JP2995809B2 - In-table character recognition device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a character recognition device for recognizing characters in a table at high speed and accurately.

(Prior Art) Conventionally, characters in a table (cell) are treated as general characters, and the character type is determined without using the vertical / horizontal relationship of the characters in the table. For this reason, there has been a defect that the determination of the character type is not accurate and the determination processing speed is slow.

(Problem to be Solved by the Invention) An object of the present invention is to provide an in-table character recognizing device capable of more accurately and quickly recognizing the character type of the in-table character.

(Means for solving the problem) Generally, when focusing on the upper and lower cell rows of the table, English characters include a plurality of character strings in the table, Japanese characters are composed of one character string, and numeric characters are in the rightmost position. And the left end positions tend not to be aligned, and it is possible to estimate the character type from the character position in the cell. Based on this knowledge, the present invention has achieved the object using the above-described characteristics of the characters in the table, and the gist of the invention is that a recognition command unit that issues a command for character recognition by an instruction of a keyboard, a mouse, and the like, An image input unit for inputting characters as images together with a table, an image memory for storing image data of the image input unit, a table structure extraction unit for extracting a table structure from the image data and decomposing the cells into cells, A character string extraction unit that extracts a character string and extracts a left end position and a right end position; a character type prediction unit that estimates a character type from the number of character strings and the character position in a cell extracted by the character string extraction unit; The character recognition unit determines the character by comparing the sample character with the image data character with priority given to the character type estimated by the character type prediction unit, the character pattern dictionary that stores the sample characters, and the character recognition unit. It has an external output unit that outputs specified characters, and a CPU that controls each of these units. In the character type prediction unit, if there are multiple character strings in a cell, it is estimated to be an English character, and otherwise, the first cell If the right end positions of character strings other than are not aligned and both the right end and left end are aligned, it is assumed to be Japanese characters, and if the right end is aligned and the left end is not aligned, the character type is assumed to be numeric The character recognition unit in the table is characterized in that the character recognition unit compares and determines the character by giving priority to the estimated character type.

(Operation) In the present invention, in character recognition in a table, attention is paid to the possibility that the same character type is used for items in the vertical direction, and a cell is extracted from image data by the table structure extraction unit. The character string extraction unit determines the number of columns of the character string in the cell and the right and left end positions, and the character type prediction unit estimates the type of the character using the characteristics of the characters in the table. The matching order is changed to the estimated character type priority, the number of times of matching with the recognition dictionary is reduced, the recognition rate of the characters in the table is increased, and the recognition speed is increased.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of the in-table character recognition device.
Reference numeral 1 denotes a recognition command unit that issues a command for performing character recognition according to an instruction from a keyboard, a mouse, or the like. 2 is an image memory for storing image data to be recognized, 3 is a table structure extraction unit for extracting a table structure from the image data and decomposing the table structure into cells.
Reference numeral 4 denotes a character string extracting unit for extracting a character string from each cell and extracting a left end position and a right end position; 5 a character type predicting unit for estimating a character type of a table row; Is performed, and if the similarity is closer than the predetermined value, the character recognition unit that outputs the character,
7 is a character pattern dictionary for storing sample characters, 8 is an external output unit for outputting characters recognized from the table, and 14 is an image input unit for inputting characters in the table as images together with the table.

FIG. 2 is a hardware configuration diagram of the in-table character recognition device.
A RAM 9 stores an image memory, a table structure, a predicted character type, and the like. A ROM 10 stores a character pattern dictionary / program. An RS232-C 11 performs a recognition command and a character output. 12 is a CPU, which controls the operation of the program. Reference numeral 13 denotes an image input unit 14 which takes an image into an image memory by a scanner.

The flow of the entire process will be described with reference to the flowchart of FIG. The table structure extraction unit 3 obtains a table format represented by horizontal and vertical line segments from the image data captured by the scanner 13 (step 1). In this table format, a rectangular area divided vertically and horizontally is called a cell. If there is no column in the table, the process ends (step 2). If there is, the character string extracting unit 4 obtains a character string for all cells except the head of the column (step 3). Next, when the character type prediction unit 5 includes at least one character string therein,
The sequence is predicted to be occupied by English characters (steps 4, 5, 6). In the case of one character string, the left end position and the right end position of the character string are obtained (step 7). If all the cells except the first cell in the column are not aligned at the right end, the character string of the cell in this column is estimated to be a Japanese character (steps 8 and 9). If the right end position is aligned and the left end position is not aligned, it is estimated as a number (step 11),
If both the left edges are aligned, it is estimated that the characters are Japanese characters (step 9). The character recognizing unit 6 recognizes characters in the cells of the column according to the estimation of the character type (step 1
2).

The explanation will be supplemented by using a specific example of the entire processing. First, in the table structure extraction unit 3, horizontal and vertical line segments are obtained from the normalized image data (see FIG. 4), and cells composed of them are extracted (FIG. 5).

Next, since these cells all have one character string, the left end position and the right end position of the character string are obtained (FIG. 6 shows the memory state of the coordinate values of the left end position and the right end position). Next, in the character type prediction unit 5, the second and third columns, except for the first cell, have the same right end position and not the same left end position. The character recognizing unit 6 changes the search order of the dictionary in accordance with these estimations and lowers the recognition level at the time of recognition, thereby obtaining "O"
And "O" are not erroneously recognized, and the number of sample characters in the dictionary to be searched is reduced, so that the character can be determined with a reduced processing time.

(Effect of the Invention) As described above, according to the present invention, the character type is determined using the number of character strings of the characters in the table and the characteristics of the right end and left end positions, and the character type is preferentially used. In addition, the recognition speed was high.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of one embodiment of the present invention, FIG. 2 is a hardware diagram of one embodiment of the present invention, FIG. 3 is a flowchart of one embodiment of the present invention, FIG. 4 is an explanatory diagram showing an example of a table for recognition, FIG. 5 is an explanatory diagram showing how each cell is called, FIG. 6 is an explanatory diagram showing the coordinates of the character string left and right ends of each cell. FIG. 7 is an explanatory diagram showing the character type predicted for each column.

Claims (1)

(57) [Claims] 1. A recognition command section for issuing a command for character recognition in accordance with an instruction from a keyboard, a mouse or the like, an image input section for inputting characters in a table as an image together with a table, and an image for storing image data of the image input section. A memory, a table structure extraction unit that extracts a table structure from image data and decomposes the cells, a character string extraction unit that extracts a character string from each cell and extracts a left end position and a right end position, and an extraction by the same character string extraction unit Character type prediction unit that estimates the character type from the number of character strings and the character position in the selected cell, and compares the sample character with the image data character with priority given to the character type estimated by the same character type prediction unit. A character pattern dictionary that stores sample characters, an external output unit that outputs characters determined by the character recognition unit, and a CPU that controls these units. If there are multiple character strings in the cell, it is assumed to be English characters, and if the right end position of the character strings other than the first cell is not aligned and if both the right and left ends are aligned, Japanese characters If the right end is aligned and the left end is not aligned, the character type is estimated to be a number, and the character is compared and determined by the character recognition unit with priority given to the estimated character type. Recognition device.