Reference mark, method for recognizing reference marks and method for object measuring

• US 6,917,720 B1
• Filed: 07/04/1997
• Issued: 07/12/2005
• Est. Priority Date: 07/04/1997
• Status: Expired due to Fees

• Current Assignee: Daimler Chrysler Company LLC (Stellantis NV)
• Original Assignee: Daimler Chrysler Company LLC (Stellantis NV)
• Inventors: Michaelis, Martin, Caesar, Torsten
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US09/445,914
• Time in Patent Office: 2,930 Days
• Field of Search: 382/181, 382/183, 382/287, 382/289, 382/291, 382/165, 382/205, 235/462.09, 235/462.11, 235/462.41, 235/494
• US Class Current: 382/287
• CPC Class Codes: G06V 10/245 by locating a pattern; Special marks for positioning

Abstract

The invention relates to a reference mark, a method for recognizing reference marks and a method for object measuring. According to the invention, the object is fitted with coded reference marks. Said marks are automatically recognized and decoded by the inventive method for reference mark recognition, whereupon a three-dimensional representation of the object is realized. According to the invention, the method compensates automatically for geometrical and radiometric interferences and generates at least one quality measurement to evaluate the quality of recognition and decoding. The invention particularly enables fast recognition and decoding of large numbers of reference marks and fast measuring of objects.

Claims

1. A process for recognition of reference marks for surveying of objects, each reference mark comprised of at least one survey character, at least one orientation character, and multiple encoded characters, comprising:

forming a digital image comprised of pixels of a supposed or actual reference mark using electromagnetic waves in the optical or an adjacent wavelength range,

automatically compensating for geometric distortions effecting the recognizability of the reference mark with the aid of the orientation character,

determining the pixel coordinates of the center of the supposed or actual reference mark within their digital image,

automatically compensating for radiometric disturbances,

automatically compensating for disturbances due to partial covering,

defining at least one quality measurement for the recognition of the reference mark,

performing at least one reassessment check for positive recognition of an actual reference mark or for rejection of a supposed reference mark on the basis of said at least one quality measurement, and

extracting from at least one of said encoded characters of an actual reference mark a characteristic character vector, which is supplied to a classification system, and

that the reference marks being recognized are comprised of spatially separated survey, orientation and encoded characters.

2. A process according to claim 1,

wherein the digital image of the supposed or actual reference mark is locally converted to binary data prior to said steps of compensation for disturbing influences, checking of at least one rejection criteria, and extraction of the character vector of an actual reference mark.

3. A process according to claim 2,

wherein said local conversion to binary data occurs by a process comprising:

determining, within a predetermined neighborhood of the pixel coordinates of the center of the supposed or actual reference mark, the maximal and the minimal color-value or gray-value of the pixels,

assigning each pixel within this neighborhood, of which the color value or gray value is closer to the maximal color value or gray value of all pixels within this neighborhood than to the minimal color value or gray value of all pixels within this neighborhood, or of which the color value or gray value lies precisely in the middle of these two values, to a group of pixels with high color value or gray value, and

assigning each pixel within this neighborhood, of which the color value or gray value is closer to the minimal color value or gray value of all pixels within this neighborhood than to the maximal color value or gray value of all pixels within this neighborhood, to a group of pixels with low color value or gray value,

determining the average of all color or gray values of all pixels of the group of pixels with high color or gray value and weighting the average with the number of the elements of the group,

determining the average of all color or gray values of all pixels of the group of pixels with low color or gray value and weighting the average with the number of elements of the group,

forming an average value from these two weighted average values, which is used as the threshold in the binary conversion,

assigning to each pixel within this neighborhood of which the color or gray value is greater than the threshold or equal to the threshold a binary value of white, and

assigning to each pixel within this neighborhood of which the color or gray value is below the threshold receives a binary value of black.

4. Process according to claim 1, comprising:

providing the surveying character, orientation character, and encoded characters of the reference mark in known, fixed numbers,

arranging the surveying characters, orientation characters, and encoded characters in known, fixed geometric relationships relative to each other,

identifying congruency areas within the digital image of the supposed or actual reference mark, and

making a decision on the basis of these congruency areas as to whether the object under examination is an erroneous or an actual reference mark, wherein the decision, as to whether the object being examined is an erroneous or actual reference mark, is made on the basis of the congruency areas in which it is checked whether the individual found congruency areas exhibit the predetermined minimum number, and exist in known geometric relation to each other within a predetermined tolerance range, and

rejecting a mark as erroneous upon assessment of a congruity as an interference, and

in the case of an assessment of the congruous area as an actual reference mark, making an unequivocal assignment for each individual congruous area, in such a manner, that it is recognized which individual congruous area is bearer of surveying characters and/or orientation characters and/or encoded characters.

5. Process according to claim 4,

wherein said surveying characters, orientation characters, and encoded characters are arranged in in known, fixed separation and positions relative to each other, and

wherein, in the decision, as to whether the object being examined is an erroneous or actual reference mark, it is checked whether the individual found congruency areas exhibit the predetermined minimum number, and exist in known geometric relation to each other within a predetermined, freely pre-selected, tolerance range.

6. A process according to claim 5, wherein the identification of the congruous areas occurs by a process comprising:

in the case of color value objects, performing a congruency analysis of the color value objects of the digital image of the supposed or actual reference mark, wherein all pixels of the digital image, of which the color-value lies within a predetermined value range, are assigned to a color value object,

in the case of gray value objects, performing a congruency analysis of gray value objects of the digital image of the supposed or actual reference mark, wherein all pixels of the digital image, of which the gray value lies within a predetermined value range, are assigned to a gray value object, and

in the case of black-and/or-white objects, performing a congruency analysis of the black or the white objects of the binary converted digital image of the supposed or actual reference mark.

7. A process according to claim 5, wherein the identification of the congruency areas occurs by a process comprising:

in the case of color-value objects, performing a congruency analysis of color-value objects of the digital image of the supposed or actual reference mark, wherein all pixels of the digital image are assigned to a color value object via a classification of color values, optionally after taking into consideration the color values in a local neighborhood,

in the case of gray value objects, performing a congruency analysis of gray value objects of the digital image of the supposed or actual reference mark, wherein all pixels of the digital image are assigned to a gray value object by means of a classification of the gray value, optionally after taking into consideration the gray value in a local neighborhood,

in the case of black-and-white objects, performing a congruency analysis of black and white objects of the binary converted digital image of the supposed or actual reference mark.

8. A process according to claim 3, further comprising:

forming a simplified representation of the detected congruency areas, preferably a contour representation using polygons—

a polygon representation,

optionally replacing the original polygon representation with an approximate polygon representation—

polygon approximation, wherein said polygon approximation is optionally prepared by exclusively extracting support points of the original polygon of the contour representation and continuing this until a predetermined maximal tolerable area deviation is exceeded.

9. A process according to claim 4, comprising:

identifying, from the number of the found congruency areas of an actual reference mark, a congruency area as the bearer of orientation characters via the recognition of the geometric relation of the various congruency areas to each other,

determining the actual orientation of the reference mark from this orientation character,

carrying out a standardization or normalization to a fixed, pre-determined standard or uniform orientation, preferably a normalization of the digital image of the reference mark or the congruent areas thereof, preferably a normalization of the representation of the congruent areas,

identifying from among the number of detected congruent areas at least one congruent area as bearer of encoded characters via the recognition of the geometric relation of the various congruent areas to each other, and

standardizing or normalizing each congruent area, which was identified as bearer of encoded characters, in size and plotting it onto a n-by-n grid, preferably a 16-by-16 grid, and

optionally subjecting each congruency area, which was identified as bearer of encoded characters, to contrast enhancement by the scaling of the color value or gray value of the individual grid elements of the congruent area over a maximal color value or gray value range, preferably over a gray value range of 0-255,

using each color value or gray value of a grid element of a congruent area, which was identified as bearer of coded characters, as a component of an n-by-n dimensional, preferably 256 dimensional, character vector of a congruent area,

subjecting this character vector to a main axis transformation in such a manner that the individual components of the transformed character vector are provided in arranged sequence according to their value or weight for a classification, and

submitting a limited number of components of the transformed character vector, counted beginning with the first and most important component, preferably the first 40 components, to an individual object classification, preferably an individual character or individual symbol classification, preferably a one-step quadratic polynomial classification,

using this classification to calculate a probability vector, of which the component number corresponds to the number of the possible encoded objects, preferably equal to the number of the possible encoded figures or encoded symbols, and of which the components indicate the probability of the identification of a particular object class, preferably the probability of the identification of a certain figure or symbol class, and

presuming that the object class, preferably figure or symbol class, of which the identification occurred with the highest probability, is valid, so long as this probability exceeds a predetermined minimal value, and

rejecting the reference marks when no object class, preferably no figure or symbol class, is identified with a probability greater than the predetermined minimal probability.

10. A process according to claim 1, wherein:

the reference marks to be recognized exhibit

surveying characters,

orientation characters including features extending in at least two directions, preferably in directions which are perpendicular to each other, and

encoded characters in the form of a fixed number of figures and/or symbols, and

the surveying characters, orientation characters, and encoded characters are provided in known, fixed number, and

the surveying characters, orientation characters, and encoded characters have a known, fixed geometric relationship to each other, preferably a known, fixed separation and relative position to each other, and are present in a known, fixed size relationship to each other.

11. A process according to claim 1, wherein:

the surveying character is in the form of a circular area,

the orientation character is formed in a U-shaped design, comprised of two side parts and a connecting segment perpendicular to the side parts and preferably significantly longer,

the encoded characters are designed in the shape of

three side-by-side numbers of a stripeless font, or

four side-by-side figures selected from the 10 numbers and the 26 capital letters of a stripeless font, and

the survey character, orientation character, and encoded characters are provided relative to each other in a known, fixed geometric relation in such a manner,

that the survey mark in the shape of a circular area is situated centrally within the U-shaped orientation character, and

that the encoded characters in the form of a fixed number of known figures are provided below the connecting segment of the U in a known, fixed separation and in a known, fixed size relationship thereto and side-by-side.

12. A process according to claim 10, comprising:

performing a search for congruency areas in the neighborhood surrounding the known pixel coordinates of the center of a supposed or actual reference mark,

defining, for each detected congruency area of a supposed or actual reference mark, a rectangle which is the smallest, axially parallel encompassing rectangle of the respective congruency area,

identifying the smallest congruency area, within the encompassing rectangle of which the known pixel coordinates of the center of the supposed or actual reference mark are situated, as the congruency area, which contains the survey mark,

identifying the next largest congruency area, within the encompassing rectangle of which the known pixel coordinates of the center of the supposed or actual reference mark are located, as the congruency area which contains the orientation character, and

making a decision, on the basis of the encompassing rectangles, whether the object being examined is a supposed or actual reference mark, said decision making comprising checking whether the geometric relationship of the two rectangles to each other is within a predetermined tolerance range, preferably in a known size relationship to each other, or the one is enclosed by the other,

rejecting the reference mark, when this tolerance range is exceeded,

determining the orientation of the mark, when this tolerance range is not exceeded, said determination of the orientation comprising,

producing a polygon approximation of the orientation character, and

determining angle of the orientation character relative to the main axis of the pixel-coordinate system by

plotting the individual polygon segments of the polygon of the polygon approximation with their lengths being weighted or averaged in an angle histogram,

determining the two highest amplitudes of this angle histogram,

assigning the highest of these two amplitudes to the connecting segment of the U, and

assigning of the second highest of these two amplitudes to the side segments of the U, or

determining the angle of the orientation character relative to the main axis of the pixel-coordinate system by

assigning the longest polygon segment as the connecting segment of the U,

assigning all polygon segments within a predetermined angle tolerance range to the connecting segment of the U,

calculating the weighted center of the angle of this polygon segment,

assigning the longest polygon segment which is not assigned to the connecting segment of the U as the side segment of the U,

assigning all other polygon segments within a predetermined angle tolerance range to the side segments of the U, and

calculating or determining of the weighted center or mean of the angle of these polygon segments, and

calculating, from the recognized angle of the connecting segment of the U relative to the main axis of the pixel-coordinate system, the rotation angle of the reference character and carrying out a rotation normalization of the image of the reference mark,

calculating, from the recognition of the angle of the side segment of the U relative to the main axis of the pixel-coordinate system, the supposed shear angle of the reference mark, and from this supposed shear angle, by means of recognition of the rotation angle, calculating the actual shear angle, and carrying out a shear normalization or correction of the image of the reference mark,

checking the congruency areas for the encoded characters in the rotation and shear normalized image of the reference mark, to verify that they are found in a known quantity below the connecting segment of the U, and that their size differences lie within a predetermined, preferably freely pre-selected, tolerance range, and

rejecting the reference mark upon non-satisfaction of this verification criteria.

13. A process for surveying an object, comprising:

providing the object to be surveyed with reference marks containing survey characters, orientation characters, and encoded characters, and

producing a digital image comprised of pixels of the object provided with reference marks, using electro-magnetic waves from the optical wave-length or an adjacent wave-length range,

identifying potential reference marks, which may be either supposed or actual reference marks, in the digital image of the object comprised of pixels, and determining the pixel-coordinates of the center of these potential reference marks,

evaluating reference marks by employing a process for recognition of reference marks for surveying of objects, each reference mark comprised of at least one survey character, at least one orientation character, and multiple encoded characters, comprising;

forming a digital image comprised of pixels of a supposed or actual reference mark using electro-magnetic waves in the optical or an adjacent wavelength range,

automatically compensating for geometric distortions effecting the recognizability of the reference mark with the aid of the orientation character,

determining the pixel coordinates of the center of the supposed or actual reference mark within their digital image,

automatically compensating for radiometric disturbances,

automatically compensating for disturbances due to partial covering,

defining at least one quality measurement for the recognition of the reference mark,

performing at least one reassessment check for positive recognition of an actual reference mark or for rejection of a supposed reference mark on the basis of said at least one quality measurement, and

extracting from at least one of said encoded characters of an actual reference mark a characteristic character vector, which is supplied to a classification system, and

that the reference marks being recognized are comprised of spatially separated survey, orientation and encoded characters

wherein the evaluation of the reference mark is carried out locally within the neighborhood surrounding the pixel coordinates of the center of these potential reference marks, and

placing the pixel coordinates of the individually recognized reference marks in reference to a 3-D coordinate system of the object to be surveyed, optionally using a 3-D correspondence analysis.