JPS63163678A - Picture automatic collating device - Google Patents

Abstract

PURPOSE:To attain correct collation even with a notch or noise or the like to two objects to be collated by extracting each collating object of two objects based on a common part of a picture of the two objects after the picture of the two objects is aligned so as to attain the collation processing based on the objects of both collations. CONSTITUTION:Alignment is applied by the various known alignment system as to the two objects to be inputted in the alignment means 11. Even when either of the two objects has a notch or noise, the two objects are aligned. Moreover, a collation object extracting means 12 extracts the collation object from each object based on the common part of the picture of the two objects to be aligned. Thus, noise caused in the two objects is eliminated and the part common to two objects is extracted as the collated object. Then the collation means 13 applies collation processing to each collation object with the noise eliminated. Thus, even with the notch or noise in the two objects to be collated, the accuracy of collation is not lowered and correct collation is applied.

Description

[Detailed Description of the Invention] [Summary] In an automatic image matching device for two objects, after performing alignment on the images of the two objects, each of the two objects is compared based on the common portion of the images of the two objects. A target is extracted, and a matching process is performed based on these two matching targets. As a result, even if the two objects to be matched have defects, noise, etc., accurate matching can be performed.

[Industrial application field]

The present invention is an automatic image matching device that has been improved so as to be able to perform a good match even if there is noise or chipping in two objects to be matched. The present invention relates to an automatic image matching device suitable for automatically matching objects that should be matched.

[Conventional technology]

When comparing two images, usually the registered images registered in advance in a dictionary etc. and the target a, U aW
A method is used that compares and matches the I+ image using matching processing such as pattern matching.

At that time, if there is a shift in the relative position of the recognized image and the registered image, especially if there is a difference in the rotation angle, matching is difficult. Therefore, in general, the two objects to be matched are aligned in pre-processing prior to matching, and both images are aligned. Correct verification results can be obtained by performing verification in a state where there is no difference in rotation angle between them.

FIG. 5 shows a seal matching device which is an example of an image matching device in which two objects to be matched are aligned in pre-processing prior to matching.

In FIG. 5, 21 is a scanner that reads the seal imprint to be verified. A first image memory 22 stores the verification seal imprint read by the scanner 21.

Reference numeral 23 is a registered seal imprint database (registered seal imprint DB), which stores registered seal imprints that serve as standards for comparison. Reference numeral 24 is a database search m (DB search section), which searches one of the registered seal impressions from the registered seal imprint DB section 23. search for. A second image memory 25 stores the registered seal imprint searched by the DB search unit 24.

Reference numeral 26 denotes a positioning unit that performs a process of positioning the verification seal imprint with the registered seal imprint. A third image memory 27 stores the aligned verification seal imprint.

Reference numeral 28 denotes a line width correction unit that performs processing to match the line width of the aligned verification seal impression with the registered seal impression. Reference numeral 29 denotes a fourth image memory in which a verification seal impression with line widths matched is stored. Reference numeral 30 denotes a collation unit that performs a collation process between the verified seal imprint and the registered seal imprint.

In this configuration, the scanner 21 reads the seal imprint to be verified and stores it in the first image memory 22. On the other hand, the DB search unit 24 searches for one of the registered seal imprints from the registered seal imprint DB 23 and stores it in the second image memory 25.

The positioning unit 26 searches the first image memory 22 for matching seal impressions. Each of the registered seal impressions is read out from the second image memory 25, the verification seal impression is aligned with the registered seal impression so that the deviation and rotation angle difference between both the seal impressions are eliminated, and the aligned verification seal impression is stored in the third image memory.

The method of detecting the rotation angle difference between both seal impressions when aligning is: ■ Rotating one of the seal impressions by a minute angle, calculating the degree of similarity between the two impressions each time, until the degree of similarity is maximized. A method in which the rotation angle of both seal impressions is the difference in rotation angle between the two seal impressions. ■ Find the angle in the direction of density change for each of the two seal impressions, generate a histogram of the angle, and calculate each histogram of both seal impressions in the direction of each angle axis. A method of comparing the two histograms while shifting them relatively and determining the rotation angle difference between the two seal impressions from the amount of shift when the two histograms are most similar (for example,
0-22281).

The method (2) has better processing efficiency than the method (2), and allows accurate positioning even if there are chips or noise in the verification seal impression.

When generating a seal impression, if the pressing force is uneven, the thickness of the verification seal impression will be uneven and the verification accuracy will be reduced. Therefore,
The line width correction unit 28 performs a process of correcting the uneven line width of the verification seal impression into a uniform line width, and storing the verification seal impression whose line width has been corrected in the fourth image memory.

The matching unit 30 reads out the matching seal impression whose line width has been corrected from the fourth image memory 29, matches it with the registered seal impression in the second image memory 25, and calculates the degree of matching.

[Problem that the invention seeks to solve]

In conventional image matching devices, if there is a positional shift between the matching seal impression and the registered seal impression, or if there is non-uniformity in line width, these are corrected as described above to perform the matching. and,
If the verification seal imprint was chipped or had noise that could not be easily removed, it was possible to complete the process up to positioning.

However, if either of the seal impressions is missing or if there is noise attached to a part of the seal impression, it is assumed that it is the shape of the seal impression and the comparison is performed, so it is not possible to determine the correct degree of matching. First, there was a problem in that the degree of similarity between seal impressions originally obtained from the same seal was reduced, and correct comparison could not be performed.

This problem is not limited to seal imprint matching, but generally occurs when one of two images to be matched has a chip or noise.

SUMMARY OF THE INVENTION An object of the present invention is to provide an improved automatic image matching device that can perform correct matching even if one of two images to be matched has a chip or noise.

[Means for solving problems]

The solution proposed by the present invention will be explained with reference to FIG. FIG. 1 is a block diagram showing the basic configuration of the present invention.

In FIG. 1, reference numeral 11 denotes a positioning means that performs positioning of two objects to be compared on the images.

Reference numeral 12 denotes a matching target extracting means, which extracts matching targets from the two objects based on the common parts of the images of the two objects that have been aligned.

Reference numeral 13 denotes a matching means that performs matching processing based on each extracted matching target.

[For production]

The alignment means 11 aligns the two input objects using various known alignment methods. As a result, even if one of the two objects has a chip or noise, the two objects can be aligned.

The matching target extracting means 12 extracts the 2
A matching target is extracted from each target based on the common parts of the target images. As a result, noise generated in the two objects is removed, and a portion common to the two objects is extracted as a matching object.

The matching means 13 performs matching processing on each of the matching objects from which noise has been removed.

By doing the above, even if there are defects or noises in the two objects to be matched, the matching accuracy will not be reduced and correct matching can be performed.

〔Example〕

Embodiments of the present invention will be described with reference to FIGS. 2 to 4. FIG. 2 is an explanatory diagram of the configuration of one embodiment of the present invention, FIG. 3 is an explanatory diagram of the detailed configuration of the verification target extraction means of the same embodiment, and FIG. 4 is a diagram of the verification seal imprint, registered seal imprint, and FIG. 3 is an explanatory diagram of a mask pattern.

An embodiment of the present invention will be described below, taking as an example a case where a verification seal impression and a registered seal impression are two objects to be verified.

(A) Configuration of Embodiment In FIG. 2, the alignment means 11, matching target extraction means 12, and matching means 13 are as described in FIG. 1.

Reference numeral 14 denotes a scanning means, which includes a scanner 141 for reading the seal impression to be verified and a first image memory 142 in which the read verification seal impression C3° is stored.

15 is a database search means (DB search means), which includes a registered seal imprint database (registered seal imprint DB) 151 in which registered seal imprints to be compared with verification seal imprints are stored;
DB search unit) 152, and a second image memory 153 in which the searched registered seal imprint R3o is stored.

In the positioning means 11, reference numeral 111 is a positioning processing unit which registers the verification seal impression C8° using various known methods.
3o to generate a verification seal imprint C81. A third image memory 112 stores the aligned verification seal imprint CS+.

In the verification target extraction means 12, reference numerals 120 to 127 are verification target extraction units, which have the configuration shown in detail in FIG. Seal impressions C32 and R3 to be checked are extracted from the seal impressions. A fourth image memory 128 stores the extracted verification seal impression C32. A fifth image memory 129 stores the extracted registered seal impression R8l.

In the verification means 13, 131 is a line width correction circuit which registers the line width of the verification seal imprint C82 using various known methods.
Processing is performed to match the line width of +, and a verification seal impression C83 is generated. Reference numeral 132 denotes a sixth image memory in which a verification seal imprint C33 with matched line widths is stored.

Reference numeral 133 is a collation unit that collates the collation seal imprint C33 whose line widths have been matched with the registered seal imprint R3, and calculates the degree of collation.

In the detailed diagram of the verification target extraction means 12 shown in FIG. 3, 120 is a DEL AND circuit, which assumes that the density value of the pixel of the verification seal imprint CS + and the registered seal imprint R3° whose density value is 1 or more is 1. A first mask pattern M P +, which is an initial mask pattern, is generated by performing a logical product of both seal impressions. 1
21 is a first mask memory in which the generated first mask pattern M P I is stored.

122 is an expansion circuit that performs processing to expand the first mask image M P + to generate a second mask pattern MP2. A second mask memory 123 stores the generated second mask pattern MP.

A comparison circuit 124 determines whether the second mask pattern M R2 includes the verification seal impression C81 or the registered seal impression R3o.

Reference numeral 125 denotes a transfer circuit which transfers the second mask pattern MP,2 to the first mask memory 121 in response to a control signal CC issued by the comparison circuit 124 when it is determined that the second mask pattern MP2 does not include the verification seal impression O8+ or the registered seal impression R3o.
Transfer and store.

126 is a first masking circuit which converts the verification seal impression C51 into a second mask pattern M P2 which is the final mask pattern.
By masking, a verification seal impression C82 is generated.

A second masking circuit 127 generates a registered seal impression R81 by masking the registered seal impression R5° with a second mask pattern M P 2 as a final mask pattern.

(B) Operation of the Embodiment The operation of the embodiment will be explained with reference to FIG. 4, which is an explanatory diagram of the verification seal imprint, registered seal imprint, and mask pattern.

In FIG. 4, (A) is an explanatory diagram of each verification stamp C80 to C83, and (B) is an explanatory diagram of each mask pattern MP1 and MP2.
FIG. 3(c) is an explanatory diagram of each registered seal imprint.

Note that each seal imprint shows only the portion where the density value of the surrounding circles is 1 or more, and illustration of the characters inside is omitted.

Below, the verification seal imprint to be verified is (a) in Figure 4 (A).
, ) is the verification seal impression C3o, and its registered seal impression is the fourth one.
An embodiment of the present invention will be described by taking as an example the case of the registered seal impression R3o shown in (c) of FIG.

The scanner 141 reads the verification seal impression, and stores the verification stamp C3o shown in (al) in FIG. 4(A) in the first pixel memory 142.
Store in. At the illustrated verification stamp C8°, the dotted line LC
The portion DF shown by is a missing portion, and N1 and N2 are noises generated when stamping. Because the pressing force is uneven, the line width of the seal impression is uneven, and the verification seal impression C3o is
As shown in the figure, there is a rotation angle difference of θ with respect to the registered seal impression R3o.

On the other hand, the DB search unit 152 reads one registered seal imprint R3 from the registered seal imprint DB 151 and stores it in the second pixel memory 153. The registered seal impression R3o is (c1) in Fig. 4(c).
As shown, it is a circular frame with uniform line width. In addition, the dotted line LC
DF' on the right side is the portion corresponding to the portion DF where the missing portion of the verification seal imprint C8.

The alignment processing unit 111 aligns the verification seal imprint C8o read from the first pixel memory 142 with the registered seal imprint R3o read from the second pixel memory 153, and generates a registered seal imprint C8, which eliminates the rotation angle difference θ. and stored in the third pixel memory 112 ((c2) in FIG. 4(A)).

The first AND circuit 120 of the verification target extraction unit calculates the density of each pixel of the verification seal impression C3 read out from the third pixel memory 112 and the registered seal impression R3 read out from the second pixel memory 153 whose density range is 1 or more. The first mask pattern MP, which is an initial mask pattern, is generated by assuming that the value is 1 and taking the logical product of both seal impressions, and the first mask pattern MP is stored in the first mask memory 12.
1 ((b+) in FIG. 4(B)). As a result, as shown in the figure, a first mask pattern M P l is generated which is made up of a common portion of the verification seal impression C81 and the registered seal impression R3 and does not include the noises N and N2.

The expansion circuit 122 sets the first mask pattern M P l to 1
The second mask pattern MP2 is generated by expanding the second mask pattern MP2 by a predetermined width (for example, one pixel) and stored in the second mask memory 123.

The comparison circuit 124 determines whether the generated second mask pattern includes the verification seal impression C3 of the third image memory 112 or the registered seal impression R3o of the second image memory 153,
If neither of them is included, a control signal CC is generated and applied to the transfer circuit 125 and the expansion circuit 122.

Upon receiving this control signal CC, the transfer circuit 125 transfers the second
Second mask pattern M P 2 from mask memory 123
is read out, transferred to the first mask memory 121, and stored. The expansion circuit 122 performs a process of expanding the stored mask pattern M P 2 by a predetermined width to generate a new second mask pattern (indicated by the same symbol M P 2 ), and stores it in the second mask memory 123. do.

Thereafter, the generated second mask pattern is used as the verification seal imprint C8l of the second image memory 153 or as the second image memory 153.
The above-described process is repeated until the registered seal imprint R3o is included, and a final second mask pattern (indicated by the same MP2) is generated and stored in the second mask memory 123 (FIG. 4(B) (b2)).

At this time, the comparison circuit 124 enables the first and second masking circuits 126 and 127.

The enabled first masking circuit 126 generates a verification seal imprint C32 by masking the verification seal imprint C8l with the second mask pattern MP2, and stores it in the fourth image memory 128 ((a3 in FIG. 4(A)). )). As a result, a verification seal imprint C82 is generated from which noises N1 and N2 are removed from the verification seal imprint C31.

The second masking circuit 127 generates a registered seal imprint R3+ by masking the registered seal imprint R3° with the second mask pattern M P 2, and stores it in the fifth image memory 129 ((c2 in FIG. 4(c)). ).

As a result, a registered seal imprint R3 having a missing part corresponding to the missing part of the verification seal imprint C82 is generated.

The line width correction circuit 131 of the matching means 13 performs processing to match the line width of the matching seal imprint C3, read from the fourth image memory 128, to the line width of the registered seal imprint R3, read from the fifth image memory 129. C33 is generated and stored in the sixth image memory 132 ((a,) in FIG. 4(A)).

As a result, when the verification seal imprint C3o is a verification seal imprint of the registered seal imprint R3o, a verification seal imprint C83 having a pattern similar to that of the registered seal imprint R3 is generated.

The matching unit 133 matches the registered seal imprint R3+ with the matching seal imprint C33 whose line widths have been matched, and calculates the degree of matching.

As mentioned above, even if there are chips or noise in the verification seal impression,
Highly accurate matching can be performed.

Although one embodiment of the present invention has been described above, the present invention can be applied not only to verification of seal impressions but also to printed character recognition, component pattern inspection, etc.

〔Effect of the invention〕

As explained above, according to the present invention, even if there are defects or noises in the two objects to be matched, the matching accuracy is not reduced and correct matching can be performed.

[Brief explanation of the drawing]

Fig. 1...Explanatory diagram of the basic configuration of the present invention, Fig. 2...
An explanatory diagram of the configuration of an embodiment of the present invention, FIG. 3: An explanatory diagram of the configuration of the verification target extraction means of the embodiment, FIG. 4: Verification seal imprint, registered seal imprint, and mask pattern of the embodiment FIG. 5: An explanatory diagram of a conventional image matching device. 1 to 3, 11... positioning means, 12... matching target extraction means, 13... matching means, 14... scanning means, 1
5... Database (DB) search means, 131...
Line width correction circuit.

Claims (3)

[Claims] (1) In an automatic image matching device that automatically matches images of two objects, (a) a positioning means (11) that performs positioning on the images of two objects to be matched; and (b) a positioning line. (c) a matching means for performing a matching process based on each of the extracted matching objects; (13) An automatic image matching device comprising: (2) The matching target extracting means (12) generates an initial mask pattern by regarding the non-zero pixels of the two aligned targets as 1, performs a logical product, and expands this initial mask pattern. generating a final mask pattern that includes at least one of the two aligned objects;
2. The automatic image matching apparatus according to claim 1, wherein the two aligned objects are masked using the final mask pattern to extract matching objects. (3) The verification means (13) is the verification target extraction means (12)
3. The automatic image matching device according to claim 1, further comprising a line width correction circuit (131) for matching the line widths of the two extracted matching objects.