Method for determining attributes using neural network and fuzzy logic

• US 5,655,031 A
• Filed: 06/16/1995
• Issued: 08/05/1997
• Est. Priority Date: 06/16/1994
• Status: Expired due to Fees

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Ishii, Shoichi, Teshima, Mitsutaka, Yukawa, Noriaki
• Primary Examiner(s): Razavi, Michael T.
• Assistant Examiner(s): NGUYEN, HA T
• Application Number:
  US08/491,070
• Time in Patent Office: 781 Days
• Field of Search: 382/181, 382/159, 382/206, 382/174, 382/194, 382/156, 382/190, 382/192, 382/161, 382/193, 382/199, 382/200
• US Class Current: 382/194
• CPC Class Codes: G06V 10/751 Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching

Abstract

An attribute decision method includes steps of making a comparison between one feature value, selected from seven feature values, of an input pattern contained in an input image and membership functions determined from 37 standard patterns including alphanumeric characters of "0" to "9" and "A" to "Z" and a hyphen of "-" with respect to the selected feature value and then obtaining output values of the membership functions for the individual standard patterns with respect to the input pattern. The seven feature values are, when the input pattern is a mesh pattern, vertical structure vector sums, horizontal structure vector sums, up-down and left-right area differences, and a vertical cross number. The method also includes steps of calculating a total value of the output value for the feature value for each of the standard patterns, and determining a standard pattern of the highest total value among the calculated total values and thus deciding one of the standard patterns which has the highest similarity to the input pattern.

Claims

1. An attribute decision method comprising steps of:

making a comparison between at least one feature value, selected from seven feature values, of an input pattern contained in an input image and membership functions determined from 37 standard patterns including alphanumeric characters of "0" to "9" and "A" to "Z" and a hyphen of "-" with respect to the at least one selected feature value and then obtaining output values of the membership functions for the individual standard patterns with respect to the input pattern,

the seven feature values being, when the input pattern is a mesh pattern, vertical structure vector sums which are obtained by determining a number of white meshes counted along each column top to bottom and bottom to top until a first black mesh is encountered in data of the mesh pattern and by summing up their results,

horizontal structure vector sums which are obtained by determining a number of white meshes counted along each row left to right and right to left until a first black mesh is encountered in the data of the mesh pattern and by summing up their results,

up-down and left-right area differences which are obtained by determining differences in areas of black meshes between upper and lower halves and between left and right halves in the data of the mesh pattern, and

a vertical cross number which is obtained by determining a number of times of crossing with black meshes counted along a center of columns in a vertical direction;

calculating a total value of the output value for the at least one feature value for each of the standard patterns; and

determining a standard pattern of the highest total value among the calculated total values and thus deciding one of the standard patterns which has the highest similarity to the input pattern.

2. The attribute decision method as claimed in claim 1, further comprising, before the comparison making step, steps of:

previously deciding one of the standard patterns which has the highest similarity by using a neural network with respect to the input pattern contained in the input image; and

selecting a plurality of upper-order standard patterns by comparing results of the preceding step with a threshold value so that only the selected standard patterns are used in the comparison making step.

3. The attribute decision method as claimed in claim 2, wherein decision results of the neural network are added to the output value or values of the feature value or values to obtain the total value in the calculating step.

4. The attribute decision method as claimed in claim 1, wherein the membership function to be created from the at least one selected feature value of each of the standard patterns is so set that a high value is outputted when the at least one feature value is inputted, and low values are outputted for values before and after the at least one feature value.

5. The attribute decision method as claimed in claim 2, wherein the membership function to be created from the at least one selected feature value of each of the standard patterns is so set that a high value is outputted when the at least one feature value is inputted, and low values are outputted for values before and after the at least one feature value.

6. The attribute decision method as claimed in claim 1, wherein when plural feature values are selected in the comparison making step, in the calculating step, the total value is obtained by adding to each other the output values of the feature values for each of the standard patterns.

7. The attribute decision method as claimed in claim 1, wherein four of the seven feature values are used which are the vertical structure vector sums and the horizontal structure vector sums.

8. The attribute decision method as claimed in claim 2, wherein four of the seven feature values are used which are the vertical structure vector sums and the horizontal structure vector sums.

9. An attribute decision method comprising steps of:

making a comparison between at least one feature value, selected from seven feature values, of an input pattern contained in an input image and membership functions determined from 26 standard patterns including alphabetic characters of "A" to "Z" with respect to the at least one selected feature value and then obtaining output values of the membership functions for the individual standard patterns with respect to the input pattern,

the seven feature values being, when the input pattern is a mesh pattern, vertical structure vector sums which are obtained by determining a number of white meshes counted along each column top to bottom and bottom to top until a first black mesh is encountered in data of the mesh pattern and by summing up their results,

horizontal structure vector sums which are obtained by determining a number of white meshes counted along each row left to right and right to left until a first black mesh is encountered in the data of the mesh pattern and by summing up their results,

up-down and left-right area differences which are obtained by determining differences in areas of black meshes between upper and lower halves and between left and right halves in the data of the mesh pattern, and

a vertical cross number which is obtained by determining a number of times of crossing with black meshes counted along a center of columns in a vertical direction;

calculating a total value of the output value for the at least one feature value for each of the standard patterns; and

determining a standard pattern of the highest total value among the calculated total values and thus deciding one of the standard patterns which has the highest similarity to the input pattern.

10. The attribute decision method as claimed in claim 9, further comprising, before the comparison making step, steps of:

previously deciding one of the standard patterns which has the highest similarity by using a neural network with respect to the input pattern contained in the input image; and

selecting a plurality of upper-order standard patterns by comparing results of the preceding step with a threshold value so that only the selected standard patterns are used in the comparison making step.

11. The attribute decision method as claimed in claim 10, wherein decision results of the neural network are added to the output value or values of the feature value or values to obtain the total value in the calculating step.

12. The attribute decision method as claimed in claim 9, wherein the membership function to be created from the at least one selected feature value of each of the standard patterns is so set that a high value is outputted when the at least one feature value is inputted, and low values are outputted for values before and after the at least one feature value.

13. The attribute decision method as claimed in claim 11, wherein the membership function to be created from the at least one selected feature value of each of the standard patterns is so set that a high value is outputted when the at least one feature value is inputted, and low values are outputted for values before and after the at least one feature value.

14. The attribute decision method as claimed in claim 9, wherein when plural feature values are selected in the comparison making step, in the calculating step, the total value is obtained by adding to each other the output values of the feature values for each of the standard patterns.

15. The attribute decision method as claimed in claim 9, wherein four of the seven feature values are used which are the vertical structure vector sums and the horizontal structure vector sums.

16. The attribute decision method as claimed in claim 10, wherein four of the seven feature values are used which are the vertical structure vector sums and the horizontal structure vector sums.

17. The attribute decision method as claimed in claim 16, further comprising, before the comparison making step, steps of:

previously deciding one of the standard patterns which has the highest similarity by using a neural network with respect to the input pattern contained in the input image; and

selecting a plurality of upper-order standard patterns by comparing results of the preceding step with a threshold value so that only the selected standard patterns are used in the comparison making step.

18. An attribute decision method comprising steps of:

making a comparison between at least one feature value, selected from seven feature values, of an input pattern contained in an input image and membership functions determined from 10 standard patterns including numerical characters of "1" to "0" with respect to the at least one selected feature value and then obtaining output values of the membership functions for the individual standard patterns with respect to the input pattern,

the seven feature values being, when the input pattern is a mesh pattern, vertical structure vector sums which are obtained by determining a number of white meshes counted along each column top to bottom and bottom to top until a first black mesh is encountered in data of the mesh pattern and by summing up their results,

horizontal structure vector sums which are obtained by determining a number of white meshes counted along each row left to right and right to left until a first black mesh is encountered in the data of the mesh pattern and by summing up their results,

up-down and left-right area differences which are obtained by determining differences in areas of black meshes between upper and lower halves and between left and right halves in the data of the mesh pattern, and

a vertical cross number which is obtained by determining a number of times of crossing with black meshes counted along a center of columns in a vertical direction;

calculating a total value of the output value for the at least one feature value for each of the standard patterns; and

determining a standard pattern of the highest total value among the calculated total values and thus deciding one of the standard patterns which has the highest similarity to the input pattern.

19. The attribute decision method as claimed in claim 18, further comprising, before the comparison making step, steps of:

previously deciding one of the standard patterns which has the highest similarity by using a neural network with respect to the input pattern contained in the input image; and

selecting a plurality of upper-order standard patterns by comparing results of the preceding step with a threshold value so that only the selected standard patterns are used in the comparison making step.

20. The attribute decision method as claimed in claim 19, wherein decision results of the neural network are added to the output value or values of the feature value or values to obtain the total value in the calculating step.

21. The attribute decision method as claimed in claim 18, wherein the membership function to be created from the at least one selected feature value of each of the standard patterns is so set that a high value is outputted when the at least one feature value is inputted, and low values are outputted for values before and after the at least one feature value.

22. The attribute decision method as claimed in claim 19, wherein the membership function to be created from the at least one selected feature value of each of the standard patterns is so set that a high value is outputted when the at least one feature value is inputted, and low values are outputted for values before and after the at least one feature value.

23. The attribute decision method as claimed in claim 18, wherein when plural feature values are selected in the comparison making step, in the calculating step, the total value is obtained by adding to each other the output values of the feature values for each of the standard patterns.

24. The attribute decision method as claimed in claim 18, wherein four of the seven feature values are used which are the vertical structure vector sums and the horizontal structure vector sums.

25. The attribute decision method as claimed in claim 19, wherein four of the seven feature values are used which are the vertical structure vector sums and the horizontal structure vector sums.

26. The attribute decision method as claimed in claim 25, further comprising, before the comparison making step, steps of:

previously deciding one of the standard patterns which has the highest similarity by using a neural network with respect to the input pattern contained in the input image; and

selecting a plurality of upper-order standard patterns by comparing results of the preceding step with a threshold value so that only the selected standard patterns are used in the comparison making step.