JPH0715365B2 - Pattern position detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of use] The present invention relates to a pattern position detecting method in an automatic machine such as a pattern cutter or a wire bonder for a printed board, which positions the printed board according to the pattern and performs various processes. Regarding

[Conventional technology]

As the density of printed boards has increased, patterns and pattern intervals have become finer, and it has become necessary to accurately detect the positions of patterns in automatic machines that perform printed board assembling work and inspection work such as wire bonding and pattern cutting.

As a conventional method of detecting the pattern position of a printed board, there is a method of recognizing the position by binarizing the image information captured from the television camera to divide it into a target pattern and a background and extracting the features of the pattern from the binary image. It was done.

[Problems to be solved by the invention]

In the conventional pattern position detecting method, the entire image including the target pattern must be binarized, which requires a long processing time, and when the image fluctuates due to disturbance, the binarization is performed at an appropriate slice level. However, there are cases where accurate position detection cannot be performed.

The present invention has been made in view of the above-mentioned drawbacks of the prior art, and it is an object of the present invention to provide a pattern detection method which has a high processing speed, is accurate, and is highly reliable.

[Means for solving problems]

To achieve this object, according to the present invention, image information of the pattern to be detected on the printed board (luminance value for each pixel)
In a pattern position detecting method for detecting a pattern shift amount based on a fixed mask pattern having no positional shift of a pre-designed value input in advance, (a) uniform illumination light is provided within a certain range including a pattern to be detected. Illuminating and forming data of the illuminating unit as image information by a television camera; (b) A / D converting the image information and storing it as digital data in a frame memory; (c) X direction A plurality of scans in which the time-series data is read from the frame memory in the horizontal direction within the preset range while incrementing the address, and the time-series data is similarly read for another scan line parallel to the read scan line. A step of reading the line, and (d) converting the time series data into frequency coefficients by fast Fourier transform. A step of transforming, (e) a step of multiplying a frequency coefficient of the input image obtained by the fast Fourier transform by a frequency coefficient of the fixed mask pattern obtained in advance, and (f) an inverse of the multiplication result. A step of obtaining a change in the area of the overlapping portion of the luminance as a position function when the image information of the detected pattern is moved pixel by pixel with respect to the fixed mask pattern by performing a fast Fourier transform and returning it to time series data. And (g) obtaining a peak position where the overlapping area is maximum from the position function, and (h) comparing the peak position and a reference position based on the fixed mask pattern to obtain a positional deviation amount in the X direction. After the position in the X direction is detected, in the step of reading the time-series data in the item (c), the frame memory is added. The same steps are repeated for the Y direction by reading the vertical direction, and the position in the Y direction is detected, and the XY in-plane deviation amount of the target pattern is obtained, so that the position correction of the target pattern is fast and highly accurate. A pattern position detection method that can be realized is provided.

[Action]

The image of the target pattern is input to the arithmetic circuit by a TV camera or the like, A / D converted, and the data relating to the position of the target pattern is read. In order to obtain the change in the area of the overlapping portion when the inspection pattern data is moved in pixel units with respect to the mask pattern data input in advance, the inspection pattern data is subjected to the fast Fourier transform and further multiplied by the frequency coefficient. An inverse fast Fourier transform is performed on the multiplication result to obtain the area of the overlapping portion as a position function. The peak position where the area is maximized is obtained and compared with the position based on the mask pattern to detect the amount of positional deviation. The above operation is X, Y
Repeat for both directions. The image information is created for two or more scan lines in each of the X and Y directions.

A known technique is used as the method of determining the reference pattern. For example, a method is used in which a plurality of sample values (values for each pixel of the brightness data of the substrate pattern) are acquired and the average is obtained.

〔Example〕

FIG. 1 is a diagram showing the configuration of a pattern position detecting device necessary for understanding the present invention. The printed board 2 is mounted on the XY table 1. A pattern 3 to be detected is formed on the printed board 2. The pattern 3 is drawn in an exaggerated size. A laser light source 4, a total reflection mirror 5 and a cylindrical lens 6 are provided to irradiate the pattern 3 with slit light. Light from the laser light source 4 (for example, He-Ne laser light) is reflected by the total reflection mirror 5 and passes through the cylindrical lens 6 to become a flat slit light 7 and a linear scan line (light cutting line) 8 at the irradiation portion. To form. A television camera 9 is provided diagonally above the direction (Y direction) perpendicular to the direction of the scan line 8 (for example, the X direction). TV camera 9
Is connected to a frame memory 10 for storing captured images, which is a central control unit (CPU) 1
Connected to 3. The CPU 13 is further connected to a data memory 11 that stores mask pattern data (fixed data) that is design value data of the pattern 3 and a fast Fourier transform board (FFT board) 12. The CPU 13 is connected to a drive control circuit 14 of a processing mechanism (for example, a pattern cutter) for performing various kinds of processing on the printed board 2.

Next, the principle and operation of pattern detection by the pattern detection device will be described.

As shown in FIG. 2A, the input image taken by the television camera 9 represents the pattern cross-sectional shape, and the height direction data corresponding to the position in the X direction is the frame memory 10.
Memorized in. Therefore, the data of the pattern shape corresponding to the position in the X direction can be read from the frame memory as time series data. A mask pattern (a pattern based on the design data of pattern 3) which has been input to the data memory 11 in advance and has no displacement is fixed data, and as shown by the solid line in FIG. Form a simple rectangle. The pattern input to the frame memory 10 with respect to such a mask pattern is displaced as shown by the dotted line in FIG. 2 (b), and the waveform is displaced or distorted due to disturbance. Therefore, the position of one pattern is fixed, the position of the other pattern is moved pixel by pixel in the X-axis direction, and the overlapping area g of both waveforms at each position t
By obtaining (t) and examining the area change, a waveform as shown in FIG. 2 (c) is obtained. Actually, the detection pattern data of the dotted line is inverted to the negative position side and then moved in the positive direction in pixel units to obtain the overlapping area. When the input pattern (dotted line) and the mask pattern (solid line) match, the overlapping area becomes maximum and the position is the position of the detection pattern to be aligned. Therefore, if the peak position tp of the overlapping area g (t) in FIG. 2 (c) is obtained, the position of the detection target pattern will be detected.

Input pattern in X direction is X (t), mask pattern is h
(T), the change of the overlapping area is It can be calculated by the convolution integral represented by.

This calculation can be performed by replacing with multiplication in the frequency domain by the convolution theorem shown below, which shortens the calculation processing time.

Convolution theorem g (t) = h (t) * X (t) H (f) X (f) FIG. 3 shows a flow chart of a pattern position detecting method using the above detection calculation principle.

Step a: A pattern is photographed by the TV camera 9 and image information is input to the apparatus.

Step b: A / D convert the input image and write it to the frame memory 10.

Step c: The height of the image data with respect to the position in the X direction is read by the frame memory 10 as time series data (X
(Determine (t)).

Step d: The read time series data X (t) is converted into frequency data X (t) by the FFT operation board 12. The mask pattern data H (f) is stored in the data memory 11 in advance.

Step e: The frequency data of the input pattern and the frequency data of the mask pattern are multiplied to calculate H (f) X (f).

Step f: The frequency function of H (f) X (f) is subjected to the inverse FFT conversion to be returned to the position function g (t).

Step g: Calculate the peak position tp, which is the maximum area, from the obtained g (t).

Step h: The difference between the peak position when there is no positional deviation (in the case of mask patterns) and the above-mentioned tp is obtained and used as the positional deviation amount in the X direction.

Next, an image in the Y direction is input and the same steps as those in the X direction are repeated to obtain the amount of positional deviation in the Y direction. As described above, the X and Y coordinates of the target pattern are detected.

In the method as described above, the greatest feature of the present invention is that time-series data is obtained for a plurality of scan lines (dotted lines) in the frame memory to perform position detection calculation, as shown in FIG. 6 (a). . As a result, the position can be detected with higher accuracy and higher reliability. FIG. 6B is an example of time series data for a fixed mask pattern. The scale represents the number of bits. The change in area when such time-series data is inverted, moved in pixel units, and superposed on the original data is shown in FIG. The bit at the peak position tp is 240.

FIG. 4 shows the configuration of an embodiment of the present invention. In this embodiment, the entire printed board 1 including the pattern 3 is illuminated by the light source 15, and the television camera 9 shoots the pattern 3 from above the pattern 3. Since the pattern 3 has a larger amount of reflected light than the printed circuit board surface, the pattern is identified as a difference in brightness. Therefore, the bright pattern corresponding portion 3'corresponding to the pattern shape is input to the frame memory 10 as shown in FIG. By scanning, for example, the dotted line portion of this frame memory, it is possible to read the time series data corresponding to the position in the X direction as shown in FIG. This corresponds to the data shown in FIG. 2 (a). Using time series data based on this brightness information, pattern position detection is performed by the same steps as those in the above-described embodiment.

Also in the embodiment shown in FIGS. 4 and 5, time-series data is obtained for a plurality of scan lines (dotted lines) in the frame memory to improve accuracy and reliability, and position detection calculation is performed (FIG. 6). Is the same.

A supplementary explanation of the concept of the area of the overlapping portion (see
(See Figures 2, 5, 6, 7).

That is, by calculating the overlapping area while moving the reference pattern one pixel at a time in the X direction, the moving amount when the area becomes maximum can be obtained (in FIG. 7, the overlapping area becomes maximum when the moving amount is δ). ).

Therefore, since the amount of movement is the amount of positional deviation of the substrate pattern, the position can be detected. The position can be similarly detected in the Y direction. However, the scanning direction of the image signal is orthogonal to the X axis.

〔The invention's effect〕

As described above, in the pattern position detecting method according to the present invention, since the pattern position is calculated by the change in the overlapping area of the detection target pattern and the mask pattern, regardless of the distortion due to the disturbance of the detection target input pattern, etc. Accurate position detection is performed. In addition, since the overlapping area is calculated by FFT conversion using the convolution theorem, the processing time is short.

Furthermore, since the image information is obtained from two or more scan lines in the same direction in the present invention, the alignment accuracy is further improved.

Further, since the pattern information to be compared is the fixed mask pattern information of the designed value, the comparison target pattern information also simplifies and speeds up the arithmetic processing as compared with the case of using the input image.

[Brief description of drawings]

FIG. 1 is an explanatory view of the structure of an apparatus necessary for understanding the present invention, FIG. 2 is an explanatory view of the principle of FIG. 1, FIG. 3 is a flow chart of the method of the present invention, and FIG. FIG. 5 is an explanatory view of the configuration of an embodiment of an apparatus for carrying out the invention, FIG. 5 is an explanatory view of input data in the embodiment of FIG. 4, FIG. 6 is an explanatory view of a characteristic portion of the method of the present invention, and FIG. 7 is an overlapping portion. It is a figure explaining the concept of area. 2 ... Printed board, 3 ... Pattern, 6 ... Cylindrical lens, 7 ... Slit light, 8 ... Scan line, 9 ... TV camera, 10 ... Frame memory, 11 ...
… Data memory, 12… Fast Fourier transform board, 13…
…CPU.

Claims (1)

[Claims] 1. A pattern for detecting a pattern shift amount based on image information (brightness value for each pixel) of a pattern to be detected on a printed board and a fixed mask pattern which is previously input and has no position shift of a designed value. In the position detection method, (a) a step of irradiating a certain range including a pattern to be detected with uniform illumination light and forming data of the irradiation section as image information by a television camera; A / D converting and storing in a frame memory as digital data, and (c) reading the time series data in the X direction horizontally from the frame memory within a preset range while incrementing the address, and further reading. Similarly, the above-mentioned time-series data is read for another scan line that is parallel to the other scan line. And (d) converting the time-series data into frequency coefficients by fast Fourier transform, and (e) frequency coefficients of the input image obtained by fast Fourier transform, and the fixed mask pattern obtained in advance. A step of performing multiplication with a frequency coefficient; and (f) moving the image information of the detected pattern in pixel units with respect to the fixed mask pattern by inverse fast Fourier transforming the multiplication result and returning it to time series data. The change of the area of the overlapping portion of the brightness in the case of making it a position function, (g) the step of obtaining the peak position where the overlapping area is the maximum from the position function, (h) the peak position and the fixing Calculating a displacement amount in the X direction by comparing with a reference position based on the mask pattern, and after detecting the position in the X direction. In the step of reading the time-series data in (c) above, the address of the frame memory is read in the vertical direction, and the same steps are repeated in the Y direction to detect the position in the Y direction and to detect the XY plane of the target pattern. A pattern position detection method that makes it possible to correct the position of a target pattern at high speed and with high accuracy by determining the amount of internal deviation.