JPS61239384A - Method for improving recognition rate in graphic recognition - Google Patents

Abstract

PURPOSE:To improve the recognition rate when graphic recognition is made, by performing the recognizing process in such a way that small-area pictures are cut out around an equidistant grid from a hand-drawn plan drawn with the grid as a reference and a mask which is approximate to a graphic form to be recognized is applied. CONSTITUTION:A whole plan containing a graphic form to be recognized is stored in a picture image memory 5 as picture data. Since the plan is drawn with an equidistant grid as a reference, part of the plan containing the graphic form to be recognized is cut out and fetched as a cut-out picture 1. The cut-out picture 1 is supplied to an AND gate 3 together with a mask 2 which has the same size as the graphic to be recognized has and is approximate to the graphic form and their AND is taken. Then the mask-processed result data 4 is subjected to an actual recognizing process.

Description

[Detailed Description of the Invention] [Summary] In a figure input processing device that automatically reads and recognizes handwritten drawings, a mask of a figure to be recognized and an approximation is created for image data of a small area cut out around a grid. After multiplication, recognition is performed, thereby preventing images other than the figure to be recognized from becoming noise and lowering the recognition rate.

[Industrial Field of Application] The present invention relates to graphic recognition in a graphic input processing device.

With the rapid progress of information processing technology, automatic design and automatic manufacturing (CAD, CAM) is becoming increasingly popular. In this system of automatic design and automatic production, technology that automatically reads and recognizes handwritten drawings and processes them for later production of CAD data is one of the technologies that occupies an important position.

The present invention relates to a method for improving the recognition rate of various symbols in drawings in a graphic input processing device that performs such processing.

[Prior Art] There are various types of original drawings that can be read by a graphic input processing device, but the explanation will be given by taking as an example the case of automatically reading a handwritten printed board pattern diagram as shown in FIG. do.

When performing figure recognition of a handwritten printed board pattern diagram, the image data read by the drawing scanner is stored in the image memory, the entire drawing in the image memory is divided into grids, and the image data is divided into rectangular shapes in units of the grid. I was cutting it out and recognizing it.

Recognition is performed by recognizing symbols within the cut out figure, recognizing line segments, measuring line widths, and synthesizing recognition data for each grid. Here, the symbol is
This refers to figures such as land patterns in printed board pattern diagrams.

[Problems to be Solved by the Invention] When cutting out the figure explained above, taking into account the size and positional deviation of the symbol, it is best to cut out the figure as a rectangle larger than the size of the symbol, as shown in Fig. 4. Normal.

Therefore, there is a problem in that line segments other than the symbol to be recognized, line segments emanating from this symbol, etc. become noise, which prevents the recognition rate from increasing.

The present invention aims to provide a method for solving these problems and improving the recognition rate.

[Means for Solving the Problems] FIG. 1 shows a block diagram of the principle of a method for improving recognition rate in figure recognition according to the present invention.

In FIG. 1, 1 is image data cut out from the image memory 5 at the center of the grid, 2 is a ROM that stores data of the figure to be recognized and proximal masks, and 3 is an AND
It is a gate.

The extracted image data 1 and mask data 2 are
The data is inputted to an AND gate 3 and subjected to a logical product, and the resultant data 4 is subjected to recognition processing.

[Operation] Figure 1 conceptually shows the masking process.
Actually, the cut out image data 1 is scanned with a mask, and for each coordinate value, the logical product with the data of the same coordinate value on the mask is taken, and the result data 4 is assembled.

In result data 4, it is outside the recognition target figure,
Most of the surrounding images that become noise during recognition processing are removed.

The mask is made slightly larger than the figure to be recognized, taking into account deviations caused by handwriting, expansion and contraction of paper, errors in determining reference points, etc., and only the types of external shapes of the figure to be recognized are prepared.

[Example] The gist of the present invention will be explained in more detail below with reference to an example shown in FIG.

In this embodiment, a case will be described in which symbols are tWtfiled in a handwritten printed board pattern diagram as shown in FIG.

In such a printed board pattern diagram, the symbol to be recognized is a land pattern, and the outer shape is, for example, a 3.5mmφΦ great circle, 2°0III
It is assumed that there are two types of small circles of Ilφ.

Among symbols with the same outer diameter, as shown in Figure 3, there are symbols that are completely filled in, half filled in, double circles, double circles, and circles with diagonal lines. There are more than 10 types of symbols, such as those with a box inside, and those with an X in a circle.

Symbol recognition involves recognizing which of these symbols a figure to be recognized is.

Processing for symbol recognition is performed, for example, in the following procedure.

■First, as preprocessing, it is detected whether a symbol exists or not in the center of the cut out image data, and if there is, it is detected whether it is a large circle or a small circle.

■If a large circle symbol is detected, mask processing is performed using a large circle mask, and if a small circle symbol is detected, mask processing is performed using a small circle mask, and the resulting image data is created.

■The image data resulting from the mask processing is recognized by performing successive comparisons with a dictionary that stores each symbol data.

Figure 2 shows cutout image data (a) containing two great circle symbols and cutout image data (for bl) containing two small circle symbols, as shown in Figure 4. This shows image data as a result of processing with large circle and small circle masks.

As shown in the figure, each mask is made slightly larger than the outer diameter of the symbol to allow for misalignment, so a very small amount of unnecessary images remain, but most of the unnecessary images have been removed.

[Effects of the Invention] As explained above, according to the present invention, the recognition rate of figure recognition can be improved by removing noise around the figure to be recognized, and the practical effects in information processing technology are extremely large. There is something.

[Brief explanation of drawings]

Figure 1 is a block diagram of the principle of the present invention. Figure 2 is a diagram showing the mask processing results of an embodiment of the present invention. Figure 3 is a printed board pattern diagram of an example of the drawing to be recognized. Figure 4 is cutout image data. FIG. In the drawings, 1 is cropped image data, 2 is mask data, 3 is AND gate, 4 is mask processing result image data, 5 is image memory, ! -1! (rl”I /l 原ff 7°a-ye thingl
Intoxication 1 Maro'nd lol Example of Mamata's fork L1 Hi 6 Kaden Hechomonryo 2r
E1 ant〉to1 anti IXa7-second eye' Mine 3 figure gob 1 conish jL hey own chiaba month i1 [condolence 4fI

Claims (1)

[Scope of Claims] In a figure input processing device that automatically reads and recognizes handwritten drawings written on the basis of grids provided at equal intervals in an orthogonal coordinate system, image memory (5 ), the image data (1
), comprising mask data (2) approximating the outer shape of the figure to be recognized, and means (3) for performing a mask operation;
) A recognition rate improvement method in figure recognition, characterized in that a recognition process is performed on the result data (4).