JPH04306771A - Automatic input processing method for drawing - Google Patents

Abstract

PURPOSE:To improve processing speed by reducing errors in the processing of separation, recognition, editing and correction of drawing elements. CONSTITUTION:The drawing made into a binary picture is vector-processed and vector information is separated into a character candidate and a graphic candidate. A man-machine interface corrects the error for the separated result. Then, recognition, editing and correction are executed. Then, separation is securely executed at high speed with a rule for separating the elements and with a small amount of rule setting even for the element with a poor consistencey of the drawing. Subsequent recognition and correction are set to be processing for the elements which are securely separated so as to reduce a processing amount.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention relates to an automatic drawing input processing method for recognizing characters and figures from drawings and documents, and in particular, character candidates and figure candidates (symbol candidates and line segment candidates) that become elements of drawings and documents. The present invention relates to a drawing element separation method for separating and extracting drawing elements and performing recognition processing on each.

0002

[Background Art] With the development of computer technology, information that has traditionally been managed using drawings has been changed to EDP (Electron
ic Data Processing) to improve the efficiency of business operations. The methods for EDP conversion are (1) using CAD and inputting each line element from the beginning while looking at the drawing; (2) reading the drawing from an image scanner and displaying the image in the background; (3) Load the drawing from an image scanner, approximate the line segments from that image to vector data, and then input the line segment elements using CAD while looking at the image. There are two methods: (1) a method of sequentially adding corrections and converting to the desired data; (4) a method of inputting a drawing from an image scanner, having a computer recognize the content of the drawing, and correcting errors in the output result.

Among the methods (1) to (4) above, the later methods are more advanced and easier to convert into EDP, and the automatic drawing input processing method that is method (4) is the processing procedure shown in FIG. become. Image processing is performed to convert a drawing containing a mixture of characters and figures (symbols and line segments) into a binary image using an image scanner or the like (step S1), and vectorization processing is performed to compress the information of the binary image and convert it into vector information. (Step S2).

There are generally two methods for this vectorization process. One method is to perform thinning processing on black pixel portions of a binary image until the line width becomes 1, and then obtain a thin line vector by approximating the thinned point sequence as a vector. The other method is to extract pixels at the boundary between white and black in a binary image, approximate this pixel sequence as a vector to extract a contour vector, and then search for a pair of contour vectors. A vector is generated at a portion corresponding to the center line of , that is, a core line vector is generated.

The vector group obtained by the above-mentioned vector processing is separated and extracted into character candidates and graphic candidates based on the isolation of the vector, the size of the rectangular area, etc. (step S3). The vector group that is considered as a character candidate is a group of vectors that correspond to one character. In addition, the vector group that is considered as a figure candidate is separated into symbol candidates and line segment candidates through structural analysis, and each of them is grouped.

Next, for each separated and extracted candidate vector group, one character candidate and symbol candidate is recognized by pattern matching with a dictionary (steps S4 and S5), and the recognition result is obtained as code data. . Line segment candidates are subjected to line segment processing to shape bends at intersections, etc. (step S6), and the processing results are obtained as start and end point data.

The character and graphic data recognized and processed in steps S4 to S6 are edited and corrected (step S7).
It is saved in the database as EDP data.

[0008]

[Problem to be Solved by the Invention] In an automatic drawing input processing method in which a computer automatically recognizes drawings, many detailed rules are set on the original drawing description to recognize characters and figures, and the content of the drawing is based on these rules. I am aware of this. For example, as shown in (a) or (b) of Figure 3, if a rule is set that a figure or number is written inside circle figure A, then (c) or (d) of the figure is set. Even when the outer figure A is in contact with the inner figure B or character C, separate recognition of figures A and B and figure A and character C can be performed.

However, since drawings that already exist are not drawn with rules in mind, even if rules set later are applied, recognition errors will occur. Furthermore, even if a newly created drawing does not follow the set rules or cannot be drawn according to the rules, recognition errors may occur.

[0010] As described above, many rules are set for separating drawing elements for automatic recognition, but unless there is a high degree of consistency between the rules and the drawing, that is, the drawing accurately follows the rules, errors may occur in element separation. This causes errors in automatic recognition after drawing element separation, and also increases processing time. Furthermore, the recognition error of the element in question causes the recognition error of surrounding elements, and the effort required to make corrections in subsequent editing and correction processing becomes extremely large.

An object of the present invention is to provide an automatic drawing input processing method that increases processing speed by reducing errors in processing from separation and recognition of drawing elements to editing and modification.

0012

[Means for Solving the Problems] In order to solve the above-mentioned problems, the present invention provides an automatic method for vectorizing drawings converted into binary images and separating vector groups into character candidates and figure candidates from this vector information. The present invention is characterized by separating elements, correcting separation errors using a man-machine interface for the separation results, automatically recognizing the corrected character candidates and figure candidates, and editing and correcting the recognition results. .

[0013]

[Operation] After separating character candidates and figure candidates by automatic element separation, errors are corrected for each candidate, and then recognized, edited, and corrected. By correcting errors in the element separation results, subsequent To shorten processing time while eliminating errors in processing from recognition to correction. Furthermore, it becomes unnecessary to prepare a huge number of rules for ensuring consistency of element separation setting rules for all drawings.

[0014]

Embodiment FIG. 1 is a processing flowchart showing an embodiment of the present invention. The difference between this figure and FIG. 2 is that element separation correction processing (step S8) is performed in the process from automatic element separation to drawing element recognition.

In the element separation correction process, errors in element separation are corrected using a man-machine interface on the result of separation of character candidates and graphic candidates by automatic element separation (step S3).

For example, for a figure in which a figure B or a character C is provided inside a circle figure A shown in FIG. 3, the figure shown in (a) or (b) in the same figure is separated by automatic element separation, and (c)
Alternatively, if the figure in (d) cannot be separated due to a defect in the rules, or if it is incorrectly recognized, candidate designations are made for circle figure A, figure B, and character C for the figure in element separation correction.

Therefore, when there is little consistency between the automatic element separation rules and the drawing to be recognized, even if the recognition is incorrect or unrecognizable, accurate element separation can be performed by element separation correction.

Element separation correction at this stage reduces the number of rule settings without preparing element separation rules for each drawing format. Furthermore, since character candidates and graphic candidates can be separated accurately, subsequent character recognition, symbol recognition, and line segment processing are made more accurate, and the time required for recognition processing is significantly reduced. Further, the amount of correction processing in subsequent editing/correction processing is extremely reduced.

[0019]

As described above, according to the present invention, after separating elements into character candidates and figure candidates from vector information in a drawing, separation errors are corrected using a man-machine interface, and the corrected elements are recognized. As a result, errors in element separation are corrected at the stage when they occur, making recognition processing after element separation and subsequent editing/correction fast and reliable. Elements can be separated reliably for drawings, and the amount of rule settings can be reduced, resulting in faster separation.

[Brief explanation of drawings]

FIG. 1 is a processing flowchart showing one embodiment of the present invention.

FIG. 2 is a conventional processing flowchart.

FIG. 3 is an illustration of a figure.

[Explanation of symbols]

A...circle shape, B...shape, C...letter.

Claims (1)

[Claims] Claim 1: A drawing that has been converted into a binary image is vectorized, automatic element separation is performed to separate the vector group into character candidates and figure candidates based on this vector information, and the separation results are separated using a man-machine interface. An automatic drawing input processing method characterized by correcting errors, automatically recognizing the corrected character candidates and figure candidates, and editing and correcting the recognition results.