JPH04198741A - Shape defect detecting device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to an apparatus for detecting geometrical defects in objects.

[Prior Art] For example, as a method for detecting partial defects and redundancies in circuit patterns on printed circuit boards,
An example of this method is the scaling method described in the paper "Flaw Extraction Apparatus Method for Complex Patterns" published in 1994-c, No. 5.

The principle of this scaling method will be explained using FIG. 1.

First, as shown in FIG. 1(a), for a target pattern with a geometrical defect such as a minute chip or a protrusion, as shown in FIG. 1(b),
), as shown in Figure 1(c), a process of removing a few surrounding pixels (reduction process) is performed, and then, as shown in Figure 1(c), a process of thickening a few surrounding pixels (enlargement process) is performed, resulting in a minute image. Remove protrusions.

The image obtained in this manner is used as a semi-normal pattern, and by calculating the difference between this semi-normal pattern and the original target pattern, minute protrusions are detected as shown in FIG. 1(d).

In addition, after the enlargement process (Fig. 1 (b')), the reduction/11 process (
By performing (C') in Fig. 1, a quasi-normal pattern is created by removing the 111/J% chipping, and by taking the difference from the original target pattern, the pattern shown in Fig. 1 (d') is created. As shown in the figure, minute chips are detected.

By performing these processes in parallel, minute chips and protrusions can be detected all at once, as shown in FIG. 1 (6).

[Problems to be Solved by the Invention] In the conventional pattern defect detection method described above, it is necessary not to lose the geometrical characteristics of the pattern and not to make contact with adjacent patterns. %The number of times of processing is limited.

As a result, there is a problem in that only minute geometrical defects in the missing and surplus portions can be detected.

SUMMARY OF THE INVENTION An object of the present invention is to provide a shape defect detection device capable of solving the above-mentioned problems and detecting a shape defect of an object regardless of its size.

[Means for Solving the Problems] In order to achieve the above object, the shape defect detection device of the present invention does not change the shape of the object in the inspection image as in the case of enlargement processing and reduction processing. A geometrical contour model is created using only the contour information of the object, and this contour model is used as a reference pattern and compared with the target image to detect geometric defects.

Accordingly, the present invention provides an object shape defect detection device comprising an imaging means for imaging an object to be inspected to obtain a target image, and an image processing means for processing the target image, wherein the image processing means is The present invention includes a contour model creation means for creating a contour model from a target image, and a shape defect detection means for detecting a shape defect on the surface of the object to be inspected by comparing this contour model with the original target image. I have to.

The present invention also provides an object shape defect detection device comprising: an imaging means for imaging an object to be inspected to obtain a target image; and an image processing means for processing the target image; A contour model creation means for creating a contour model from a target image; and a shape defect detection means for detecting a shape defect on the surface of the object to be inspected by measuring the contour model in the horizontal and vertical directions. We are trying to have the following.

The contour model creation means creates the contour model by, for example, approximating the target image with a straight line and a circular arc.

Further, the contour model creation means takes, for example, a projection distribution of contour pixels in the same direction as a previously predicted straight line, and sets a pixel point sequence near the maximum frequency position as a straight line that approximates the target image.

Furthermore, the contour model creation means, for example, determines the center position of a circle that passes through a pixel point sequence distant from two adjacent straight lines and is inscribed in the two straight lines, and calculates the inscribed circle near the maximum frequency position. Let it be a circular arc that approximates the target image.

[Operation] In the shape defect detection apparatus of the present invention, when the image capturing means images the object to be inspected to obtain a target image, the image processing means operates as follows.

First, the contour model creation means of the image processing means approximates the target image with a straight line, and approximates curved portions such as corners with circular arcs.

In this way, a contour model is created by approximating the target image with straight lines and circular arcs.

Then, the shape defect detection means of the image processing means detects a shape defect on the surface of the object to be inspected by comparing the created contour model with the original object image.

Therefore, it is possible to deal with similar changes in the shape of the same object, and to detect defects and redundant geometric defects regardless of size.

The shape defect detection means of the image processing means quantitatively detects the shape defects on the surface of the object to be inspected by measuring the area of the created contour model in the horizontal/vertical directions.

[Example 1] An example of the present invention will be described below with reference to FIGS. 2 to 7.

FIG. 2 is a block diagram showing the system configuration of a shape defect detection device according to an embodiment of the present invention.
201 is an object to be inspected, 202 is a camera, 203 is an image processing unit, and 204 is an output monitor.

First, the surface of the object to be inspected 201 is imaged by the camera 202,
The obtained image signal is sent to the image processing section 203.

The image processing unit 203 creates a contour model from the sent image signal. This contour model is then compared with the original target image to detect geometric defects in the missing and surplus portions of the inspection object 201, and the results are displayed on the output monitor 204.

FIG. 3 is a diagram showing an outline of defect detection using a contour model.

FIG. 3(a) shows the result of imaging the object to be inspected 201.
The obtained target image is shown.

In FIG. 3(a), the shaded area represents the background, and the other areas represent the object.

To create a contour model, first, as shown in FIG. 3(b), a window is provided in advance in a region where a contour is thought to exist. Using the contour pixels in this window area, the contour can be drawn using the straight line equation (Y=aX
+b).

For each window, define the equation of the straight line in the same way, and
As shown in figure (d), the slope, Y-intercept, starting point, and ending point are
Record it as a table.

In addition, as shown in FIG. 3(e), shape defects are quantitatively detected by measuring the distance Ω in the parallel direction or the distance in the perpendicular direction to the defined straight line. be able to.

4 to 6 are flowcharts showing specific processing for creating a contour model performed by the one-image processing unit 203, and a processing process based on the flowchart.

In step 401, the image processing unit 203 converts the target image into two
It is converted into a value and converted so that one pixel has a value of O (black) or 1 (white).

Here, in Figure 5 (,), the diagonal line is 0 (black).
, the other parts are set to 1 (white).

Next, in step 402, a fixed window is provided as shown in FIG. 5(b) for each part where the boundary between black and white is defined by a straight line in advance, as explained in FIG. , limit the area for image processing. Then, processing is performed on the area of this fixed window.

Here, the description will focus on the fixed window shown in the shaded area in FIG. 5(b).

First, in step 403, a boundary pixel (hereinafter referred to as a change point) where a pixel changes from black to white with respect to a fixed window area is determined as shown in FIG. 5(c). Change points are detected by searching in the vertical direction.

Next, in step 404, a projection distribution is taken in the horizontal direction of the fixed window, as shown in FIG. 5(f), for the extracted change points, as indicated by the x marks shown in FIG. 5(d). Then, the maximum degree value is determined, and in step 405, the change point is limited to the point of change near that position.

Perform linear approximation using the least squares method.

In this way, by performing linear approximation by focusing on the predicted changing points on the straight line, it is possible to accurately extract the outline of the object.

After linear approximation, portions where corners and the like are previously defined as curved lines are approximated with circular arcs.

Regarding the method of arc approximation, Ll in FIG. 6(a).

The explanation will focus on the corner portion between the two straight lines of Ll.

First, in step 406, the image processing unit 203 moves the two straight lines LL and Ll inward by W pixels, as shown in the sixth Wi (b), in order to extract the contour pixels of the corner part, and then the straight lines and the target Find the points of intersection with the boundary of the object, and perform boundary line tracing between these points of intersection.

Next, in step 407, as shown in FIG. 6(c), the extracted individual contour pixels are transformed by generalized Hough transform.
Parameter conversion is performed to the radius Re of the circle inscribed in the two straight lines LL and Ll. Then, for the converted inscribed circle radius Re, a 1 degree distribution is taken as shown in Figure 6 (f), the inscribed circle radius Rc (■ax) of the maximum degree value is determined, and Figure 6 (d
), the inscribed circle with radius Rc (wax) is
Define as the outline of the corner.

By performing the same process on other curved parts and summing up the points of intersection and contact with the previously determined straight line, it is possible to create a contour model using straight lines and circular arcs, as shown in Figure 6(e). .

FIG. 7 is a diagram showing the principle of generalized Hough transform, which is an important element necessary for performing circular arc approximation.

In Fig. 7(a), the two straight lines are Ll, Ll, the straight line that bisects the angle of the two straight lines is Y=eX+f, the angle between the straight line and the bisector is θ, and the intersection of the two straight lines is (Xcrs).

Y crs), the center point of the inscribed circle is (Xc, Yc), the arbitrary contour pixel coordinates are (Xi, Yi), the distance between the intersection of the two straight lines and the center point of the inscribed circle is D, the arbitrary contour pixel Let Ri be the distance between the coordinates and the center point of the inscribed circle, and let Rc be the radius of the inscribed circle.

Here, it is assumed that the two straight lines Ll and Ll are defined in advance by linear approximation.

2 straight lines Ll, the center point of the circle inscribed in Ll (Xc).

The locus of Yc) can be considered to exist on a straight line passing through the middle of the two straight lines, in other words, on the bisector Y=eX+f.

If we define the center point (Xc, Yc) of an arbitrary inscribed circle, we can calculate the distance between the intersection of two straight lines and the center point of the inscribed circle, and the distance between the arbitrary contour pixel coordinates and the center point of the inscribed circle. The distance R1 and the radius Rc of the inscribed circle are respectively D = Xc-Xcrs
+Yc-Ycrs)"-(1)Ri=Xc-X
i + Yc-Yi)" - (2) Rc=Dsi
It can be expressed as nθ·(3). Any contour pixel coordinates (Xi, Yi) are 2
In order to be a point on a circle inscribed in straight lines LL and Ll, Ri= Re...(4)
You need to be satisfied that

The inscribed circle that satisfies this condition is determined for all the extracted contour pixels, and as shown in FIG. 7(b), the maximum frequency radius Rc
(+max) is determined, and the contour is approximated by an inscribed circle with this radius.

When the bisector is defined as Y=eX+f, the above (1) ~
The result of solving equation (4) for the center point (Xc, Yc) of the inscribed circle is: Xc = (B + f':T)/A -
(5) Convert any contour pixel to the radius of a circle inscribed in two straight lines using the above conversion formula, where Yc = e be able to.

[Effects of the Invention] As explained above, according to the present invention, there is an effect that a shape defect of an object can be detected regardless of the size.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram showing the principle of pattern defect detection using the conventional scaling method, Fig. 2 is a block diagram showing the system configuration of a shape defect detection device according to an embodiment of the present invention, and Fig. 3 is a contour model. Explanatory diagram showing an overview of defect detection by
The figure is a process flowchart for contour model creation, Figure 5 is an explanatory diagram showing the process of linear approximation, Figure 6 is an explanatory diagram showing the process of arc approximation, and Figure 7 is an explanatory diagram showing the principle of generalized Hough transform. 201... object to be inspected, 202... camera, 203...
...Image processing unit, 204...Output monitor. Applicant Hitachi, Ltd. (and 1 other person) Agent Patent attorney Tomi 1) Kazuko Figure 4 Figure 3

Claims (5)

[Claims] 1. In an object shape defect detection device comprising an imaging means for imaging an object to be inspected to obtain a target image, and an image processing means for processing the target image, the image processing means generates a contour model from the target image. A shape defect characterized by having a contour model creation means for creating a contour model, and a shape defect detection means for detecting a shape defect on the surface of an object to be inspected by comparing this contour model with an original target image. Detection device. 2. In an object shape defect detection device comprising an imaging means for imaging an object to be inspected to obtain a target image, and an image processing means for processing the target image, the image processing means generates a contour model from the target image. It is characterized by having a contour model creation means for creating a contour model, and a shape defect detection means for detecting a shape defect on the surface of the object to be inspected by measuring the area of the contour model in the horizontal/vertical direction. Shape defect detection device. 3. The shape defect detection device according to claim 1 or 2, wherein the contour model creation means creates a contour model by approximating the target image with a straight line and a circular arc. 4. 4. The shape defect detection device according to claim 3, wherein the contour model creation means takes a projection distribution of contour pixels in the same direction as a previously predicted straight line, and converts the pixel point sequence near the maximum frequency position to the target image. A shape defect detection device characterized by using an approximate straight line. 5. The shape defect detection device according to claim 4. The contour model creation means determines the center position of a circle that passes through a pixel point sequence apart from two adjacent straight lines and is inscribed in the two straight lines, and calculates the inscribed circle near the maximum frequency position from the target image. A shape defect detection device characterized by detecting approximate circular arcs.