JPS62148838A - Defect recognizing method - Google Patents

Abstract

PURPOSE:To recognize a quantity indicating features of a defect by reducing the line width of a binary-coded contour part, further tracking this contour part and finding a shape feature curve, and comparing it with a predetermined value and recognizing the contour part. CONSTITUTION:When an image is inputted, an object is lighted and reflected light is photodetected by an image pickup means, and the two-dimensional video signal of the object circuit pattern is obtained through photoelectric conversion. Then, the input image is further converted into a binary signal. Light is reflected irregularly owing to the inclination of the contour part 7 of the circuit pattern, so a part 17 is different in density from a part 15 corresponding to a base material part 5 and a part 16 corresponding to the circuit pattern part 6. The binary image is obtained on the basis of the density. Line thinning processing 2 reduce the line width of the binary pattern of the contour part 7 to one picture element as preprocessing for tracking the shape of the contour and converting it into feature data. Data conversion processing 3 detects the tracking start address of the binary pattern of the contour part 7 reduced in line width to one picture element and calculates the feature data while tracking the contour, thereby generating a linear shape feature curve.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a defect detection method for detecting defects in the appearance of a circuit pattern.

(Prior Art) A conventional device for recognizing defects in the shape of an object to be inspected is disclosed in, for example, Japanese Patent Laid-Open No. 59-214984. In the device of this document, the object to be inspected is
Obtain a binary image of the object by capturing it with a camera, etc., sequentially trace the boundary line between the object and the background in the binary image, detect the addresses of all boundary points on the boundary line, and then A method is used to find the angular change of the tangent to the boundary line and detect defects in the shape of the object from its peak value.

(Problems to be Solved by the Invention) However, with the conventional defect recognition methods described above, it is not possible to know the height and depth of defects that should be noted in, for example, circuit pattern defect inspection. There was a problem that it was difficult to apply it as a device.

The present invention has been made in order to solve the problems of the prior art. The purpose of this invention is to provide a defect recognition method that can recognize defects.

(Means for Solving the Problems) In order to solve the problems of the conventional mud storage technology, the first method of the present invention binarizes a two-dimensional image of a circuit pattern to be inspected obtained by photoelectric conversion, and Obtain a binary image of the outline of the pattern part and other parts, then thin the line width of the binarized outline, and trace the thinned outline to determine the characteristics of the outline shape. The data is calculated to obtain a shape characteristic curve, and the value indicated by the shape characteristic curve is compared with a predetermined threshold value, and when the former value is greater than or equal to the latter value, it is recognized as a defect.

In addition, in the second method of the present invention, instead of recognizing defects by comparing the value indicated by the shape characteristic curve with a threshold value in the first method, the shape characteristic curve of the circuit pattern and a predetermined normal Defects are detected by comparing the shape characteristic curve of the contour.

(Operation) According to the first method of the present invention, first, a circuit pattern to be inspected is illuminated, and the reflected light is photoelectrically converted to obtain a two-dimensional video signal, which is input as a human image. This human image is then binarized, but at this time, light is diffusely reflected at the contour of the pattern, so that a contour pattern is obtained with a density that is different from that of the other portions. Next, the line width of this contour pattern is thinned for later processing. This line thinning process is performed, for example, by reducing a line width consisting of a plurality of pixels to a line width of one pixel. Then, contour tracing is performed sequentially from a certain point on the thinned contour, and a one-dimensional shape characteristic curve reflecting the characteristics of the contour shape is obtained. Next, the value indicated by the obtained shape characteristic curve is compared with a predetermined threshold value, and if the former value is larger than the predetermined threshold value, it is recognized as a defect. At this time, since the value indicated by the shape characteristic curve corresponds to the height, depth, etc. of the defect, quantitative defect recognition becomes possible.

On the other hand, according to the second method of the present invention, the operations up to contour tracing and obtaining the shape characteristic curve are the same, but the operations thereafter are as follows. That is, the obtained shape feature curve is compared with a standard feature curve without defects. In this case, a comparison can be easily made simply by one-dimensional positioning of the two, and defects can be detected quantitatively based on the difference between the two.

(Example) An embodiment of the present invention will be described in detail below based on the drawings.

FIG. 1 is a block diagram showing the processing procedure of the defect recognition method of this embodiment, including image input/binarization processing 1, thinning processing 2
, data conversion processing 3 and comparison processing 4. Each process will be explained in detail later. FIG. 2(a) is a diagram showing an example of a circuit pattern to which the method of the present embodiment is applied. In the figure, 5 is a base material portion, 6 is a circuit pattern portion, and 7 is an outline portion of the circuit pattern. In the example shown in FIG. 2(a), there is no difference in reflection characteristics between the base material part 5 and the circuit pattern part 6, and the circuit pattern can only be distinguished by its three-dimensionality, so it cannot be used in a general defect recognition method. This is an example that is difficult to apply. Figure 2(b) shows the second
This is a diagram showing a human image of the target circuit pattern shown in FIG. It is.

First, the manual image binarization process 1 will be described.

Image input is performed by illuminating the object as shown by the arrow in Fig. 2 (a) and capturing the reflected light by the imaging means (not shown) of the TV left camera.
This is done by receiving light and obtaining a two-dimensional image signal of the target circuit pattern through photoelectric conversion. The input image is further subjected to binarization processing. In the object shown in FIG. 2(a), light is diffusely reflected due to the slope existing in the contour part 7 of the circuit pattern, so in the input image, the density of the part 17 corresponding to the contour part 7 is different from that of other parts. That is, a portion 15 corresponding to the base material portion 5 and a portion 16 corresponding to the circuit pattern portion 6
(Fig. 2(b)). A binary image of the target circuit pattern is obtained based on this density.

In the thinning process 2, the above process 1 is performed as a preprocess for tracking the contour shape of the circuit pattern and converting it into feature data.
Processing is performed to set the line width of the binary pattern of the contour portion 7 obtained in step 1 to 1 pixel.

In data conversion process 3, the outline portion 7 whose line width is 1 pixel is
The tracking start address of the binary pattern is detected, and feature data is calculated while contour tracking is performed using the address position as a starting point, thereby obtaining a one-dimensional shape feature curve.

Here, a method for calculating the above feature data will be explained with reference to FIG. FIG. 3 shows a binary pattern of the contour part 7 including the defective part A, and in the figure P1 is the latest position obtained by tracking, P
2 is a position d before Pl, P3 is further d before P2
B is the tracking direction. Now, consider a string P, l- that connects P and P3, and let the distance 2 between this string [[ and P2 be the feature data for P2. And (1) in Figure 4
As shown in (4), tracing is performed sequentially and the values of the feature data sentences at that time are obtained one after another to obtain a one-dimensional shape feature curve as shown in FIG. FIG. 5 shows a shape characteristic curve when d is set longer than the circumferential length f of the defective portion A when the defective portion A occurs in a straight line portion as shown in FIG.

(1) to (4) and 12. it4 corresponds to that used in FIG. In the curve shown in FIG. 5, 24 corresponds to the height of the defective part A, and the curve (:l)-(4)-(5) corresponds to the shape of the defective part A. Also, the curve (1) −(2
) −(3) and (5) −(6) −(7) are p,
It can be seen as a virtual image generated when p3 passes through the defective part A. The height 12.1a of these curves is the defect area A
Because it is small, about % of the true height f14, later comparison process 4
It's not a big hindrance.

The above example deals with a case where a defect occurs in a straight portion, but next, a case where a defect occurs in a curved portion will be explained with reference to FIGS. 6 and 7. FIG. 6 shows a binary pattern of the contour portion 7 when a defective portion C occurs in the curved portion. The dotted line in the figure shows the contour when normal. In this case, the feature data L is the actual height of the defect C, that is, the height Lo of the defect C with respect to the normal contour, plus the offset due to the curvature of the normal contour. This is a shape characteristic curve obtained in the example of FIG. 6.

Next, in comparison process 4, defects are recognized using the shape characteristic curve obtained in data conversion process 3. To recognize this defect, the first method is to compare the value indicated by the obtained shape characteristic curve with a predetermined threshold, or the first method is to compare the obtained shape characteristic curve with a normal shape characteristic curve. A comparative method is used.

The first method is as shown in FIG.
This is a method of checking whether the shape characteristic curve of the circuit pattern obtained in 1 is within the threshold value α determined by the allowable defect size and the offset error due to the curvature of the normal pattern.

In the case of the figure, D and E that exceed the threshold value α are recognized as defects. This first method is effective for small sharp defects.

The second method is to compare the shape characteristic curves of a circuit pattern and a defect-free standard circuit pattern.

That is, the shape characteristic curve F of the circuit pattern to be inspected shown in FIG. 9(a) is one-dimensionally aligned with the shape characteristic curve G of the standard pattern shown in FIG. 9(b), and the difference between the two is calculated. Defects are recognized by finding them as indicated by H in FIG. 9(C).

This second method is effective for inspecting smooth defects over a wide range and the shape of the circuit pattern itself.

(Effects of the Invention) As described above in detail, according to the present invention, defects are recognized by focusing only on the outline of the circuit pattern, so that only the outline of the circuit pattern appears clearly. It can be applied to inspection in a process (for example, after photolithography development). In addition, we use a method that generates data representing the characteristics of defects and uses them to recognize defects.
Small sharp defects can be directly recognized by their height and depth. Furthermore, by preparing a shape characteristic curve of the contour of a normal circuit pattern and comparing it with the shape characteristic curve of the circuit pattern to be inspected, it is now possible to detect even large and smooth defects. Precise mutual positioning as in the comparison method is not required.

[Brief explanation of drawings]

Figure 1 is a block diagram showing the processing procedure of the defect recognition method according to the present invention, Figures 2 (a) and (b) are diagrams showing an example of a circuit pattern to be inspected and its human image, and Figure 3 is a data An explanatory diagram of the method of calculating feature data in the conversion process, FIG. 4 is a diagram showing the state of contour tracking in the data conversion process,
Fig. 5 is a diagram showing a shape characteristic curve when a defect occurs in a straight section of the object to be inspected, Fig. 6 is a diagram showing an example of a pattern when a defect occurs in a curved section of the object to be inspected, and Fig. 7 9 is a diagram showing a shape characteristic curve when a defect occurs in a curved portion of the object to be inspected, FIG. 8 is an explanatory diagram of the first method in comparison processing, and FIG.
The figure is an explanatory diagram of the second method in comparison processing. 1... Image manual/binarization processing, 2... Thinning processing,
3...Data conversion processing 4.--Comparison processing, 5
...Base material part, 6...Circuit pattern part, 7...Contour part, A, C-...Defect part,
l...Characteristic data.

Claims (2)

[Claims] (1) Binarize the two-dimensional image of the circuit pattern to be inspected obtained by photoelectric conversion, obtain a binary image of the outline and other parts of the circuit pattern, and calculate the line width of the binarized outline. The line is thinned, the thinned contour is tracked, the feature data of the contour shape is calculated to obtain a shape feature curve, the value indicated by the shape feature curve is compared with a predetermined threshold, and the former value is calculated. A defect recognition method characterized in that a case where the value is greater than or equal to the latter value is recognized as a defect. (2) Binarize the two-dimensional image of the circuit pattern to be inspected obtained by photoelectric conversion, obtain a binary image of the outline and other parts of the circuit pattern, and calculate the line width of the binarized outline. The line is thinned, the thinned contour is traced, the characteristic data of the contour is calculated, a shape characteristic curve is obtained, and the calculated shape characteristic curve is compared with the shape characteristic curve of the normal contour determined in advance. A defect recognition method characterized by detecting a defective part.