CH654608A5 - Method for detecting defects in sets of fibres and apparatus for implementing this method - Google Patents

Abstract

The defects in a bundle of fibres, such as lengths of wool, are detected during passage of the web of fibres (13) under at least one photosensitive detector (11 or 21), by directing the light towards the length so that at least a portion of this light is received by the detector, detecting signals produced by this detector. The bundle of fibres or length is preferably stretched, for example by a system with a sleeve (26) and cylinders (27 and 28) and passes successively under at least two detectors, of which the first detector (11) is placed in a manner such that it receives the light transmitted by the source (14) through the length, the other detector (21) is placed in order to respond to the light reflected by the length and coming from the source (24). The signals produced by the first detector being synchronised with those of the second detector corresponding to the same portion of the bundle of fibres. The signals are analysed electronically in order to detect the presence or the size of a defect, and in order to discriminate using the differences between reflectance and transmittance, between defects of different types. <IMAGE>

Description

** ATTENTION ** start of the DESC field may contain end of CLMS **.

CLAIMS 1. Method for detecting and classifying faults appearing in a network of fibers, which consists in passing the network of fibers under at least two photosensitive sensors, directing light towards the network of fibers, one of the sensors being placed to respond to the light transmitted through the fiber network and the other being placed to respond to the light reflected by the fiber network, the signals produced by the first sensor being synchronized with those of the second sensor corresponding to the same portion of the fiber network and detecting the presence and nature of a fault from the signals produced by the sensors during the passage of a fault, characterized in that, when the analog signal emitted by at least one sensor exceeds the smallest of a series of thresholds (THI, TH2, TH3, ...), a digital signal composed, for each sensor, of a series of digital words identifying the largest threshold exceeded by the analog signal as well as the lateral (X) and longitudinal (Y) positions of the location of the network where the overshooting took place is stored, and the values stored in memory from the two sensors and corresponding to the same portion of the fiber network are then analyzed.

2. Method according to claim 1, in which the digital signals stored in memory are analyzed to detect the presence or the size of a fault.

3. Method according to claim 2, in which, for each sensor, a reference signal is automatically created and represents the average value of said signals in the absence of a fault, and this reference signal is subtracted from a video signal produced by this sensor to reveal a fault as an impulse after subtraction.

4. Method according to claims 2 and 3, wherein the sensor comprises a network of photoelectric cells placed transversely and which are used to scan the network of fibers during its passage, and in which the output of the sensors feeds a computer which provides a output indicating the presence or size of a fault.

5. Method according to one of claims 1 to 4, wherein the digital signals stored in memory are analyzed to determine the nature of a fault.

6. The method of claim 5, wherein the fiber network is illuminated on the same side and also on the side opposite to the sensor, and the signals produced by the sensor are analyzed with regard to the transparency and reflectance of the fiber network.

7. The method of claim 6, wherein the relative intensity of the illumination from the same side and the opposite side is adjusted to a value such that the intensity detected by the sensor in the presence of the fiber network and in absence of defect is equivalent to light intensity in the absence of the fiber network.

8. The method of claims 6 and 7, wherein the illumination from one side of the fiber network is provided by a reflected light passing through the fiber network and from a source placed on the other side .

9. Apparatus for implementing the method according to claim 1 comprising a device for transporting the fibers, a light source directed towards the network of fibers and thus in relation to one or the other sensor so that the light transmitted or reflected by the network of fibers falls respectively on the first or the second sensor, and the means for analyzing signals produced by the sensors in response to the light falling on them, characterized in that the means (32-37 ) to analyze the signals include means for storing, when the signal emitted by at least one of the sensors exceeds the smallest of the thresholds (THI, TH2, ...), digital signals describing a fault and composed, for each sensor, of a series of words characterizing the largest threshold exceeded as well as the lateral and longitudinal positions of the exceeding, as well as means (36) for identifying the nature of the fault from the signals stored in memory.

10. Apparatus according to claim 9. wherein two sensors (11 and 21) are placed along the fiber path (13), a light source (14) is placed on the side opposite this path relative to one of these sensors (II), and second (24) and third (25) light sources are placed respectively on the same side and on the side opposite to the fiber network with respect to the other sensor (21).

 He. The apparatus of claim 10, wherein the third light source (25) is a reflector positioned to reflect light from the second source (24) toward the second sensor (21).

12. Apparatus according to claims 9, 10 or 11, wherein each sensor comprises a network of photoelectric cells oriented transversely to the path of the fiber network, and the analysis means is adapted so that each sensor scans the network of fibers during its passage under the sensor.

13. Apparatus according to one of claims 9 to 12, comprising a computer adapted to produce, from the sensor output signals, indications of the presence, size or nature of any defect.

14. Apparatus capable of examining the material in the form of a wick or a network of fibers according to one of claims 9 to 13, comprising drawing means which precede the transport device.

The present invention relates to the detection and classification of defects appearing in fiber assemblies, and more particularly to the methods and apparatuses intended for the detection, classification and measurement of the dimensions of defects in rabans and other textile fabrics or webs. .

In the ribbons or veils of fibers, in particular in the case of natural textile fibers, such as wool, various types of defects are found in variable quantities such as bundles of twisted fibers (buttons), vegetable materials (thistles, straws) , etc. An excessive quantity of these defects causes difficulties during manufacture (increase in the frequency of breaks in the spinning machine for example) and decreases the quality of the yarn (increase in the number of thickness defects for example), as well as the knitting or fabric (too many visible defects lead to rejecting a knitting panel or a certain length of a piece of fabric). In combing and spinning, especially those working with long fibers, we therefore carry out a routine check on all batches of textile materials, in order to determine the frequency and nature of the defects, as well as their dimensions (these three factors each having a significant influence on the manufacture and quality of textile products). At present, this control is entrusted to an observer who performs a visual inspection on samples of ribbon or veil of fibers. This subjective control is unfortunately not very reliable and very little reproducible from one firm to another (we observed, for defect counts of a certain type, ratios of 1 to 10 between the extreme observers). These results are therefore very difficult to use in commercial relations.

In order to remedy these difficulties, it is necessary to resort to an automatic objective measurement, independent of the operator.

The present invention describes a method and an apparatus, characterized respectively by claims 1 and 9, for performing such a measurement.

The analysis of the fiber network by the system described must be carried out on a veil of fibers of relatively low density (of the order of 10 g / m2) and sufficiently homogeneous. When the material to be examined is in the form of a ribbon or a wick, it is therefore necessary to thin it and spread it, until forming a homogeneous veil having the required density, which requires a process and a stretching device meeting these requirements.

In connection with this invention. light radiation is directed at the fiber sample as it passes under the

minus two sensors, transmitted and reflected light is detected and, from the resulting signals, the presence, size or type of any defect is determined.

Preferably, the invention uses two sensors, one intended to receive mainly the light by transmission and the other intended to receive mainly the reflected light.

The means used to analyze the signal from the sensors, such as a computer, can be adapted to synchronize the signal produced by one sensor with that produced by the other sensor analyzing the same portion of the sample of fibers.

The reflected light detection preferably uses two light sources, the first placed on the same side as the sensor and the other, which can be a diffusing reflector, placed on the opposite side.

The sensor concerned detects, in this case, the light coming from the first source and reflected directly by the set of fibers and light from the second source after transmission to the tra verses of this set. This configuration makes it possible to differentiate defects of origin or of a different nature.

The purpose of this diffusing reflector is actually to create an illu mined in such a way that the fiber sheet seems to disappear when of observation by reflection, seen from the same side as the sensor reflection.

Another possibility is to use a frosted glass as background or other translucent element with an auxiliary light source, which can be adjusted in intensity to obtain the disappearance of the fiber sheet vis-à-vis the light background.

Preferably, each sensor is a network of photo cells electric placed transversely to the fiber sheet to analyze. If the signal analysis process is suitable for pro a scanning of the material by sensors, the size of a die can be determined easily.

The invention will be described below, only by way of example, with reference to the drawings below: fig. 1 is a representation of the arrangement of the stretching and of the video unit of a possible configuration of the invention, fig. 2 is a detailed representation of the transparency detection system of FIG. 1, fig. 3 is a detailed representation of the detection system by reflection of fig. 1, fig. 4 is a block diagram of an interface system connecting the video unit of FIG. 1 to a computer used for analysis, fig. 5 is another block diagram illustrating certain functions of the apparatus of fig. 1 to 4, fig. 6 is a drawing showing the levels of detection on the video signal, fig. 7 is a flowchart illustrating the criteria used for the classification of faults.

The preferred version of this invention is based on an optical analysis of the image of the ribbon or other set of fibers, projected onto a set of two linear arrays of photodiodes.

This type of sensor integrates on the same support for example 1024 photodiodes, which are spaced from each other by 25 .mu.m and form a strip with a width of 15 to 30 ttm. A linear array of CCD cells, or a CCD (Charge Coupled Device) array, can also be used as a sensor.

The ribbon is preferably illuminated in two different ways.

In the first illumination system, the source is placed so that the light passes through the ribbon.

This is illustrated in fig. 2, in which the transparency sensor 11, with its associated optical system 12, is placed above the ribbon or other set of fibers 13, and a tubular light source 14 is placed below the ribbon.

The two optical systems 12 and 22 form on the sensors an image of a transverse line of the sheet of fibers.

In the second case, the light source is placed so that the light is reflected by the ribbon.

This is illustrated in fig. 3, in which the reflection sensor 21 and its associated optical system 22 are placed above the ribbon 13, while two tubular sources 24 are placed symmetrically above the ribbon and act in cooperation with a reflecting background or a diffusing screen 25, attached below the ribbon.

In the stretching and video system shown in fig. 1, the ribbon 13 is delivered from a stretching system comprising a double sleeve 26 and stretching cylinders 27, and is driven under the sensors 11 and 21 by another pair of cylinders 28.

The transparency sensor 11 is placed near the reflection sensor 21 at a fixed distance, and the two sensors periodically scan the ribbon 13 perpendicular to the direction of movement and parallel to its surface at the exit from the stretching zone 7, 28 The two sensors and their respective light sources are separated by a black panel 29 provided with a slot allowing the passage of the ribbon.

By means of appropriate circuits, a train of 1024 pulses is obtained for each sensor, the amplitude of each pulse being proportional to the quantity of light received by each photodiode. The pulse train forms a video signal.

These video signals are digitized into a series of 16-bit words and stored in a FIFO (first in, first out) memory before transmission to the computer.

The electronic circuits required to convert video signals to 16-bit words are contained in an interface system shown in detail in FIG. 4 and described below.

The interface is under the control of a microprocessor which empties the FIFO memory and transmits the information to the computer via serial transmission lines.

The computer calculates the parameters of the detected faults, classifies the faults and prints a summary report at the end of the analysis.

The elements of a preferred embodiment of the apparatus of the invention and their operating mode will now be described in more detail.

The stretching system is shown schematically in FIG. 1 and is required to prepare the fibrous material before analysis.

The material should preferably be stretched to approximately 1 g / m (1 ktex) over a maximum width of 128 mm.

To stretch this material, a sleeve 26 is used, in order to reduce the number of adjustments necessary to allow the analysis of different types of material.

In the area under the sensors, the material must remain horizontal and stretched and must advance at constant speed. Information concerning the speed and displacement of the ribbon is transmitted to the interface system.

The video system shown in fig. 1 includes the transport of the ribbon, the light sources 14, 24 for its illumination, the optical systems 11, 22 and the accessories for projection on the linear arrays of photodiodes, as well as the linear arrays of photodiodes 11 and 21.

The associated control circuits are contained in a front unit 31.

The ribbon passes at constant speed horizontally below the sensors and passes through an area which is uniformly illuminated by incandescent lamps of tubular shape.

Using the optical system, an image is projected onto the respective sensors.

The light image generates on each photodiode a voltage proportional to the quantity of light received; this voltage is extracted for successive photodiodes, in the form of a pulse train, by an appropriate electronic circuit.

The intensity of the light sources can be controlled by the electronic interface and the sources are supplied by a regulated DC voltage.

The ribbon movement and the scanning pulses required by the sensors are synchronized.

The main purpose of the transparency sensor 1 1 is to determine the real size of each defect.

The fiber ribbon is stretched in such a way that it is almost invisible to the transparency sensor and the video signal is therefore a constant corresponding to the amount of light projected by the tubular light source on the sensor.

The purpose of the reflection sensor 21 is to distinguish faults on the basis of the brightness characteristics.

The analysis must be carried out near the transparency sensor in order to avoid displacement of the defects in the ribbon, between the two sensors.

The reflection sensor receives light from two sources: the light reflected by the stretched ribbon and the light reflected by the reflector diffusing from the background (or emitted by the translucent glass covering an auxiliary light source, serving as background in another possible embodiment).

The diffusing reflector is made of a material whose color and reflectance are close to those of the fibers, so that the stretched ribbon merges with its background.

In the absence of a fault, the light intensity of the sources is adjusted in such a way that the signal from the photodiodes is located halfway between the black level and the saturation level.

The signals coming from each of the 1024 photodiodes are almost identical, with the exception of variations due to small differences in sensitivity between photodiodes and a systematic variation between the intensity of the light received at the center or at the ends, from the difference in distance from the source.

This systematic difference, compensated in the electronic interface, an almost constant video signal appearing as a horizontal line at 50% between the black level (0%) and the light saturation (100%), can be viewed on the screen d 'an oscilloscope.

If a fibrous defect (for example a button, consisting of a tangle of fibers with a dense core) appears, an increase in the amount of reflected light relative to the background is detected at this point, and this results in an increase in the video signal from 50 to 75% for example.

If a plant defect (e.g. straw, thistle or piece of thistle, etc.) appears in the stretched ribbon, a decrease in the amount of reflected light is observed at this point and the video signal has a decrease of 50 to 25% for example.

It is therefore possible by this method to separate the fibrous defects from the plant defects.

This results from the fact that a fibrous defect is both very bright and semi-opaque, which allows part of the light reflected by the background to be transmitted by the fibrous defect, adding to the light reflected by the defect. On the contrary, a yellow or brown vetal defect is both less luminous than the veil of fibers which surrounds it, therefore reflects less light, and is moreover perceived completely opaque, and consequently transmits only a very small part of the light reflected by the background scattering detector, which in total corresponds to much less light on the reflection sensor.

The video signal from the transparency sensor is mainly used for determining the shape and size of the defect, but also for partial identification of faults (fibrous or vegetable), based on the reduction in the level of transmitted light (corresponding to the 'opacity) and on the slope of this reduction on the edges of the defect.

To this end, five detection levels TH 1 to TH5 are fixed to evaluate the opacity of each defect (fig. 6).

For a fibrous defect, a button for example, the opacity will be average (for example reaching level 3 in the core) and the slope between the passage to level I and the passage to level 3 relatively low.

For a plant defect, the opacity will be greater (exceeding level 4 or 5) and the steeper slope, because the edges of the defect are sharper for the transmitted light.

The additional criterion, namely the increase or decrease of the video signal from the reflection sensor, is used to confirm identification in borderline cases.

A software program applying a weighting to each criterion provides a final classification.

The determination of the dimensions of the defects is best carried out on the basis of the signal from the transparency sensor because the contrast and the signal-to-noise or signal / background ratio is much better than that of the reflection sensor.

On the latter, a small button and the fibrous defects of low density disappear completely, and the transition on the edges of the other defects are blurred and ill-defined.

It is possible in certain cases to obtain an even greater difference between the quantity of light reflected by the fibrous defects and that reflected by the plant defects, by inserting on the path of the light rays reaching the reflection sensor, an ordinary filter or an interference filter in the blue region of the spectrum.

As described above, the transparency sensor receives an image of greater clarity and therefore a more precise indication of the presence, size and shape of the defect.

The reflection sensor allows the identification of plant defects as such, because they have a lower reflectance.

Fibrous defects can be recognized by their greater reflectance; however, this does not exhibit as significant a change in signal as in the case of plant defects.

Furthermore, the defects identified and not noticed by the reflection sensor as being plant will be assumed to be fibrous. Plant defects not identified by the reflection sensor for any reason can nevertheless be identified either by their shape or by the slope of the transparent video signal.

Further identification, such as the type of defect (straw, thistle, etc.), can be carried out on the basis of the signal from the transparency sensor, the form of the defect indicating a particular type of plant or fibrous defect.

An example of software analysis of video signals allowing the classification of the various types of faults is given in fig. 7.

The R / T ratio between the reflection signal R (taking the gray level at 50% as a reference) and the transparency signal T (taking the saturation level 100% as a reference) allows the separation between the fibrous defects and plants.

Indeed, for a plant defect, the signal R affected by this defect will be large and negative (close to the black level) and the signal T large and negative (close to the black level); in this case the ratio will be positive and close to 1.

For a fibrous defect, R will be positive and T will be negative and important; the ratio will therefore be negative and small, close to 0 in most cases. By setting a positive level of 0.5 for the separation of the R / T ratios, we will obtain a classification where a value greater than +0.5 indicates a plant defect.

Other classes of defects are identified on the basis of their shape, the gray level and their projected length Lp.

The interface system, as shown in fig. 4, can be divided as follows: signal processing 32 and FIFO memory 33, associated with a particular sensor; a microprocessor 34, a volatile memory (RAM) 35 and a specialized program; transmission channels to computer 36 via serial lines 37. and counters and control logic.

The video signal from the linear array of photodiodes is amplified in a differential amplifier 41 and is used to generate a reference video signal 42, which reflects the average value of each photodiode when the ribbon is faultless.

The subtraction in the amplifier 43 of the primitive video signal Vl (T), present at the output of 41, and of the reference video signal 42, gives the output of 43 (therefore also at the input of the comparator and of the logic threshold 44) an analog video signal V2 (T) which represents only the detected fault, and resembles a pulse (see fig. 6). The scanning being linear, the position X in the transverse line (that is to say the number of each photodiode, from 0 to 1023 for example) is proportional to the time T during a scanning.

Fig. 6 represents the analog video signal V2 (T), corresponding to V2 (X) for a particular line, with on the abscissa the sequence number X of each photodiode. This diagram corresponds to a particular longitudinal position Y, and one can for example imagine that the Y axis is perpendicular to the drawing. The signal V2 (T) of fig. 6 corresponds to a fault; the thresholds THi to TH5 are shown.

The signal V2 is then compared to the threshold levels and its amplitude is digitized.

The reference signal can be used to perform automatic control of the brightness of the sources. It is also possible to provide an automatic adjustment of the detection thresholds in response to changes in circumstances.

To reduce the amount of information transmitted to the computer, the fault under analysis is reduced in the form of a series of 16-bit words, each word providing the following information: 10 bits for the photodiode number: X = 0 to 1023 (X = transverse coordinate of a point of the defect in the water table fiber), 3 bits for the number of the detection threshold which has just been passed (THI to TH5), 1 bit to identify information of type X.

During scanning, when a fault is detected, the integrator digital 45 calculates the volume of opacity E that the fault presents and this information is transmitted with a special flag such as a word of type X, just after the end of the detected fault.

At the end of the line scan, if a fault has been detected and information X and z have been stored, information Y is added.

This information Y (corresponding to the longitudi coordinate nal of a point of the defect in the sheet of fibers) is also a word 16 bits which contains the following data: 12 bits for the number of the line scanned (= longitudi position nale), 1 bit to identify type Y information, 3 bits for indicators such as END of analysis.

Type Y information must always follow a series of information type X messages concerning the line analyzed.

Information Y can be alone in order to transmit to the computer the information contained in the flags.

The line number is transmitted only when a fault has been met in line.

The END of analysis indicator is transmitted after receipt of a microprocessor command, which indicates to the computer that this is the latest information.

A computer is used to calculate a series of parameters for each fault from the data from each sensor.

To increase the processing speed, the computer never stores in memory a complete image of the defect, but only a certain number of vectors which it completes using the information received.

The end of a fault is recognized by the computer by the presence of the GAP indicator in type Y information.

At this time, the computer will superimpose the vectors of each sensor to obtain the set of information from all the sensors about the same fault and it will calculate and compare a series of parameters to deduce the characteristics of the fault, which will then be classified according to well-defined rules (for example according to the software program, the organizational chart of which is given in fig. 7). When the END indicator is encountered, the computer will print a bulletin summarizing this classification.

The classification can lead to the calculation of parameters such as: size, shape, opacity (gray level), projected length, discontinuity and ratio between sensors (for example R / T) for these parameters.

The sensors can scan the ribbon at a speed and with a sensitivity such that a photodiode will see a square surface whose sides are 14 mm on a 128 mm wide ribbon.

The speed of the tape under the sensors can be greater than 1.5 m / min, and even 3 m / min with more recent sensors.

When a FIFO memory overflow condition arrives, the information already analyzed is not lost; a bulletin is given automatically.

The tape can pass under the sensors without being analyzed; this is used for proper preparation of the material before the start of the analysis.

The summary bulletin can be sent to another computer for data archiving. The bulletin can give information on faults per meter or per unit of weight of the tape analyzed.

In this specification and in the claims, reference is made to light; it is obvious, however, that a form of electromagnetic radiation other than that located in the visible spectrum can be used in accordance with this invention in association with the appropriate sensors, and the term light should be understood in this way.

Claims (14)

1. Method for detecting and classifying faults appearing in a network of fibers, which consists in passing the network of fibers under at least two photosensitive sensors, directing light towards the network of fibers, one of the sensors being placed to respond to the light transmitted through the fiber network and the other being placed to respond to the light reflected by the fiber network, the signals produced by the first sensor being synchronized with those of the second sensor corresponding to the same portion of the fiber network and detecting the presence and nature of a fault from the signals produced by the sensors during the passage of a fault, characterized in that, when the analog signal emitted by at least one sensor exceeds the smallest of a series of thresholds (THI, TH2, TH3, ...), a digital signal composed, for each sensor, of a series of digital words identifying the largest threshold exceeded by the analog signal as well as the lateral (X) and longitudinal (Y) positions of the location of the network where the overshooting took place is stored, and the values stored in memory from the two sensors and corresponding to the same portion of the fiber network are then analyzed. 2. Method according to claim 1, in which the digital signals stored in memory are analyzed to detect the presence or the size of a fault. 3. Method according to claim 2, in which, for each sensor, a reference signal is automatically created and represents the average value of said signals in the absence of a fault, and this reference signal is subtracted from a video signal produced by this sensor to reveal a fault as an impulse after subtraction. 4. Method according to claims 2 and 3, wherein the sensor comprises a network of photoelectric cells placed transversely and which are used to scan the network of fibers during its passage, and in which the output of the sensors feeds a computer which provides a output indicating the presence or size of a fault. 5. Method according to one of claims 1 to 4, wherein the digital signals stored in memory are analyzed to determine the nature of a fault. 6. The method of claim 5, wherein the fiber network is illuminated on the same side and also on the side opposite to the sensor, and the signals produced by the sensor are analyzed with regard to the transparency and reflectance of the fiber network. 7. The method of claim 6, wherein the relative intensity of the illumination from the same side and the opposite side is adjusted to a value such that the intensity detected by the sensor in the presence of the fiber network and in absence of defect is equivalent to light intensity in the absence of the fiber network. 8. The method of claims 6 and 7, wherein the illumination from one side of the fiber network is provided by a reflected light passing through the fiber network and from a source placed on the other side . 9. Apparatus for implementing the method according to claim 1 comprising a device for transporting the fibers, a light source directed towards the network of fibers and thus in relation to one or the other sensor so that the light transmitted or reflected by the network of fibers falls respectively on the first or the second sensor, and the means for analyzing signals produced by the sensors in response to the light falling on them, characterized in that the means (32-37 ) to analyze the signals include means for storing, when the signal emitted by at least one of the sensors exceeds the smallest of the thresholds (THI, TH2, ...), digital signals describing a fault and composed, for each sensor, of a series of words characterizing the largest threshold exceeded as well as the lateral and longitudinal positions of the exceeding, as well as means (36) for identifying the nature of the fault from the signals stored in memory. 10. Apparatus according to claim 9. wherein two sensors ( 11 and 21) are placed along the fiber path (13), a light source (14) is placed on the opposite side to this path relative to one of these sensors (II), and second (24) and third (25) light sources are placed respectively on the same side and on the side opposite the fiber network with respect to the other sensor (21).  He. The apparatus of claim 10, wherein the third light source (25) is a reflector positioned to reflect light from the second source (24) toward the second sensor (21). 12. Apparatus according to claims 9, 10 or 11, wherein each sensor comprises a network of photoelectric cells oriented transversely to the path of the fiber network, and the analysis means is adapted so that each sensor scans the network of fibers during its passage under the sensor. 13. Apparatus according to one of claims 9 to 12, comprising a computer adapted to produce, from the sensor output signals, indications of the presence, size or nature of any defect. 14. Apparatus capable of examining the material in the form of a wick or a network of fibers according to one of claims 9 to 13, comprising drawing means which precede the transport device. The present invention relates to the detection and classification of defects appearing in fiber assemblies, and more particularly to the methods and apparatuses intended for the detection, classification and measurement of the dimensions of defects in rabans and other textile fabrics or webs. . In the ribbons or veils of fibers, in particular in the case of natural textile fibers, such as wool, various types of defects are found in variable quantities such as bundles of twisted fibers (buttons), vegetable materials (thistles, straws) , etc. An excessive quantity of these defects causes difficulties during manufacture (increase in the frequency of breaks in the spinning machine for example) and decreases the quality of the yarn (increase in the number of thickness defects for example), as well as the knitting or fabric (too many visible defects lead to rejecting a knitting panel or a certain length of a piece of fabric). In combing and spinning, especially those working with long fibers, we therefore carry out a routine check on all batches of textile materials, in order to determine the frequency and nature of the defects, as well as their dimensions (these three factors each having a significant influence on the manufacture and quality of textile products). At present, this control is entrusted to an observer who performs a visual inspection on samples of ribbon or veil of fibers. This subjective control is unfortunately not very reliable and very little reproducible from one firm to another (we observed, for defect counts of a certain type, ratios of 1 to 10 between the extreme observers). These results are therefore very difficult to use in commercial relations. In order to remedy these difficulties, it is necessary to resort to an automatic objective measurement, independent of the operator. The present invention describes a method and an apparatus, characterized respectively by claims 1 and 9, for performing such a measurement. The analysis of the fiber network by the system described must be carried out on a veil of fibers of relatively low density (of the order of 10 g / m2) and sufficiently homogeneous. When the material to be examined is in the form of a ribbon or a wick, it is therefore necessary to thin it and spread it, until forming a homogeneous veil having the required density, which requires a process and a stretching device meeting these requirements. In connection with this invention. light radiation is directed at the fiber sample as it passes under the ** ATTENTION ** end of the CLMS field may contain start of DESC **.