JPS6290781A - Character recognition system - Google Patents

Abstract

PURPOSE:To shorten a processing time by obtaining a true quantity of dislocation in position every printing character on characters of plural lines, calculating a statistic quantity every same column of the respective lines and carrying out a template matching making an average value correction value of a collation start position if the quantity is within a prescribed quantitative range. CONSTITUTION:The quantity that a central position 2 of a printing character is dislocated from a central position 3 of an ideal printing line is a sum of the quantity 4 of a skew and the quantity 5 of dislocation in position peculiar to the printing character. Accordingly, the quantity 4 of the skew of respective lines is obtained and the quantity 5 of dislocation inposition peculiar to the printing character is added thereto, a statistic quantity every same column of the respective lines is obtained. If the statistic quantity is situated within a proper quantitative range from a character line width forming a character pattern and the number of scanning lines and a within a standard deviation 2sigma, an average value can be used as a correction value for deciding a collating start position of a template matching. Thereby, a processing time is shortened and a rejection rate is reduced.

Description

[Detailed Description of the Invention] [Summary] This is a character recognition method that quantifies the printing condition by determining the statistical value of the amount of character position shift in the same digit of a printed character string of multiple lines, and uses this data as a basis. A character recognition processing method has been proposed that determines the optimal matching position using template matching, thereby shortening the matching processing time and making it possible to significantly reduce the jet rate.

[Industrial application field]

The present invention relates to a character recognition method for a character recognition device that can efficiently recognize even low-quality printed characters.More specifically, the present invention relates to a character recognition method for a character recognition device that can efficiently recognize even low-quality printed characters. Otherwise, the position of the printed character string may become irregular, resulting in parts of the printed characters being missing or being crushed. In addition to shortening the processing time for matching character patterns and templates,
This invention relates to a character recognition method that makes it possible to significantly reduce the rejection rate.

As information processing becomes more computerized and all processing becomes faster, much of the data input still relies on humans.
Research and development of character reading technology has been progressing with the aim of solving this bottleneck and enabling high-speed data processing by humans.

OCR (optical character recognition) technology has been highly developed based on these needs, and its results include the well-known postal code reading devices, telephone frequency reading devices, and barcode reading devices in the supermarket industry. There is a practical use of leagues.

However, with the development of information processing using electronic computers, popular low-cost printers have entered the market, and even low-quality printed materials output by these printers can be recognized quickly and optically. There is a need for the development of a character recognition method that can do this.

[Conventional technology]

FIG. 4 shows preprocessing in a conventional character recognition system.

To cut out characters in accordance with the amount of skew (tilt) due to vertical positional deviation in character line units of printed matter or relative positional deviation of paper, horizontal direction search range 41 and vertical direction search for characters are required. By moving a mask with a width of range 42, a histogram 43 of the black points of the character part is obtained, the coordinates of the upper and lower ends and left and right ends of the character are detected, the position of the circumscribing frame of the character is determined, and the character is cut out. There is.

In general, each line of printed matter output by a line printer has the same tendency, and it is possible to determine the tendency of several lines and use this data to reflect in cutting out characters.

However, in low-priced printers that are popular on the market, the printed characters are often arranged in an irregular manner, and as a result, the tops of the characters are randomly chipped. In order to accommodate such character patterns, an increase in processing time was unavoidable.

[Problem that the invention seeks to solve]

As described above, in the conventional character recognition method, it is not possible to perform cutting corresponding to the missing characters for objects in which missing printed characters appear randomly.

Therefore, taking the template matching method as an example, the matching process had to be handled by increasing the shift amount. For this reason, there is a problem in that the processing time inevitably increases.

The present invention aims to solve the above-mentioned problems by adding data in units of digits in which defective characters are located in addition to the conventional method of cutting out characters based on the skew amount in units of lines.

The purpose of this is to calculate a more accurate collation position even for a printed character in which a portion of the character is missing, thereby shortening the character recognition processing time.

[Means for solving problems]

In the present invention, the true positional deviation of a character is determined for each character digit position in a line of printed matter, and the average value of this positional deviation amount is determined for each same digit.

By determining the comparison start position based on this average value, it is possible to start the comparison at a more accurate position, and the above-mentioned problem can be solved.

[Effect]

In FIG. 1, the amount by which the center position 2 of the printed character deviates from the ideal center position 3 of the printed line is the sum of the skew amount 4 of the printed character line and the inherent positional deviation amount 5 of the printed character.

Therefore, the skew amount 4 for each line is determined in the same manner as in the conventional method, and the unique positional deviation amount 5 of the printed characters is added to this to determine the statistical amount for each same digit in each line.

As long as this statistic is within an appropriate quantitative range based on the character line width and number of scanning lines that form the character pattern, and is within a standard deviation of 2σ, the matching start position for template matching will be determined using the average value. It can be used as a correction value.

In this way, by determining an accurate and reasonable verification start position, processing time can be shortened and the rejection rate can be reduced.

〔Example〕

FIG. 2 is an enlarged diagram for explanation of printed characters often seen on printed matter output by a low-cost held printer.

This printed character has chipping at the top and bottom of the character due to vertical positional misalignment and the shading of the ink ribbon.

If you observe the printer output in the line direction (left and right) as in this example, it is irregular and the same tendency cannot be observed, but if you observe the digits of each line in the up and down direction, although the character types differ depending on each line, the characters in each column Regularity can be seen in the positional deviation.

Therefore, the true positional deviation amount 21 of each character position is calculated from the tanθ of the skew amount 22 of the printed character line and the unique positional deviation amount 23 of the character for each digit. Find statistical values.

Note that tan θ can be calculated by, for example, printing skew detection marks (generally using a minus sign “-”) at the beginning and end of a character line, and examining their relative positions on the image.

As long as the statistical values obtained in this way are within the above-mentioned appropriate quantitative range, the extracted pattern is corrected using the average value, and the matching start position for template matching is determined.

FIG. 3 is an example of a pattern after the cutout process in which the matching position was determined by the above average value, 31 is the cut out pattern, 32 is the position corrected by the average value (optimum matching position), and 33 is the amount of correction. It is.

In this way, even if the printed characters are considerably chipped due to misalignment as shown in FIG. 3, if reasonable alignment is taken, it is possible to judge as long as there is no distinguishing element in the chipped portion of the characters.

Therefore, it is possible to deal with low-quality printed characters, thereby shortening the verification processing time and significantly reducing the rejection rate where verification is impossible.

In addition, in printed matter output from a normal line printer, the entire - line has a similar skew, and in order to solve this problem, the above-mentioned function that collects and calculates data in units of digits across multiple lines is used to The optimum position information is accumulated every time the printed characters are compared, and as the line progresses to the bottom, the comparison can be started from a position closer to the optimum position.

Note that it is also possible to retry the first line based on the optimal position information on the last line and perform character recognition through more accurate matching processing.

〔Effect of the invention〕

The present invention makes it possible to obtain the optimal matching position even for low-quality printed characters, thereby shortening the template matching processing time and making it possible to significantly reduce the rejection rate due to inability to match. It is an effective character recognition method.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of the method of the present invention. FIG. 2 is an explanatory diagram of the implementation of the present invention, Figure 3 is a diagram illustrating the optimal matching position, FIG. 4 is an example of a conventional character extraction process. In Figure 1, 1 is the true positional deviation amount of printed characters, 2 is the center position of the printed character, 3 is the ideal center position of the print line, 4 is the skew amount of the printed character line, Yang Ryo aH Pesimo Koko Figure 1 Hand invention - Kabuto immersion Meichu Figure 2

Claims (1)

[Claims] The true positional deviation amount (1) for each printed character is determined for multiple lines of characters, and the positional deviation amount (1) for each same digit in each line is calculated.
As long as this statistic is within a predetermined quantitative range, template matching is performed by using the average value of the above positional deviation amount as the correction value for the matching start position for each same digit in each row. A character recognition method that is characterized by