JPH057748B2 - - Google Patents

Description

[Detailed Description of the Invention] <Technical Field of the Invention> The present invention, for example, when determining the authenticity of a seal stamped on a check, etc., reads the stamp imprint from the check and converts it into an image, and detects a steep outline from the stamp image. The present invention relates to a method for extracting a singular point of an image, which extracts a point that changes to a specific point (hereinafter referred to as a singular point) and performs a matching process for a seal imprint or the like.

<Background of the Invention> Conventionally, this type of singularity extraction method has been used, for example, to image a seal imprint, extract data on the contour line of the seal imprint from the entire image, and then trace the contour line with data, so that the contour line is sharp. Detect the part that changes to
This is considered a singularity. This has resulted in disadvantages such as the complexity of singular point extraction processing, poor processing efficiency, and increased memory capacity.

Therefore, the inventor scanned the screen by setting a mask with a rectangular field of view for the imaged figure, and also scanned the four sides of the mask to detect only one side of the mask or only two adjacent sides. After detecting intersecting partial patterns of images, we have developed a singular point extraction method that determines the existence of singular points by making conditional judgments on the partial patterns within the field of view of the mask, and then calculates the location of the singular points. did.

According to this method, there is no need to extract the contour line of the entire image, and there is no need to track data on the contour line, so the singular point extraction process is simplified, processing efficiency is improved, and memory capacity is saved. can be realized. However, in this method, in addition to the stroke of interest, peripheral strokes may be included in the mask. In such a case, the stroke portion intersects with many sides of the mask, making it impossible to extract singular points, which has the disadvantage of reducing the number of singular points extracted and lowering the accuracy of matching seals, etc. .

<Objective of the Invention> The present invention improves the singular point extraction method using a rectangular mask and increases the number of singular points extracted, thereby improving the efficiency of singular point extraction processing and reducing memory capacity. It is an object of the present invention to provide a method for extracting singular points in images that achieves savings and improves the accuracy of matching seal stamps and the like.

<Configuration and Effects of the Invention> In order to achieve the above object, the present invention scans an image of an object using a large mask having a rectangular field of view and a small mask having a rectangular field of view included within the field of view of the large mask. By scanning at least one of the four sides of the large mask and the small mask at each scanning position, when a partial pattern of the image that intersects only one side or only two adjacent sides of either mask is detected, an estimation step of estimating that a singular point exists in a partial pattern within the mask; and when the existence of a singular point is estimated in the estimation step, a predetermined method that represents a shape feature of the partial pattern from a partial pattern within the mask; a condition determination step in which a condition is determined to determine the existence of a singular point based on the measured parameters by measuring the parameters of the parameter, and when the existence of a singular point is determined in the condition determination step, the position of the singular point is measured. A series of position measurement steps were performed to extract singular points in the image.

According to the present invention, compared to the conventional method of extracting data from the outline of the entire image, the singular point extraction process is simplified, processing efficiency is improved, memory capacity is reduced, and the singular point extraction process is simplified. This invention achieves the excellent effects of achieving the purpose of the invention, such as being able to prevent a decline in the verification accuracy of seals, etc. due to a decrease in the number of extractions.

<Description of Embodiments> FIG. 1 shows an example of a device for determining whether a stamp stamped on a check is a genuine stamp or a counterfeit stamp.

The illustrated device takes out a plurality of checks 1 stored in a case 10 one by one from the bottom, and
It is sent onto the conveyor belt 11.

During transportation, each check 1 is checked for the stamp imprinted on the check, and a sorting device 2 downstream of the transportation path determines whether a genuine stamp was used, a check with a counterfeit stamp, or a check with a fake stamp. The checks are sorted into three types of unidentifiable checks and sent to collection cases 21, 22, and 23, respectively.

A reading device 50 that reads the code number printed on the check 1 using an optical or magnetic method, and a camera device 3 that images the stamp imprinted on the check 1 are installed along the conveyance path. The output of the reading device 50 is sent to the external memory 5, and the output of the camera device 3 is sent to the image processor 4 via the data bus 40, respectively. The external memory 5 stores a plurality of reference seal impressions, and based on the code number read by the reading device 50, the image data of the corresponding reference seal impression is extracted and sent to the image processor 4.

The camera device 3 includes a reading line sensor 31 and a binarization circuit 32, as shown in FIG. An optical signal corresponding to the image configuration is extracted, and this optical signal is photoelectrically converted to output an analog image signal _i1 . Further, the binarization circuit 32 binarizes the image signal and sends this binary signal _i2 to the image processor 4 via the data path 40 in a bit serial format corresponding to columns. In the figure, the oscillator 33 is the reading sensor 31.
an X-axis coordinate counter 34 for supplying readout pulses and calculating point coordinates of the current signal;
Supplies clock pulses to Furthermore, a carry signal _i3 is output from this X-axis coordinate counter 34, and this carry signal _i3 is supplied to a Y-axis coordinate counter 35 for calculating the Y-axis coordinate, and these X-axis coordinate counters 34 and Y Each output of the axis coordinate counter 35 is input to the image processor 4 via a data path 40.

The image processor 4 has a program memory 41 that stores the singular point extraction program shown in FIGS. 6 and 7, as well as various programs necessary for collating seal impressions and determining authenticity. Area A (shown in Figure 10) stores data for specifying the mask, areas B and C store data related to the size of the mask and the coordinates of each side, and area D stores data related to singular points. an image memory 43 having E, and an arithmetic control unit 42 that reads data from the image memory 43, decodes and executes a program, extracts singular points, and controls various operations and a series of processes related to seal imprint verification.
(hereinafter simply referred to as CPU), consists of CPU4
3 further issues a control command to the sorting device 2 via the output device 24 based on the result of determining the authenticity of the seal.

Figure 3 shows an example of the singular point extraction method.
The present invention has a large mask 6A having a rectangular field of view of A bits vertically and B bits horizontally on the screen 60 containing the seal image.
Then, a small mask 6B with a rectangular visual field of vertical a bits (a<A) and horizontal b bits (b<B) included in the visual field of the large mask 6A is set, and both masks 6A, 6B are
are scanned together in the horizontal direction, and during the scanning process, as shown in FIG. The position of the singular point is detected using the method.

FIG. 5 shows an example of a pattern in which a stroke portion 7 having a singular point P included in the mask is used. In the first to fourth patterns shown in FIGS. 1 to 4, the singular point P is at the right end of the stroke portion 7. Located at the bottom, left, and top, respectively.
The fifth to eighth patterns shown in FIGS. 5 to 8 are located at the lower right end, lower left end, upper left end, and upper right end of the stroke portion 7, respectively.

In each of the above patterns, in the first to fourth patterns, the stroke portion 7 intersects only one side of the mask,
In the fifth to eighth patterns, the stroke portion 7 intersects only two adjacent sides of the mask.

FIGS. 6 and 7 show specific examples of singular point extraction operations.

First, in step 80 of FIG. 6, various initial values such as the field of view size of the large mask 6A and the small mask 6B are set in area A of the image memory 43, and then in step 81, the large mask 6A including the small mask 6B is set. (hereinafter simply referred to as large mask 6A) is screen 60
Check whether the top has been scanned one column. In the illustrated example, by setting a column counter M in the memory area 44 (shown in FIG. 9) in the CPU 42 and determining whether or not the contents of the column counter M have reached the set value m, one column on the screen is It is determined that the scan for the minute has been completed. In this case, since the determination at step 81 is "NO", the process advances to step 82, where the horizontal flag F _H
Check whether it has been set. This horizontal flag
_FH is a flag that is set when any of the first to eighth patterns is included in the mask,
In this case, the determination of "F _H = 1" in step 82 is "NO", and the process proceeds to step 83, where the large mask 6A is moved laterally by 12 masks (1/2B bits), and further in step 84, it is moved horizontally. The contents of counter M are incremented by 1.

The next step 85 is to check whether the large mask 6A has completed scanning all the lines on the screen 60.
A judgment is made based on whether the contents of the row counter N set within have reached the set value n. In this case, the determination at step 85 is "NO", so step 86 is performed.
The process then proceeds to a process of determining whether any of the first to eighth patterns is included in the large mask 6A. In this process, first, each side of the large mask 6A is sequentially scanned, and as shown in FIG. is present, mask 6A
It is estimated that a stroke portion 7 having a singular point is included within the stroke. When this condition is satisfied, the determination in step 876 becomes "YES" and the process proceeds to step 90. On the other hand, as shown in FIG. 11, other stroke portions 7 around the stroke portion 7 of interest within the large mask 6A.
0,71 is included, which is why the large mask 6A 4
If there are constituent points of the seal impression on three of the sides, the determination at step 87 is "NO" and the process proceeds to step 88, where each side of the small mask 6B is sequentially scanned. And one of the four sides of the small mask 6B
If the constituent points of the seal impression exist only on the side or two adjacent sides, the determination at step 89 becomes "YES" and the process proceeds to step 90.

If the determination in both steps 87 and 89 is "NO", the process returns to step 81 and the same operation is repeated. When scanning for one column of the screen is completed, the content of the column counter M reaches the set value m, the determination in step 81 becomes "YES", and the process proceeds to step 92.
Next, it is checked whether the vertical flag _Fv is set. This vertical flag _Fv is a flag that is set when a singular point P is extracted in scanning each row, and is set as "Fv _" in step 92.
= 1” is “NO”, proceed to step 93, and large mask 6A is used for 1/2 mask (1/2A bit).
After vertical movement, the contents of the row counter N are incremented by 1 in step 94.

In the scanning process of the large mask 6A, the judgment in either step 87 or step 89 is "YES".
Therefore, when it is determined that there is a possibility that there is a stroke part with a singular point in either mask 6A or 6B, the process proceeds to step 90, and if the conditions for a singular point are satisfied, that singular point is detected. After calculating the coordinate data, in the next step 91, a horizontal flag F _H and a vertical flag F _v are respectively set. As a result, the determination of "F _H = 1" in step 82 becomes "YES", and after the horizontal flag F _H is cleared in step 95,
In step 96, the large mask 6A is moved laterally by one mask (b bits), and in step 97, the contents of the column counter M are incremented by two. The processing in steps 96 and 97 takes into consideration that each mask 6A and 6B does not extract the same singular point twice. Also, since the vertical flag _Fv is turned off in step 91,
"F _v = 1" in step 92 when scanning advances to the next row
The result is "YES", and for the same purpose, after clearing the vertical flag _Fv in step 98, the large mask 6A is moved vertically by one mask (A bit) in step 99, and the column counter N is cleared in step 100. Contents 2
to add.

The flowchart in FIG. 7 shows details of the singular point extraction operation. First, in step 110, as shown in FIG.
After detecting black pixel points Q and R located at both ends of one side or two adjacent sides of (6B), line segment
Calculate the length S between QRs. Next step 111
compares the length S and the set value T _H ,
When the condition of “S × T _H ” is satisfied, step 111
becomes "YES" and the process proceeds to the next step 112, where the relative coordinates (x ₀ , y ₀ ) of the center point C of the line segment QR are calculated. On the other hand, when the condition of step 111 is not satisfied, the determination becomes "NO", and the singular point extraction operation is terminated, assuming that no singular point exists. Next, in step 113, the relative coordinates of each contour point of the stroke portion 7 are determined as (xi, yi) (where i = 1, 2, ..., n).
Then, the distance di from the center point C to each contour point is calculated using the following equation and stored in area E of the image memory 43.

Next, in step 114, the maximum value T is extracted from among the above distance values, and in the next step 115, the maximum value T is extracted.
, the T/S value and the set value T _H2 are compared. When the determination of "T/S>T _H22 " in step 115 is "NO", the singular point extraction operation is completed assuming that no singular point exists; on the other hand, when the determination in step 115 is "YES", the process proceeds to step 116. advance, maximum value T
The contour point on the mask 6A is defined as the singular point P.
Calculate the relative coordinates in (6B) (step
116). Now, assuming that the relative coordinates of the singular point P are (xi, yi), in the next step 117, these relative coordinates are converted into absolute coordinates (Xi, Yi) on the screen 60.
These absolute coordinates (Xi, Yi) are the absolute coordinates of the starting point Z (shown in FIG. 8) of the mask 6A (6B) on the screen 60 (i, j), and the absolute coordinates of the origin 0 of the screen 60 (1 , 1), it is given by the following equation.

Xi=i+xi-1 Yi=j+y-1 By the above calculation, the absolute coordinates of the singular point P are calculated, and this coordinate data is stored in area D of the image memory 43.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram showing the configuration of the seal verification device, Fig. 2 is a circuit block diagram of the camera device, Figs. 3 and 4 (1) and (2) are explanatory diagrams showing the singular point extraction method, and Fig. 5 Figures (1) to (8) are explanatory diagrams showing examples of patterns within the mask, Figures 6 and 7 are flowcharts showing the singular point extraction operation, Figure 8 is an explanatory diagram showing the principle of coordinate transformation, 9 and 10 are explanatory diagrams showing the memory configuration, and FIGS. 11 and 11 are explanatory diagrams showing the singular point extraction method according to the present invention. 6A...Large mask, 6B...Small mask, P...
Singularity.

Claims (1)

[Claims] 1. An image singularity extraction method for extracting a point where the outline changes sharply from an image of an object as a singularity, which method comprises: By scanning four sides of at least one of the large mask and the small mask at each scanning position while scanning with a small mask having a rectangular field of view included in the field of view of the mask, it is possible to scan only one side of either mask or two adjacent sides. an estimation step of estimating that a singular point exists in the partial pattern within the mask when a partial pattern of the image that intersects only an edge is detected; and when the existence of a singular point is estimated in the estimation step; a condition determination step of measuring a predetermined parameter representing a shape feature of the partial pattern from the partial pattern in the mask, and determining a condition for determining the existence of a singular point from the measured value; 1. A method for extracting a singular point in an image, which comprises extracting a singular point in an image by sequentially performing a position measurement step of measuring the position of the singular point when the existence of the point is recognized.