JPS6027964B2 - Directional high-cut spatial frequency filter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a laser beam diffraction pattern spatial frequency filtering method (hereinafter referred to as spatial filtering method) that detects defects in a regular pattern by using the diffraction phenomenon of light.
relates to a spatial frequency filter (hereinafter referred to as spatial filter) formed by photographically recording an optical Fourier transform spectral intensity distribution of a regular regular pattern.

In recent years, a spatial filtering method using laser light has been put into practical use as a method for detecting defects in regular patterns such as various metal filters and IC masks.

When processing two-dimensional images, this spatial filtering method requires a conventional method of scanning the two-dimensional image with a small light beam and processing the obtained output signal with an electronic computer, which is complicated and takes a long processing time. However, spatially parallel processing of two-dimensional images is easy, can be done with an inexpensive optical system, and can be achieved at high speed, so it is mainly used for processing regular two-dimensional patterns. It has been put to practical use as a defect inspection device for industrial products such as metal filters, IC masks, and textiles. The basic configuration of an optical system of a defect inspection apparatus using a spatial filtering method will be explained with reference to FIG.

The laser beam 2 emitted from the laser oscillator 1 passes through the collimator 3
The parallel light beam 4 is expanded and is applied to the object 5 to be inspected.

The object to be inspected 5 is placed on the front focal plane of the Fourier transform lens 7, and the Fourier transform spectrum of the object to be inspected 5 appears on the back focal plane due to the light 6 that is diffracted when passing through the object to be inspected. On the back focal plane of the Fourier transform lens 7, there is placed a negative photographic film, that is, a spatial filter 8, which records the Fourier transform spectral intensity distribution of the normal pattern of the inspected object 5 (by the Fourier transform lens 7). Of the Fourier transformed spectra of the inspected object 5, only the spectra corresponding to normal patterns are absorbed by the spatial filter 8, and the spectra corresponding to defective patterns are transmitted. By the way, since this spatial filter 8 is placed on the front focal plane of the inverse Fourier transform lens 10, the light 9 transmitted through the spatial filter 8 is transmitted through the inverse Fourier transform lens 101.
From this, the inverse Fourier transform is performed, and the inverse Fourier transform lens 10
A spatially filtered inverse Fourier transform image, that is, an image of only the defective portion of the object to be inspected 5 appears on the back focal plane and is detected by the photodetector 11. If a screen is placed in place of the photodetector 11 on the back focal plane of the inverse Fourier transform lens 10, the defect will appear as a bright spot on the screen, making it possible to visually recognize the defect. The above is the basic configuration of the spatial filtering method. FIG. 2 shows, as an example, a conventional spatial film 8 in the case of two-dimensional grid cotton.

FIG. 2a shows a two-dimensional lattice stripe, and FIG. 2b shows a spatial filter 8 formed by photographically recording the optical Fourier transform spectral intensity distribution of the two-dimensional lattice stripe. It has a transparent part 13 that does not change.

FIG. 2c shows the relationship between the geometry of the two-dimensional grid cotton and the Fourier transform spectral intensity distribution in the x direction. The axis represents the spatial frequency x = appetizer (M wavelength, f' focal length), and the vertical axis represents the intensity distribution 1 (X) = (poisonous ramie) 2 (pot poisonous ramie) 2.

The intensity distribution 1 (X) is such that the interference fringes (non-toxic light) 2 due to multiple slits are modulated by the diffraction image (reverberation) 2 (dotted line in the figure) of one slit. In addition, A, B
, C indicate the length of one side of the grating shown in a of FIG. 2, the grating interval, and the total length. When inspecting industrial products with regular patterns using a spatial filtering method, generally a spatial filter is used in which the optical Fourier transform spectral intensity distribution of a normal pattern is photographically recorded on a 35-rib film or the like. do.

However, although photo-recorded spatial filters are easy to produce and inexpensive, they require strict mechanical accuracy of the optical system, such as light misalignment and rotation around the optical axis, and it is difficult to use a general convex lens as a Fourier transform lens. In this case, the focal plane becomes a curved surface, causing the higher-order diffracted light to become out-of-focus, and the spectral intensity of the higher-order diffracted light to be too small, resulting in the 1 to 1 pi (1 skin osec) of the photographic film. Since it cannot record in a dynamic range, it has a practical problem that high-order diffracted light cannot be filtered and a sufficient S/N ratio cannot be obtained. Conventionally, methods for improving these problems include using a directional filter or a special Fourier transform lens with a flat Fourier transform surface.

FIG. 3 is an example of a spatial filter called a directional filter, which targets a pattern such as a two-dimensional lattice pattern in which the main components of the Fourier transform spectrum are distributed in the 00 and 900 directions. The hatched opaque section 14 and the transparent section 15 outside the opaque section 14 reduce the mechanical precision required for the optical system and improve its practicality. This method takes advantage of the fact that defects do not have a specific directionality on the spectral plane, and the components in the oo and 9p directions, that is, the linear components of the pattern, are treated as normal components and are filtered out, and the other components are are all regarded as defective components and are allowed to pass through. therefore,
It is impossible to filter out pseudo-defect spectra with no directionality, such as corner R (see FIG. 6), and a sufficient S/N ratio cannot be obtained. On the other hand, a special high-resolution Fourier transform lens with a flat Fourier transform surface is expensive, costing several million yen, and is not suitable for practical use.Even if it is used, problems with the dynamic range of photographic film remain, so it You can't expect much of a filter effect. This invention was made in consideration of the above points, and it is possible to easily perform spatial parallel processing of two-dimensional images, which is the feature of the spatial filtering method, and to perform it using an inexpensive optical system. It solves the problems of the previous model and further improves its practicality. The present invention will be explained in detail below with reference to the drawings. Figures 4a and 4b show a conventional spatial filter photographically recording the optical Fourier transform spectral intensity distribution of an industrial product such as the metal filter with two-dimensional lattice stripes shown in Figure 2, and an enlarged view of its main parts. Yes, Figure 5 a, b
5 shows an embodiment of the present invention, and FIG. 5a shows a combination of the spatial filter shown in FIG. FIG. 5b is an enlarged view of the main part.

In FIG. 4, the optical Fourier transform spectral intensity distribution of the two-dimensional lattice stripes outside the secondary spatial fill is recorded and blackened opaque areas 16 and transparent areas 17-1, 17 whose spectral intensity was weak and were not blackened. -2.

The main components of the spectrum of the two-dimensional lattice are oo and 900
There is a very small component in the a direction between them.

Therefore, even on a photographic film, the blackening distribution directly reflects the intensity distribution state of the spectrum, and in the 00 and 900 directions, which are the main components, even the high-order diffracted light region is blackened, and in the 0 direction, only the low-order diffracted light region is blackened. In addition, the diameter of the blackened spot is different from the diffraction-limited spot diameter determined by the size of the light source (for example, several peaks), and in the region of near zero-order diffracted light with high intensity, overexposure during recording may occur. Due to the effect of halation, the spot diameter becomes large, on the order of 100 to 150 degrees, and the spot diameter becomes smaller as the region of higher order diffracted light is reached. Now, when the inspection object 5 having the defect 18 shown in FIG. 6 is spatially filtered by the method shown in FIG. 1 using the spatial filter shown in FIG. The high-order spectrum passes through the transparent part 17-1, and the low-order spectrum passes through the transparent part 17-2, and appears as a bright spot on the inverse Fourier transform surface.

However, when considering industrial products such as various metal filters, corner RI9 always exists at the corner of Sekiguchi, so the high-order spectrum in eight directions of this corner RI9 also passes through the transparent part 17-1. However, it appears as a bright spot on the inverse Fourier transform surface, resulting in a so-called pseudo-defect output, which significantly reduces the S/N ratio. Considering the noise components of conventional spatial filtering methods, optical system noise (light scattering due to irregularities on the lens surface, dust, etc.) and higher-order spectra where the photographic film is not sufficiently darkened due to insufficient spectral intensity and out-of-focus. There is high-order diffracted light that passes through the area, and most of both are transparent parts 17.
-1 is transmitted. Since the transparent portion 17-1 occupies most of the area of the spatial filter, the amount of transmitted light that is area-integrated is quite large, even if it is weak light such as optical system noise or high-order diffracted light. Novel findings based on this experiment form the basis of this invention. Some of the higher-order diffracted light that passes through the spatial filter comes from around the straight portions of the grating apertures. This is based on the marginal wave theory (Hiroshi Kubota: Wave Optics; Iwanami Shoten, 1971), which states that higher-order diffracted light is diffracted light from the periphery of a closed aperture.
, P. 257), which transmits the non-blackened portion of the spatial filter in the oo and 900 directions. In this way, although higher-order diffracted light is certainly an important component of the spectrum of defects, when considered as a whole, it is nothing but the main component of noise. Therefore, in order to improve the S/N ratio of a spatial filter, it is necessary to consider a spatial filter that transmits as much as possible of the diffracted light components of defects and cuts as much as possible of high-order diffracted light that becomes noise. As mentioned above, FIG. 5 shows a spatial filter according to the present invention, in which a directional high-cut filter 20 having a hatched area is added to the spatial filter formed by photographic recording shown in FIG. . As the shape of the shaded area shows, the directional high-cut filter 2 masters the 0 direction, where no high-order diffraction light region is recorded, to the low-order diffraction light region where the spectrum is recorded, and on the other hand, The principal component directions of oo and 9p are as follows.
Since the spectrum is recorded up to the higher-order diffracted light region and has a filtering ability, it closes to the no-pass tendency and is shaped almost in accordance with the directional component ratio of the optical Fourier transform spectrum intensity distribution of the two-dimensional lattice stripe. By adding a directional cut filter 20 with such a shape to a conventional spatial filter formed by photographic recording, it is possible to cut most of the high-order diffracted light centered in the 0 direction, which is the main component of noise. Optical system noise, corner radius, and peripheral diffracted light can be almost completely removed.On the other hand, in the 0 direction, only the low-order diffracted light with strong light intensity is transmitted, and in the oo and 900 directions, the spectral components of defect 18 are transmitted. Depending on the filtering ability of the recorded spatial filter, it is possible to transmit as much as high-order diffraction light as possible to prevent a decrease in the optical power of the defect 18 as much as possible, so that the S/N ratio can be significantly improved. The effect of the directivity cut filter is not simply to increase the defect output and improve the S/N ratio.
Looking at the image quality, in the case of 0, which combines a circular low-pass filter and a photo-recorded spatial filter, the S/
Although there is an improvement in N, there is significant image deterioration due to circular hole diffraction, whereas with the directional / -cut filter, the direction of the main component of the spectrum is made similar to high-pass, so the influence of diffraction is almost negligible. Therefore, a very good inverse Fourier transform group is obtained. In addition, looking at machine precision, in the low-order diffraction light region that is mainly used, the black spot diameter of the recorded spectrum is 10 to 2 times larger than the diffraction limit spot diameter, which is within the size of the light source. Because of this, the tolerance range for rotation around the optical axis, optical axis deviation, etc. is wide, and the mechanical precision of the spatial filter can be significantly reduced. This directional cut filter can be made of any material that absorbs light, so it can be made by painting a spatial filter made of photographic recording black in a predetermined shape, or
It can be made by simply pasting black paper cut out in a predetermined shape, and is extremely simple and inexpensive. A piece of black paper cut out into a predetermined shape may be placed several rows away from the spatial filter in the optical axis direction. In this way, the spatial filter according to the present invention is formed by photographically recording the optical Fourier transform spectral intensity distribution of a regular pattern, and the direction in which high-pass characteristics are imparted to the direction with many components according to the directional component ratio of the spectrum. By using a directional high-cut spatial filter in combination with a high-cut filter, the spatial parallel processing of two-dimensional images, which the spatial filtering method has, can be done easily and with an inexpensive optical system. It is possible to realize a high performance defect inspection device for industrial products having regular two-dimensional patterns without any loss in quality and without requiring mechanical precision.

Next, the effects of this invention will be explained.

Figures 7a and b show regularity 2 used to confirm the effects of this invention.
FIG. 3 is an explanatory diagram showing an open square portion 21 and a defect 22 of a metal filter on the 0.2-thickness side having a dimensional pattern; and FIG.

Figure 8 shows the optical properties of the metal filter shown in Figure 7 when a He-Ne laser is used as the light source and a convex lens with a focal length of 25 mm and an aperture of 10 mm is used as the Fourier transform lens. Spatial filter 23 photographically records the Fourier transform spectral intensity 3 degree distribution of the target
FIG. 4 is an enlarged view of the directional/cutting spatial filter according to the present invention in combination with FIG.

The principal component direction of the Fourier transform spectrum is the oo direction and 90
However, since the component in the oo direction is the largest, the high-pass property of the directional filter 24 in the oo direction is made the strongest, and the oo direction/9 direction = F/E = 2. There is. FIG. 9 shows a case where the directional/cut spatial filter shown in FIG. The ratio of the light intensity of the defect 22 and the corneroid light intensity measured on the inverse Fourier transform plane when a spatial filter is used (curve B), that is, S/N
This is a graph drawn by changing the DJ in FIG. 8.

Note that the conventional spatial filter has an S/N=5. 1st
Figure 0 shows the directionality/light intensity of the defect 22 measured in Figure 9.
・This is a graph drawn by comparing a low-pass spatial filter (curve A) with a circular low-pass spatial filter (curve B).

Note that the light intensity of the conventional spatial filter is 2.3. As is clear from FIG. 9 and FIG. 10, if the directional / cut spatial filter according to the present invention is used, the S/N will be approximately 5 times higher than that of the conventional spatial filter made by photographic recording.
The signal-to-noise ratio is improved and the light intensity at the defect 22 can be higher than when using a circular low-pass spatial filter. S
The light intensity of the defect 22 with D=3J peaking at /N is only 20% lower than that obtained from a conventional spatial filter with a directional high-cut spatial filter, but with a circular low-pass spatial filter. That would be an 80% reduction. Figure 11 shows the influence of optical axis misalignment on S/N. Considering that the effective S/N limit for an inspection machine is S/N = 4,
In the spatial filter (curve B) formed by the photographic record of the subordinate, 1
The directional high-cut spatial filter according to the present invention (curve A)
According to this, a deviation of the optical axis of up to 35 degrees is allowed. Figure 12 shows the influence of rotation around the optical axis on S/N. Considering S/N = 4 as the effective limit, the spatial filter (curve B) formed by conventional photographic recording is While it becomes unusable when the optical axis is rotated by 0.2 degrees or more, according to the directional high-cut spatial filter (curve A) according to the present invention,
．． Optical axis rotation up to 0 degrees is allowed.

As explained in detail above, the spatial filtering method using the directional high-cut spatial filter of the present invention has the feature of simple, inexpensive, and high-speed spatial parallel processing of two-dimensional images that the spatial filtering method has. High S/N, high defect output, high quality images, and easy machine accuracy can be achieved in defect inspection of industrial products with two-dimensional regular patterns without damage, making it extremely effective in practice. can do.

[Brief Description of the Drawings] Figure 1 is a basic configuration diagram of a defect inspection device using a spatial filtering method, Figure 2a is an explanatory diagram showing two-dimensional lattice stripes,
Fig. 2b is a plan view of a spatial filter formed by photographic recording of two-dimensional lattice stripes, Fig. 2c is a graph showing the Fourier transform spectral intensity distribution, Fig. 3 is a plan view of a directional filter, and Fig. 4 Fig. 4a is a plan view of a spatial filter formed by photographic recording of an industrial product having two-dimensional grid cotton, Fig. 4b is an enlarged view of the main part of Fig. 4a, and Figs. 5a and b are one implementation of this invention. As an example, Figure 5a shows the orientation of an industrial product with two-dimensional checkered stripes.
Fig. 5b is an enlarged view of the main part of Fig. 5a, Fig. 6 is an explanatory diagram showing closed parts and defects of an industrial product with two-dimensional lattice cotton, Fig. 7a is a plan view of the cut spatial filter. An explanatory diagram showing the entrance part and defects of the metal filter. Figure 7b is an enlarged view of the main part of Figure 7a. Figure 8 is an enlarged view of the directionality of the metal filter. An explanatory diagram, Fig. 9 is a graph showing the S/N of the directional high-cut spatial filter and the circular low-pass spatial filter, and Fig. 10 is a graph showing the relative light intensity of the defective part of the directional high-cut spatial filter and the circular low-pass spatial filter. Figure 11 is a graph showing the relationship between the optical axis deviation and S/N of a directional/-cut spatial filter and a conventional spatial filter. It is a graph showing the relationship between the optical axis rotation of the spatial filter of the pot and the S/N. In the figure, 1 is a laser oscillator, 2 is a laser beam, 3 is a collimator, 4 is parallel light, 5 is the object to be inspected, 6 is diffracted light,
7 is a Fourier transform lens, 8 is a spatial filter, 9 is light transmitted through the spatial filter, 1 is an inverse Fourier transform lens,
11 is a photodetector, 12 and 14 are opaque parts, 13 and 15 are transparent parts, 16 is a spectrally blackened opaque part, 17-
1, 17-2 is a transparent part that is not blackened, 18 is a defect, 1
9 is a corner R, 20 is a directional cut filter, 2
Reference numeral 1 indicates a gate part of a metal filter, 22 a defect, 23 a spatial filter, and 24 a directional high-cut filter. Figure 1 Figure 2 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12

Claims (1)

[Claims] 1. A spatial frequency filter formed by photographically recording the optical Fourier transform spectral intensity distribution of a regular pattern, and a direction in which a high-pass property is given to the direction with many components according to the directional component ratio of the optical Fourier transform spectral intensity distribution. A directional high-cut spatial frequency filter for recognizing regular pattern defects, which is combined with a directional high-cut filter.