JPH04340674A - Edge information extracting device - Google Patents

Abstract

PURPOSE:To extract an exact edge feature point by a simple operation without distinguishing whether an objective edge is one part of a straight line or one part of a circular are in the edge information extracting device to extract specified edge information from the monitor image, etc., of an object. CONSTITUTION:This device is provided with a means displaying a multilevel image, a means setting more than one windows in the multilevel image, a means superimposing the window set by this means to the multilevel image and to display the image, a means extracting the edge information at a part instructed by the window in this multilevel image, a means detecting the coordinates of an edge representative point in the window twice at least based on the edge information while virtually switching window width to a central part, and a means calculating true edge representative point coordinates based on the first edge representative point coordinates, second edge representative point coordinates and window width change rate.

Description

[Detailed description of the invention]

0001

[Field of Industrial Application] The present invention relates to an edge information extraction device for extracting specific edge information from a monitor image of an object in a measuring device for measuring the dimensions and coordinates of an object, and in particular, the present invention relates to an edge information extraction device for extracting specific edge information from a monitor image of the object. The present invention relates to an edge information extraction device that can accurately extract edge information without distinguishing between straight lines and circular arcs.

0002

[Background Art] Conventionally, in the field of object measurement, for example, a microscopic image of an object to be measured is taken with a TV camera, and this is displayed on a monitor screen, and several points on the screen are specified. A measuring device is known that calculates the coordinates, width, length, radius, etc. of the object to be measured. In this type of apparatus, measurement is performed by specifying a point on an edge in an image that defines the width, length, etc. to be measured of the object to be measured. Therefore, edge information extraction processing is required to extract information on specified edges on the monitor image. For this edge information extraction process, for example, there is a method of binarizing multivalued image data using the intermediate value of screen brightness (density) as a threshold level, clarifying edge parts with this preprocessing, and then extracting edge information. . However, depending on the multilevel image, it is difficult to obtain an appropriate threshold level that can be applied to the entire image. In order to improve this point, the present inventors previously proposed a method of setting an arbitrary number of windows in a multilevel image (Japanese Patent Application No. 2-35799). With this method, it is possible to set a local threshold level within a window of limited area, so the entire multilevel image can have multiple threshold levels, and the entire multilevel image can be binarized using a single threshold level. There is no such ambiguity.

0003

[Problem to be Solved by the Invention] In the above-mentioned window setting type edge information extraction device, the coordinates of the black-white transition point are determined for each pixel column from the binarized data in each window, and The square method is applied to determine representative coordinates within the window (for example, the coordinates of the midpoint) from the arithmetic mean. According to this method, if the target edge is part of a straight line, there will be little error, but if the target edge is part of a circular arc, an error will occur between it and the actual edge feature point (the calculated coordinates will be (deviated from the feature point toward the center of the arc). However, when considered pixel by pixel, there is a problem in that the above-mentioned error is unavoidable because there is no way to determine whether an edge within a window is part of a circular arc or a part of a straight line.

The present invention solves this problem and calculates accurate edge feature points (true edge representative point coordinates) with simple operations without distinguishing whether the target edge is a part of a straight line or a part of a circular arc. ) is intended to be extracted.

[0005]

[Means for Solving the Problems] In order to achieve the above object, the present invention provides means for displaying a multi-valued image, means for setting one or more windows in the multi-valued image, and a means for setting one or more windows in the multi-valued image. means for displaying a window superimposed on the multi-valued image, means for extracting edge information of a portion designated by the window in the multi-valued image, and means for extracting edge representative points within the window based on the edge information. Means for detecting the coordinates at least twice by virtually switching the window width toward the center, and detecting the true edge based on the coordinates of the first edge representative point, the second edge representative point coordinates, and the window width change rate. The present invention is characterized by comprising means for calculating representative point coordinates.

[0006]

[Operation] If the edge in the window is part of a straight line, the first and second edge representative points do not change even if the window width is changed. On the other hand, if the edge within the window is part of a circular arc, changing the window width will cause an error between the first and second edge representative points. This error is closely related to the window width change rate, and makes it possible to calculate the true edge representative point coordinates using Equation 1 below. That is, if the first edge representative point coordinate is P0', the second edge representative point coordinate is P0'', and the window width change rate from the first to the second time is 1/m, then the true edge representative point coordinate P0 is is roughly as follows.

[0007]

[Math. 1] P0 = (m2・P0''-P0')/(m2-
1)

The above equation 1 can be applied regardless of whether the target edge is a part of a straight line or a part of a circular arc.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. FIG. 1 is a block diagram of a main part showing an embodiment in which the present invention is applied to a size measuring device for a minute object, and FIG. 2 is a perspective view showing the overall configuration thereof. This measuring device can be broadly classified into
As shown in FIG. 2, there is a measuring microscope 2 that magnifies the optical image of the object to be measured 1, a CCD camera 3 attached to the measuring microscope 2, and a dimension measurement process by processing the image signal from the CCD camera 3. The data processing device 4 is configured to perform the following steps. The measuring microscope 2 is equipped with a stage 12 on which the object to be measured 1 is placed. This stage 12 is movable in the X direction, Y direction, and rotational direction by an X direction stage adjustment handle 13, a Y direction adjustment handle 14, and a stage rotation knob 15, respectively. Further, a lens support rail 16 extending upward is fixed to the back of the main body 11, and a lens assembly 17 is supported on this lens support rail 16 so as to be movable up and down. The lens assembly 17 includes four objective lenses 19a, 19b, 19c, and 19d with different magnifications attached to a four-hole revolver 18 on the side facing the object to be measured 1. Further, the lens assembly 17 includes an eyepiece 20 on its upper part, and the eyepiece 20 and the objective lens 19a.
-19d are optically coupled via the erect trinocular tube 21. This lens assembly 17 is moved up and down by a focus adjustment handle 22. Further, the erecting trinocular tube 21 is moved up and down by the erecting trinocular tube vertical movement knob 23. The CCD camera 3 attached to the top of the erect trinocular tube 21 of this measuring microscope 2 has a size of, for example, 768×49
The number of pixels is 3.

The data processing device 4 is configured as shown in FIG. Note that even if each functional block inside the processing device main body 30 is realized by hardware,
It may be realized by a CPU and software. An analog image signal of an enlarged image of the object to be measured 1 captured by the CCD camera 3 is input to the A/D converter 31 . The A/D converter 31 converts the analog image signal into, for example, 8
Convert to bit multivalued image data. A/D converter 31
The multivalued image data output from the multivalued image memory 32
It is designed to be stored in . Further, the data processing device 4 is provided with a keyboard 33, through which the number, position and size of windows, calculation conditions for dimension measurement, etc. can be set. Data for setting the window from the keyboard 33 (hereinafter referred to as window data) is stored in the window setting memory 34. The window data stored in the window setting memory 34 and the multi-value image data of the object to be measured 1 stored in the multi-value image memory 32 are superimposed in the synthesizing means 35 and displayed on the CRT display 36. There is.

On the other hand, the multi-value image data stored in the multi-value image memory 32 and the window data stored in the window setting memory 34 are also supplied to an in-window binarization means 37. The in-window binarization means 37 binarizes only the multivalued image data within the specified window to clarify edge portions within the window. The image data binarized by the in-window binarization means 37 is supplied to the edge coordinate detection means 38. The edge coordinate detection means 38 extracts edge information from the binarized data, and detects the coordinates of a representative point of the edge, for example, the midpoint (hereinafter referred to as edge coordinates). The in-window binarization means 37 and the edge coordinate detection means 38 appropriately access the in-window processing memory 42 and execute the processing. The edge coordinates detected by the edge coordinate detection means 38 are stored in the coordinate memory 39. The calculating means 40 calculates dimensions such as the width, length, radius, etc. of the object to be measured 1 based on the edge coordinates of each window stored in the coordinate memory 39 and the calculation conditions specified by the keyboard 33. This calculation result is displayed on the liquid crystal display 41 as a measurement result.

Next, the operation of the apparatus configured as described above will be explained. Measured object 1 placed on stage 12
After being magnified by the measuring microscope 2, the optical image of CC
The image is captured by the D camera 3. The analog image signal from the CCD camera 3 is A/D converted into 8-bit multi-value image data by an A/D converter 31 and stored in a multi-value image memory 32. When the multi-value image data is stored in the multi-value image memory 32, the multi-value image data including halftones is displayed on the CRT display 36. An arbitrary number of windows of size M×N (for example, 20×14) are set within the multivalued image displayed on the display 36, and a threshold level is set within each window to binarize the multivalued image. do. FIG. 3 shows an example in which straight edges are included in the window binarized in this manner. The illustrated binarized data has a step-like shape because it has been enlarged, but when the N points obtained as edge information (14 black points in the illustrated example) are arithmetic averaged using the least squares method, the illustrated An approximate straight line L1 like this can be obtained. For example, the midpoint P0 (point N/2) of this straight line L1 becomes the representative coordinates (edge coordinates) of the binarized data within the window. In this way, when the edge within the window is part of a straight line, the obtained edge coordinates almost match the actual edge feature point.

On the other hand, the situation is different when the edge of a circular arc is included within the window. Arc P1P0P in Figure 4
2 is included on the circumference of a circle having a center O and a radius R0, and the center is rotated counterclockwise by an angle θ from the right horizontal line H. Here, the case where R0=20 and M×N=20×14 is taken as an example. If similar edge extraction processing is performed within this window, 14 points will be obtained, but if these points are arithmetic averaged using the method of least squares, the obtained approximate straight line L2 will connect points P1' and P2'. street, its midpoint P0
' is a position shifted from the midpoint of the arc P1P0P2, that is, the true edge feature point P0. The present invention attempts to reduce this error as much as possible.

Points P1 and P2 at both ends of arc P1P0P2
When the midpoint of the straight line L3 connecting the lines is Q, the approximate straight line L2
The deviation Δ from the midpoint P0' is equal to the area S of the arc P1P0P2 divided by the width N of the window. In other words, the area of the parallelogram P1P1'P2P2' is equal to S. The area S of the arc P1P0P2 is fan-shaped OP
It is expressed by Equation 2 below, which subtracts the area of triangle OP1P2 from the area of 1P2.

[0015]

[Formula 2] S=R02 {(θ2-θ1)-sin(θ2-θ1)}
/2

On the other hand, since points P0' and P0 are both on the window midline T, their Y coordinates are both R0
・Sin θ. At this time, the Y coordinates of points P1 and P2 are shifted up and down by N/2 from this value. Therefore, the following relationship holds true.

[0017]

[Math. 3] sin θ1=sin θ−N/2R0sin θ2
= sinθ+N/2R0

Using these relational expressions, the points P0 and P0
'The ratio between the distance E, that is, the error E, and the radius R0 is expressed as follows.

[0019]

[Equation 4] E/R0 = cos θ-(cos θ2-cos θ1
)/2 +{(θ2−θ1)−
sin(θ2-θ1)}×R0/2N

[0020] The first term on the right side of the above equation is the X coordinate of P0, and the second term on the right side is P1, P2.
The X coordinate of the midpoint between them and the third term on the right side are related to the aforementioned Δ, which is the arc area S divided by the window width N. The error E mentioned above is closely related to the window width N. That is, E/R of number 4
The relationship between 0 and N/R0 is illustrated in FIG. In this figure, the inclination θ of the radius R0 of the circular arc is used as a parameter, and is set at every 5° from 0° to 45°. As is clear from this figure, if the slope θ is constant, the error E decreases as the window width N becomes smaller. It also shows that if the window width N is constant, the error E decreases as the slope θ becomes smaller. The extent to which the error E decreases with respect to a decrease in the window width N becomes clear when displayed on a logarithmic scale as shown in FIG. This Figure 6 shows the same content as Figure 5, but each characteristic is a straight line, and from its slope, E/R0 is the reduction rate of N/R0 (m)
This shows that the value decreases as the square of .

The present invention focuses on this point and performs the following signal processing. That is, when an M×N window is set as shown in FIG. 7, the first edge feature point detection is performed using this window. The feature point detected at this time is the aforementioned P0'. Next, the window width is virtually set to 1/m toward the center. The dashed-dotted lines V1 and V2 in FIG. 7 indicate the range of this virtual window. Once a virtual M×(N/m) window is set in this way, the second edge feature point P0" is found again in the same manner as described above. This second feature point P0" is As shown in Figure 8, parallel to straight line L2, midpoint P of the circular arc
This is the midpoint of straight line L4, which is closer to 0. This point P0”
also has an error with respect to the true point P0. However, this error has decreased compared to the first point P0'. As shown in Figures 5 and 6, the degree of decrease in this error is related to the window width decrease rate m, but in general, the error is approximately proportional to the square of the decrease rate m2. Decrease. Therefore, in the present invention, the true edge feature point P0 is calculated using the above-mentioned equation (1). This number 1 shows that the distance (error) between the points P0 and P0'' shown in Figure 8 is the point P0 and P0'
This takes advantage of the fact that the distance (error) is 1/m2. This calculation formula can be applied to both straight lines and circular arcs. Therefore, the operator can find true edge feature points without distinguishing between them.

The present invention can be implemented with various modifications. As an example, if m = 2, the error will be 1/4, and m
=3, the error is reduced to 1/9, but according to the formula of Equation 1, the same result can be obtained no matter how the value of m is selected. Further, in the embodiment, the window width is reduced by N, but if the arc crosses in the horizontal direction, the window width may be reduced by M.

[0023]

As described above, according to the present invention, in an edge information extraction device that extracts specific edge information from a monitor image of an object, the target edge is a part of a straight line or a part of an arc. Accurate edge feature points can be extracted with simple operations without distinguishing between

[Brief explanation of drawings]

FIG. 1 is a block diagram of main parts showing an embodiment of the present invention.

FIG. 2 is an external perspective view of a device according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram of a straight edge.

FIG. 4 is an explanatory diagram of an arcuate edge.

FIG. 5 is a characteristic diagram of window width and feature point error.

FIG. 6 is a logarithmic display characteristic diagram of FIG. 5.

FIG. 7 is an explanatory diagram of window width switching.

FIG. 8 is a relationship diagram between each degree feature point and true feature point.

[Explanation of symbols]

1...Object to be measured, 2...Measuring microscope, 3...CCD camera,
4...Data processing device, 32...Multi-valued image memory, 34...Window setting memory, 35...Composition means, 36...CRT display, 37...In-window binarization means, 38...Edge coordinate detection means, 39...Coordinate memory, 40... Arithmetic means, 42... Memory for processing within the window.

Claims (1)

[Claims] 1. Means for displaying a multivalued image, means for setting one or more windows in the multivalued image, and means for displaying the window set by this means superimposed on the multivalued image. a means for extracting edge information of a portion designated by a window in the multivalued image; and a means for virtually switching the coordinates of an edge representative point within the window based on the edge information so that the window width is directed toward the center. and means for calculating the true edge representative point coordinates based on the first edge representative point coordinates, the second edge representative point coordinates, and the window width change rate. edge information extraction device.