IR measurement spectrum of a measurement textile is recorded. The invention is particularly advantageously applicable to the detection of a textile and any stains present thereon, particularly in a household.

DE 37 060 56 A1 discloses a method for generating and recognizing optical spectra and switching and sensor systems, in particular for sewing and textile automation.

EP 1 242 665 B1 discloses a washing machine, tumble dryer, spin dryer or machine for chemical cleaning or for dyeing textiles in a drum with a device for recognizing properties of a textile, the device providing at least one transmitting and at least one receiving element for transmitting or receiving electromagnetic radiation and an evaluation circuit connected to the reception element, the radiation transmitted and / or transmitted by the transmission element and reflected by the textile being received by the reception element and evaluable in the evaluation circuit.

It is the object of the present invention to at least partially overcome the disadvantages of the prior art and, in particular, to provide a particularly quick and reliable identification of textiles with structurally simple means.

This object is achieved in accordance with the features of the independent claims. Advantageous embodiments are the subject of the dependent claims, the description and the drawings. The object is achieved by a textile recognition device having an IR scanner and a data processing device, the data processing device being set up to recognize a textile type of the measuring textile on the basis of an IR measuring spectrum recorded by a measuring textile by means of the IR scanner. wherein the data processing device is set up to classify the recorded IR measurement spectrum or an IR measurement spectrum calculated therefrom on the basis of comparisons with IR reference spectra which correspond to different reference textile types and as the textile type of the measurement textile that reference textile type whose IR reference spectrum shows a best match with the IR measurement spectrum.

The use of a classification gives the advantage that a particularly quick and reliable identification of textiles can be achieved with structurally simple means, which can also be easily used by an end user.

In particular, the textile recognition device is a household textile recognition device that can be used in a household.

The IR scanner can irradiate a textile or a textile with IR light and record or measure reflected IR light. This textile, the type of textile to be determined, is referred to below as "measurement textile". The measuring textile can be a piece of laundry, for example a piece of clothing. The type of textile can correspond to a composition of the textile from one or more pure types of fibers, as indicated, for example, on laundry labels. The IR scanner works not only monofrequency, but in a predetermined wavelength band, so that it records wavelength-resolved IR reflection measurement spectra. The wavelength band can in particular be an NIR (near infrared) band and / or a MIR (middle infrared) band. For example, the wavelength band can have a bandwidth between 1300 and 3000 nm, for example from 1500 nm to 2500 nm, from 1750 nm to 2250 nm, from 1800 nm to 2200 nm, from 1550 nm to 1950 nm, from 2000 nm to 2500 nm, from 2250 nm to 2650 nm etc. A spectral resolution or step size can be between 5 nm and 20 nm, for example. A wavelength band can have, for example, between 50 and 500 spectral measuring points. The recorded IR measurement spectrum can in principle be compared directly with IR reference spectra. For a particularly reliable determination of the type of textile, however, it is advantageous if the recorded (“original”) IR measurement spectrum is further processed to an “Dar calculated” IR measurement spectrum. The IR measurement spectrum calculated from this remains a spectrum. The IR reference spectra can also be referred to as IR comparison spectra.

The fact that the IR measurement spectrum is classified includes, in particular, that it is compared as a spectrum with the IR reference spectra and that the type of textile is adopted from the most suitable or most matching IR reference spectrum. This is e.g. in contrast to evaluating only characteristic quantities such as peak values (peak heights) or positions of certain peaks etc. from the IR measurement spectrum. In particular, the IR measurement spectrum does not need to be derived, Fourier transformed, etc. in the mathematical sense.

The IR reference spectra can be determined experimentally and / or calculated. The experimental determination is advantageously carried out by means of an IR scanner of the same type, by means of which the IR measurement spectrum is also recorded. The experimental determination can be made in the factory. Additionally or alternatively, the experimental determination can be carried out by an end user, for example using a calibration sample set which has several fabric samples with reference or reference textiles of a known type of textile. The fabric samples can include several different processing types for a textile type, for example single-layer and double-layer samples. In particular, the samples can only comprise pure fiber types, with IR data spectra for fiber mixtures then being calculated from the IR reference spectra of the pure fiber types using the data processing device.

In a further development, the samples do not have a dedicated background. This has the advantage that the fabric samples can also be measured if they rest on a body part of a user. This allows an effect of the spectral influence of the skin to be taken into account.

A best match can be found, for example, using the least squares method or other correlation methods. It is one embodiment that the best match examination is a multi-stage examination or classification. This has the advantage that the textile type can be determined or recognized particularly quickly. In particular, the classification can comprise a first classification ("rough classification") which is carried out first, followed by at least one further classification ("fine classification"). The multi-level classification enables a hierarchy of classification groups and a classification based on decision trees or decision paths to be implemented. For example, a rough classification can be carried out on the basis of a determination of a best match with an IR reference spectrum from a group of IR reference spectra, each of the IR reference spectra representing a group of several types of fibers whose IR spectra are similar.

For example, in the course of a rough classification for determining a pure type of fiber, an IR reference spectrum can first serve as a representative for the groups "cotton and linen", "wool and silk", or "polyacrylic, polyamide and polyethylene". If it is determined by comparing the (original or derived) IR measurement spectrum that the textile type of the measurement textile belongs to one of these groups, the IR measurement spectrum can be compared with the IR reference spectra of the individual fiber types in this group in a subsequent fine classification. For example, if it was recognized during the rough classification that the textile type of the measurement textile belongs to the "Cotton and Linen" group, the following measurement can correspond to the IR measurement spectrum with the two textile types "Cotton" and "Linen" IR reference spectra are compared to identify the final type of textile, for example "cotton". In the example above, only five spectral comparisons need to be carried out, whereas a direct comparison with the individual textile types would require seven spectral comparisons. This multi-level classification thus results in a reduction in calculation effort with the same accuracy.

An even coarser classification can, for example, be carried out beforehand by means of a classification as a pure fiber type or fiber mixture.

Pure fiber types can, for example, natural fibers of vegetable and / or animal origin such as cotton, linen, hemp fiber, silk, wool etc. and / or chemical fibers such as polymer fibers (e.g. polyamides, polyethylenes, polyacrylonitriles, polyurethanes (possibly with a variant elastomer component, elastane), polypropylene) etc. and / or regenerated fibers such as viscose, modal, lyocell or cupro, but are not limited to this. Fiber mixtures can in particular be mixtures of two or more of these pure types of fibers with respective mixture proportions.

It is an embodiment that the IR reference spectra include spectra of pure fibers and / or spectra of fiber mixtures. This enables particularly diverse textiles to be recognized. At least the following IR reference spectra are preferably provided:

It is an embodiment that the data processing device is set up to classify the IR measurement spectrum on the basis of comparisons with IR reference spectra, which correspond to different mixture fractions of the identified fiber mixture, when the fiber type has been recognized as the textile type, and the Recognize mixture proportions on the basis of the IR reference spectrum, which shows the best agreement with the IR measurement spectrum. This embodiment can also correspond to a multi-stage classification in that the fiber mixture is first determined and then the mixture proportions of the pure fibers or fiber types on which the underlying fibers are based are determined with knowledge of the fiber mixture. The IR reference spectra that are used to determine the mixture proportions correspond to the spectra of the corresponding fiber mixtures to determine the mixture proportions.

Methods such as SIMCA, PLS and PCA are often used in accordance with the prior art. These methods are linear modeling methods. A major disadvantage of these common methods in the NIR wavelength range between 1000 - 2000 nm is the poor separability of the following exemplary mixtures X / Y from the main textile X (Y is the textile material with a smaller proportion): cotton / elastane (CO + EL) , Polyamides / elastane (EL + PA), silk / elastane (EL + SE), cotton / wool (CO + WO), cotton / viscose (CO + CV) and polyester / elastane (EL + PES).

In order to enable a classification of a textile material with a percentage Y in the single-digit percentage range, a non-linear separation method is preferably used. In particular, the use of PLS density, bagging trees and / or random forrest comes into consideration.

At PLS Density, the pure textiles and blend classes are preferably modeled separately. Each point in the point cloud in the PLS subspace can correspond to a potential, which is then fitted into a distribution statistic. A class affiliation can be defined by the skilful choice of termination criteria.

With the random-forrest method, several non-correlated decision trees can be trained to enable a separation of the classes (pure textiles and a class of all mixtures). Bagging trees allow the different model results to be merged and weighted and may provide an improved overall forecast result. In addition to the advantage of classifying difficult textile blends, the use of the PLS density method can increase the robustness of the model, particularly in relation to the variation of the sample spectra, the variation in device handling by the customer, the sensor-sensor variance (production-related hardware variance) and further negative environmental influences. This can be done, for example, through the implicit modeling in the training data set and / or the statistical consideration of the influences in modeled pure textile or blend classes.

In embodiments, the PLS Density, Bagging Trees and / or Ran dom Forrest methods can provide a probability value that a measured spectrum belongs to one of the classes in the table above with spectra of pure fibers and spectra of fiber mixtures. In one embodiment, there is an IR reference spectrum for each class. There are preferably classes for both pure textiles and fiber blends. A non-linear separation method, e.g. PLS Density, Bagging Trees or Random Forrest, can provide a probability W for each of these classes that the IR measurement spectrum corresponds to this class. Thus, each measured spectrum per class can get a probability. The values can vary, for example, between 0 ... 1, where 0 indicates a completely improbable class membership and 1 a completely reliable class membership. If e.g. the class CO gets the value 1 and all other classes get the value 0, it means that the model is very sure that the measured spectrum is cotton.

In some embodiments, all classes may have values between 0 and 1. The data processing device can be set up to evaluate a large number of probabilities W on the basis of threshold values S, preferably 0 <S0 <S1 <S2 <1. In one embodiment, the following non-linear separation method is used:

• If a class has a probability W of greater than S2, this class is recognized as the textile type of the measurement textile. For example, if the recognized class is a pure textile, ie pure fiber, the user can be shown 100% cotton. If the recognized class is a fiber mixture, for example a PLS regression with the IR measurement spectrum can be carried out in a subsequent step. The percentages of the respective fiber are preferably used as a proportion determined and issued to the customer. For example, 50% cotton and 50% polyester can be issued to the user.

• If all classes have a probability W of less than SO, a completely unknown material is assumed. Then, for example, the instruction “please scan a textile” could be issued to the user.

If some classes have a probability of W less than SO (W <S0), some classes have a probability W between SO and S1 (S0 <W <S1) and some classes have a probability W between S1 and S2 (S1 <W <S2) an unknown fiber mixture or a fiber mixture with 3 or even more components can be assumed. As a result, the user could then be shown, for example, “fiber mixture”. In this case, further regression is preferably dispensed with, since a percentage recognition of the fiber components could possibly be very difficult.

• If some classes have a probability of W less than SO (W <S0), some classes have a probability W between SO and S2

(S0 <W <S2) and a class has a probability W greater than S2

(W> S2), then an unknown fiber mixture or a fiber mixture with 3 or more components can be assumed, with one type of fiber being known. The result can then be, for example, "fiber mixture with X", where X e.g. is a pure textile from the table above. Here too, a further regression is preferably dispensed with, so that for reasons of efficiency no percentage recognition of the fiber components is attempted.

It is a further development that the IR reference spectra, which are used to determine the mixture proportions, have been calculated or simulated from IR reference spectra of pure fibers or fiber types.

For example, an IR reference spectrum Ir, mix of a fiber mixture consisting of i (i> 1) pure fiber types according to

Ir, mix = S, Ai lr, mat_i can be calculated with S, Ai = 1. Lr, mat_i is the IR reference spectrum of the i-th pure fiber type and Ai is the percentage of the i-th pure fiber type in the fiber shung. To select certain IR reference spectra lr, mat_i pure fiber types from the totality of all IR reference spectra of pure fiber types, the selection criterion can be, for example, the minimum Mahalanobis distance to the data center in the feature space of the first two main components after specific main component analysis of the respective textile type.

Alternatively or additionally, the IR reference spectra, which are used to determine the mixture proportions of fiber mixtures, can be determined experimentally. This can e.g. by measuring IR spectra on mixed textile samples, e.g. on mixed textile samples with percentage steps of Ai = 1, 2, 5, 10, 20, 30 and 50% proportion of a pure fiber in a fiber mixture.

In addition, a test can be carried out before the classification is used to determine whether the measured spectrum is suitable for use with the classification. This allows incorrect measurements to be recognized, e.g. by identifying spectral artefacts, for example by an outlier detection. If it is recognized that a measured spectrum is unsuitable for use with the classification, a message can be issued to repeat the measurement. Unsuitable samples can also be identified by comparing a measured IR spectrum with model samples (plausibility / significance test). A message can then be given to a user that the IR measurement spectrum is not suitable for use in the classification or the hierarchical model.

It is an embodiment that the data processing device is set up so that when the type of textile has been recognized as a pure fiber or fiber type, the IR measurement spectrum additionally follows, based on comparisons with IR reference spectra, the mixtures of the recognized pure fiber with low Share proportions of fibers of a different type, classify and recognize a mixture share based on the IR reference spectrum, which shows the best agreement with the IR measurement spectrum. This has the advantage that the type of textile of textiles with a strongly dominant fiber content and only small admixtures of other types of fiber can be determined quickly and reliably. Here, for example, a rough classification can first of all be carried out with respect to a specific fiber group (for example “cotton and linen” etc., as described above) and then in the following io

Fine classification the suitable pure fiber type can be determined. If the textile has the recognized "pure" fiber type as a strongly dominant fiber component, the fine classification also leads to a result if there are small additions (e.g. of <1%) of other fiber types. In order to reliably recognize these other types of fiber as well, following the detection of the dominant "pure" type of fiber, a new classification can be carried out to determine whether and if so in what small amount other types of fiber are present ("very fine classification").

It is an embodiment that the IR spectra used for the classification (measurement spectra and reference spectra) are absorbance spectra. An IR measurement absorbance spectrum represents an IR measurement spectrum calculated from an IR measurement reflection spectrum. This has the advantage that the amount of the shares of different pure fiber types in a fiber mixture can be determined particularly precisely from the size of the tips of the absorbance spectrum . In particular, the proportions of different pure fiber types in a fiber mixture can be linearly reduced to the strength of the associated absorbance spectrum. If, for example, an IR absorbance spectrum for a fiber mixture is to be calculated from the IR absorbance spectra of pure fiber types, the mixture proportions can be set linearly via the relative strength of the IR absorbance spectra of the pure fiber types. This is not easily possible when using IR reflection or reflectance spectra.

The absorbance spectra can be calculated from the measured reflectance spectrum using the formula la = - log Ir, with la an absorbance spectrum and Ir a reflectance spectrum. Alternatively, other methods can also be used, e.g. a Kubelka-Munk transformation.

It is a further development that the absorbance spectra are smoothed absorbance spectra. This increases the reliability of the spectrum comparison. The smoothing can be done for example using a Savitzky-Golay filter. It is a further development that the absorbance spectra are SNV-corrected absorbance spectra. This further increases the reliability of the spectral comparison. The previously smoothed absorbance spectra can be subjected to an SNV (standard normal variant) correction.

It is an embodiment that the absorbance spectra are calculated from measured and subsequently standardized reflectance spectra. This has the advantage that the reliability of determining the type of textile increases even further.

The standardization of a directly measured, original IR reflectance measurement spectrum lr, raw, which has reflection values as measured values (e.g. so-called "counts"), can be done, for example, using the formula

Ir, norm = (lr, raw - Ir, dark) / (rest - Ir, dark) or the formula

Ir, norm = lr, raw / (rest - Ir, dark) can be converted into a standardized IR reflectance measurement spectrum Ir, norm, where rest is the reflectance spectrum of a given standard material such as Spectralon and Ir, dark is the reflectance dark spectrum. The mathematical connections can be carried out for each spectral channel or spectral point of the spectra.

The reflectance dark spectrum Ir, dark can be automatically picked up by the IR scanner and advantageously enables a noise component inherent in the IR scanner to be suppressed. The use of the reflectance spectrum INST of the standard material advantageously enables a "normalization" of the IR measurement spectrum to percentage values.

It is an embodiment that the data processing device is set up to

- to recognize a textile type of the measuring textile as described above on the basis of the IR measuring spectrum recorded by the measuring textile by means of the IR scanner; to recognize a type of stain of the stain on the measurement textile on the basis of the IR measurement spectrum recorded by the measurement textile in the area of the stain using the IR scanner.

The detection of the type of stain can be carried out analogously to the detection of the type of textile, e.g. using IR reference spectra for different stain types. In particular, the type of textile and then the type of stain can be determined from at least one original IR measurement spectrum. To determine the type of textile, the IR scanner can be aimed at a spotless area of the textile, and then the IR scanner can be pointed at the spot to determine the type of spot. This means that the type of textile and type of stain can be determined particularly reliably. Alternatively, the IR scanner can only be aimed at the stain, and the type of textile and the type of stain are determined from the same IR measurement spectrum. The number of measurements can advantageously be reduced in this way.

The standardization of a directly measured, original IR measurement spectrum lr, raw, which has reflection values as measured values on the area of a textile provided with a spot, can be done, for example, using the formula

Ir, norm = (lr, raw - Ir, dark) / (lr, tex - Ir, dark) or the formula

Ir, norm = lr, raw / (lr, tex - Ir, dark) are converted into a standardized reflectance spectrum Ir, norm of the spot, where lr, text is the reflectance spectrum of the measurement textile without a spot. Ir, tex can e.g. Ir, raw for the textile case described above.

It is an embodiment that the textile recognition device additionally has a visual sensor that is sensitive in the visual spectral range and the data processing device is set up to use an IR classification of the type of stain as described above and by means of the visual textile in the area by means of the visual sensor of the visual spectra values recorded (eg color signals) to recognize a type of stain of the stain on the measurement textile by comparison with visual reference spectral values. The visual spectral values belong in particular to different colors, so that the visual spectral values can form a color spectrum (for example an RGB color spectrum) in order to be able to determine a color of the textile at the location of the spot.

By combining the thermochemical recognition of the type of textile as described above by means of the IR classification together with a visual optical evaluation of the stain, a particularly precise determination of the type or material composition of the stain can be achieved. Is e.g. the amount of stain material is very small, the resulting uncertainty in the determination of the stain type by means of IR classification can be improved by an additional color analysis. For example, a stain that comes from a carrot can be distinguished from other-colored stain types by recognizing its orange-colored hue, and thus the stain type can be recognized more reliably.

The visual sensor can be integrated in the IR scanner (combined IR-vis scanner) and in particular operated simultaneously with it. The visual sensor can e.g. one or more photo diodes (e.g. RGB sensitive photo diodes), a CCD sensor, etc.

It is an embodiment that the IR scanner is a hand-held scanner. This gives the advantage that the determination of the type of textile is particularly simple and can be carried out regardless of location, even by an end user.

It is an embodiment that the data processing device is integrated in the IR scanner. This achieves the advantage that the textile recognition device can be operated autonomously, in particular even without an Internet connection or the like.

It is an embodiment that the data processing device is integrated in an external entity connected to the IR scanner for data processing purposes. This has the advantage that the computing power for carrying out the textile recognition is made available by the external entity and the IR scanner can be kept simple and inexpensive. The external instance can be, for example, a network server, for example a manufacturer or representative of the textile recognition device, or a computer network such as the so-called "cloud".

It is an embodiment that the textile recognition device is set up to issue at least one laundry care instruction based on the recognized textile type, for example at least one wash instruction (e.g. comprising a maximum washing temperature, recommended detergent etc.), at least one cleaning instruction (e.g. comprehensively recommended or not recommended) Methods of chemical cleaning etc.), at least one drying instruction (e.g. comprising suitability for a drying process, a maximum drying temperature etc.), at least one stain removal instruction (e.g. comprising a stain remover suitable for removing a recognized stain on the recognized textile type etc.).

The object is also achieved by a method for recognizing a type of textile, in which at least one IR measurement spectrum of a measurement textile is recorded; the recorded IR measurement spectrum or an IR measurement spectrum calculated from it is classified by comparing it with IR reference spectra which correspond to different reference textile types; and as the type of textile of the measuring textile that reference type of textile whose IR reference spectrum shows a best match with the IR measuring spectrum is adopted.

The method can be designed analogously to the textile recognition device and gives the same advantages. For example, the method can also be further developed into a method for detecting stains, etc.

The object is further achieved by a computer program product which, when it runs on a data processing device, carries out the above method. The computer program product can be designed analogously to the method and to the textile detection device and has the same advantages

The above-described properties, features and advantages of this invention, as well as the manner in which they are achieved, will become clearer and more clearly understood in connection with the following schematic description of an exemplary embodiment, which is explained in more detail in connection with the drawings. 1 shows a possible sequence for recognizing a type of textile on the basis of the textile detection device according to a first exemplary embodiment; and FIG. 2 shows a possible classification structure used in step S6 of FIG. 1.

1 shows a possible sequence for recognizing a type of textile on the basis of the textile detection device 1. The textile detection device 1 has an IR scanner 2 and a data processing device 3. The data processing device 3 can be integrated in the IR scanner 2 or be an external entity, e.g. a network computer. In particular in the event that the data processing device 3 is an external entity, the IR scanner 2 is network-compatible. For this he can e.g. have a wireless communication module (not shown) such as a Bluetooth module or a WLAN module.

In a step S1, an original IR reflectance measurement spectrum of a measurement textile (e.g. a piece of laundry of an end user, not shown) is recorded by means of the IR scanner 2 and transmitted to the data processing device 3.

In a step S2, the original IR reflectance measurement spectrum is converted into a standardized IR reflectance measurement spectrum, e.g. using an IR spectrum of a standard material and / or a dark spectrum.

In a step S3, the standardized IR reflectance measurement spectrum is converted into an IR absorbance measurement spectrum, e.g. by logarithmization.

In a step S4, the IR absorbance measurement spectrum is smoothed, e.g. using a Savitzky-Golay filter.

In a step S5, the smoothed IR absorbance measurement spectrum is SNV corrected.

In a step S6, the SNV-corrected smoothed IR absorbance measurement spectrum is subjected to a classification by comparison with reference spectra, as a result of which the type of textile, in the case of mixed textiles, including the proportions of pure fiber types, is determined. The textile type of the measuring textile is the reference textile type whose IR Reference spectrum shows a best match with the IR measurement spectrum, adopted as a textile type or recognized.

In a step S7, the recognized type of textile is displayed (e.g. on the textile recognition device 1 and / or a device that can be connected to it in terms of data technology, such as a smartphone, etc.) and laundry care instructions are output if necessary.

2 shows a possible classification structure used in step S6 of FIG. 1 with different comparison or decision blocks B1 to B8.

In a first decision block B1, by comparing the SNV-corrected smoothed IR absorbance measurement spectrum with corresponding IR absorbance reference spectra, it is examined whether the IR measurement spectrum belongs to a textile ("TEX") or not ("NONTEX") . A correlation method determines whether the IR measurement spectrum belongs to a textile or not, whereby a best match of the IR measurement spectrum with one of the IR reference spectra is evaluated as a match with this reference spectrum. The lack of belonging to a textile can be determined by the fact that the deviation from all IR reference spectra belonging to a textile exceeds a predetermined threshold value, that is to say there is not a sufficiently good agreement with any of these IR reference spectra.

If the IR measurement spectrum has been classified as belonging to a textile, a decision block B2 examines whether this belongs to the IR measurement spectrum by comparing the SNV-corrected smoothed IR absorbance measurement spectrum with (further) IR absorbance reference spectra Textile consists of a single type of pure fiber ("P") or consists of a mixture of fibers ("MIX").

Decision blocks B1 and B2 can also be regarded as belonging to a rough classification.

If it is determined in decision block B2 that a pure textile is present, the process branches to decision block B3, in which the IR measurement spectrum is compared with IR reference spectra which either correspond to a pure fiber type or a group from correspond to several pure fiber types. The pure fiber types and groups of several pure fiber types are symbolized here as rounded boxes.

The groups of several pure fiber types used in decision block B3 can e.g. include:

- group of two or more cellulosic synthetic fibers,

- group of two or more vegetable fibers,

- group of two or more fibers of animal origin,

- Group of two or more polymer fibers, but are not limited to this.

The IR reference spectrum of a group can e.g. be used when the IR reference spectra of the individual pure fiber types in this group are similar. The IR reference spectrum of the group can then also be regarded as a coarsened or generalized IR reference spectrum for this group, which on this hierarchy level sufficiently approximates all IR reference spectra of the individual pure fiber types in this group.

If it is determined in decision block B3 that the IR measurement spectrum belongs to a single type of fiber (e.g. polyester), a branch is made to decision block B5. If it is determined in decision block B3 that the IR measurement spectrum belongs to a group of pure fiber types, a branch is made to decision block B4.

In decision block B4, the IR measurement spectrum is compared with the individual IR reference spectra of the pure fiber types in the group and the association of the textile with one of these fiber types is determined.

The pure fiber types are symbolized as rounded boxes.

Decision blocks B3 and B4 can be viewed as "fine classification". In decision block B5, a comparison of the IR measurement spectrum with correspondingly fine IR reference spectra is used to qualitatively investigate whether the pure fiber type (e.g. wool) previously identified in decision blocks B3 or B4 is not a slight addition of at least one other fiber type (e.g. Polyamide and / or polyacrylonitrile) (ie there is a mixture of fibers), and which this is at least one other type of fiber. If not, the pure fiber type classified as textile type is displayed to a user. The qualitative blends of dominant (almost pure) fiber types are symbolized here as rounded boxes.

If the dominant "pure" fiber type previously identified in decision block B5 has a low admixture of other fiber types, then in decision block B6 a comparison of the IR measurement spectrum with correspondingly fine IR reference spectra is used to classify how high the proportions are (eg 98% Wool + 2% polyamide). The result of the classification is shown to a user below. The quantitative mixtures of dominant (almost pure) fiber types are symbolized here as rounded boxes, e.g. Comprehensive IR reference spectra for fiber blends 99.5% wool + 0.5% polyamide, 99% wool + 1% polyamide, 98% wool + 2% polyamide, etc.

If it is determined in decision block B2 that there is a fiber mixture, a branch is made to decision block B7, in which the IR measurement spectrum is compared with IR reference spectra, the mixtures of different types of fibers. The blends of different types of fibers are symbolized here as rounded boxes.

Possible fiber mixtures can e.g. comprise a mixture of two or more of the pure fiber types above for decision blocks B3 and B4.

If the fiber mixture has been qualitatively classified in decision block B7, the IR measurement spectrum is subsequently compared in decision block B8 with IR reference spectra which correspond to different mixture proportions of the previously determined fiber mixture. The result of the classification is then displayed to a user quantitatively and qualitatively. The different blending proportions for basically all possible fiber blends are symbolized here as rounded boxes. The classification based on decision blocks B7 and B8 is basically similar to the classification based on decision blocks B5 and B6, whereby IR reference spectra of other fiber mixtures and / or fiber components can also be used, e.g. 50% wool + 50% polyamide, 60% wool + 40% polyamide, 70% wool +30% polyamide, etc .

The above method can be present as a computer program product in the data processing device 3, e.g. as "embedded" software.

Of course, the present invention is not limited to the exemplary embodiment shown.

In this way, step S6 or S7 can also be followed by a sequence for detecting a type of stain. This process can be carried out analogously to steps S1 to S7 or S2 to S7, the IR reference spectra then representing spectra of different stain types or stain materials. Only in step S2 is the original IR reflectance measurement spectrum (which also includes spectral components of the stain material) not standardized using a standard material, but the type of textile. For example, an IR reflectance spectrum of the spot-free textile can be used as the IR standard spectrum.

The IR classification can also be followed by visual / optical stain detection, for which the textile detection device can also have an optical sensor 4, which is indicated in FIG. 1.

In general, "a", "a" etc. can be understood to mean a single number or a plurality, in particular in the sense of "at least one" or "one or more" etc., as long as this is not explicitly excluded, e.g. by the expression "exactly one" etc.

A number can also include the specified number as well as a customary tolerance range, as long as this is not explicitly excluded. Reference symbol list

1 textile recognition device

2 IR scanners

3 data processing device

4 Visual sensor

S1-S7 procedural steps

B1-B8 decision blocks

TEX textile

NONTEX non-textile

P Pure type of fiber

MIX fiber blend

Claims

1. Textile recognition device (1), comprising an IR scanner (2) and a data processing device (3), the data processing device (3) being set up by means of an IR picked up by a measurement textile using the IR scanner (2) -Measuring spectrum to recognize a textile type of the measuring textile, the data processing device (3) being set up to compare the recorded IR measuring spectrum or an IR measuring spectrum calculated therefrom on the basis of comparisons with IR reference spectra which correspond to different reference textile types classify and adopt as the textile type of the measurement textile that reference textile type whose IR reference spectrum shows the best match with the IR measurement spectrum, characterized in that the IR reference spectra include spectra of pure fibers and spectra of fiber mixtures and the data processing device (3) is set up for this purpose when a fiber mixture is recognized as the type of textile has been to classify the IR measurement spectrum on the basis of comparisons with IR reference spectra, which correspond to different mixture fractions of the identified fiber mixture, and to recognize the mixture fractions on the basis of the IR reference spectrum, which shows the best agreement with the IR measurement spectrum.

2. Textile recognition device (1) according to claim 1, characterized in that the data processing device (3) is set up to use a non-linear separation method, in particular using PLS density, bagging trees or random forrest, in order to use the IR measurement spectrum to classify.

3. Textile recognition device (1) according to one of the preceding claims, characterized in that the examination for a best match is a multi-stage examination.

4. Textile recognition device (1) according to one of the preceding claims, characterized in that the data processing device (3) is set up for this purpose, when the type of textile has been recognized as a pure fiber, the IR Classify the measurement spectrum on the basis of comparisons with IR reference spectra, which correspond to mixtures of the detected pure fiber with small fractions of fibers of another type, and to recognize a mixture fraction based on the IR reference spectrum, which shows the best match with the IR measurement spectrum.

5. Textile recognition device (1) according to one of the preceding claims, characterized in that the IR spectra used for the classification are absorbance spectra, in particular smoothed and SNV-corrected absorbance spectra.

6. Textile recognition device (1) according to claim 5, characterized in that the data processing device (3) is set up to calculate the absorbance spectra from measured and subsequently standardized reflectance spectra.

7. Textile recognition device (1) according to one of the preceding claims, characterized in that the data processing device (3) is additionally set up to use an IR measurement spectrum recorded by means of the IR scanner (2) of the measurement textile in the area of a spot Detect the type of stain on the measuring textile.

8. The textile recognition device (1) according to claim 7, characterized in that the textile recognition device (1) additionally has a visual sensor (4) which is sensitive in the visual spectral range and the data processing device (3) is set up to additionally by means of the visual sensor (4 ) to recognize a type of stain of the stain on the measurement textile by comparing it with visual reference spectral values of the measurement textile in the area of the stain.

9. Textile recognition device (1) according to one of the preceding claims, characterized in that the IR scanner (2) is a hand-held scanner.

10. Textile recognition device (1) according to one of the preceding claims, characterized in that the data processing device (3) is integrated in the IR scanner (2).

1 1. Textile recognition device (1) according to any one of the preceding claims, characterized in that the data processing device (3) is integrated in an external entity connected to the IR scanner (2) for data processing purposes.

12. Textile recognition device (1) according to one of the preceding claims, characterized in that the textile recognition device (1) is set up to output at least one laundry care instruction based on the recognized material composition.

13. Method (S1-S7) for recognizing a type of textile, in which

at least one IR measurement spectrum of a measurement textile is recorded (S1); the recorded IR measurement spectrum or an IR measurement spectrum calculated therefrom is classified by comparison with IR reference spectra which correspond to different reference textile types (S6), the IR reference spectra comprising spectra of pure fibers and spectra of fiber mixtures; and

the reference textile type whose IR reference spectrum shows the best match with the IR measurement spectrum is adopted as the textile type of the measuring textile (S6),

characterized in that

when a fiber mixture has been recognized as the textile type, the IR measurement spectrum is classified on the basis of comparisons with IR reference spectra which correspond to the recognized fiber mixture under different blending proportions and

the mixture proportions are recognized on the basis of the IR reference spectrum, which shows the best agreement with the IR measurement spectrum.

14. The method according to claim 13, characterized in that a non-linear separation method, in particular using PLS density, bagging Trees or Random Forrest, is used to classify the IR measurement spectrum.