Optics: measuring and testing – Shape or surface configuration
Reexamination Certificate
2002-03-20
2003-07-08
Font, Frank G. (Department: 2877)
Optics: measuring and testing
Shape or surface configuration
C358S003050, C358S461000
Reexamination Certificate
active
06590669
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to a stereo optical method for recording the shape of three-dimensional objects.
In many industrial production fields and in the multimedia applications field, methods are now required in order to convert geometric, three-dimensional data relating to an object to numerical data on a computer, by means of suitable measurement devices. This may be done for quality control purposes, or else with the aim of displaying these objects realistically and three-dimensionally to a human observer. There is also a large amount of interest in recording objects automatically, and of transmitting them and visualizing them using the Internet.
Optical methods for recording the shape of objects are increasingly replacing the previously used mechanical scanning methods. A major advantage of the optical methods is that the measurement does not involve any contact and thus has no mechanical influence on the object. A further advantage is that a large number of object points can be recorded simultaneously, thus shortening the measurement time.
Known shape recording methods are generally based on the triangulation principle, the stereo principle or interferometric techniques.
In one known triangular method, a light point is projected onto the surface of the object to be measured, and is observed from a direction other than the illumination direction. The coordinates of the illuminated points can be calculated from the knowledge of the orientation of the projection beam in three dimensions, and of the direction from which the observed point is perceived. Although the method is accurate and unambiguous, it is slow, however, since the surface of the object to be measured must be scanned point-by-point. Furthermore, the only points on the surface which can be recorded are those which are visible directly both from the location of the light source and from an observing camera. A data record containing such a measurement is thus never complete. Although a number of data records can be obtained by repeated measurements using different observation and illumination perspectives, it is necessary, however, in order to record the shape of the object in its totality, to relate these data records to one another geometrically (matching), and this frequently still requires action by a human user. Furthermore, during matching, the interfaces between the data records also have an unpleasant appearance since the individual data records can rarely be made to coincide perfectly. Edges and sudden changes can occur as artifacts, which not only adversely affect the accuracy of the data but, in particular, also cause visual disturbance to a viewer. The human eye is able to identify even very small projections and indentations in the surface of a visualized or actual object. In addition to the position of a point in three dimensions, people can also deduce the inclination of the surface from the illumination conditions. Even minor variations in position can cause a major change in the inclination, as a result of which even very small irregularities are susceptible to a human observer. This is generally a fundamental problem in most methods for three-dimensional recording of shapes. In most cases, the recording of measured data is not matched to this situation, so that even a small amount of noise in the data has a very disturbing effect on the viewer. This also applies to the known methods described in the following text.
Further-developed methods based on triangulation include the light section technique and strip projection. In the former, a line is projected onto the surface of the object to be measured, rather than an individual point. This line is observed from a direction that is different from the illumination direction. The three-dimensional coordinates of the illuminated points are obtained in the same way as that mentioned above. Although this method is faster than point-by-point triangulation, it is, however, still slower than other methods which can record an entire surface in one step. In this case as well, a number of measurements are required from different perspectives, which are then matched in order to produce a complete representation of the object.
Strip projection is a further development of the light section technique, in which a number of lines are projected simultaneously onto the surface of the object to be measured. The intensity of these lines varies cyclically in the lateral direction, and makes it possible for the observation camera to distinguish between the individual lines. Although the method is fast, it is also necessary to join a number of measurements together by matching, so that the edges and sudden changes mentioned above can also occur here.
Interferometric methods are frequently used for high-precision measurements. These methods are also subject to the problem that the results of a number of individual measurements must be jointed together in order to produce a complete three-dimensional representation of the object being measured. Furthermore, these methods are very sensitive to very minor vibration, and can generally be used only in laboratory conditions.
A further group of methods is based on the stereo principle. These make use of the fact that two views of an object which have been recorded from different viewing angles contain information about the three-dimensional shape. These are referred to as binocular stereo methods. Software algorithms are used to identify corresponding features of the object in the two records. The different position of the feature in the two images represents a measure of the depth of the feature in three-dimensional space. The main object of binocular stereo is to determine the correspondence between features. One method is to compare small image details with one another on the basis of their brightness structure. Two difficulties occur in the process. If image details have no significant brightness structures, they cannot be associated with one another. This means that the three-dimensional depth of object points can be determined only in structured areas of the object. Furthermore, the brightness, to be more precise the light intensity, of an object is not the same from different viewing angles. This can likewise lead to it being impossible to determine depth.
The binocular stereo principle can be extended from two views to a number of views. This provides further information and makes the correspondence analysis process more reliable, but in many cases this is still not sufficient.
A further group of stereo methods uses different illumination conditions to determine the shape of objects. In contrast to the binocular stereo method, the viewing angle remains fixed, and the illumination direction is varied. This is thus referred to as a photometric stereo method. The brightness levels from the individual lighting directions make it possible to deduce the inclination of the object surface. In this case, a variable which forms the derivative of the three-dimensional depth is measured rather than the three-dimensional depth itself. Photometric stereo methods are highly suitable for measuring local object structures, but global structural measurements are still subject to errors. A global object structure can be established better by using a method which measures the three-dimensional depth itself, that is to say, for example, a binocular stereo method.
The methods mentioned above thus have the disadvantage that it is not always possible to uniquely associate associated image points on different images with one another. This is referred to as the correspondence problem.
SUMMARY OF THE INVENTION
The object of the invention is thus to specify a method for optical recording of shapes, in which the correspondence problem is at least very largely overcome.
This object is achieved by a method described below. This method allows both global and local object structures to be recorded accurately. These two principles are normally also used to a major extent in the way that humans viewing an object detect its shape.
Font Frank G.
Nguyen Sang H.
Ostrolenk Faber Gerb & Soffen, LLP
LandOfFree
Method for optically detecting the shape of objects does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for optically detecting the shape of objects, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for optically detecting the shape of objects will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3101461