Method for character-and/or-object recognition

• US 5,033,101 A
• Filed: 06/17/1988
• Issued: 07/16/1991
• Est. Priority Date: 06/20/1987
• Status: Expired due to Fees

• Current Assignee: Ralf-A Sood
• Original Assignee: Ralf-A Sood
• Inventors: Sood, Ralf-A.
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): Mancuso, Joseph
• Application Number:
  US07/208,217
• Time in Patent Office: 1,124 Days
• Field of Search: 382/30, 382/33, 382/34, 382/9, 382/10, 382/8, 382/42, 382/38, 382/25
• US Class Current: 382/209
• CPC Class Codes: G06V 30/2504 Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches

Abstract

Unknown characters and objects are read opto-electronically and the read signals are digitized to generate a binary image of the characters and objects, said image being compared with binary reference images of known reference characters and reference objects stored in reference memories, the reference raster image(s) with the highest coincidence with the read characters or objects being selected to identify the read-in characters or objects. The stored binary reference images form a first binary comparison plane. At least one more binary comparison plane is provided, which consists of stored reference raster images generated by the binary imaging of characteristics or of characteristics parts of the known characters and objects. The actual character or object image read-in in binary manner is compared with at least one reference raster image of a further comparison plane for coincidence if the comparison with the reference raster image of the first comparison plane, or of the first comparison plane combined with a further comparison plane, results in a degree of coincidence below a predetermined threshold.

Claims

1. A method for recognizing objects, comprising the steps of:

(a) opto-electronically reading the objects;

(b) digitizing the read signals to generate a binary image of the objects;

(c) comparing the objects with a first binary comparison plane formed of known stored reference raster images of objects;

(d) determining a first correlation value and comparing with a predetermined threshold value;

(e) selecting the reference raster images where the first correlation value exceeds the threshold value to identify the read actual objects;

(f) if the correlation value in step (d) is below the threshold value, comparing the read objects with a second binary comparison plane formed of stored reference raster images of characteristic parts of the reference objects;

(g) determining a second correlation value and combining with the first correlation value and comparing the resulting combined correlation value with the predetermined threshold value;

(h) selecting the reference raster images where the combined correlation value exceeds the threshold value to identify the read actual objects;

(i) if the combined correlation value in step (g) is below the threshold value, comparing the read objects with at least one further binary comparison plane formed of stored reference raster images of other characteristic parts of the reference objects;

(j) determining at least one further correlation value and combining with the first and second correlation values and comparing the resulting combined correlation value with the predetermined threshold value; and

(k) selecting the reference raster images where the combined correlation value from step (j) exceeds the threshold value to identify the read actual objects.

2. A method as in claim 1, and including the step of:

(a) associating the binary images of the features of the known reference objects with weighting areas of predetermined bit numbers.

3. A method as in claim 1, and including the step of:

(a) selecting between 55% and 100% degree of coincidence as the predetermined threshold value for identification.

4. A method as in claim 3, and including the steps of:

(a) comparing with at least one next-best reference objects and determining a correlation value;

(b) determining the difference between the first correlation value from the next-best correlation value; and

(c) comparing the difference with a predetermined minimum value.

5. A method as in claim 4, and including the step of:

(a) choosing a value between 30% and 100% of the deviation from identity as the predetermined minimum value.

6. A method as in claim 1, and including the step of:

(a) comparing pixel-wise the reference objects reproduced in pixels and the read objects imaged in pixels for coincidence; and

(b) counting the number of coincidences or deviations from the comparison results.

7. A method as in claim 1, and including the step of:

(a) terminating the comparison with the reference raster images when the predetermined degree of coincidence has been reached.

8. A method as in claim 1, and including the steps of:

(a) prior to testing for the degree of coincidence, determining the height and the width of the read actual object imaged in binary manner and comparing with the height and the width of the binary reference raster images; and

(b) proceeding with the testing with those reference raster images for which the comparison lies within a predetermined range of tolerances.

9. A method as in claim 8, and including the step of:

(a) choosing the least area determined by the height and width of the mutually compared actual objects and reference objects as the weighting area.

10. A method as in claim 8, and including the step of:

(a) choosing the largest area determined by the height and width of the mutually compared actual objects and reference objects as the weighting area.

11. A method as in claim 10, and including the steps of:

(a) scanning pixel-wise the outer contour of each object;

(b) entering the scanning results into a table having a number of lines depending on the height of the objects; and

(c) copying the objects according to the tabulated values into a target zone for further processing for recognition.

12. A method as in claim 11, and including the step of:

(a) erasing the read objects imaged in binary form according to the tabulated values after copying.

13. A method as in claim 11, and including the step of:

(a) scanning the outer contour pixel only in one direction of right-rotary or left-rotary.

14. A method as in claim 11, and including the steps of:

(a) processing first the highest projecting object that is first contacted during scanning of the binary image over the entire width of the x-axis; and

(b) processing the objects to the left or right of the first processed object.