Optics: measuring and testing – Shape or surface configuration
Reexamination Certificate
2000-06-21
2003-04-01
Dang, Hung Xuan (Department: 2873)
Optics: measuring and testing
Shape or surface configuration
C356S603000, C356S607000
Reexamination Certificate
active
06542250
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention, in general, relates to a method of taking three-dimensional measurements of an object surface and, more particularly, to a method of projecting patterns onto an object surface and of defining the object coordinates by analyzing the similarity of the grey scale gradation in images of the object surface taken by at least two matrix cameras.
2. The Prior Art
Methods heretofore known of optically measuring three-dimensional objects at close range are, for instance, photogrammetric methods or stripe-projection methods. Whereas in the stripe-projection method striped patterns which are geometrically as exact as possible are projected onto the object to be measured, the patterns used in photogrammetric methods may be more or less arbitrary.
The projector for projecting patterns onto an object surface will hereinafter sometimes be referred to a pattern projector. The device for generating an image of the object surface and the sensor for recording the intensity distribution, pixel by pixel, in the image will hereafter sometimes be referred to as matrix camera. More specifically, in the present context matrix cameras are understood to include CCD cameras and other apparatus suitable for recording an intensity distribution in a pixel by pixel manner. A point in the sub-pixel range is to understood as a position on the (camera) sensor which is positioned between pixel centers. Since it is generally assumed that the measured grey scale values are valid only for the pixel centers, the grey scale value for a point in the sub-pixel range is interpolated from surrounding pixels. For instance, bilinear interpolation is one method by which the grey value may be calculated on the basis of the surrounding four pixels. The position on the sensor of the second matrix camera, hereinafter sometimes referred to as the corresponding point, may be determined relative to a selected pixel of the sensor of a first camera by projecting the same portion of the surface of the body to be measured on the selected pixel and on the corresponding point.
In the simplest case, the measurement structure for practicing a stripe-projection method consists of a pattern projector for projecting a striped pattern, a matrix camera of the kind disclosed, for instance, by German patent specification 3,938,714, and the object to be measured. The form or shape of the stripes projected onto the surface and affected in a determined manner by the object, is recorded by the sensor and may be evaluated by a computer connected therewith.
The so-called phase shift method is a special method which ensures relatively high accuracy of the measured 3D-coordinates. Essential aspects of the evaluation of the intensity distributions recorded by the matrix cameras relate to defining the absolute order of the individual stripes as well as the phase position of each image point. For the latter, N≧3 individual exposures of the object surface are sequentially taken, with the periodic striped pattern projected onto the object surface being shifted by an nth part of the stripe period between each exposure. Accordingly, a set of N individual images is obtained for each image point.
However, measuring coordinates in this manner suffers from a disadvantage inherent in the phase-shift method, i.e. the need for calibrating the pattern projector. This is because of the principle which requires defining the 3D coordinates by means of the matrix camera and the projector, i.e. by triangulation generally. Accordingly, the projector has to perform a function which is usually assumed by the camera. In general it is, however, more difficult or complex to calibrate a projector than it is to calibrate a camera. Moreover, the distortion parameters are generally larger and cannot be as easily determined as in the case of a camera. This, in turn, may result in a greater uncertainty in the calculation of the spatial coordinates.
A further problem arises from the fact that oftentimes the grey scale gradation of the stripes is not sinusoidal in a way which may be considered to be ideal. Any such deviations may have a direct distorting effect on the results of the measurement and lead to a lower accuracy of the measurements.
In photogrammetric methods, the calculation of the coordinates of the surface is based upon images of the object surface taken by at least two cameras. One advantage of photogrammetry is its high measuring accuracy. Yet where only one pair of images is used, measurements may become critical at or near strong curvatures or other irregularities in the object surface. The reason for this is that the position of corresponding image points is usually determined by analyzing similarities in the grey scale gradation of two camera images within a defined measuring area of but a few, e.g. 15×15, pixels. Usually, it is the maximum of the correlation function or the minimum of the sum of the square of errors of these two grey scale gradations which is used as a measure of the similarities.
Accordingly, the accuracy of measurements obtained by this method is strongly dependent upon the size of the measured surface and critical surfaces, i.e. strongly curved or highly irregular surfaces generally cause systematic measurement errors. While this disadvantage could be lowered by reducing the size of the measurement surface, it would result in a reduced number of pixels for analyzing the grey scale gradation and, therefore, in a less exact position of the corresponding points determined by an analysis of the similarities. Accordingly, seeking to solve the problem in this manner would seem at best to be of limited usefulness.
An alternative measuring method has been disclosed by German Patent specification DE 196 23 172 in which the correlation surface consists of but one pixel. The method can be realized by initially projecting a sequence of n (for instance 12) different patterns onto an object surface and by recording a corresponding number of image pairs. Accordingly, a sequence of n grey values becomes available of a pixel selected by one camera. The correlation coefficient between this sequence of grey values and the sequence of grey values of a point of the other camera is drawn upon to determine the corresponding point. The algorithm is based upon a search for the correlation maximum along the epipolar line. The method suffers from the drawback that it is not possible within the coordinate system of the object to measure either equidistantly or along predetermined directional vectors within the object coordinate system.
OBJECTS OF THE INVENTION
Accordingly, it is an object of the invention to provide for equidistant measurements in an object coordinate system in a method of three-dimensionally measuring object surfaces in which patterns are projected onto the object surface and of determining the object coordinates by analyzing the similarity of the grey scale gradation in images of the object surface recorded by at least two cameras.
Another object is to provide for measuring along directional vectors in an object coordinate system in the context of a method of the kind referred to supra.
BRIEF SUMMARY OF THE PREFERRED EMBODIMENTS
In the accomplishment of these and other objects, the invention in a currently preferred embodiment thereof provides for a method of three-dimensionally measuring an object surface in which predetermined patterns are projected onto the object surface and spatial coordinates of measuring points on the object surface are determined by analysis of the similarity of the grey scale gradation in images of the object surface recorded by at least two matrix cameras, and including the steps of:
a) each matrix camera generates a set of N sequential images whose pattern is changed by predetermined values between individual exposures,
b) a first approximation point is determined for calculating a first measuring point,
c) coordinates in the sensor surfaces of the two matrix cameras are correlated by mathematical projection to the approximation point or with a number
Albrecht Peter
Michaelis Bernd
Dang Hung Xuan
Hormann Karl
LandOfFree
Method of three-dimensionally measuring object surfaces 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 of three-dimensionally measuring object surfaces, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of three-dimensionally measuring object surfaces will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3050874