JP2001344632A - Boundary line detecting method of paper sheets and feature value extracting method of the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the feature value extracting method of paper sheets capable of stably extracting the feature value of the paper sheets by improving detecting precision without receiving the influence of deviation. SOLUTION: A boundary line being that of lightness/darkness is detected as the feature place of the paper sheets, the position of the detected feature place is compared with a reference value to calculate a deviation and the extracting position of the feature value is corrected by using the deviation. Concerning the boundary line, the difference of pixel values with the adjacent row of a specific column is calculated with respect to the pixel value of the column of the paper sheets and the boundary line is detected by using a row having the largest difference, Otherwise, the sum of pixel values is taken by the unit of the row of the paper sheets to calculate the difference of the total values between with the adjacent row to detect the boundary line by using the row having the largest difference between the total values. Furthermore, the extracting position of the feature value is corrected by deviating the position of the measuring section of feature desired to extract by the portion of the value of the calculated deviation based on the deviation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting a boundary line of paper sheets, which stabilizes the extraction of characteristic amounts in a discriminating machine for discriminating paper sheets such as bills, for example, provided with an optical sensor, and a feature amount extraction. About the method.

[0002]

2. Description of the Related Art Conventionally, for example, Japanese Patent Laid-Open Publication No. Hei 6-84043 has provided a plurality of measurement sections for detecting the characteristics of a bill printing pattern. In the measurement section to detect, the section maximum value, in the measurement section to detect the features of the dark part of the bill printing, the section minimum value, in the measurement section of the part without printing such as the watermark part of the bill, the section average value, A technique is disclosed in which at least one or more pieces of data are obtained and used as a feature amount to remove the influence of a slippage or a slippage of a bill.

[0003] For example, Japanese Patent Application Laid-Open No. 9-180025 discloses that when a detected waveform of a print pattern is compared with a reference waveform to determine a sheet such as a bill, a deviation width of the detected waveform from the reference waveform may occur. A plurality of representative waveforms in which the detected waveform is shifted by varying between the minimum deviation width and the maximum deviation width are created, and a representative waveform at a plurality of predetermined positions of the representative waveform is compared with a reference waveform. There is disclosed a technique in which a small amount is used as a shift amount.

Further, for example, Japanese Patent Application Laid-Open No. 10-21442 discloses that in order to mitigate the influence of paper sheet feeding and printing misregistration, pattern matching is changed from a pixel unit to an area unit and the trueness of the paper sheet is reduced. A technology for discriminating fake has been disclosed.

[0005]

However, the technique disclosed in the above-mentioned Japanese Patent Application Laid-Open No. 6-84043 does not
A measurement section is provided, and the influence of the deviation is reduced using the section maximum value, section minimum value, and section average value as feature values. However, since the deviation itself is not corrected, extraction is required to realize high discrimination performance. The accuracy of the feature value to be obtained is insufficient.

Further, in the technique disclosed in Japanese Patent Application Laid-Open No. 9-180025, it is necessary to create a plurality of waveforms whose detection waveforms are shifted, which requires a long processing time.

Further, in the technique disclosed in Japanese Patent Application Laid-Open No. 10-21442, only pattern matching is performed on an area basis, and as in Japanese Patent Application Laid-Open No. 6-84043, it is extracted to realize high discrimination performance. The accuracy of the feature amount is insufficient.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems of the related art, and is capable of improving the accuracy of detecting a characteristic portion and extracting a stable characteristic amount without being affected by a shift. An object of the present invention is to provide a class boundary detection method and a feature amount extraction method.

[0009]

In order to achieve the above object, a method for detecting a boundary line of a paper sheet according to the present invention includes a paper sheet for detecting a boundary line serving as a light-dark boundary as a characteristic amount of the paper sheet. In the class boundary detection method, a region including the boundary of the paper sheet is imaged, and a difference between a pixel value of a specific column of the paper sheet and a pixel value of an adjacent row of the column is calculated. , A boundary line is detected using a row having the largest difference.

Further, the paper sheet boundary line detecting method according to the present invention is the paper sheet boundary line detecting method for detecting a boundary line which is a boundary between light and dark as a characteristic amount of the paper sheet. Take an image of the area including the boundary line, take the sum of the pixel values for each row of paper sheets, calculate the difference between the total value of the adjacent rows,
The method is characterized in that a boundary line is detected using a row having the largest difference between the total values.

Further, in the paper sheet feature amount extracting method according to the present invention, the paper sheet boundary line according to claim 1 or 2, wherein a boundary line serving as a light-dark boundary is used as a paper sheet characteristic portion. , A deviation amount is calculated by comparing the position of the detected characteristic portion with a reference value, and the characteristic amount extraction position is corrected using the deviation amount.

Further, in the paper sheet feature amount extracting method according to the present invention, the feature amount extracting position is based on the calculated shift amount, and the position of the measurement section of the feature to be extracted is shifted by that value. It is characterized in that it is corrected by

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First, an outline of the present invention will be described. In the feature extraction method for paper sheets according to the present invention,
A boundary line, which is a characteristic portion of a sheet such as a banknote, is detected from a captured image, and the position of the detected boundary line is compared with a reference value calculated statistically in advance to calculate a shift amount. Is used to correct the feature amount extraction position.

Hereinafter, specific embodiments will be described with reference to the drawings. FIG. 1 illustrates the concept of the present invention. As shown in FIG. 1, the present invention uses a bill
Take an image with a camera or the like, using the captured image, as a characteristic portion of the bill, a boundary where the brightness changes significantly, for example, a band, a picture, a watermark, a boundary of an area such as a portrait is detected,
The shift amount is calculated by comparing the shift amount with a reference value statistically obtained from all data, and the feature amount extraction position is corrected using the shift amount. As an implementation method, two algorithms can be considered. They will be described below.

First Embodiment (Algorithm 1) FIG. 2 shows the basic principle of Algorithm 1, and FIG. 3 is an explanatory diagram showing a method of determining a boundary line candidate. In the embodiment described below, in the imaging screen, each straight line formed of pixels arranged in a two-dimensional horizontal direction forms each column, and each straight line formed of pixels arranged in a two-dimensional vertical direction forms a row. It is assumed. As shown in FIG. 2, as a characteristic portion of the banknote, the target is within the possible range of the boundary where the brightness changes greatly, and the difference between the pixel value of the specific column and the adjacent pixel value (the adjacent row is The difference between the pixel values is calculated, and as shown in FIG. 3, the row having the largest pixel value difference is set as the boundary line in the specific column.

The specific column to be searched is determined for the center of the banknotes in the column direction and several columns spaced at equal intervals (for example, 10 pixels) from the center. As a result of the search, the boundary lines exist by the number of the specific columns. In the preceding stage, the direction of the denomination is determined, and the possible range of the boundary can be determined. Then, an average value of the values (the number of rows) of these boundary lines is obtained and used as the position of the boundary line to be obtained. Finally, by comparing the detected boundaries with reference values obtained statistically from all data,
The shift amount is calculated, and the feature amount extraction position is corrected.

FIG. 4 shows a flowchart of the algorithm 1. That is, all the pixel values are obtained (step S1), the difference between the pixel values of the specific column and the adjacent row is obtained (step S2), and this process is performed for each specific column. That is, a difference between pixel values is calculated between adjacent rows on each specific column (step S3). Next, the row having the largest pixel value difference is set as a candidate for a boundary line (step S4), and a boundary to be detected is determined by averaging the candidates for the boundary line (step S5), and the boundary is compared with a reference value. Then, for example, the amount of cutting deviation is calculated (step S6). Then, the position of the characteristic amount is corrected using the calculated shift amount as described later.

Second Embodiment (Algorithm 2) FIG. 5 shows the basic principle of Algorithm 2, and FIGS. 6 and 7 are explanatory diagrams showing boundary line candidates and a method of determining a boundary line. is there. As shown in FIG. 5, as a characteristic portion of a bill, a possible range of a boundary where the brightness changes greatly is targeted, and the sum of the pixel values is calculated for each line of the bill within the range. It is assumed that the denomination direction is determined in the previous stage, and the possible range of the boundary can be determined. Next, the difference between the total values of adjacent rows is calculated, and FIG.
As shown in (1), the row having the largest difference between the total pixel values in row units is determined as a boundary line candidate.

Then, the difference between the total value of the line of the boundary line candidate and the previous line and the difference of the total value of the previous line and the previous line are compared. When the difference between the former and the latter is almost the same (see FIG. 7A), it is determined that the previous row contains a large number of pixels including the intermediate color, and that row is set as a boundary. When the latter difference is small (see FIG. 7B), it is determined that the previous pixel includes many pixels that are brighter than the intermediate color, and the line of the boundary candidate is set as the boundary. Finally, by comparing the detected boundaries with reference values obtained statistically from all data,
The shift amount is calculated and the feature amount extraction position is corrected.

FIG. 8 shows a flowchart of the algorithm 2. That is, all pixel values are acquired (step S11), the pixel values are totaled for each row (step S12), and the difference between the total values is calculated between adjacent rows (step S13). Next, the line having the largest difference in the total value is set as a candidate for the boundary line (step S14), and the line is compared with the previous line using the difference in the total value (step S15), and the boundary line to be detected is determined. (Step S16). And
For example, the amount of cutting deviation is calculated (step S17), and the position of the characteristic amount is corrected using the calculated deviation amount as described later.

FIG. 9 shows the algorithm 1 or 2 described above.
The method of calculating and correcting the shift amount based on the detection value of the boundary line detected according to the following will be described. The shift amount (Δy) is obtained by subtracting a reference value from the detected position (detected value) of the boundary line. Then, assuming that the coordinates of the feature to be extracted in the reference data are (x ₀ , y ₀ ), the position of the feature (x ₁ , y ₁ ) in the actual data is the correction amount (= shift amount (△ y)).
By using (x ₀ , y ₀ + △ y).

As described above, according to the present embodiment,
It is affected by the deviation by detecting the boundary line that is the boundary between light and dark as the characteristic point of the paper sheet, comparing the position of the detected characteristic point with the reference value, calculating the deviation amount, and performing correction. Without this, a stable feature amount can be extracted, and improvement in discrimination performance can be expected.

In the detection of a boundary line as a characteristic portion, an algorithm 2 for calculating the sum of pixel values in units of rows is particularly used.
According to the method according to the above, than the method of detecting the feature mark,
There is an effect that detection accuracy is high without being influenced by local light and darkness such as stains and stains.

Further, by utilizing the above-described present invention, two feature points are detected, a difference between the detected positions is calculated, and the difference between the detected positions is compared with a difference between reference values.
It is also possible to cope with speed fluctuations in conveyance and expansion and contraction of images due to skew correction.

[0025]

As described above, according to the present invention, since a boundary line serving as a boundary between light and dark is detected as a characteristic portion of a paper sheet, the characteristic portion can be detected accurately and easily. Can be used in the feature value extraction method, and the position of the detected feature point is compared with the reference value to calculate the amount of shift and make a correction. There is an effect that a feature amount can be extracted, and improvement in discrimination performance can be expected.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the concept of the present invention.

FIG. 2 relates to a first embodiment of the present invention, and is an algorithm 1 for correcting a feature amount extraction position using a shift amount.
FIG.

FIG. 3 is an explanatory diagram of a method for determining a boundary line candidate based on Algorithm 1.

FIG. 4 is a flowchart based on algorithm 1.

FIG. 5 relates to a second embodiment of the present invention, and is an algorithm 2 for correcting a feature amount extraction position using a shift amount.
FIG.

FIG. 6 is an explanatory diagram of a method of determining a boundary line candidate based on Algorithm 2.

FIG. 7 is an explanatory diagram of a method for determining a boundary line based on Algorithm 2.

FIG. 8 is a flowchart based on Algorithm 2.

FIG. 9 is an explanatory diagram related to a method for calculating and correcting a shift amount according to the present invention.

[Explanation of symbols]

 S1 to S6 Processing steps based on algorithm 1 S11 to S17 Processing steps based on algorithm 2

Continuing from the front page (72) Inventor Shinichi Nakajima 1-1, Tanabe-Nitta, Kawasaki-ku, Kawasaki-shi, Kanagawa Prefecture F-term in Fuji Electric Co., Ltd. 3E041 AA02 BA11 BB03 BC03 CA01 CA09 CB03 5B057 AA11 DA07 DA12 DB02 DC02 DC16 DC19 DC22 5L096 BA03 BA18 FA03 FA38 GA07 GA19

Claims (4)

[Claims] 1. A method for detecting a boundary line of a paper sheet which detects a boundary line serving as a boundary between light and dark as a characteristic portion of the paper sheet, wherein an area including the boundary line of the paper sheet is imaged, and A border line of paper sheets, wherein a pixel value difference between a pixel value in a specific column and an adjacent row in the column is calculated, and a border line is detected using a row having the largest difference. Detection method. 2. A method for detecting a boundary line of a paper sheet, which detects a boundary line serving as a light-dark boundary as a characteristic portion of the paper sheet. The sum of pixel values is calculated for each line, the difference between the total values of adjacent lines is calculated, and the boundary line is detected using the line having the largest difference between the total values. Boundary detection method. 3. A method for detecting a boundary between light and dark as a characteristic portion of a paper sheet using the method for detecting a boundary line of a paper sheet according to claim 1 or 2, and detecting the detected characteristic portion. A feature amount extraction method for a paper sheet, comprising: calculating a shift amount by comparing the position of the sheet with a reference value; and correcting the feature amount extraction position using the shift amount. 4. The method according to claim 3, wherein the extraction position of the characteristic value is a position of a measurement section of the characteristic to be extracted by the value based on the calculated shift amount. A feature amount extraction method for paper sheets, wherein the feature amount is corrected by shifting.