JPH0394385A - Drawing reader - Google Patents

Abstract

PURPOSE:To easily extract a straight line intersecting with many lines or a circular graphic intersecting with many lines by accurately vectorizing a linear component or a circular graphic component. CONSTITUTION:The drawing reader is constituted of an image input means 1, a binary image storing means 2, a character separating means 3, a thinning processing means 4, a thinned image storing means 5, a feature point extracting means 6, a feature point storing means 7, a straight line extracting means 8, a vector data storing means 9, a circular graphic extracting means 10, and a vectorizing processing means 11. Thereby, highly accurate vectorized data are obtained from a linear component and a circular graphic component. Consequently, the drawing reader with a high recognition ratio for a machine design drawing in which many linear and circular graphic components are written can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a drawing reading device that reads drawings such as handwritten mechanical drawings using a scanner and converts them into CAD data.

[Conventional technology]

A conventional drawing reading device will be explained with reference to the drawings.

FIG. 8 is a block diagram of a conventional drawing reading device. In the figure, the conventional drawing reading device includes binary image storage means 2 for storing binary image data input from a scanner;
A line thinning process 4 for thinning the value image data, a character separation means 3 for detecting and separating character components from the binary image data, and converting the image data thinned by the line thinning process means into vector data. The vectorization processing means 9 includes vectorization processing means 9 for converting vector data into vector data, and vector data storage means 11 for storing the above-mentioned vector data. Here, a vector refers to a short line segment that approximates a thinned image, and vectorization processing refers to processing that approximates a thinned image by a set of vectors.

Next, a procedure for converting an image into vector data using a conventional drawing reading device will be explained. The binary image data inputted from the scanner is thinned in a 1,000-stage thinning process 4 to obtain thinned image data with a line width of 1 pixel. Note that the Hilditch method, Deutch method, Yokoi method, Tamura method, etc. have been known for the thinning process.
Either method is used. Next, the character separation means 3 measures the size of each continuous component of the thinned image data, and if this measured value satisfies a preset standard value for determining that it is a character, The continuous components are regarded as characters and separated from the binary image. Next, the vectorization processing means 1.1 traces the image represented by the black dot in the thin line image data pixel by pixel, and processes the traced trajectory by successively approximating it with shorter line segments. Get a vector image.

[Problem to be solved by the invention]

However, in the conventional drawing reading device described above, distortion often occurs due to line thinning, so it is difficult to obtain highly accurate vector data.

In particular, it is difficult to extract a single straight line that intersects with many lines, or a circular figure that intersects with many lines.

SUMMARY OF THE INVENTION An object of the present invention is to solve these problems and provide a drawing reading device that can vectorize linear components and circular diagram components with high accuracy.

[Means to solve the problem]

The figure reading device of the present invention includes: an image input means for reading a drawing and creating a binary image having a binary density level; a binary image storage means for storing the binary image; and a binary image storage means for storing the binary image. a character separating means for detecting and separating character components from an image in a storage means; a thinning processing means for creating a thinned image by thinning the binary image from which the character components have been separated; and a thinning processing means for creating a thinned image. a thin line image storage means for storing, a feature point extracting means for inputting the thin line image and detecting feature points from end points, branch points, and intersections; and storing feature point data such as coordinates of the feature points. A length exceeding a preset threshold value is obtained from the thinned image in the thinned image storage means by using a feature point data storage means to store the feature point data, and a HOUGH transformation that limits an image area using the feature point data. a straight line extracting means for extracting a straight line component, converting it into vector data and outputting it; a vector data storage means for storing the output of the straight line extracting means; circular figure extracting means for detecting the center point of a circular figure component, extracting the circle figure forming component based on position data of the center point, and outputting the extracted circle figure component to the vector data storage means; the straight line extracting means; and the circular figure extracting means. and vectorization processing means for extracting the graphic components that could not be extracted in the above steps, vectorizing them, and outputting the vectorized vectors to the vector data storage means.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a structural diagram showing an embodiment of a drawing reading device according to the present invention. The drawings to be read by this device are mechanical design drawings. Furthermore, characters used in this drawing are limited to alphanumeric characters.

The image input means 1 reads a drawing with a scanner and outputs binary image data with density levels of 0 and 1.

Binary image storage means 2 stores image data output from image input means 1.

The character separation means 3 reads out the binary image (image data) stored in the binary image storage means 2, measures the length of the minimum rectangle surrounding the continuous component for each continuous component, and determines the length of the rectangle. Determine whether the threshold value for determining it as a character is met.Then, determine whether the continuous segment is a character or a line figure, and delete the character from the image data.Here, character recognition is performed. will be omitted.

The thinning processing means 4 thins a binary image and outputs the thinned image to the thinned image storage means 5, and its contents are based on the conventional technology.

The feature point extraction stage 6 inputs the thin line image and tracks pixels with a pixel value of 1 representing a figure for each continuous component in the thin line image. At this time, the pattern of pixel values of 8 pixels around the tracking pixel is examined for each pixel, and the pixel at the end of the figure that is the end point and the pixel that is branching in three directions in the figure that is the branch point are determined. Pixels branching in four directions in a figure at an intersection are detected. These points are collectively referred to as feature points. The coordinate positions and types of the feature points and information on which direction they are continuous are stored in the feature point storage means 7 in a table format as shown in FIG. Note that, as shown in FIG. 3, the directions here are defined as seven directions centered on the pixel of interest.

After all feature points have been detected, the straight line extraction means 8 extracts straight line components from the image. The straight line extracting means 8 extracts straight lines from the image in the thinned image storing means 5 at high speed and with high precision using the data in the 1,000-stage feature point memory 7. The flow of this process is shown in FIG.

In FIG. 4, first, the feature point data is read from the feature point storage means 7, and a set of two feature points whose feature points are as far apart as possible and whose consecutive directions are close are searched for in step 21.
22>. There is a possibility that a straight line exists between these two feature points. This straight line is extracted by HOUGH transformation in which the area is limited by the feature point data.

Next, a partial area is set (step 23), and this setting method will be explained using FIG. Now, two feature points A and B
It is assumed that the following is retrieved from the data in the feature point storage means 7. A rectangular area surrounding points A and B and matching the slope of the straight line connecting points A and B is defined as a H O UGH conversion area. straight line AB
The length of the side in the direction perpendicular to is set to an appropriate length in consideration of the possibility that the branch point B is shifted due to thinning distortion.

Next, HOUGH conversion is performed (step 24).

In the HoucR transformation, 2 expressed in x-y coordinates
A straight line m in dimensional space is expressed by equation (1), where θ is the angle that a straight line that is perpendicular to the straight line and passes through the origin makes with the X axis, and ρ is the distance between the straight line m and the origin, and p =x-cos θ + y-sin θ
・・・・・・K1〉 Points on the straight line are expressed as ρ by formula (1〉)
When converted to the parameter space represented by and θ, it is represented by a single locus.

From the properties of the HOUGH transformation described above, the trajectories in the ρ-θ space of points existing on a straight line in the x-y space all intersect at one point, so a straight line can be extracted by inversely transforming this intersection ( Step 25).

By the way, since straight line extraction using HOUGH transformation requires a huge amount of calculation, this problem is solved by region restriction (step 23 above).

Next, the starting point and ending point of the vector are determined from the coordinate position of the feature point, and these values are output as vector data (step 26).

The circle figure extracting means 10 utilizes a drafting rule that a machine design drawing is described so that the intersection of center lines is located at the center position of a circle figure (FIG. 6). Therefore, it cannot be applied to drawings that do not follow this drafting rule.

Next, the flow of circular figure extraction processing will be explained. First, the vector data storage stage 9 is examined to find the coordinates of the intersection of two orthogonal lines. Next, a histogram is taken using the distance from the center position as a parameter (FIG. 7), and its peak value is detected. If the peak value exceeds a threshold value set in proportion to the distance from the center position, it is determined that a circular figure exists. Vector data is generated from the center position and radial distance obtained in this manner, and is output to the vector 1 to data storage means 9.

The vectorization processing means 11 performs vectorization processing on the figure components that could not be detected by the straight line extraction means 8 and the circular figure extraction means 10 in a manner similar to the conventional method, and outputs the result to the vector data storage means 9.

〔Effect of the invention〕

As described in detail above, according to the present invention, it is possible to obtain highly accurate vectorized data of straight line components and circular figure components, which improves the conventional drawbacks, and many straight line components and circle figure parts are described. This has the effect of providing a drawing reading device with a high recognition rate for mechanical design drawings.

It is a composition diagram showing an example of the device.

1... Image input means, 2... Binary image storage means, 3
... Character separation means, 4... Thinning processing means, 5...
・Thinning image storage means, 6...Feature point extraction means, 7.
...Feature point storage means, 8...Line extraction means, 9...
Vector data storage means, 10... circular figure extraction means,
11... Vectorization processing means.

Claims (1)

[Claims] image input means for reading a drawing and creating a binary image having a binary density level; binary image storage means for storing the binary image; and detecting character components from the image in the binary image storage means. a thinning processing means for creating a thinned image by thinning the binary image from which the character components have been separated; a thinning image storage means for storing the thinned image; a feature point extraction means for inputting a thin line image and detecting feature points consisting of end points, branch points, and intersections; a feature point data storage means for storing feature point data such as coordinates of the feature points; and a feature point data storage means for storing feature point data such as coordinates of the feature points. By using HOUGH transformation that limits the image area using data, straight line components with a length exceeding a preset threshold are extracted from the thinned image in the thinned image storage means and converted into vector data. a straight line extracting means for outputting the straight line extracting means; a vector data storing means for storing the output of the straight line extracting means; and a center point of the circle diagram component is detected from the thinned image using the data of the vector data storing means. circular figure extracting means for extracting circular figure components based on point position data and outputting them to the vector data storage means; and extracting figure components that could not be extracted by the straight line extracting means and the circular figure extracting means. A drawing reading device comprising vectorization processing means for vectorizing and outputting the vectorized data to the vector data storage means.