Optics: measuring and testing – By configuration comparison – With comparison to master – desired shape – or reference voltage
Patent
1986-07-16
1988-11-01
Laroche, Eugene R.
Optics: measuring and testing
By configuration comparison
With comparison to master, desired shape, or reference voltage
358107, G01B 1100, H04N 718
Patent
active
047814632
DESCRIPTION:
BRIEF SUMMARY
This invention relates to a method and apparatus for use in measurement. The invention has particular, though not exclusive, application in the measurement of flat objects or flat surfaces of bulky objects when viewed by means of a digitizing television-type system. It is possible to embody the invention in such a way that it may be used, in one application, to monitor the size and shape of objects passing down a production line viewed by a television camera, from above.
There is a growing requirement for automated image analysis and in many applications this involves checking the size and shape of an object constituting an image. Checks of this sort are suitable for computer-control and, to present the image in a form suitable for analysis by computer, the image must be digitized. The digitization may be effected by taking the image acquired by a television camera or similar device and dividing it into picture elements (hereafter called pixels). The pixels are arranged ideally in some regular pattern, of which rectangular or hexagonal arrays are the most common, though others are possible. In an embodiment to be described, a rectangular array of pixels is employed, but the arrangement can readily be modified to apply to other geometries to define a measuring grid or lattice.
The resolution available in a digitized image depends to some extent on the density of the pixel array. The most common systems use square arrays of 256.times.256 or 512.times.512 pixels, but some have rectangular windows or square arrays of up to 4096.times.4096 pixels; however, with normally practicable lenses and cameras it is doubtful whether the true optical resolution is better than can be achieved with a 512.times.512 pixel raster. Since, in the ideal case of perfect contrast between an object and background on the field of view, any particular pixel would be either "on" or "off", i.e. register as being either inside or outside the object, it may be considered that the resolution of a digitized picture is directly proportional to the inverse of the pixel density and that, for example, the length of an object detected by a 512.times.512 pixel array could be determined to only about 1 part in 500. This conclusion results from the practice of determining lengths by a point-to-point calculation with the positions of the two points in question each carrying, in this instance, an error of up to unit inter-pixel distance.
It will be seen from the particular embodiments of the present invention to be described that it is possible to determine the position of a line or edge in a digitized picture to a much greater degree of accuracy than plus or minus one unit of distance (the inter-pixel distance) by the use of at least one of three procedures:
(a) by employing many points to make a statistical analysis of the position of the line,
(b) in the case of a straight line (or edge), by positioning the line so that it makes a very small angle with one of the coordinate axes and
(c) by repeating the measurement an appropriate number of times (e.g. 10 times) with small relative displacements of the camera, the object or the measuring grid or lattice (equivalent to fractions of the inter-pixel distance) between successive measurements.
The distance between two straight lines or edges, which might be the length or breadth of a rectangular object in a digitized picture, can then be obtained as the difference between their respective positions within an error which is a compound of the two separate small errors. When the positions of lines or edges are determined in one of these ways, secondary features such as angles or positions of corners can be determined to a high degree of accuracy. Furthermore, the calculations of line position can be simplified by making use of the distinction in an image between the lattice, on the points of which the pixels are centered, and a grid, which exists between the pixels. This simplification applies particularly to positioning curved lines or edges. In the case of a straight line or edge it is straightforward to
REFERENCES:
patent: 4238780 (1980-12-01), Doemens
patent: 4379308 (1983-04-01), Kosmowski et al.
patent: 4575751 (1986-03-01), Duschl
patent: 4613749 (1986-09-01), Vali
Microelectronics and Reliability, vol. 15, pp. 613 to 618, Pergamon Press, 1976, GB, "The Alignment of Graphic Images in Solid-State Technology", Z. M. Wojcik.
Pattern Recognition, vol. 16, No. 4, pp. 413-420, 1983, G.B., "Subpixel Edge Estimations", P. D. Hyde and L. S. Davis.
Lee John W.
Rosen Dennis
Birkbeck College
LaRoche Eugene R.
Mis David
LandOfFree
Method and apparatus for use in measurement of the position of a 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 and apparatus for use in measurement of the position of a, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for use in measurement of the position of a will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-887167