WO2012049370A1

WO2012049370A1 - System for monitoring a web and a corresponding method for monitoring the web

Info

Publication number: WO2012049370A1
Application number: PCT/FI2011/050889
Authority: WO
Inventors: Jukka Paananen; Hannes Kalaniemi
Original assignee: Metso Automation Oy
Current assignee: Valmet Automation Oy
Priority date: 2010-10-13
Filing date: 2011-10-13
Publication date: 2012-04-19
Anticipated expiration: 2013-04-13
Also published as: FI20135480A7; FI127055B; FI20106053A0

Abstract

The invention relates to a system for monitoring a web, for example a paper web, which system includes - at least one set of lighting elements, general cameras (10) for imaging the web (16) over essentially the entire width of the web (16), at least one processing unit (30) with image- processing software (37) for analysing the images taken using the general cameras (10), at least two precision cameras (10') for imaging the web (16) in a specific illumination geometry at a substantially greater precision relative to the surface area than the general cameras (10), to detect small deviations, - at least one processing unit (30) with image-processing software for analysing the images taken by the precision cameras (10'), - a camera beam (28), - a host unit (34), - a user interface (32), and - a data-transfer network (43). The system further includes software means (65) for statistically analysing the deviations of the images taken by the precision cameras (10'), to create a result covering the entire web. The invention also relates to a corresponding method.

Description

SYSTEM FOR MONITORING A WEB AND A CORRESPONDING METHOD FOR MONITORING THE WEB

The invention relates to a system for monitoring a web, for example a paper web, which system includes

at least one set of lighting elements for implementing an illumination geometry for illuminating the web for imaging, general cameras installed parallel to each other as a series in the width direction of the web, for imaging the web over essentially the entire width of the web, using which same general cameras the web is arranged to be imaged in at least one illumination geometry,

at least one processing unit with image-processing software, to analyse the images taken using the general cameras, using a selected algorithm to detect deviations,

at least two precision cameras for imaging the web in a specific illumination geometry at a substantially greater precision relative to the surface area than the general cameras, to detect small deviations,

- at least one processing unit with image-processing software, to analyse the images taken using the precision cameras, using a selected algorithm to detect deviations,

a camera beam for supporting the general cameras and precision cameras in connection with the web,

- a host unit for controlling the lighting and imaging, a user interface for illustrating the analysis results and for controlling the system, and

a data-transfer network for transferring data between the parts of the system.

The invention also relates to a corresponding method for monitoring the web.

In web formation, the monitoring of deviations appearing in the moving web is of very great importance in avoiding web breaks. Web breaks can be avoided by detecting deviations and the process parameters can be altered to avoid web breaks. Though the costs of individual web breaks may even be reasonably small, but at an annual level web breaks lead to a large cost item. In addition, web monitoring is extremely important in terms of quality control.

Generally, web deviations are monitored with the aid of cameras installed parallel to each other in the cross-direction of the web. The cameras image a specific area and deviations are detected from the area with the aid of an analysis apparatus. User demands have led to monitoring systems using an ever greater resolution, i.e. precision, in the cameras, relative to the surface area, in order to detect smaller deviations. However, this increases the investment costs of the system and also increases the data flow.

Publication US 2006/0239510 Al, which discloses a web- monitoring system, in which the web is images using two different kinds of camera, is known from the prior art. The first cameras image the web generally over a larger area while at least one second camera traverses the web to image it in greater detail. However, it is expensive to implement such a system, because separate devices are required in it to arrange the traversing of the second camera. The system according to the invention is intended to create a new type of web monitoring system, by means of which more and smaller deviations in the web can be detected with lower costs than previously, compared to the systems according to the prior art. The characteristic features of the of the system according to the present invention are stated in the accompanying Claim 1. The invention is also intended to create a new type of method, by means of which more and smaller deviations can be detected at lower costs than previously, compared to the methods of the prior art. The characteristic features of the method according to the invention are stated in the accompanying Claim 12. A camera imaging a web has a specific maximum resolution, which is the greatest number of pixels forming the image, by means of which the camera can image a specific area. To be able to detect small deviations more precisely from the web, the imaging area imaged by the camera must be reduced, so that the monitoring pixels of the imaging area will become smaller. In order to find small web deviations, such as dirt and needle holes, the imaging area must, however, be reduced in size to such an extent that the camera's imaging area will cover only a very small area of the web. As the web moves the whole time at a set speed, the camera's imaging frequency will not be sufficient to image the web continuously. Thus, the camera does not take images sufficiently frequently for images from each very small imaging area to accumulate continuously. The cameras generally operate at maximum frequently, so that the imaging frequency cannot be increased.

Achieving a sufficient precision demands an increase in the number of cameras, or cameras equipped with a higher resolution or imaging frequency. Both of these solutions create additional costs in equipment investment. However, it is widely known that dirt and needle holes are distributed evenly over the entire width of the web. The intention of the invention can be achieved by means of a system for monitoring a web, which includes at least one set of lighting elements for implementing an illumination geometry for illuminating the web for imaging, general cameras installed parallel to each other as a series in the width direction of the web, for imaging the web over essentially its entire width, by means of which same general cameras the web is arranged to be imaged in at least one illumination geometry, and at least one processing unit with image-processing software for analysing the images taken by the general cameras using a selected algorithm to find deviations. In addition, the system includes at least two precision cameras for imaging the web in a specific illumination geometry at a precision substantially greater than that of the general cameras relative to the surface area, to find small deviations, and at least one processing unit with image-processing software for analysing the images taken by the precision cameras, using a selected algorithm, to find deviations. Further, the system includes a camera beam for supporting the general and precision cameras in connection with the web, a user interface for illustrating the analysis results and controlling the system, a host unit for controlling the lighting and imaging, as well as a data- transfer network for transferring data between the parts of the system. An essential feature is that the system further includes software means for analysing statistically the deviations of the images taken by the precision cameras, to form a result covering the entire web.

By means of the system according to the invention, the general cameras can monitor the web over its entire width with sufficient precision to find normal deviations. In addition to this, the precision cameras are preferably used for the precise examination of small areas, so that the size of the pixels being examined will be small enough to find small deviations too. The areas between the areas imaged by the precision cameras are covered with the aid of statistical analysis, based on the knowledge that the small deviations are distributed evenly .

Preferably, the images taken by both the general cameras and the precision cameras are arranged to be analysed using essentially the same image-processing software. Thus, the system can be implemented using one set of image-processing software and the tuning of the image analysis will be simpler than when using two sets of software. The processing unit is preferably arranged to use essentially the same computation algorithm to calculate the deviations of the images taken using both the general cameras and the precision cameras. Thus, the system's precision cameras can be in the same system as the general cameras, i.e. the general cameras use the same data-transfer means, servers, displays, and processing unit.

The processing unit is preferably arranged statistically to analyse the deviations calculated using the computation algorithm from the images taken using the precision cameras and to interpolate the intermediate areas either linearly or non- linearly. Thus, the number of small deviations over the entire width and length of the web can be decided on the basis of the images taken from small areas by the precision cameras. With the aid of statistical analysis, sufficient information on the small deviations appearing in the web can be obtained with a small increase in the number of cameras.

The system can have 3 - 10-times more general cameras than precision cameras. The general cameras cover the entire web width, whereas the precision cameras cover only part of the web .

Preferably both the general cameras and the precision cameras are attached to the same camera beam. The system's space requirement is then minimized and the beam is as simple as possible. Thus, the system can also be implemented without traversing .

According to one embodiment, the processing unit is arranged to use the same image-analysis processor card for analysing the images taken by both several general cameras and at least one precision camera. Thus, the system is economical to implement, as the entire data flow runs through the same image-analysis processor card or cards. The precision cameras are preferably arranged to cover at least the edges and central part of the paper web. This will cover the most important and statistically significant areas of the web in terms of deviations.

According to one embodiment, the precision cameras have the same electrical properties as the general cameras, so that a unified set of accessories and processing units can be used with the cameras. The lighting elements can be strobe-light elements.

The images taken by the precision camera are preferably arranged to be analysed with the aid of the same processing unit and the analysis results are arranged to be displayed with the aid of the same user interface. Thus, the analyses of all the images can be performed with the aid of the same processing unit and all result can be displayed in the same user interface . The essentially greater precision of the precision cameras relative to the general cameras means that the ratio of the pixels of the images taken by a general camera to the images taken by a precision camera can be 3 - 12, preferably 4 - 6. This will then achieve sufficient precision for monitoring dirt, hole-defects, and fibre orientation, for example, but will nevertheless allow similar electrical properties to be used in both the general cameras and precision cameras.

The precision cameras are preferably arranged to image a smaller imaging area than that of the general cameras. Thus, the number of pixels of both the general cameras and the precision cameras can be the same.

Each processing unit preferably includes at least one image- analysis processor card, with the aid of which the system is arranged to analyse both the images taken by several general cameras and the images taken by at least one precision camera. All the images can be analysed using similar types of image- analysis processor cards, or, in some cases, even using the same image-analysis processor card.

The processing unit with image-processing software for analysing the images taken by the general cameras is preferably the same as the said processing unit with image-processing software for analysing the images taken by the precision cameras. The system can then be implemented in the simplest possible form.

According to one embodiment, each image-analysis processor card includes several input channels, at least one FPGA processor for all the input channels and at least one dedicated DSP processor for each input channel. The use of this type of image-analysis architecture will achieve extremely efficient and rapid image analysis. According to one embodiment, both the general cameras' processing unit and the precision cameras' processing unit are arranged in the same processing unit. This will increase the system's simplicity. In one embodiment, the system includes at least two lighting elements for implementing different illumination geometries at different times, in order to find many types of deviation and both the general cameras and the precision cameras are arranged to image the web in at least two different illumination geometries at different times. In practice, when using two illumination geometries, four separate images are taken of one point on the web, i.e., for example, base images of the web taken by the general cameras and the precision cameras and surface images of the web taken by the general cameras and precision cameras. In the method according to the invention for monitoring a web, for example a paper web, the web is illuminated by lighting elements and images of the web are taken, in at least one illumination geometry, by general cameras installed parallel to each other to form a series over the entire width of the web, in order to find deviations. In addition, images of the web are taken by at least two precision cameras in a specific illumination geometry, at a greater precision relative to the surface area, to find small deviations, and both the images taken with the general cameras and those taken with the precision cameras are analysed to find deviations. The essential aspect is that the deviations of the images taken by the precision cameras are additionally analysed statistically to create a result covering the entire web. Thus, the system can be implemented without having to increase the imaging precision of the precision cameras enormously in order to achieve a sufficiently precise search for a deviation.

The images taken by both the general cameras and the precision cameras are preferably analysed using essentially the same image-processing software. When using the same image-processing software, the examination of the image can be performed very rapidly and simply, and when combined with statistical analysis the method is also comprehensive.

The method is preferably used in the system, referred to above, according to the invention.

The images taken by the precision cameras are preferably analysed together with the images taken by the general cameras, using the same computation algorithm. This further increases the simplicity of the image analysis, as all the images can be processed without their having to be sorted before the image analysis . The images taken by the precision cameras are preferably analysed using statistical analysis and the intermediate areas are interpolated either linearly or non-linearly . The information obtained using the precision cameras, for example on the amount of dirt, can be analysed statistically, as dirt particles are generally distributed evenly over the entire width of the web.

The images can be analysed immediately using automatic image analysis, which automatic image analysis can take place at at least the imaging frequency. The automatic analysis of the images is performed completely without analysis stages that take place using human vision. Automatic image analysis, in which the images are analysed at at least the imaging frequency, makes the process fully real-time. All of the deviations in the web are then analysed in real time, thus considerably increasing the information on hazardous and non- hazardous events. In the system according to the invention, the precision cameras preferably use completely the same peripheral devices as the general cameras, making it possible to use the same camera beams, light sources, image-analysis processor cards, and other equipment for the precision cameras too.

The analysis results obtained using the two types of camera can be displayed as their own virtual bars in the user interface. The operator can monitor various deviations from separate virtual bars. This facilitates practical work and perception, as the deviations detected using the precision cameras cause problems in the process or final product that are different to those caused by the deviations detected using the general cameras . Thanks to the statistical analysis, only a few precision cameras are required over the width of the web, for example, if there are 16 general cameras there can be 4 precision cameras. The web being monitored in the method can move at a speed of 100 - 10000 m/min, preferably 400 - 3500 m/min. It is very important to detect deviations in the web, as deviations in the web can cause problems in the web- manufacturing process itself, or later in the use of the web. In the present patent application, the term web refers to a moving product web being manufactured. Such a product web being manufactured can be, for example, paper or plastic film. The term paper refers to board, chemical pulp, tissue papers, and coated or otherwise further processed papers. Narrow streaks or other deviations that are difficult to detect can appear in plastic films or in papers.

With the aid of the system, detections can be performed in the same observation position in several different illumination geometries, and now also at different precisions. The web is then illuminated at different times in different illumination geometries and is imaged using the same cameras. The web is imaged in at least two illumination geometries using one and the same camera while there are cameras for two precision ranges. The contrast differences of the areas being imaged can be great, or the details to be detected can be of very different types, in which case the particular features of each type of defect should be recorded precisely when imaging. There can be more than two imaging illumination geometries.

When using the same cameras in each observation position for imaging the web in at least two illumination geometries, the number of cameras can be made considerably small. In each observation position, the cameras take up a considerable amount of space, of which there is a very limited amount available. The placing of the cameras, i.e. the adding new illumination geometries to the observation positions, is limited in part by space. The invention permits the comprehensive monitoring of a web using general cameras, as well as precision cameras that are more precise than the general cameras, as there are considerably fewer more precise precision cameras than general cameras .

A reasonable total number of cameras permits them to be placed in positions, in which they will be less liable to dirtying. The cameras' lenses can, of course be cleaned, but while they are being cleaned the cameras cannot be used. Less cleaning permits more comprehensive monitoring of the web. The cameras are also preferably placed in such a way that they will remain clean. Apparatuses containing fewer cameras than previously will be smaller and can be placed in very many locations. In addition, the cameras form a considerable cost item in the web- monitoring system. In terms of cost savings too, the system according to the invention is very cheap to operate. The small number of cameras also affect the process equipment required, as a specific amount of data-processing capacity is required for each camera.

The same type of camera, with different optics, can be used for both general imaging of the web and for more precise imaging. In other words, the precision cameras can have similar electrical properties to those of the general cameras, only the lens will be different. With the aid of different optics, the cameras' imaging areas are of different sizes, so that when all the cameras use the same resolution (the number of pixels forming the image) , smaller monitoring pixels will appear on a smaller area. Smaller deviations, such as dirt, can be found from a smaller monitoring area.

By means of the system and method according to the invention, the web can be monitored sufficiently precisely to find dirt and other small deviations, without radically increasing the number of cameras. The use of precision cameras along with general cameras considerably increases the possibility of performing various kinds of monitoring. With the aid of the precision cameras, additional information is obtained on the web, such as the web's dirtying, the number of needle holes, and fibre formation. The information obtained from the system can also be used at different points in a web-forming machine for repairing detected defects and taking them into account in the process.

In the following, the invention is described in detail with reference to the accompanying drawings depicting some embodiments of the invention, in which

Figure 1 shows a schematic diagram of the system according to the invention,

Figure 2 shows a simplified view of the camera beam of the system according to the invention, Figure 3a shows the principle of the system with different illumination geometries, when detecting deviations from the web surface when illuminated from above,

Figure 3b shows the principle of the system with different illumination geometries, when detecting deviations from the web frame structure when illuminated from below,

Figure 4 shows the positioning of the cameras and light sources of the system in one embodiment,

Figure 5 shows the image-analysis processor card used in one embodiment of the system,

Figure 6 shows a view of the user interface according to one embodiment of the system according to the invention,

Figure 7 shows an image taken by a general camera, in which there is a thin point in the web, Figure 8 shows an image taken by a general camera, in which there is a needle hole in the web, Figure 9 shows an image of the thin point of Figure 7, taken by a precision camera,

Figure 10 shows an image of the needle hole of Figure

8, taken by a precision camera,

Figure 11 shows an image, in which there is dirt in the web, taken by a precision camera.

Figures 1 - 6 show one preferred embodiment of the system according to the invention for monitoring a web. The embodiment presents a system and method for use with at least two illumination geometries, but they can also be implemented using a single illumination geometry. The system according to Figure 1 includes at least two strobe-lighting elements 36, 36' for implementing different illumination geometries at different times, in order to find many types of deviation. The system also includes cameras 100 for imaging the web 16 in each illumination geometry. The cameras are of two types, general cameras 10, which image the whole web generally, over its full width, and precision cameras 10', which image specific points in the web in greater detail from small areas, in considerably greater detail than the general cameras. The general cameras 10 are installed on a camera beam to form a camera series, the width of the imaging area of which covers essentially the entire width of the web in a unified manner.

In a preferred embodiment, the general cameras are intended to act as both defect detectors (WIS) and in break monitoring (WRM) . In the preferred embodiment, the precision cameras image discontinuously the same area at a considerably higher resolution (the number of observation points to the surface area) than the general cameras, but the precision cameras act only as defect detectors. Each camera is preferably arranged to image the web in at least two illumination geometries at different times. In addition, the system includes at least one processing unit 30 for analysing the images taken by both cameras 100. The system can includes several processing units 30. The system further includes software means 65 for the statistical analysis of deviations in the images taken by the precision cameras 10', in order to form a result covering the entire web. In addition, the system includes a user interface 32 for displaying the analysis results and for controlling the system. A host unit 34 is used to control the lighting and imaging .

In this connection, the term processing unit 30 refers to a hardware and software totality intended for image analysis, which performs the analysis of the images. The processing unit 30 comprises at least one image-analysis processor card 74 together with image-processing software 37 and a hardware rack 33, in which there are power supplies for the image-analysis processor card or cards 74 and other possible components of the processing unit. The other components can be, for example, the buffer memories referred to later, or software means for statistical analysis. In the system shown in Figure 1, the general cameras are used to take images of the web at such a frequency that the web can be imaged continuously with the aid of the general cameras. The illumination geometries are preferably arranged to be changeable, in such a way that the web is imaged continuously in at least two illumination geometries. At momentary break can occur in the imaging, for example, for washing a camera. Such an interruption is, however, temporary and the web is essentially imaged completely. The precision cameras 10' preferably image a smaller area of the web, in which case the precision they use relative to the surface area will be greater. The same imaging frequency is preferably used in the precision cameras 10' and in the general cameras 10, so that, due to the smaller imaging area, the precision cameras 10' image the web only intermittently. Generally, the images do not cover the entire surface area of the web, but instead the intermediate areas are covered by statistical analysis. In the system according to the invention, the monitoring of the web using a corresponding method takes place as follows. In the system shown in Figure 1, the processing unit 30 and the host unit 34 are controlled with the aid of the user interface 32 through a data network 58, which in turn controls the cameras 100 and the strobe-lighting elements 36 and 36'. In one preferred embodiment, the host unit 34 first commends the cameras 100 to open the shutters (ERS = electronic rolling shutter) . After this comes a safety time-lag, which ensures that all the cameras' shutters open. After the safety time-lag, the shutters of the cameras 100 are sure to be open and the strobe-lighting elements 36 or 36' receive a command to light up for the desired time. When the strobe-lighting elements 36 or 36' switch off after the desired time, the shutters of the cameras 100 remain open for a safety time-lag, after which they close and the data is read to the processing unit 30. The data imaged by the cameras 100 travels over a bus 42 to the processing unit 30, where the data is processed. The processing unit 30 performs the same image analysis, in which all deviations are defined, for the images taken by both the precision cameras 10' and the general cameras 10. The processing unit 30 performing the image analysis preferably consists of image-analysis processor cards 74 (Figure 5) and a hardware rack 33 for securing them. The processing unit and thus the image-analysis cards can also be in connection with the cameras. The analysis results of the images of the general cameras, obtained from the processing unit after the image analysis, are transferred using a data-transfer network 58 to the user interface 32, to be displayed to the operator as a basis for decision-making. The images meeting the selected criteria are shown on deviation displays.

According to one preferred embodiment, the images of the precision cameras are analysed using the same analysis equipment and essentially the same computation algorithm as the images of the general cameras. The computation algorithm is designed to filter the information it obtains from the images when searching for specific deviations, which are defined in the parameters of the computation algorithm. Irrespective of whether the images are taken by a general camera or a precision camera, the computation algorithm searches the images for various deviations in the same manner. Computation algorithms known from the literature can be used as the computation algorithm, in order to find deviations. In its simplest form, the computation algorithm can be, for example, an algorithm using the threshold technique, which search for differences in light in the pixels.

The deviations found by the computation algorithm are classified with the aid of a classifier, which identifies a deviation as a specific defect. The largest deviations are best found from the images taken by the general cameras, whereas smaller deviations cannot be properly distinguished from these images. The defects generally found using the precision cameras can be defined separately in the parameters of the computation algorithm, as can the defects found using the general cameras. Figure 7 shows an image taken with a general camera, in which there is a thin point 200 in an 80-g/m² web 16. As the dimension of one side of a monitoring pixel of a general camera is 0,4 mm, the thin point 200 is extremely difficult to detect. For its part, Figure 8 shows a needle hole 202 with a 0,1-mm diameter, detected using a corresponding general camera to that in Figure 7. Even the smallest deviations in the images taken by the precision camera can be detected reliably, because the resolution, i.e. precision, of the precision cameras is considerably better than that of the general cameras. In addition to the defects detected by the general camera, the analysis can also find even smaller defects (ultra-small dirt spots, needle holes, etc.), which generally appear evenly over the entire web. Figure 9 shows an image taken of an 80-g/m² web 16, in which a thin spot 200 appears, which is the same as that in Figure 7. As the dimension of one side of a monitoring pixel of the precision camera is 0,05 mm, the thin spot can be clearly detected in Figure 9. Figure 10 shows the same needle hole 202, which was previously shown in Figure 8. The needle hole 202 in the precision-camera image can be clearly detected. Figure 11 shows dirt 204, which could not be detected at all by the general camera. However, the precision cameras found it. The same type of precision camera as in Figure 9 was also used in Figure 10 and 11.

In addition to the defect images shown in Figures 7 - 11, data of the defect is displayed to the operator in the form of numerical information. The system preferably includes means for creating numerical information, with the aid of which the type and size of the defect, and in the case of defects larger than a selected threshold also location data, are displayed to the operator. Each defect can have a defect-specific threshold value, defects above which can be displayed to the operator together with their location value. Defects smaller than this can be shown only by their type, size, and number.

Imaging only part of the web with the aid of precision cameras is sufficient to provide reliable information on the appearance of small deviations. On the basis of statistical analysis it is possible to perform, for example, dirt detection according to the standard, or fibre-orientation detection. Information, for example on deviations in the fibre orientation, can be forwarded to the formation sensors.

In this connection, the definition "essentially the same image- processing software" refers to the fact that software performing similar operations, which detects deviations based on the software's internal computation algorithm, is used for the analysis of images taken by both the precision cameras and the general cameras. The image-processing softwares can differ in the case of the parameters used in the computation algorithms and in the case of other defect-specific definitions. Further, the definition "essentially the same algorithm" refers to the fact that the computation algorithms are similar in operating principle, but can differ from each other in the case of the tuning parameters depending on the resolution of the images of the algorithms. This means that parameters tuned to the resolution of the precision camera can be used in the threshold algorithm intended for analysing the images of the precision camera.

Unlike the images from a general camera, the deviations detected from the precision cameras and the defects classified from the deviations are preferably sent to the software means 65 as numerical values (for example, the defects in a specific period of time) . The software means can be implemented on the analysis-process card, in the processing unit, or also as a separate unit outside the processing unit, as in Figure 1. The separate unit can be, for example, a calculation circuit or similar on the user-interface or server side, which performs interpolation. The computation power required for statistical analysis is only a fraction of that required for image analysis, so that the software means for statistical analysis can be located more freely. The software means 65 perform statistical analysis for defects detected from the images of the precision camera, such as for example needle holes and dirt, on the based on the surface-area data imaged by the precision camera. Statistical analysis preferably means that the deviations detected from the precision-camera images are interpolated exploiting the surface-area data either linearly or non-linearly to cover the intermediate areas, so that an estimate covering the entire web is obtained from the defects detected by the precision camera. With the aid of the estimate, the values of small defects can be presented as a trend relative to time, so that the development of dirtying can be monitored as a function of time.

Figure 2 shows a simplified view of the placing of the system's cameras. In the figure, the cameras 1 00 are shown in a side view of the camera beam 28 , so that the distribution of the cameras 1 00 over the entire width of the web 1 6 can be seen. The general cameras 1 0 are permanently attached to the camera beam 28 parallel to each other at even intervals as a series of cameras over the entire width of the camera beam 28 , so that the entire width of the web 1 6 can be imaged continuously in real time. In this connection, the term camera series refers to the fact that the general cameras are parallel to each other and that the images taken by them form an image series in the cross-direction of the web, which covers the entire width of the web. The precision cameras 10 ' are also placed at even intervals permanently on the camera beam 28 over the width of the web 1 6 . As the imaging area 90 ' of the precision cameras 10 ' is considerably smaller than the imaging area 90 of the general cameras 1 0 , the precision cameras 1 0 ' are able to image only small areas of the web 1 6 . In addition, there are considerably fewer precision cameras 10 ' over the width of the web. In terms of the deviations detected by the precision cameras, the intermediate areas 92 remaining between the imaging areas in both the longitudinal and transverse direction of the web can be covered with the aid of statistical analysis, as the small deviations, such as the dirtying of the web, spread evenly over the width of the web, so that measurement of only the density of the dirtying will be sufficient.

With the aid of statistical analysis performed on the deviations detected from the precision-camera image after the image analysis, the intermediate area remaining between the images in both the transverse and longitudinal directions of the web can be interpolated either linearly or non-linearly . This means that an estimate of the density of the deviations remaining in the intermediate areas is sought with the aid of statistical analysis, so that the entire web is covered.

According to Figure 2, the precision cameras 10' can be placed permanently on the camera beam 28 in a corresponding manner to the general cameras 10 and they can share the same hardware, for example light sources and logic cards, with the general cameras. Functionally, the operation of the precision cameras corresponds to that of the general cameras. In practical terms, it is extremely important that the precision cameras can be located on the same camera beam as the general cameras. The system can then be implemented without a separate camera beam for the precision cameras, the installation of which in the cramped spaces inside the confines of the web-forming machine would be very difficult and lead to additional costs. The installation of the precision cameras on the same camera beam also permits existing systems to be updated with precision cameras and the software changes demanded by the system, after which it will be possible to start using the method according to the invention for web monitoring. The fixed installation of the cameras eliminates traversing apparatuses, which add to the investment and maintenance costs of the system. In addition, the analysis of images attached to a traversing measuring head would be considerably more difficult. Images of precision cameras attached to a traversing measuring head could not be analysed using the same computation algorithm as the images of general cameras.

The cameras used as precision cameras are preferably the same type of camera as the general cameras, so that they can be easily connected to the system alongside the general cameras. The greater precision of the precision cameras is achieved by giving the cameras a different type of optics, with the aid of which a smaller imaging area of the web can be monitored. The web's cross-direction length of an individual pixel to be monitored can be, for example, 0,1 mm, whereas in the general cameras it can be, for example, 0,44 mm. When using a 0,1-mm pixel size with a 640 x 480 resolution, the surface area of the imaging area imaged by an individual precision camera would have a size of 6,4 cm x 4,8 cm. The surface area of the imaging area imaged by the precision cameras is so small, that the imaging frequency of the precision cameras would be insufficient for continuous web monitoring. Thus, the intermediate area remaining between the images is covered with the aid of statistical analysis. Using modern camera technology, the pixel size can be reduced down to a cross- direction length of 0,01 mm. Generally, the ratio of the individual pixel width of images taken by general purpose cameras to a pixel of the precision cameras can be 3 - 12, preferably 4 - 6, in which case a sufficiently small imaging area can be achieved for finding small deviations for statistical analysis.

According to one embodiment, in connection with the images of the precision cameras it is possible to use in the machine direction the so-called binning function of the cameras, in which individual consecutive pixels in the longitudinal direction are combined and read as a series.

According to a second embodiment, in the precision cameras it is possible to use a lens, by means of which the web can be imaged in its direction of travel for as long a distance as that with a general cameras, but the width of which is extremely small. In this case, statistical analysis will only be required in the machine's cross direction.

The precision cameras can also have different electrical properties to those of the general cameras, but in that case the pulsing of the precision cameras must correspond to that of the general cameras, so that the joint timing of the light source and the cameras will work properly. The deviations found on the basis of the image analysis performed on the images taken by the precision cameras can be easily aligned in the images taken by the general cameras. This is due to the fact that the precision cameras and general cameras have the same position in the longitudinal direction of the web-forming machine. In addition, the deviations detected by the precision cameras are easy to scale with the deviations detected by the general cameras.

Figures 3a - 5 show the operation of the cameras on a general level, when the operation of both the general cameras and the precision cameras is depicted, unless otherwise stated.

Figures 3a and 3b show an application, in which the web 16 is imaged and illuminated in different illumination geometries. The different illumination geometries are formed, when the web 16 is illuminated by lighting elements 11 and 12 illuminating at different times. The lighting elements 11 and 12 illuminating at different times can be on both sides of the web 16. The camera 100 is located on only one side of the web 16.

In Figures 3a and 3b the light coming from a switched-on lighting element 11 or 12 is shown with solid lines. The light that would come from a switched-off lighting element 11 or 12, were it to be switched on, is shown with broken lines. The area 18 of the line illuminated by the illuminated elements 11 and 12 is shown with cross-hatching. By detecting the web 16 in the same observation position 20 in different illumination geometries at different times provides, from a single observation position 20 addition information, compared to a situation in which the web is illuminated by several lighting elements simultaneously. If the illumination geometries shown in Figures 3a and 3b are used separately from each other, several observation positions will be needed. Besides the increase in the observation positions, more cameras will then be needed. The increase in the number of cameras will increase costs considerably. Figure 3a shows a situation, in which the lighting element 12 on the camera 100 side of the web 16 is being used to illuminate the web 16 as desired. When using the lighting element 12 on the camera 100 side of the web 16, the web 16 is imaged by reflected light. Using such an illumination geometry, defects particularly in the surface of the web can be detected. For example, an orange-skin pattern in the surface of a paper web will be revealed in this illuminated, as will impurities in the surface of the web.

Figure 3b shows a situation, in which the lighting element 11 on the opposite side of the web 16 from the camera 100 is in use, illuminating the web 16 as desired. When the camera 100 and the lighting element 11 are on opposite sides, the web 16 is illuminated by transillumination. Using such an illumination geometry, defects particularly in the structure of the web 16 can be detected. Gel accumulations in a paper web are examples of such defects. The detection of gel accumulations in a paper web is based on their greater light transmittance than paper.

In the illumination geometry shown in Figure 3b, the web 16 is imaged and illuminated from different sides. When lighting take place through the web 16 by means of the lighting element 11, the structure of the web 16 is detected in its entirety. Such a detection of the whole structure that takes place together with observation of the web surface structure in Figure 3a is very practical, as information is then obtained on very different types of defect. Using such transillumination, the web passes light through it. A precondition for transillumination is that the web passes at least some of the light through it.

Figure 4 shows the system's cameras and light sources for monitoring a web. On each side of the web there are lighting elements 11 and 12 as well as cameras 100 on one side of the web. The placing of the cameras 100 above the web 16 is advantageous, as it significantly reduces the dirtying of the cameras. There are considerably fewer precision cameras than general cameras. The lighting elements 11 and 12 can consist of many different types of light, but a lighting element preferably consists of LED lights 22. The lighting elements 11 and 12 are thus light beams 24 and 25 formed of LED lights 22. The LED lights permit the production of a rapid strobe light and have a long service life. The lighting elements are preferably LED-strobe lighting elements. It can be clearly seen from Figure 4 that, even through only a single camera is drawn in the other figures 3a and 3b, there are preferably several cameras 100 parallel to each other in the cross direction of the web 16, forming a camera series. According to one embodiment, deviations can be detected in the web at a single observation position using four illumination geometries. Some of the illumination geometries permit only individual images of the web to be taken, but the illumination geometries are preferably changed in such a way that all the areas of the web are imaged in at least two illumination geometries. On the other hand, using some of the illumination geometries, images may be taken only in order to monitor the general level. Less frequent imaging will permit a reduction in the amount of data.

In Figures 3a - 4, two illumination geometries alternate, but in the method according to the invention more illumination geometries can be used. Images of the web can preferably be taken at an imaging frequency of 42 images per second using both illumination geometries, i.e. the illumination geometry is changed 84 times each second. It will then be possible to detect each point on a web moving at 500 m/min using both illumination geometries over a 200-mm long observation area. As the web speed increases and the length of the observation area decreases, the illumination geometries should be alternated even more rapidly. In addition, there should be more than two alternating illumination geometries. In such a case, the illumination geometries can alternate as much as thousands of times each second. In Figure 4, the camera beam 28 comprises the necessary number of cameras 100, as the observation area is typically so large that it is impossible to detect entirely using a single camera, when seeking a high precision. The camera beam can be, for example, a beam made from steel, which is arranged in the immediate vicinity of the web over the entire width of the web. The system can be implemented using line or matrix cameras, but its implementation using matrix cameras is preferable, as when using matrix cameras an area can be imaged, unlike when using, for example, line cameras. When imaging a larger area much more information is collected than from a line. Matrix cameras can be used to image faster moving webs than line cameras, as when using line cameras the imaging frequency will become very high. In the future, when the number of pixels in the cells increases, present precisions will be achieved using fewer cameras. A processing unit designed to calculate the matrices is preferably connected to the matrix camera. During the entire process, it is possible to use cameras based on the same architecture. The imaging parameters and computation algorithms will then be adapted to this task.

It is possible to perform a Z transform in connection with the cameras used in the system, which will permit a more effective utilization of the bits of the cameras. Using a Z transform, a linear scale is converted to a non-linear one, when the precision of an image presented by a specific number of bits can be guided to a desired density range. In addition, the cameras' S/N value can be 60 - 120, preferably 70 - 100, dB. If the cameras' S/N value is less than 60 dB, problems will arise in the measurement of dense areas, due to noise. In theory, the camera beam can also comprise a single camera (not shown) , if the observation area is small, or the required precision is low. Such an application could be, for example, the imaging of a narrow web, when only a single camera will suffice. However, in usual embodiments there are cameras parallel to each other, forming a camera series. In the figure, the differently timed lighting elements are on both sides of the web, but they can also be on one side. When it is desired to monitor a web in specular-reflected and scattered light, there should be at least two lighting elements on the camera side of the web. The web is preferably monitored using general cameras over its entire width and using precision cameras at even intervals for specific points. The monitoring of the web edges is important, as they affect the runnability of the web, even though the edges may be left unused in the final product.

The web can also be monitored using a system, in which the lighting elements illuminating at different times are on the same side of the web as the cameras. If the lighting elements include a high-angle lighting element and a low-angle lighting element that are on at different times, the illumination geometry can be used to detect different types of defect on the web surface. The detectable defects will depend on the angle, at which the web is illuminated. The use of differently timed lighting elements can correspond to that in the applicant's previous Finnish patent application FI 20065570.

According to one embodiment, the web can be illuminated with diffuse light from the same side of the web from which it is imaged. If the web is illuminated with diffuse slanting light from the cameras' side, an area of specular-reflected light and an area of scattered light will be formed on the web surface. The specular-reflected light leaves the web surface to the cameras at essentially the same angle at which it have arrived on the web from the lighting elements. For its part, the scattered light arrives from the web at the cameras at essentially a different angle to that at which it has arrived from the lighting elements at the cameras. Monitoring the web in specular-reflected and scattered light is highly advantageous, as very different types of defect are seen in these.

In the system according to Figure 1, the exposure time is adjusted as desired by altering the flash time of the strobe- lighting elements 36 and 36'. In the adjustment of the exposure, it is preferable to use strobe-lighting elements, by means of which really short exposures can be achieved, for example 5 - 10 microseconds. The light must be strobe light in order to change the illumination geometries, so that the use of the same strobe light to control the exposure is very practicable. Buses 38 and 38' run to the strobe-lighting elements 36 and 36' from the host unit 34, in order to give an exposure command. The strobe-lighting elements 36 or 36' can be switched on at different times. The lighting will then take place in different lighting geometries at different times. Once lighting has taken place using strobe-lighting element 36, the next lighting will take place using strobe-lighting element 36'. A bus 40 runs to the cameras 100 in order to transmit an imaging command. The lighting elements 11 - 15 are preferably strobe-lighting elements 36 and 36', so that the lighting event will be sufficiently rapid.

The cameras' exposures can also be adjusted using the cameras. At the present moment, the reading of data from cameras is still slow, so that the adjustment of the exposure is preferably performed using the lighting elements. In the future, when data-reading speeds increase, the cameras can be used to adjust the exposures.

In the system, one imaging command to the cameras is sufficient, on the basis of which the cameras are programmed to open the shutters and close them after the desired time, as well as to read the data from the imaging element for transmission to the processing unit. A safety time-lag is used to ensure that all the cameras' shutters will be open when the lighting elements flash. In such a very high-speed system, in which hundreds of images are taken each second, delays can easily become significant. Safety time-lags, by means of which the detrimental effects of the delays are minimized, are used due to the disturbing significance of the delays. The synchronization of the strobe-lighting elements 36 and 36' is preferably arranged for the imaging period of the cameras 100. The imaging period of all the cameras 100 is essentially the same. The timing of the strobe-lighting elements 36 and 36' and the cameras 100 is controlled centrally by the host unit 34. When imaging with several illumination geometries and two different precisions, an enormous amount of data comes from the cameras to the processing unit. In the processing unit 30 there is computing power of 0,1 - 100 teraflops per camera 100. With such a computing power hundreds of images each second can be analysed using automatic image-analysis from start to finish and the results displayed to the operator. In the analysis of the web, methods can be used, by means of which web defects can be found, the detection of which has not been possible, for example using methods based on thresholding. The web can then run at 100 - 1Q000 m/min, preferably 400 - 3500 m/min. Web imaging can take place in observation positions, in which only a short portion of the web is visible. The imaged length of the web in the machine direction can be 0, 1 - 300 mm, preferably 50 - 150 mm. When the processing unit contains power or more than 0, 1 teraflops, it can be used to process a continuous flow of images taken from a web moving at more than 100 m/min, in which the web is recorded at a precision in the order of millimetres. The analysis of the images is preferably performed at the imaging frequency without recording before analysis, so that the storage capacity will be minimized. When the image processing takes place at the imaging frequency, it takes place in practice in real time.

The system according to Figure 1 can include at least one compressed buffer memory 64, in which the latest images of the web taken by the cameras 100 for a period of 0,5 - 30 minutes will fit. If necessary, the images can be saved for a longer time too, but 5 - 30 minutes is generally enough. The buffer memory is preferably a circular buffer, in which the new data are always saved on top of the older data. Thus a circular buffer always contains the latest images from a defined period of time. The system also preferably includes an uncompressed buffer memory 60, in which all the images are momentarily saved in an uncompressed form. It is unnecessary to increase the size of the uncompressed buffer memory excessively, so that the uncompressed images from a period of 0,5 - 5 minutes will fit into it. Some image can be retrieved from the uncompressed buffer memory for displaying through the data-transfer network at any time that the system is in use. The storage of the images in the buffer memory terminates if a predefined disturbance takes place on the machine. Such a disturbance can be, for example, a web break taking place on a paper machine. There will then be images in an uncompressed form in the buffer memory, from the time preceding the disturbance. The uncompressed very accurate image will give a better point of departure than previously for detecting the reasons for the deviation. Both buffer memories 60 and 64 can be implemented on the image-analysis processor card 74.

The system also preferably includes permanent storage means 62, in which deviation images are stored. A defect or deviation image is an image, in which there is a deviation meeting set conditions. The number of deviation images is extremely small compared to the total number of images taken. Thus, the deviation images can be stored in permanent storage means 62 over the data network 58. The deviation images are saved as such uncompressed in the permanent storage means, for later examination. Any deviation whatever can be retrieved over the data network 58 to the user interface 32 from the permanent storage means 62 for later examination. The host unit 34 shown in Figure 1 includes control means 35, with which the imaging parameters of the cameras 100 are changed to the values determined by the illumination geometry being used. When using each illumination geometry with imaging parameters optimized precisely for the illumination geometry, the camera bits can be targeted at precisely the density area where they are needed. It will then be possible to use cameras, the number of bits in which will be less than would otherwise be needed. In turn, the reduced number of bits of the cameras will permit a reduced need to transfer data and a reduced processing capacity.

Imaging is preferably implemented in a 10 - 50-bit, preferably 12 - 24-bit form. The dynamics, i.e. depth of the imaging is highly dependent on the number of bits being used. At less than 10 bits it is impossible to display images in such a way that many kinds of defect can be detected accurately from them. It is preferable to use more than 12 bits, so that one and the same camera can be used to image defects detected in different types of lighting. The term different types of lighting refers, for example, to imaging with transillumination and reflected light. In one embodiment, the bits are targeted at different density areas in different illumination geometries. If the light source is on the same side of the web, different types of lighting can be achieved, for example, by imaging scattered and specular-reflected light, when the contrast difference will be great. On the other hand, as the number of bits increases greatly, the amount of data will grow unnecessarily. Generally, 24 bits will be sufficient, as the contrast difference is seldom so great that this would not suffice for showing images.

For their part, the system operator too can analyse the images. However the images are preferably analysed entirely using automatic image analysis with highly-developed image-analysis methods, as analysis using the human eye is slow. In addition, analysis using the human eye is laborious, so that it may easily be left without being performed. Even though every operator were to analyse the images taken of the web to the best of their ability, there are considerable differences between operators. Thus, the analyses performed by operators are always to some extent subjective. On the other hand, analysis using a machine always takes place the same way. In addition, the analysis of images is preferably performed immediately after imaging, when it is possible to minimize the storage capacity required. In addition, immediate automatic image analysis is essential if the monitoring process is to be made real-time.

As already stated earlier, when imaging webs very precisely at a high imaging speed, the amount of data grows to become very large. In the system according to the system shown in Figure 1, a fibre-optic data-transfer network 43 is used between the cameras 100 and the processing unit 30, by means of which the images are transferred from the cameras 100 to the processing unit 30. It is nearly impossible to use copper conductors to implement the transfer speeds required, as the data-transfer capacity required is in the order of gigabits a second. By using fibre-optics, the transfer speeds can be implemented more cheaply than by using, for example, copper conductors. The processing unit 30 is connected to the data network 58 for statistical analysis. The permanent storage means 62, the user interface 32, and the host unit 34 are also connected to the data network 58. By means of such a structure, a system can be implemented, in which the data-transfer capacity is optimized for each operating purpose. By using fibre-optics, the transfer of large amounts of data can be cheaply implemented, as when using fibre optics a data-transfer speed of, for example 10 Gb/s, can be used. The speed of a normal data-transfer network, for example a LAN network, can be, for example, 1 Gb/s. A normal data-transfer network is sufficient for post-processing data transfer. The amount of data to be transferred decreases considerably during processing, as, when the web production process is operating, the deviation images are typically less than 1 % of the total number of images.

Figure 5 shows an image-analysis processor according to one embodiment, which is optimized for the analysis of images coming from four cameras. The card can contain, for example, four input connections, i.e. input channels 77, in which case the images coming from four cameras can be led to the image- analysis processor card 74, through their own input connection 77. After image analysis, the information is led out through one output connection 79. In the implementation of the image-analysis processor card 74, special processors 69 are used, by means of which a capacity suitable for the computation of images is achieved of at least 0,1, preferably 0,25 teraflops per camera, the image flow coming from which is 600 Mb/s. Matrix operators are preferably using in the analysis of the images, as these can be used to analyse bit maps effectively.

The image-analysis processor card 74 has one processor 75 especially intended for handling matrix operations, which is preferably an FPGA 75' (Field-Programmable Gate Array). The FPGA processor 75' is preferably common to all the channels coming to the image-analysis processor card in question from all of the various cameras. In addition, the image-analysis processor card has nine processors 73 with firmware, which are preferably DSPs 73' (Digital Signal Processors). There is at least one dedicated DSP processor 73' for each individual camera input channel, which performs the image analysis. The same or similar image-analysis processor cards can be used for the analysis of the images of both the general cameras and the precision cameras, i.e. all the images are run to similar image-analysis processor cards, and a similar image analysis is performed on all the images. The number of image-analysis processor cards varies according to the number of cameras used. A particularly efficient image-analysis speed is achieved by means of such an architecture.

According to one embodiment, the system includes several image- analysis cards, which are all located in a single hardware rack. Further, an individual image-analysis processor card comprises, according to Figure 5, many input channels, i.e. the images from several cameras are processed using a single image- analysis card. The images of all the cameras, irrespective of whether they are taken by general or precision cameras, are fed to these image-analysis processor cards. Preferably, the same image-processing software is run on all the images. The image- analysis processor cards are preferably remotely programmable, i.e. their configuration takes place as remote operation. An embodiment of this kind is extremely cost-effective in implementation, as the image analysis of all the cameras takes place using the same apparatus and software, so that the system can be implemented without changes to the equipment. Only the necessary parameters changes are performed on the software and software means are added, by means of which statistical analysis is performed on the image of the precision cameras.

According to another embodiment, for each camera in the system there is a dedicated image-analysis processor card.

Figure 6 shows one example of the user interface used in the system. Through the user interface, it is easy for the user to control the monitoring of the web and make observations of the deviations appearing in the web. The deviations in the web detected by the precision cameras are preferably displayed in the same user interface as the deviations of the general cameras. There can be individual virtual bars in the user interface for the deviations of both camera types, so that the number of deviations in a single bar will be advantageous in visual terms. Different types of illumination geometry too can be displayed in different virtual bars. This will be emphasized particularly, if several illumination geometries are is use simultaneously, in which case one of the physical camera beams can have at least four virtual bars, in which the deviations are displayed. In Figure 6 the deviations detected by both the precision cameras and the general cameras are shown in the same deviation map 300.

Figure 6 shows a view of the user interface 21 according to one embodiment of the system according to the invention, on a paper machine. The user interface 21 functions at any work, i.e. operating station 23 (Figure 1) whatever connected to a data- transfer, i.e. local area network. The display of the operating station using the user interface must have a sufficient resolution. The user interface is preferably programmed using internet-browser technology, so that the operating station can be implemented without special-application software. An operating station, to which a display is connected, and which is connected to a 1-Gb/s data-transfer network, can be used as the main operating station. Dual-display technology can also be used in the operating station. In the basic setup of the user interface 21, the left-hand side 85 is reserved for displaying individual deviations and for general information, which are displayed automatically. A display 26 of the latest deviation is displayed at the upper edge on the left-hand side 85. The latest deviation is shown automatically in this display. The deviation arrives in the latest-deviation display immediately it is detected. The image shown in the latest-deviation display is uncompressed and is shown in its natural size. If the image of the deviation is too large for the latest-deviation display 26, the deviation is reduced automatically in a suitable ratio. On the left-hand side of the display 26 is a display 31, in which information on the deviation shown in display 26 can be displayed.

In Figure 6, below the display 26 there can be, for example, a history browsing display 124, from which previously detected deviations can be browsed. Below the browsing display 124 there can be deviation-counter displays 116, in which the number of deviations in the reeling drum in production can be presented by deviation class. Other values too can be presented in connection with the deviation-counter display, such as the reeling drum's number, the grade code, and the speed. At the lower edge of the user interface 21, there can be a general information display 122, in which basic operating data can be found, such as the time, the grade, and the web speed. In addition, at the upper edge of the left-hand side 85 of the user interface 21 there can be fast-select buttons 118, from which separate additional information pages can be opened, for example, for formation, trends, reports, alarms, and settings. If desired, the web's formation, which is determined when imaging the web in detail with lighting taking place as transillumination by strobe light, can also be shown on the same display 26. The formation display can be in natural size and its scale can be alter using the zoom function. The web is seen in its entirety by scrolling the view laterally over the web. At the left-hand side next to the formation display a formation number display can be shown in the display 31, in which the characteristics of the formation calculated from the image area, as well as the characteristics of the formation of the entire web, are shown. The characteristics are shown as a formation index, i.e. total variation, mean floe size, and skewness .

A 19" TFT display, for example, can be used as the user- interface display. By seeing deviations in real time, the operator can make decisions to correct the problems. The process and quality-disturbance image come to the display automatically, without manual work to retrieve the images. The images of the event chain are typically shown to the operator in an uncompressed form. The operator can be offered information on, for example, formation, statistics, reports, trends, and profiles.

In its entirety, the operation of the system permits real-time monitoring of events. When the system detects a deviation in the web, the deviation is shown immediately to the operator. All the deviations found by the system are shown to the operator automatically.

The right-hand side 83 of the user interface 21 is reserved for a deviation map (defect map) 300. Colours are preferably used in the illustration of the images. Different symbols are used for normal individual defects in the deviation map 300 while background colours are used to depict statistically calculated defects. A length of the web defined by the operator is shown in the defect, i.e. deviation map 300. The deviation map can be imaged immediately prior to reeling, so that the defect map will show the quality of the paper web manufactured on the paper machine. In the deviation map 300, the paper web travels downwards representing the defined paper web, so that new defects, i.e. deviations come into view at the upper edge of the deviation map 300. The deviations are marked on top map by symbols 101, which depict very well the basic type and size of the deviation. If the deviation is of long duration, for example in the coating, the symbol 101 on the deviation map 300 is scaled to a corresponding length. At the right-hand side of the deviation map 300 is a trend display 104, which shows the distributions in the machine direction of the defects on the deviation map 300 as a trend graph. Underneath the deviation map 300 is a profile display 108 of the defects on the deviation map, from which the distribution of the deviations in the cross direction of the web can be seen. Defects detected by the precision cameras can be shown with the aid of different colours, which depict the density of a specific small deviation, for example dirtying, in the area. The system according to the invention is very advantageous for use in detecting deviations in a paper web. Here, the term paper machine refers to paper, board, tissue, and chemical pulp machines. The method can be used in the wire, press, or drying sections of a paper machine. The use of the system in paper machines is extremely advantageous, as the imaging areas on a paper machine, in which web can be monitored, are typically very short. The monitoring distance in the machine direction of a paper machine is 0,1 - 300 mm, preferably 50 - 150 mm. When the speed of the paper machine is constant, the imaging frequency is defined to be such that all the areas of the web are imaged using the general cameras, in which case they can be presented as a continuous image of the web. The precision cameras image more precise imaging areas of the web, from which, with the aid of statistical analysis, information on small deviations can be obtained. The system is suitable for use in high-speed processes, such as on paper machines, in which the paper speed is, for example 2100 m/min, and the area being imaged is 200-mm long. In such a case, 175 images per second should be taken in each illumination geometry, in which it is desired to create a continuous image of the entire web.

The method according to the invention is also advantageous for use in web monitoring in the different stages of plastic manufacture. The quality of a plastic web can be monitored immediately after extrusion or later, for example after longitudinal and lateral stretching. After extrusion, the plastic web is examined, to detect deviations in it, such as holes and impurities. After stretching, different lighting angles are used to bring orientation into view. Orientation has an effect on, for example, the barrier and strength properties of plastic, making it quite essential to know the orientation. In addition, the evenness of the surface of the plastic film and the reflective properties of the plastic film can be studied. These are of considerable significance in terms of the end products and, in addition, tell a great deal about the state of the manufacturing process. In plastic, the orientation can be in either direction, depending on stretching. In addition, there can be other narrow transverse defects too in plastic. The system according to the invention is extremely practical for finding defects of this kind, as it permits the web to be observed also in the cross direction in many illumination geometries. Observation can take place in other directions too than the machine/cross directions. A suitable angle can be selected on the basis of the defects to be detected.

There can also be cross-direction defects in other webs than plastic webs. For example in paper, cross-direction narrow lines will appear, if the doctor blade has shaken during coating and a thin narrow cross-direction coating streak has remained on the paper.

In the present application, the term "deviation" refers to a disturbance in the process. Such a deviation can be a defect or a break. Defects are, for example, holes, impurities, or streaks. The term "small deviation" refers, for example, to dirt or a needle holes, or a deviation in the fibre orientation. When using several illumination geometries simultaneously at two different imaging precisions, the main purpose is to find defects in the web, but if, however, a break occurs in the process, it too will be noticed.

The web speed is typically more than 100 m/min, as stated above. The method can also be applied without problems in higher-speed processes. One of the central areas of application is paper machines, the speeds of which are 400 - 2400 m/min. In a few years' time 3500 m/min. The method can also be applied with considerably faster webs. Such faster webs can move at as much as 10000 m/min, but nevertheless their monitoring can take place using the method according to the invention. If the web moves at 100 - 10000 m/min, preferably 400 - 3500 m/min, the computation power required for the analysis of the images will be 0,1 - 100 teraflops per camera.

Claims

1. System for monitoring a web, for example a paper web, which system includes

- at least one set of lighting elements (11, 12) for implementing an illumination geometry for illuminating the web for imaging,

general cameras (10) installed parallel to each other as a series in the width direction of the web (16) for imaging the web (16) over essentially the entire width of the web (16) , using which same general (10) cameras the web (16) is arranged to be imaged in at least one illumination geometry,

at least one processing unit (30) with image- processing software (37) , to analyse the images taken using the general cameras (10) , using a selected algorithm to detect deviations ,

at least two precision cameras (10' ) for imaging the web (16) in a specific illumination geometry at a substantially greater precision relative to the surface area than the general cameras (10) , to detect small deviations,

at least one processing unit (30) with image- processing software (37) , to analyse the images taken using the precision cameras (10' ) , using a selected algorithm to detect deviations,

- a camera beam (28) for supporting the general cameras

(10) and precision cameras (10' ) in connection with the web (16) ,

a host unit (34) for controlling the lighting and imaging,

- a user interface (32) for illustrating the analysis results and for controlling the system, and

a data-transfer network (43) for transferring data between the parts of the system,

characterized in that the system further includes software means (65) for statistically analysing the deviations of the images taken by the precision cameras (10' ) , to create a result covering the entire web.

2. System according to Claim 1, characterized in that the images taken by both the general cameras (10) and the precision cameras (10') are arranged to be analysed using essentially the same image-processing software (37).

3. System according to Claim 1 or 2, characterized in that both the processing unit (30) of the general cameras (10) and the processing unit (30) of the precision cameras (10') is arranged to use essentially the same computation algorithm for computing the deviations of the images taken by both the general cameras (10) and the precision cameras (10').

4. System according to any of Claims 1 - 3, characterized in that the software means (65) are arranged to statistically analyse the deviations computed by the computation algorithm from the images taken by the precision cameras (10') and interpolate the intermediate areas (92) either linearly or non- linearly.

5. System according to any of Claims 1 - 4, characterized in that the system includes 3 - 10-times more general cameras (10) than precision cameras (10').

6. System according to any of Claims 1 - 5, characterized in that both the general cameras (10) and the precision cameras (10') are permanently attached to the same camera beam (28).

7. System according to any of Claims 1 - 6, characterized in that the ratio of the pixels of images taken by a general camera to those of images taken by a precision camera is 3 - 12, preferably 4 - 6.

8. System according to any of Claims 1 - 7, characterized in that each processing unit (30) includes at least one image- analysis processor card (74), with the aid of which the system is arranged to analyse the images taken by several general cameras (10) and the images taken by at least one precision camera (10 ' ) .

9. System according to Claim 8, characterized in that each image-analysis processor card (74) includes several input channels (77), at least one FPGA processor (75') for all the input channels (77), and at least one dedicated DSP processor (73') for each input channel (77).

10. System according to any of Claims 1 - 9, characterized in that the precision cameras (10') have similar electrical properties to those of the general cameras (10) .

11. System according to any of Claims 1 - 10, characterized in that the said processing unit (30) with image- processing software (37) for analysing the images taken by the general cameras (10) is the same as the said processing unit (30) with image-processing software (37) for analysing the images taken by the precision cameras (10').

12. Method for monitoring a web, for example a paper web, in which method

- the web (16) is illuminated by lighting elements (11,

12) ,

images of the web (16) are taken by general cameras (10) installed parallel to each other as a series over the entire width of the web (16) in at least one illumination geometry, to find deviations,

images of the web (16) are taken in addition by at least two precision cameras (10') in a specific illumination geometry at a substantially greater precision relative to the surface area than the general cameras (10), to find small deviations,

the images taken by both the general cameras (10) and the precision cameras (10') are analysed, to find deviations, characterized in that the deviations of the images taken by the precision cameras (10') are in addition analysed statistically to form a result covering the entire web.

13. Method according to Claim 12, characterized in that the images taken by both the general cameras (10) and the precision cameras (10') are analysed using the same image- processing software (37).

14. Method according to Claim 12 or 13, characterized in that the method is used in the system according to any of Claims 1 - 11.

15. Method according to any of Claims 12 - 14, characterized in that the images taken by the precision cameras (10') are analysed using essentially the same computation algorithm with the images taken by the general cameras (10) .

16. Method according to any of Claims 12 - 15, characterized in that the images taken by the precision cameras (10') are analysed using statistical analysis and the intermediate areas (92) are interpolated either linearly or non-linearly .

17. Method according to any of Claims 12 - 16, characterized in that the images are analysed immediately using automatic image-analysis, which automatic image-analysis takes place at at least the imaging frequency.