The versatile use of 3-D measuring systems improves productivityPhotogrammetric true 3-D measuring systems improve quality and productivity.Introduction Generally, photogrammetry is the art, science and technology of obtaining reliable geometric information about physical objects and the environment through the process of recording, measuring, and interpreting photographic images. Within photogrammetry, accurate 3-D measuring has been studied over 100 years. Primarily the general means used for topographic mapping, photogrammetry has already proven its merit even in extreme machine vision applications. At present, real-time photogrammetry is becoming increasingly important in industrial machine vision. Most non-contact 3-D measuring systems found in the automotive and other industries are not true 3-D. Measurement of the third dimension is based on a priori information, like the knowll distances to a conveyor belt or the known sizes of the object details. This makes the systems quite inflexible. Very small changes in production may require substantial reprogramming or a totally new system. In recent years, the rapid development of digital imaging and image processing techniques has led to close-range photogrammetric solutions becoming a part of the most effective machine vision systems. The continuing rapid evolution oI modern microelectronics has developed these older single camera concepts into more general and sophisticated ones, which fully exploit photogrammetry through the simultaneous operation of several cameras. Consequently, it is now possible to apply true optical 3-D measuring systems that provide a high degree of flexibility with regard to production changes. The 3-D data obtained can be utilized by robots, milling machines, and process and quality control systems. Basically the same 3-D measuring machine concept can be used in all stages of production, from the design studio to the conveyor belt. The technical background of modern 3-D measuring technology The basic component of a modern optical 3-D measuring system is an electronic camera, which can be a standard or special video camera. All the geometric information necessary for the three-dimensional determination of object details is primarily stored in the twodimensional images. A true on-line 3-D vision system needs at least two cameras. The three space co-ordinates of object details are derived by triangulation from the overlapping images produced by the cameras. Machine vision is not restricted to stereo vision. Human vision is based on a stereo approach with two imaging components (eyes). Machine vision can include any number of cameras set at arbitrary angles to each other. The actual transformation parameters for real-time photogrammetric systems are determined by system calibration. After the system is set up, a set of control points with accurately known positions or accurately known mutual distances is first identified and located on the images. The two-dimensional image co-ordinates are then used for the spatial resection of each image, after which the transformation parameters are solved with no a priori information on the cameras or their orientations. Essentially, the mathematical model of the resection must also cope with all the systematic geometric distortions of the information flow during the projection. After calibration of the system, it is possible to locate every object space detail derived from images and viewed by at least two cameras at a time. This is done by an intersection, which is a reverse and simultaneous transformation of the two-dimensional image co-ordinate observations to the three-dimensional object space co-ordinate system. The transformation parameters are those determined during system calibration. The details to be measured on the object or in the object space during the measuring stage are ordinary object features like bolt holes, comers, edges, etc. Details to be measured can also be first signalized with a light spot or with specific targets. Photogrammetry as a measuring technique is well suited to demands for great precision. The relative accuracy of real-time photogrammetric systems is mainly limited by image resolution. Using standard black and white video cameras, the accuracy rate is presently 1:20,000, i.e. 0.1 mm in 2 metres. Almost any level of accuracy level can be achieved by using a sufficient number of cameras, which may have a high resolution. Design criteria for an industrial system To be practical, an industrial 3-D measuring system should be designed according to the following rules.
Though there are many applications in the automotive industry, we will only deal with two of them, glass and tool measuring, here. Compared to the basic measuring
system, the glass measuring system also includes a scanner and an ultraviolet
(UV) laser light source. A UV light spot produced by the light source is
reflected onto the glass surface by scanner mirrors. The UV light generates
a visible fluorescent spot on the surface, which can be seen by the cameras
and measured threedimensionally. (Figures 1 and 2).
 The scanner can be used to project any number of spots or stripes onto the glass surface, providing great flexibility in measuring. Thanks to its quick feedback to the process control, the system effectively minimizes production losses and increases throughput. There will be less need for manual measuring and any damaged moulds will be easier to trace. The original 3-D co-ordinates of the glass are measured from the original sample glass before the actual on-line measuring. This procedure, carried out at the measuring station, takes about five minutes per glass model.  For on-line process control, the convex glass is transferred to the measuring station. First of all, a positioning measurement is made, to locate the actual co-ordinates of the glass. The cameras then take the necessary images of the glass sheet. Next, the system calculates the actual 3-D position of the glass with reference to the nominal position. A UV light spot is moved to the chosen nominal points for automatic camera measurement, thus eliminating the need for a mechanical positioning device. As a result, the camera measurement provides a chosen number of 3-D points. The nominal points can be defined in many ways, for instance directly from the CAD-data of the glass. This system is also self monitoring. After the initial calibration, all subsequent calibrations will be carried out automatically. If a good quality UV laser is used, the repeatability of the measurements is 0,1 mm (3 6). Repeatability is here the main factor to be studied, since the basic measuring platform provides sophisticated and easy calibration procedure to compensate systematic errors. Consequently, a glass can be measured in a global co-ordinate system providing global precision values or in a local co-ordinate system. The latter option gives only relative measuring values compared to some nominal value. However, this kind of information is also valuable in process control, as it reveals possible process trends or other malfunctions. The technology described earlier is also suitable for dimensional checks on automotive tools. Measurements can be carried out in two ways. In the direct method, the tool is scanned periodically by a light spot during production. In addition, each product made by the tool can be checked, with the results of the measurement of the products also indirectly indicating the condition of the tool itself. Combinations of both methods can also be used. For design studio purposes,
a non-contact method is greatly superior to other methods, since the materials
are soft in the early stage of design. Photogrammetric methods allow the
same measuring system to be used throughout the design process, from soft
to hard materials, at reasonable cost.
|
