JPS633243B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an apparatus for recognizing a target pattern position within a pattern with extremely low contrast with high accuracy.

2 of low contrast pattern signals detected by a television (hereinafter abbreviated as TV) camera
Conventionally, a binarization method using a difference signal of a pattern signal has been used for digitization. FIG. 1 is an explanatory diagram of the differential signal binarization method according to the prior art.
The figure shows an image of a pattern with low contrast, FIG. 1-b shows the electric signal of the pattern image, and FIGS. 1-c and 1-d show the process of obtaining the binarized signal 7. In Figure 1-a, 1 is an image of a pattern with low contrast detected by a TV camera.

The electrical signal 3 of the scanning line 2 is as shown in Figure 1-b, and in order to binarize it, a differential signal 4 as shown in Figure 1-c is created, and thresholds 5 and 6 are applied to this.
A method is known in which the binarized signal 7 shown in FIG. 1-d is generated by multiplying the An example of a signal processing circuit is shown in Part 2-
Shown in Figure a. Figure 2-b is a diagram showing a TV signal and a delayed signal, Figure 2-c is a diagram showing the process of forming a binary signal, Figure 2-d is a diagram showing a binary signal, and Figure 2-c is a diagram showing the process of forming a binary signal. Figure e is a diagram showing an effective area to be binarized.

In FIG. 2-a, the TV signal 8 is amplified by an amplifier 9, delayed for a certain period of time by a delay line 10, and then amplified by an amplifier 11. The difference between the TV signal 8 and the delayed signal 12 is amplified by an amplifier 13 to produce a differential signal 16 as shown in FIG. 2-c. This is multiplied by thresholds 14 and 15 and binarized by comparator 17, and then led to flip-flop 19, where the binarized signal obtained by threshold 14 is used as a set signal, and the binarized signal obtained by threshold 15 is used as a reset signal. Then, the second
A binarized signal such as 20 as shown in figure -a is obtained. Note that 18 shown in FIG. 2-e is a signal in an effective area to be binarized.

However, this method has the following drawbacks. 1st
- If there is a pattern parallel to the scanning line 25 on the scanning line, such as scanning line 19 in the figure 1-a,
The boundaries of the pattern become unclear as in the electrical signal 20 shown in Figure e, and the difference signal 21 shown in Figures 1-f.
A point occurs where the output signal of becomes small. When this signal is binarized by multiplying it by the threshold values 22 and 23, it becomes a binarized signal 24 shown in FIG. 1-g, and some parts are not binarized accurately.

Generally, the binarization pattern 26 obtained by the above-mentioned binarization method (referred to as the differential signal binarization method) is the third
The result was as shown in Figure 2-a, and there was a drawback that whiskers such as those shown in numbers 27 to 31 were generated.

Next, a conventional method for determining the coordinates of a target circular pattern from within the binarized pattern shown in FIG. 3-a will be described. Fig. 3-b shows the circular pattern 32 to be found, and a so-called pattern matching method is used to scan the entire TV screen using this as a dictionary pattern and find the coordinate position with the best matching on the TV screen. ing.

In this conventional method, the TV screen is, for example, 320 x 240.
Since it uses 64x64 sampled memory for dictionary patterns, it has the disadvantage of requiring a huge amount of memory.

The purpose of the present invention is to eliminate the above-mentioned drawbacks of the prior art, to binarize a low contrast pattern signal, and to
An object of the present invention is to provide a small-scale pattern position recognition device that extracts a target pattern from digitized signal patterns and recognizes its coordinates with high accuracy.

To achieve the above object, the present invention first performs coarse position recognition by coarsely sampling the TV screen. For example, if the TV screen is sampled at 40 x 30 and the dictionary pattern is also sampled at 8 x 8, the pattern matching circuit will be extremely small as will be described later. This circuit recognizes the approximate position of the target pattern. Next, a frame of a certain size is set around this coarse position, and the flip-flop of the differential signal binarization circuit described above is reset in this frame. As a result, the whiskers as shown in FIGS. 3, 27 to 31 can be erased in the frame area.

The precise position of the target pattern is found within this frame using the pattern boundaries.

The present invention will be explained in detail below based on the drawings.
FIG. 4 is a diagram illustrating a pattern alignment method for liquid crystal display elements as an embodiment of the present invention. In the figure, when assembling the liquid crystal display element, align the alignment marks 35 and 36 on the upper glass plate 33 and the alignment marks 37 and 38 on the lower glass plate 34 accurately, and then assemble the two glass plates. It needs to be pasted. Reference numerals 39 and 40 denote lamps for illuminating the mark from directly above, and the illuminated mark is magnified by objective lenses 41 and 42 and imaged by a TV camera.

Figure 5-a is an enlarged view of the liquid crystal pattern. The alignment mark 48 is a circle. horizontal scanning line 50
The TV signal 51 detected by the is shown in FIG. 5-b. Since the liquid crystal pattern is formed using an extremely thin Nesa film, the contrast when detected by a TV camera is extremely low. Furthermore, since the purpose of the glass plates for liquid crystal display elements is to align the two glass plates, the marks are detected with a TV camera after the two glass plates are brought close to each other in advance. Therefore, when the upper mark is detected, the lower pattern remains as a ghost, and conversely, when the lower mark is detected, the upper pattern remains as a ghost, resulting in large undulations being superimposed on the TV signal. As a result, a signal such as a TV signal 51 as shown in FIG. 5-b is detected.

The reason for using a circular pattern as the alignment mark here is as follows. The liquid crystal pattern is formed by printing a resist, and this is because a circular pattern can be formed with the highest accuracy. FIG. 6 shows a coarse pattern matching circuit for recognizing the coarse position of a circular pattern on the top glass or the bottom glass. A difference signal is created from the output of the TV camera 52 in a difference circuit 53, and this is multiplied by a threshold value in 54 to yield 2
After being converted into a value, the flip-flop circuit 55 converts the video signal into a binary value. 56 is a setting signal for a binarization effective area. 57 is a pattern matching circuit.
This circuit performs coarse pattern matching.
TV screen sampling is coarse at 40 x 30 samples. Therefore, it is sufficient to use seven 40-bit serial-in/serial-out shift registers. In addition, eight 8-bit serial-in, parallel-out shift registers are used for pattern matching, an 8 x 8 bit memory is used for dictionary patterns, and exclusive OR is performed between the 8 x 8 shift register and the 8 x 8 bit memory. Pattern matching is completed by finding the point where the output is minimum. In this way, rough position recognition is completed.

FIG. 7 shows a binarized image 65 and a coarse recognition position 66 determined by rough pattern matching. A frame 67 of a certain size is generated around this rough recognition position 66, and the flip-flop is reset at position 68. That is, if the reset signal generating circuit 63 in FIG. 6 generates the reset signal 64 corresponding to the signal at the reset position 68 to reset the flip-flop, all the binary signals at the position 68 in FIG. 7 become "0". become.

For example, whiskers such as 69 in FIG. 7 are also reset at the reset position 68, so that they do not affect the intended circular pattern. Figure 8 is the 7th
This is an enlarged view of a portion of the frame 67 in the figure. In the figure, the precise center position of the circular pattern is found within the frame 67. The coordinates of the rough recognition position 66 are (x ₁ , y ₁ )
shall be. Further, the size of the frame 67 is assumed to be m×m samples. First, we will describe how to obtain the fine center coordinates in the x direction.

In the frame 67, focus on the m/2 sample scanning line 73 and find the intersections a and b between this and the circular pattern.
seek. At this time, the TV screen is sampled (320 x 240 samples). The fine center coordinates in the x direction are a+b/2.

Next, focusing on the vertical line 72 whose x coordinate is the a+b/2nd sample, the intersection points c and d between this and the circular pattern are determined. The fine center coordinate in the y direction is c+d/2.

In this method, the boundaries of the circular pattern are used to find the fine center of the circular pattern, so it is necessary that the horizontal scanning line is not obstructed by whiskers 27 or the like.

Therefore, since the flip-flop is reset at 68, there is no fear that whiskers will enter the frame 67 from the outside, and high reliability in detection is ensured.

As explained above, according to the present invention, the position recognition of low-contrast patterns such as those of liquid crystal display elements, for which automatic position detection was difficult in the past, can be performed with high precision using a small-scale electric circuit, and the following effects can be achieved. arise.

(1) By adopting a coarse/fine switching type pattern detection method, the shift register (320 bits x 63 lines) required for conventional pattern matching circuits has been replaced with a coarse pattern matching circuit shift register (40 bits).
x7).

(2) By adopting a pattern boundary detection method for precise position detection, it has become possible to obtain position recognition accuracy within ±1 sample.

(3) With the invention of the flip-flop preset method, position recognition errors have been completely eliminated.

[Brief explanation of the drawing]

Figures 1a to 1g are explanatory diagrams of a differential signal binarization method according to the prior art, Figures 2a to e are differential signal binarization circuits according to the prior art, and Figures 3a and b are differential signal binarization patterns according to the prior art. FIG. 4 is an explanatory diagram of a liquid crystal display element pattern alignment method taken up as an embodiment of the present invention, FIG. 5 is an enlarged view of the vicinity of the alignment pattern, and FIG. 6 is a diagram showing pattern position recognition according to the present invention. FIG. 7 is an explanatory diagram of the circuit, FIG. 7 is an explanatory diagram of a method for recognizing the fine center position of a pattern, and FIG. 8 is an enlarged view of the frame 67 in FIG. 8: TV signal, 10: Delay line, 9, 11, 1
3: amplifier, 12: delayed signal, 14, 15: threshold, 16: difference signal, 19: flip-flop,
20: Binarized signal.

Claims (1)

[Claims] 1 Convert the binarized pattern into an electrical signal using a TV camera, and calculate the difference between the original video signal of the binarized signal pattern of this electrical signal and a delayed video signal that is delayed by a certain period of time. Create a differential signal,
A function of setting fixed fixed threshold values above and below the amplitude of the difference signal from a reference amplitude position, shaping the difference signal, and inputting the shaped signal into a flip-flop to binarize the video signal of the signal pattern to be binarized. a mechanism having a function of recognizing the coarse position of the center of a target pattern in the pattern to be binarized by coarse pattern matching using the binarized signal of the signal pattern to be binarized; A frame of a certain size is provided in the pattern to be binarized around the center position, and a mechanism having a function of resetting the flip-flop of the differential signal binarization circuit at the leading edge position of the frame and a frame are provided within the frame. 1. A pattern position recognition device comprising: a mechanism having a function of recognizing the precise position of the center of a target pattern in a binarized pattern.