WO2001086228A1

WO2001086228A1 - A method to measure surface structure

Info

Publication number: WO2001086228A1
Application number: PCT/SE2001/000974
Authority: WO
Inventors: Roger Tuomas; Matti Rantatalo
Original assignee: Roger Tuomas; Matti Rantatalo
Priority date: 2000-05-05
Filing date: 2001-05-04
Publication date: 2001-11-15
Also published as: AU5896201A; SE0003904L; SE0003904D0

Abstract

For measuring a surface structure, one takes two or more pictures with a progressive-scan CCD camera (15) and drag light from different directions for different pictures. A two-dimensional Fourier transform is then carried out for each picture and the two spectra are added to form a spectrum that is an approximation of the correct spectrum for the surface that is photographed. This spectrum illustrates the surface structure. The method can be carried out on-line in a paper machine and the result will be as good as the result of conventional laboratory tests.

Description

A method to measure surface structure

Technical field

This invention relates to a method of measuring a surface structure of a material by lighting up the surface with drag light and photographing it electronically with a CCD- camera and then making a 2-dimensional Fourier transformation in order to get a spectrum of the intensity in the obtained picture.

Background of the invention and short description of prior art

The knowledge in the paper making industry of the parameters that influence the quality of the paper is crucial for the possibility of producing a paper with the right quality. A poor knowledge of these parameters may in worst case lead to the production of a paper with a non-wanted property. It is therefore of great importance to know as much as possible about these influencing variables. The surface structure of the paper is one of these variables and it is particularly interesting to know when it comes to the ability of the paper to get a good printing. Heavy periodic disturbances in the surface of the paper may result in defects in the print and it may derive for example from worn press felts.

It is necessary to predict when press felts and wire cloths have served its time so as to avoid producing defect paper and to avoid unnecessary stoppage of production. It would be desirable to know which one of the press felts and wire cloths that causes a non-wanted periodic disturbance in the surface structure so that the defective felt/wire cloth can be exchanged. This would make it possible to optimise the utilisation of felts/wire cloths.

Today, the surface of the paper and the influence of the periodic cloths (felts/wire cloths) are tested and measured with laser equipment in the laboratory. Paper specimens are cut out and tested off-line. This testing is a refined and accurate measuring method, but very time consuming. It can also be difficult to understand the result from these tests since the specimens are small and few as compared with the amount of produced paper. The most common testing method for surface roughness is however leakage tests of the PPS and Bendtsen type. Such testing gives surface roughness parameters based on the volume of air that passes between a mouthpiece and the surface of the paper. A large amount of air passes when the surface is rough and less amount passes when the surface is fine. In production, this test is carried out each time a new paper roll is mounted, which means that only one test is carried out for each 2.5 km of produced paper. Such occasional tests are not necessarily representative for such a large amount of produced paper and the likelihood for discovering periodic defects is low.

At present, there is a great demand for surface roughness measurement equipment with a higher frequency of the testing in order to get test results that have an improved statistical validity. A large amount of test data gives more experience and more knowledge of the nature of the paper surface and will facilitate the analysis of the paper quality. Today, there is no method available with test data comparable with the quality of the test data from the testing in the laboratory.

From EP-0586 795-A, it is known to light up a track of material, for example a sand paper track, with a drag light and to take pictures with a CCD-camera and then to carry out a 2-dimensional Fourier transformation in order to get a frequency spectrum. The method may be sufficiently accurate for certain purposes, but the accurateness is not of the kind necessary for example in papermaking. If it can be supposed that the surface to be analysed has the same structure in all directions, for example a sandpaper surface, then this method may be sufficient. This is, however, not the case in paper making when it very seldom is predictable in which direction for example shrinks and periodic disturbances will occur. The method in the patent lacks the possibility to fully analyse a 3-dimensional surface. From SE-508822, it is known to light up a surface with more than one light source. The aim of the method is to create a 3-dimensional representation of the correct surface by means of two pictures with 180 degrees between the light sources. The two pictures must be pictures of exactly the same surface otherwise the method will not work. Two different surfaces will give an incorrect representation. With this method with two light sources located in directions opposite each other, no information is achieved about structures that are oriented in line with these light sources. It is not an aim of the method to create an approximation of the correct frequency spectra and it cannot show the distribution of the surface structures in all directions of the 3-dimensional surface.

From SE 511985, it is known to take photos of a surface with light from different directions and to use the difference between the pictures in order to discover structures in the surface that appear in different ways in different light. The method demands that exactly the same surface illuminates in order to make the creation of the new differential picture possible and it will not work if the pictures are taken on different areas or are overlapping.

Object of invention and brief description of the invention

It is an object of the invention to test the surface structure of a surface quickly and with great accurateness, and this object is fulfilled principally in that one takes at least two pictures in a row with drag light from various directions, , and makes a Fourier transformation of each picture, and then takes a part of the spectrum or the entire spectrum of each picture and adds them to a spectrum in order to create an approximation of the accurate spectrum of the surface. The invention has been given the characteristics defined in the claims.

Particularly in on-line measurements of a passing material track, for example in a paper making machine, this method has the advantage of providing a good approximation of the spectrum of a surface element also if the two pictures, the spectra of which are added, are not taken of the same surface element but taken of adjacent surface elements or overlapping surface elements.

Brief description of the figures

• Figure 1 is a schematic representation of a part of a device for taking pictures of a surface; a device that can be used for carrying out the method according to the invention,

• Figure 2 is a flow chart of an image processing of pictures taken with the device of figure 1, this image processing being in accordance with the inventive method.

• Figure 3 is a diagram with surface roughness on the y-axis and a number of paper samples on the x-axis; this diagram showing one testing in accordance with the invention and one PPS testing.

• Figure 4 shows in the same manner as figure 3 a comparison between one testing in accordance with the invention and one Bendtsen testing.

• Figure 5 shows in the same manner as figures 3 and 4 a comparison between two tests, but both of them being in accordance with the invention.

• Figure 6,7, and 8 show three different set-ups of lights sources as seen from above.

• Figure 9 shows the camera view indicated in figure 1.

Description of a preferred embodiment

Figure 1 shows a surface 11 , the structure of which is to be defined. The surface 11 can be the surface of a passing paper track in a paper machine, the track moving fast in the direction indicated by the arrow. Such a paper track moves usually at a velocity exceeding 10 m/s. A source of light 12 lights up the surface 11 through a lens system 13 that provides a parallel beam that is reflected in a mirror 14 so that it falls obliquely on the surface 11 as a drag light. A CCD-camera 15 or a corresponding device with an objective 16 is directed at a right angle to the surface 11 for taking pictures of the enlightened part of the surface. The parallel beam 17 falls on the surface along with the material track. A similar device with light source, lens system, and mirror is arranged to give a similarly obliquely falling light transverse to the material track, that is, at right angle to the direction of movement of the material track. Since the two devices for directing light to the surface 11 are exactly alike, only one of them is shown. A part of the light from each light source is led in a fibre 18 to the camera view as shown in figure 9. This part 19 of the camera view is then used to measure the intensity of the light source. The remaining part of the camera view is used to calculate the structure of the surface, for example the roughness.

The light sources 12 can preferably be stroboscope lamps with xenon. The CCD- camera 15 is preferably a progressive scan camera, that is, a CCD-camera in which all the pixel sensors reads individually the light intensity at the same instant. These sensed intensities provide together an instantaneous image of the surface. The camera 15 and the two stroboscope lamps 12 are coupled to a computer that has a video card, a frame grabber, that makes it possible to couple the camera to the computer. The card trigs the lamps one at the time and trigs simultaneously the camera so that the camera takes pictures with one lamp flashing at the time.

Figure 2 illustrates the analysis of two pictures taken in a row. The surface elements that are sensed have been denoted 11a and 11b. The light beams that fall onto the paper track transverse to the track have been denoted 30 and the light beams that falls onto the paper track along the track have been denoted 31. They enlighten the respective surface elements 11a and 11b and cause one image each. The images are processed the same way and cause two frequency spectra denoted 33,34. Drag light will provide an approximation of the surface structure and such an approximation lacks information about the structure properties along the beam. Important information will thus be missing if only one direction is used to analyse a 3- dimensional surface. The combination of information from pictures taken with drag light from various directions will provide more information about the surface than one picture. The image processing of the information from the images in the frequency plane is special for the inventive method. When the information from different images are added in the frequency plane and not in the original distance plane in which the light intensity in the image depend on the distance co-ordinates x and y (called time plane when the signal depends on the time), it is not necessary that the same surface element is photographed. The advantage of not being forced to photograph the same surface element makes the sensor design easier and the components can be cheaper and simpler.

Every other picture is taken with light from one source and every other with light from another source. The frame grabber trigs simultaneously one of the light sources and all the pixel sensors of the camera and then simultaneously the other light source and all the pixel sensors of the camera so that two pictures are provided one taken with light at right angle to the light of the other. The pictures can for example be taken of rectangles of 2-4 cm and it can be 40 ms between the pictures. The paper track can then have moved 40 cm in the direction of the track. In order to cover the whole track, the camera can be traversing so that it moves a few mm between the shots. The shots will then be taken in a zigzag path along the paper track. In a paper machine in which the paper track moves in a speed of 10-15 m/s, it can be suitable to take shots with less than 100 ms between the shots.

In the flow chart, figure 2, the squares 40 represent a compensation for uneven light. Such compensation should not be necessary when the light is parallel. Alternatively, one could also high pass filter the image in order to get rid of the uneven light. Then, the big variations in the image disappear. A high pass filtering and a low pass filtering leave a frequency band but are time consuming. The function in this square is not necessary when the light is good and completely parallel. A compensation for variations in the light between the pictures is however usually necessary since the flashes from different lamps and also the flashes from one and the same lamp can vary in intensity. Such compensation is carried out by the aid of the picture in surface 19 in figure 9. Thus, a reference of the light source that lightens the surface with a drag light is created in the camera view and the energy in the reference is sensed. The energy in the image is compensated for the difference between the actual energy in the reference and a comparison value. Then, the compensated energies in the two drag light images can be added. Every source of light should have a separate part in the image for the reference sensing. The references provide also information about whether or not the sources of light are in function.

The function in the squares 41 is a subtraction of the mean value of all the light values (power) from the light value of each pixel. In this way, the power (the light value) , which depends on the general light level of the picture, is removed.

The function in the squares 42 is a 2-dimensional Fourier transformation. The power distribution in the image is provided by this 2 D Fourier transform. The 2 D Fourier transformed image comprises a matrix with the same dimensions as the original picture. Each element in the matrix comprises a complex number consisting of a real part and an imaginary part. Since it is the intensity (the power) of each element that is of interest, the absolute values of the intensities are used. Each element (pixel) in the 2-dimensional matrix represents a structure in the paper surface. The power of this structure (that is, how rough the structure is) is given by the absolute value. The phase can also be determined, but it is not interesting if one only wants to know how rough the structure is.

The two frequency spectra 33,34 provided from the original pictures are added to a spectrum 35 that is an approximation of the correct spectrum for the surface of the paper. It has been experienced that it is of little importance if the two pictures are not from exactly the same part of the paper but are from parts a few decimetres from each other. Since each picture has drag light from only one direction, there will be empty "sectors" in each image as will be described in spectra 33 and 34. Alternatively to adding the whole spectra, one may cut pieces from them and add the pieces so that they do not overlap. Cutting pieces in this way can be particularly advantageous when one uses more than two pictures and adds their spectra. One may for example take three pictures with 120 degrees angle between the light beams. By adding the frequency spectra 33,34, one will get a description of the anisotropy/isotropy in the image, which is not possible with prior art methods. In order to receive as correct measure of the structure; e.g. the surface roughness, as possible, the component of colour variation should be as small as possible. If a paper with uneven colour on its surface is to be analysed, the variation in colour will affect the surface measures 33,34. If two surfaces with the same roughness are compared, and one of them has an even colour and the other has a strong surface colour variation, the surface with the variation in colour will generate larger values for the measures 33,34 than for the one with even colour. In order to control the variation in surface colour, in connection with taking a drag light picture, one can also take a picture with common light, that is, a light that enlightens all parts of the surface to the same extent. This light should not contain any drag light that causes shadows in the picture. By sensing the energy in the intensity variation in this picture, for the frequency band in question, one gets a measure of the colour variation. This measure can then be used to weight the measures 33,34 so as to compensate for the influence of the colour variation on the measures of surface roughness. Thus, one senses the energy in the common light pictures for a frequency band and compensates the drag light pictures for variations in this energy between various drag light pictures.

Figures 6,7, and 8 show three different set-ups of light sources over a surface 70 as seen from above, that is, as seen at right angle to the surface so that the light directions are projected on the surface.

The set-up according to figure 6 comprises two light sources. The idea is to place the number of light sources evenly on a circle. In this set-up, the light source 71 lights up the surface 70 from an angle of 0 degrees. Since the additional information that can be received is practically negligible for the analysis of for example paper surfaces if the light source 72 is placed at 180 degrees, this light source should instead be placed at an angle of 90 degrees in order to provide maximum additional information. In other words, two light sources take up four positions on the circle and it is no advantage of having four light sources instead of two. In this context, one can consider two light beams directed at 180 degrees to each other to have the same direction.

The set-up according to figure 8 comprises three light sources evenly spread on the circle. According to the discussion above, the three light sources could as well be placed at 60 degrees angle according to figure 7 as at 120 degrees angle according to figure 8. Independently of the number of light sources, the angle between the light sources as projected on the plane of the surface should be between 30 degrees and 120 degrees. If there is many light sources, the angle between two adjacent light sources can be less than 30 degrees but the angle between two light sources that are not adjacent each other will still be in the interval 30-120 degrees.

In paper making, the computer can compare the added spectra which one receives at short intervals (about every second) and one can in this way detect defects and disturbances in the surface structure and also correlate them to their positions along the paper track which may be several km long. It is also possible to detect tendencies of deteriorating quality and correct the errors at an early stage and in this way avoid producing paper of wrong quality. With the considerable bank of knowledge one can build up as a result of these on-line measurements, the process can be trimmed and worn press felts and wire cloths can be exchanged before they cause non-wanted structures in the paper surface. This prevents unnecessary production stops and provides for a considerable increase of quality and a reduction of the costs for lack of quality.

In a test equipment with a speed of 20 m/s, 28 paper samples have been tested online with the described method and the result has been compared with PPS testing and Bendtsen testing. In the diagram shown as figure 3, the structure testing according to Bendtsen is shown as a full line and the on-line testing in accordance with the invention is shown as a dotted line. The x-axis shows 28 different paper samples. The diagram shows a very good correspondence between the two methods of measuring structure. The diagram shown as figure 4 shows in a similar way that the correspondence between Bendtsen tests and the on-line testing is equally good.

The figures 4 and 5 shows that the correspondence between the on-line testing and each one of the two laboratory test methods is about as good as between the two laboratory test methods.

The diagram shown as figure 5 shows a comparison between the inventive structure testing at the velocity of 20 m/s and testing with non-running samples. It shows a surprisingly good correspondence between the method of taking two pictures on the same surface element and the method of taking the two pictures of two adjacent surface elements.

Although the invention is exemplified as used in paper production, it can have many other uses both for testing moving material and non-moving material. Instead of one camera as shown, two cameras in line can be used and triggered in response to the velocity of the track of material so that the cameras take pictures of the same surface elements, but as shown by the tests, this is normally not necessary. A single camera or its lens system could alternatively be movable so that it takes two pictures of the same surface element in a row.

Claims

1. A method of measuring a surface structure of a material by lighting up the surface (11) with drag light and photographing it electronically with a CCD-camera or the like (15) and then making a Fourier transformation in order to get a spectrum of the intensity in the picture obtained, characterised in that one takes at least two pictures in a row with drag light from various directions, with an angle of 30-120 degrees between the light directions as projected on the plane of the surface, and makes a Fourier transformation on each picture, and then takes the entire spectrum or a part of the spectrum (33,34) of each picture and adds them to a spectrum in order to create an approximation of the accurate spectrum of the surface.

2. A method according to claim 1 , characterised in that, from each picture, one subtracts the mean value of the light of the picture before one makes the Fourier transformation.

3. A method according to claim 1 or 2, characterised in that one carries out the measurement on-line on a track of material (11) in fast movement and compares consecutively received added spectra in order to indicate possible defects in the passing material track.

4. A method according to claim 3, characterised in that one measures all the pixels at the same time.

5. A method according to claim 3 or 4, characterised in that one traverses the camera (15) over the material track.

6. A method according to any one of claims 3-5, characterised in that one takes pictures with less than 100 ms between the pictures, the spectra of which are added together.

7. A method according to any one of the preceding claims, characterised in that, in connection with taking a drag light picture, one takes also a picture without drag light and measures the energy in the later picture for the frequency interval and compensates the drag light pictures for variations in this energy between the various drag light pictures.