CMM 3_7 PROGRESS_Layout 1 05/11/2010 15:58 Page 45 feature METROLOGY Don Braggins, UK Industrial Vision Association INDUSTRIAL VISION AND MICRO MANUFACTURING MAIN IMAGE: Figure 4. Effects on an image of varying the angle of incident illumination, courtesy of STEMMER IMAGING. M achine vision is now a well-established tool on the production line for making critical dimensional measurements as part of the quality control process in both high volume and low volume manufacturing, and provides a real return on investment. Quantifiable measurements that can be obtained from an image include distances, angles, centre of mass, area and circular fit, to name but a few. Industrial vision measurements can also be directly linked into statistical process control (SPC) methods to improve product quality, reduce wastage, improve productivity and streamline the process. The speed and accuracy of current industrial vision systems means that in many applications 100% inspection can be carried out, and each and every product or component can be measured. By feeding these data into the SPC system, not only can trends be identified earlier, but random and sudden defects can also be identified, which is not possible using more traditional sampling methods for SPC data acquisition. The challenge for using industrial vision in micro manufacturing is being able to make measurements at the required scale in a production line environment. As with any measurement system, the key factors of accuracy, precision and repeatability are of vital importance in machine vision systems. Accuracy is an indication of how close the actual measurement is to true value. Precision is the number of digits to which the measurement can be read. Repeatability shows the closeness of a number of repeated measurements. This is illustrated in Figure 1, where it can be seen that a group of measurements could have poor accuracy and poor repeatability, or good repeatability but cmm 3/7 www.micromanu.com 45 CMM 3_7 PROGRESS_Layout 1 05/11/2010 15:58 Page 47 feature METROLOGY Fifth Anniversary of the Calibration Laboratory for Laser Power Meters: http://bit.ly/lasercomponents Figure 3. Telecentric imaging, courtesy of STEMMER IMAGING. poor accuracy, or good accuracy but poor repeatability, as well as the desired combination of good accuracy and good repeatability. Precision is generally limited by the repeatability and accuracy characteristics. Figure 1. Measurement accuracy and repeatability, courtesy of STEMMER IMAGING. Figure 2. Fundamental Imaging parameters, courtesy of National Instruments. Factors affecting accuracy and repeatability in machine vision A number of factors affect accuracy and repeatability in machine vision systems, including the camera, optics, illumination, object positioning and measurement algorithms. Fundamental imaging parameters are field of view, working distance, resolution, depth of field and sensor size (Figure 2). Measurements are only as good as the information that can be extracted from the image, and image quality is dependent on resolution, contrast, depth of field, perspective and distortion. Resolution is the amount of object detail reproduced by the imaging system and contrast is the difference between the object and background greyscale values. Machine vision measuring systems specify accuracy and repeatability in terms of physical units, such as microns, provided that the image field-of-view (the dimensions of the area seen by the camera) and the number of elements in the camera’s image sensor are known and fixed. This calibration procedure defines the measurement dimension for each individual pixel in the camera sensor and is dependent not only on the number of pixels in the sensor, but on the quality of the lens used as well. For micro manufacturing applications, even with optimised lens/sensor combinations, the measurement dimension of the individual pixel may still be too large for the measurements required, however, measurement algorithms which use interpolation and fitting techniques to allow measurements to be made to a fraction of a pixel can be used to improve the measurement range. There is a wide choice of sensor resolutions available allowing the most appropriate choice for the dimensions being measured, however the lenses used in the optical system are also of critical importance. A perfect lens would fully reproduce an image from an object with absolutely no degradation. However sharpness, contrast, illumination, spectral transmission and distortion all affect the ability of a lens to reproduce an image. A lens can only resolve so much detail in terms of spatial frequency — the finer the image detail passing through the lens, the harder it is for the lens to reproduce a good image; the image becomes less distinguishable. Another consideration is the performance of the lens across the entire field of view. For some lenses, the resolution capability deteriorates the further away from the centre axis, which is an important consideration if the resolution is required across the entire image or alternatively means that the object must be accurately positioned in the centre of the field of view. Depending on the size of the object and the distance to the camera, there may also be issues with depth of field or off-axis distortion. Telecentric imaging A telecentric lens is a compound lens used in machine vision systems to eliminate dimensional and geometric variations of images within a range of cmm 3/7 www.micromanu.com 47 CMM 3_7 PROGRESS_Layout 1 05/11/2010 15:59 Page 49 feature METROLOGY Figure 5. Inspection of solar cell manufacture, courtesy of STEMMER IMAGING. different distances from the lens and across the whole field of view, by collimating the light entering the lens. Telecentric lenses have the same magnification at all distances. An object-space telecentric lens creates images of the same size for objects at any distance and has constant angle of view across the entire field of view. An object that is too close or too far from the lens may still be out of focus, but the resulting blurred image will be the same size as the correctly-focused image would be. This makes them ideal for metrology applications, especially in the automotive and electronics industries, for high accuracy measurement of parts with complicated three-dimensional shapes. Figure 3 shows an image of a thick drilled block using a conventional lens and telecentric lens. The conventional lens shows the walls of the holes, making measurement of the diameters difficult. With the telecentric lens, the walls are not seen and the diameters can readily be measured. Off-axis imaging In many applications it is not possible to position the imaging system so that it is mounted perpendicular to the object being measured. If the imaging system has to be mounted off-axis, then distortion and foreshortening effects can result and these must be corrected in order to ensure accurate measurements. Illumination Machine vision illumination controls how the object appears to the camera. Adequate illumination can often make the difference between a system that works reliably and one that does not. After all, if the appropriate image is not presented to the camera for measurement, considerations of accuracy and repeatability become largely irrelevant. At the most basic level, there must be ‘enough’ light so that the camera can acquire a good image. Beyond this, it is almost always necessary to use the orientation, geometry or colour of light to highlight relevant details or minimise the appearance of unhelpful parts of the image, such as glare. Probably the most important factor that governs how the image appears is the angle at which the light falls on the object (Figure 4). For example, light approaching an object from a low angle will tend to create highlights on raised edges. Choosing the appropriate wavelength of illumination can play a major role in revealing or masking specific features on the object, while the ‘quality’ of the light describes whether the light is diffuse or not. Lighting control is also important. For example, the light can be strobed or pulsed, the intensity can be raised or lowered or, in a more complex set-up, different lighting scenarios can be preprogrammed using a lighting controller. Speed Another important factor to be considered is that of speed. Once all of the dimensional measurement criteria have been met, the system needs to be able to make the measurements at the appropriate speed for the production line. In addition to the ability to acquire and process the images at the appropriate rate, it also means that the imaging system needs to be interfaced into the production line, with camera triggers linked into product delivery system, and if necessary into a reject mechanism to allow ‘out of spec’ components to be removed. Real world measurements Three practical examples of the use of industrial vision systems in micro manufacturing applications are illustrated here. The first is the use of a vision system in the manufacture of solar cells. ECKELMANN AG, located in Wiesbaden, Germany, has developed a vision system based on line scan cameras for laser edge detection as part of the edge isolation process in solar cell manufacture (Figure 5). It was designed for the ASYS Group cmm 3/7 www.micromanu.com 49 CMM 3_7 PROGRESS_Layout 1 05/11/2010 15:59 Page 51 Floyd Exhibits at Metalex 2010 in Bangkok: http://bit.ly/floydbangkok (Dornstadt, Germany), a leading manufacturer of handling systems, process machines and special machines for the electronic and solar industries, and is fully integrated to provide feedback control to the production process. Edge isolation provides electrical separation between the active front side of a solar cell and the rear side. A laser cuts a small groove along the cell edges, the depth of the groove depending on the cell doping. The difficulty lies in positioning the groove as close as possible to the outer contour of the cell in order to maximise the active surface and thus the efficiency. The edge isolation control system features a line scan camera with 4096 pixels, optics and customised LED illumination supplied by STEMMER IMAGING. The image processing system measures the outer contours of the cell and feeds them back to the control system of the laser equipped with a deflection mirror to provide an active feedback system. If the edge damage is within tolerance levels the laser will ignore it and proceed with the cutting process. Image acquisition and analysis take place in just 800 ms and the resolution of the system makes it possible to ensure that the distance to the edge during laser cutting is below 100 µm. The calibration and qualification of the laser and camera have been automated so the system can easily be commissioned or recalibrated after maintenance work. The second application uses vision to help with handling components in the micro assembly and testing industry. Since these components are tiny, almost weightless, and highly sensitive to electrostatic charge, inspecting and sorting them ready for assembly can be a painstaking task. Vision has been incorporated into a system that combines the functions of feeding, orientation and inspection of parts. IMS (Almelo, The Netherlands) develops and builds high-tech production equipment for the high-precision, electronics and medical industry. Working together with the University of Twente (Enschede, The Netherlands) and Bosch Rexroth (Lorh am Main, Germany), they developed the Vision Inspection Feeding System feature METROLOGY (ProVIS), a multi-purpose, modular system for the supply and recognition, inspection, handling and placing of parts. Recognition and inspection, ProVIS (Figure 6) uses two separate cameras, one for recognition, and another for inspection. The Matrox Imaging Library software development kit performs all the product recognition and inspection tasks. To use the system, a technician calibrates the ProVIS with a part that is within tolerance to create what’s known as the Golden Template. Then the camera takes pictures of the parts on the inspection stage. Finally, specific processing modules analyse the parts. First the Geometric Model Finder (GMF) module locates the parts in the image, so the Metrology module can measure the features of each part. The results, both good and bad parts, are displayed on the monitor. Parts that pass inspection can be used for assembly; parts that cannot be recognised are most likely lying on their sides or too close to another part, so they are re-fed into the system by a vibratory tray. If the inspection shows a part to be out of tolerance, the system tags it; if the system is feeding parts for assembly, the non-conforming parts will be kept out of the assembly step. The system can also be programmed to find surface defects. The Metrology module figures prominently in the solution, and is used for finding dimensions and checking tolerances, complex operations that are processing-intensive. With the appropriate optical system the measurement results are accurate to +/- 0.01 mm. Without appropriate lighting, the camera is unable to produce usable images. The ProVIS features a dome with blue light above the tray where the products are fed to the system. The inspection of the parts’ dimensions, the fundamental task of the system, is backlit. The right combination of illumination and zoom lenses provides the accuracy needed for such tiny parts. As with most assembly applications time is important, and all the visual inspections and data processing has to be completed within the allowed cycle Figure 6. ProVIS inspection system, courtesy of Matrox Imaging. cmm 3/7 www.micromanu.com 51 CMM 3_7 PROGRESS_Layout 1 05/11/2010 15:59 Page 52 feature METROLOGY Figure 7. Inspection of SMDs, courtesy of Cognex UK. time, typically 1 to 5 seconds, depending on the complexity of the inspection operation. The final application also makes use of another machine vision capability — that of optical character recognition to ensure that the correct batch number is marked on a product. At Micro Crystal in Switzerland, a smart camera from Cognex is used to inspect the packaging of ceramic surface-mounted devices (SMDs) which contain the quartz oscillator used in watches and mobile phones. A smart camera is one in which the image processing is carried out within the camera itself, rather than transferring the image to a PC for processing. These SMDs are sold on rolls of polyester tape in quantities up to 16,000. SMD packages can be as small as 2 x 1.2 mm. The company has been using an Optical Tape End-Controller on this production line. Similar to a film-cutting table, the roll of uninspected SMD tape is guided over a worktable where the smart camera performs an automatic optical inspection (Figure 7). The inspection process checks the following criteria: 52 cmm 3/7 www.micromanu.com • That there is an SMD in the package. • That the position of the SMD is correct. • That the batch number is present on the ceramic housing of the SMD and that it can be read perfectly. The image processing is achieved using pattern-matching technology which locates objects reliably even if they are of different sizes, differently aligned, if their appearance is poor or even if they are partly covered. By analysing the geometrical information, it is able to determine the position of the object clearly. LED lighting is used to give the optimum illumination to ensure reliable detection of the laser marked batch numbers, even though they are not always sharply contoured. According to the packaging specifications of Micro Crystal, the gold contact surfaces of the ceramic housing must always be on the side facing away from the camera, and so they should be invisible to the vision system. If the smart camera detects fluctuations in brightness triggered by the gold areas on the grey ceramic surface, the Optical Tape EndController sounds the alarm. The faulty section is then moved to a pre-determined processing point and removed by hand. About the UKIVA The UKIVA is a Special Interest Group of the PPMA (Processing & Packaging Machinery Association), and its prime objective is promoting the use of vision by manufacturing industry. The Association’s members are involved in the supply of vision systems and components for use in a wide range of industrial imaging applications. Thanks are due to UKIVA members Cognex UK (www.cognex.co.uk), Matrox Imaging (www.matrox.com), National Instruments (www.ni.com) and STEMMER IMAGING (www.stemmerimaging.co.uk) for their contributions to this article. www.ukiva.org