Paper web inspection with
intelligent line scan cameras
Automatic inspection of production
is a standard practice in most paper mills. During the latest 40 years
the development of sen-sor technology used for this application area has
advanced from electromechanical devices, via laser scanners and phototran-sistor
detectors, to CCD cameras. In this article, a modern paper web inspection
system, ULMA NTi, which has been developed by ABB Industry Oy, is described.
Arguments for using line scan cameras are considered, and some specific
features of the intel-ligent single sensor camera of ULMA NTi are presented.
Introduction Paper web inspection is necessary for detection and reporting of harmful surface quality deviations such as holes, spots, wrinkles, edge cracks, and coating streaks. Concerning these defects, the customers of paper mills have their requirements, which a paper maker should meet to avoid complaints or returns. In addition, certain defects may cause production problems already at the paper mill e.g. web breaks or damages of soft calender rolls. Therefore, web inspection systems are used, not only for making paper quality reports, but also for identifying and removing the defect causes. In these tasks, advanced defect detection and classification techniques are of benefit. Web inspection system Figure 1 shows the main parts of ULMA NTi, which is a web inspection system developed by ABB Industry Oy, Helsinki. The measuring frame is typically installed in a paper machine’s dry end. The frame consists of a sensor beam with intelligent cameras and a light source beam. Depending on the paper type, transmitting or reflecting measurement, or both, can be used. The light source beam is used to enhance the defect signal strength to an adequate level needed for fast imaging. Depending on the application, either wide spectrum tungsten lamps or fluorescent lamps are used. Tungsten lamps with DC power supply provide strong, flicker-free, and controllable illumination matching well with the sensitivity of CCD cameras. The system adjusts light intensity automatically with varying paper grades, dust build-up, and lamp ageing. These specific tubular lamps have also long lifetime, i.e. on average over 5 years. The PC-based operator station receives defect data from the cameras, produces reports, initiates controls, functions as the user interface, and communicates with other systems. The electronic cabinet contains equipment needed for power supply and interfac-ing with the process. In addition, specific alarm devices, colour markers and pulse encoders are often needed.
Figure 1. Web inspection system CCD cameras for web inspection Today, solid state cameras are most common sensors used in web inspection systems. The light sensitive solid state device of a camera is either a linear or matrix CCD (charge coupled device) array. An array consists of cells (or pixels) that generate electrical charges, which are proportional to the amount of light energy they receive during the exposure. Thus, the cell charges produce an electrical image of the object. A linear CCD chip is “1-dimensional” device because it consists of 512 … 4096 cells in a row. However, a line scan camera utilising such a chip is a natural selection for web inspection, since the second dimension of imaging (the machine direction, MD) is rendered by movement of the web. Continuous illumination and electronic shuttering is used with line scan cameras. Successive line images of the web are taken with very short exposure times. When an exposure is complete, all the cell charges are instantane-ously moved to parallel transfer registers, from which they are read out in serial mode. The next exposure starts immediately. After A/D-conversion, the successive line values are combined by the data processor to make up a complete, continuous image. Another choice for web inspection could be a CCD matrix camera with stroboscopic lighting for freezing the web movement during the exposure. The common analog matrix cameras are off-the-shelf products, which are mainly designed for surveillance or similar applications. Therefore, their analog output signal conforms to the TV-standards of CCIR or EIA, which makes their interfacing with computers simple. However, the standard line frequency (or double of it, if available) limits the read-out rate to a low value. In con-trast, with specialised line scan cameras, higher pixel frequencies are possible providing larger information flow. This also makes it possible to use large linear CCD chips (up to 4096 pixels), when a high resolution is required by the application. This is a remarkable benefit of the line scan cameras, because in contrast e.g. with EIA standard matrix cameras, the CD resolution is only 768 pixels. The MD resolution of line scan cameras is the same or better than those of standard matrix cameras, depending on the speed of the web and the pixel clock frequency. The MD resolution of matrix cameras is often impaired by a factor of two, since to gain speed they are typically used in field mode, where two adjacent lines are com-bined together before shifting out. There are several differences in solid state technologies utilised in CCD arrays. Different types of arrays have been developed also for various application areas. Line scan cameras are designed for industrial or research use, whereas the focus of standard matrix cameras is often on the consumer markets. This difference reflects also on the camera specifications e.g. on photoresponse non-uniformity and dynamic range values. In addition, the CCD matrix arrays are most frequently utilising so-called interline transfer technique, which means that certain parts of the CCD chip surface area are used only for charge transfer, which makes the percentage of the light sensitive area smaller than with line arrays. As a result, also the sensitivity and the definition accuracy of line scan cameras are typically better.
Figure 2. Digital camera of ULMA NTi
Intelligent line scan camera The requirements of real-time inspection on modern paper ma-chines are e.g. demonstrated by the speed of production that can be 300 square metres of paper in a second. The imaging task is allocated to several or tens of cameras, depending on the required resolution. The digitised data flow coming from each camera can be 40 MB/s. It is obvious that each camera requires dedicated processing for its data. If processing involves sophisticated image analysis in real time, a standard PC is not powerful enough for the task, and accordingly, a different solution has to be found. The solution developed for the ULMA NTi is an intelligent multi-function line scan camera. The considerations presented in the previous chapter have among other things contributed to the decision to rely on the line scan principle. The camera is based on modern digital signal processing and processor technology utilis-ing DSPs and PLDs. It is also called the digital single sensor camera, since it combines hole, spot, subtle defect, and streak detection into one camera architecture. An advantage of this special camera is that in it a desired number of truly parallel computations and algorithms can be run at a high speed. This is in contrast to PCs and other conventional computers, which make their calculations mostly in serial mode and in addition suffer from the overhead of an operating system. The camera is depicted in Figure 2. The object (the paper surface) is imaged on the CCD chip by a standard lens. The signal-processing unit makes 12-bit A/D-conversion and subsequent processing of CCD output signal. The unit contains several elec-tronic boards with different tasks. The boards operate in parallel to get as high throughput as possible. PLDs are used for high speed detection and classification tasks. PLD is a programmable logic array, which contains tens of thousands gates that can be connected by software to perform various serial and parallel computations at the speed of hardware. However, algorithm modifications are easy to realise by downloading a new “soft wiring”. A DSP in the camera is used to extract continuous flaws like streaks and scratches in the paper web. The processing is done in parallel with the handling of other defects, which enables all the flaws to be processed in a single camera. Digital integration with 16-bit resolution makes it possible to detect very subtle coating streaks. In addition, a RISC processor
is contained in the camera to process the data after defect detection by
the PLD. Its tasks include defect shape and size extraction, skipping the
defects thaare smaller than a selected minimum size, formation index calculation,
and recog-nition and elimination of known repeating patterns (watermarks
etc.) in the sheet.
Single-PC solution Using intelligent cameras means that the processing power of the inspection system is distributed where it is most needed. The basic decisions concerning the defect detection and classification are made in the cameras. The results of all cameras are sent to a PC, which combines the information to cover the whole width of the web. An intelligent camera interface board in the PC makes all the real time decisions concerning process I/O signals. The PC also hosts an operator interface, reporting and net communication tasks. One PC is capable to perform these tasks due to the distribution of front-end processing into the cameras, Accordingly, the system is also a single-PC solution. Additional PCs are installed only if extra process stations or data base servers are needed. Figure 3 shows a layout of web inspection system with several operator stations and data base servers.
Contact information Juhani Rauhamaa
.
|
