JPS5926074B2 - Pattern information processing device - Google Patents

Description

[Detailed description of the invention] The present invention relates to a pattern information processing device.

Conventionally, various pattern information processing methods and devices have been provided for defect inspection of articles.

In such conventional methods, the pattern information of the inspected object is converted into black and white binary data, which is used as data such as the shape and area of the inspected object, and this data is combined with data of a standard pattern of a preset reference object. By comparison, it is determined whether the inspected object is acceptable or not.
However, there are few actual objects to be inspected that are clearly divided into two values, black and white, and many have a certain degree of density.

Therefore, the conventional methods as described above have the disadvantage that they are difficult to use for inspecting articles with complex patterns. In conventional pattern information processing methods, when a television camera or the like is used, one screen of the television camera is usually
For example, the degree of brightness of the image of the inspected object divided into 500 x 500 pixels and projected onto each pixel is determined by A-D(
There are many methods in which the characteristics of the object to be inspected are extracted by converting the data into numerical values using an analog-to-digital (analog-to-digital) converter and processing this data using a computer or the like.

In this case, the amount of data becomes enormous and the processing time becomes long. Therefore, when inspecting goods for defects, etc., the screen of the TV camera is not divided into pixels, but a certain threshold value is set and the amount of exposure from the TV camera is measured. There is also a method in which, for example, only the portion of the image signal of the inspection object that exceeds this threshold is converted into binary values, and this value is counted to extract the characteristics of the inspection object or detect defects such as scratches. be. This latter method will be explained with reference to FIGS. 1A and B.

The method related to FIG.
The product is inspected to determine whether or not it exceeds the limit. This method does not detect defects in complex patterned articles;
This method is effective for inspecting articles with relatively plain substrates or articles with simple patterns such as scratches, but is not suitable for inspecting articles with complex patterns. Another conventional method shown in FIG. 1B is used for inspecting articles with a more complex pattern than the method described with respect to FIG. 1A, such as measuring the area of the article.

The method shown in FIG. 1B measures the length (time) of a portion in which the image signal S3, which is more complex than that shown in FIG. 1A, exceeds a certain threshold value S4. Therefore, using the clock signal, AND the portion of the image signal S3 that exceeds the threshold S4 and the clock signal, and use the number of measured clock signals to calculate the amount of the portion of the image signal S3 that exceeds the threshold S4. The area of the article corresponding to the image signal S3 is calculated. In any case, the conventional methods described above have the disadvantage that when the pattern of an article becomes complex, it is not possible to accurately inspect the article.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a pattern information processing method and an apparatus suitable for use in defect inspection of articles, pattern recognition, etc.

The pattern information processing device according to the present invention is characterized by a television camera that forms an image signal of an article, and a means for creating at least three different threshold values within the level of one cycle of the image signal. and means for dividing one cycle of the image signal by the at least three respective predetermined threshold values, and an AND operation between each of the divided at least three signals from the means and a clock pulse from the clock pulse generating means. at least three AND means each taken;
The invention comprises at least three means for counting the number of clock pulses from the at least three AND means, respectively, and creating a number sequence representing one period of the image signal.

Hereinafter, an example of the image signal processing technology of the pattern information processing device according to the present invention will be explained with reference to FIGS. 2A, B, C, and D.
Let me explain first.

In FIG. 2A, Sp indicates a waveform for one period of an image signal of an article from, for example, a television camera.

In the present invention, as shown in the figure, a plurality of threshold values A, b, c, d and e are set in advance for the image signal S.
Set it so that it is within the level of p. In addition, each threshold value a to e
The interval (level difference) between them and the number of each threshold need not be limited to specific values and numbers, and may be arbitrary. If the portions where the image signal Sp crosses each threshold value a to e are extracted, waveforms Al, bl, cl, dl, and e1 as shown in FIG. 2C are obtained. Here, a clock signal of a constant frequency, that is, a pulse Sc as shown in FIG. 2B, and signals a1 to e as shown in FIG.
When ANDed with 1, the signal A as shown in Fig. 2D is obtained.
2, b2, c2, d2 and E2 are obtained. This signal A
The number of clock pulses in 2 to E2 is counted and formed into a number sequence, and this number sequence is used to form a pattern representing this image signal Sp, and thus the article corresponding to this image signal Sp. Incidentally, as shown in FIG. 2, if each threshold value a to e is selected and the frequency of the clock pulse Sc is selected, the signal A2
The number of clock pulses in E2 is 4+3=7, respectively.
14,3+17=20,2+22=24 and 34, therefore, the sequence corresponding to the image signal Sp is {7,14,
20, 24, 34}. For the first time, if the object photographed by an imaging body such as a television camera differs, the image signal from the television camera will also vary accordingly, and therefore the sequence of numbers obtained by processing this image signal as described above will also vary. Of course, it depends on the item.

That is, by processing the image signal of each article in this way, each article can be represented by a series of numbers pattern. For example, a case will be described in which the above-described present invention is applied to article recognition (pattern recognition).

In this case, a reference article is imaged with a television camera, the image signal is processed as described above to create a corresponding number sequence, and this number sequence is stored in, for example, a computer as a reference pattern. Next, the item to be recognized is
A TV camera photographs the item and similarly creates a number sequence pattern corresponding to this item, and a computer compares this number sequence pattern with a pre-stored standard number sequence pattern to determine whether or not the two match. However, it may be determined whether the recognized article is the same as the reference article. It will be obvious that the image signal processing technology of the present invention described above can be used for inspecting the quality of articles, inspecting the shape and presence of defects, and measuring the area, in addition to the above-mentioned examples.

In these cases, the number and size of threshold values to be set are changed depending on the symmetry to be inspected, and a necessary sequence pattern is formed to achieve the desired purpose. According to the present invention described above, the number of threshold values to be set is changed to create a number sequence according to the pattern of the article to be recognized or inspected, so that articles with complex patterns can be accurately recognized or inspected. It is something that can be done.

Furthermore, for example, in the case where the image signal is in phase with the image signal Sp as shown by the dotted line Sp'' in FIG. Sp' is T1 earlier than Sp, and the falling portion is also T1 earlier.

Therefore, as shown by the dotted line in waveform e1 in FIG. 2C, there is no change in the length (time) of the rising portion of the waveform as a whole, and as a result, in waveform E2 in FIG. As shown in , there is no change in the number of clock pulses, and the image signals Sp and Sp2 are both the same as a numerical sequence, so there is a risk that the image signals Sp and Sp' corresponding to different articles will be judged to be the same, and that they will be immediately different. There is a risk that even if the items are different, they may not be judged as the same. In such a case, for example, the constant frequency clock pulse Sc shown in FIG. 2B is not used, but a frequency modulated clock pulse is used. Otherwise, the number of clock pulses will not be constant within a unit time, but will change over time, so even within the same time T1 at the beginning and end,
Since the number of clock pulses varies, the sequence of image signals Sp and Sp' naturally differs. Therefore, the risk of erroneous determination as described above can be avoided. Incidentally, by using the frequency-modulated clock pulse in this manner, the same effect can be obtained for other threshold portions.

Next, with reference to FIG. 3, an example of the pattern information processing apparatus according to the present invention described above will be explained.

In the figure, T is an imaging body such as a television camera,
As a result, when an article (not shown) as a reference object is imaged, for example, an image signal Sp as shown in FIG. 2A is obtained. This image signal Sp is supplied to a low-pass filter 2 via an amplifier 1 such as an operational amplifier to remove unnecessary noise components from the image signal. The output signal SPl of the low-pass filter 2 is input to the level comparators 31, 32, 33, .
...Supplied to one input terminal of 3n. On the other hand, the voltage from the reference voltage source B is applied to the potentiometers Pl, P2,
P3, . . . applied to Pn, and the level that has gone up from this, and the image signal SPl from the low-pass filter 2.
The level thresholds THl, TH2, T that exist within the level of
H3,...THn (which corresponds to the thresholds a to e in Figure 2A) are obtained, and these thresholds THl...
・THn, level comparator 31...3n
The signal is supplied to the other input terminal of the terminal. Therefore, each level comparator 31...3n outputs the image signal from the low-pass filter 2 to each potentiometer p1...
...Threshold value THl set by Pn...T
Output is output while Hn is exceeded. Schmitt triggers 41, 42, 43...4n that shape the rise of the waveform of the output of the level comparators 31...3n
(Thus, signals corresponding to a1 to e1 as shown in FIG. 2C are obtained.)
, 63, . supply. Therefore, these AND circuits 61...6n output clock pulses as shown in FIG. 2D. These AND circuits 61...
6n output clock pulse to counters 71, 72, 7.
3, . Memorize it as a number sequence pattern. Next, the object to be inspected is photographed with a television camera, and the same processing is performed to create a number sequence pattern.This is compared with the number sequence pattern of the reference body using a computer, and the quality of the object to be inspected is determined.

It should be noted that the technique of processing the sequence pattern with a computer is a well-known technique for such an inspection apparatus, so a detailed explanation thereof will be omitted. In addition, in Fig. 3, 9 is a modulator,
This controls the clock pulse generator 5 to obtain the above-mentioned frequency modulated clock pulses, thereby implementing the latter image processing technique.

Based on the embodiments of the invention described above, it will be obvious that many variations and modifications may be effected by one skilled in the art.

[Brief explanation of the drawing]

1A and 1B are waveform diagrams for explaining the conventional information processing method, FIGS. 2A to 2D are waveform diagrams for explaining the pattern information processing of the present invention, and FIG. 3 is a waveform diagram for explaining the pattern information processing of the present invention. It is a system diagram showing an example of a device. In the figure, Sp and Sp' are image signals, Sc is a tick pulse, a to e are threshold values, T is a television camera, 1 is an amplifier, 2 is a low-pass filter, 3, to 3n are level comparators, 41 to 4n is a Schmitt trigger, 5 is a clock pulse generator, 61 to 6n are AND circuits, 71 to 7n are counters, 8 is a computer, 9 is a modulator, B
is a voltage source, P1 to Pn are potentiometers, and TH1 to THn are threshold values.

Claims (1)

[Scope of Claims] 1. A television camera that forms an image signal of an article, means for creating at least three different threshold values within a level of one period of the image signal, and a television camera that forms an image signal of an article, at least three means for respectively discriminating the divided signals from the means using at least three different threshold values; A pattern information processing device comprising an AND means and at least three means for respectively counting the number of clock pulses from the at least three AND means and creating a number sequence representing one period of the image signal. 2. The pattern information processing device according to claim 1, wherein the clock pulse generating means includes a frequency modulator, and frequency modulates the clock pulse.