DE2912559C2 - Pattern information processing method and means - Google Patents

Description

characterized by the step of frequency modulating the clock pulses (SC).

2. Pattern information processing device for performing the method according to claim 1, comprising:

a) an image sensor (T) that records an object,

so that it generates a video signal (SP, SP ') of this object;

b) means (P ₁ -P _n ) for generating a plurality of threshold levels (ae; THi-TH _n ) which differ from one another;

c) a device (3, -3 ") for subdividing the video signal (SP; SP ₁ ) by the plurality of threshold levels (ae; TH _{1 -TH n} ) so that gate signals (fli-e _; ) are generated corresponding to the video signals divided by the plurality of threshold levels (ae; TH _{x -TH n} );

d) a device (5) for generating a clock pulse signal or a clock pulse series (SC);

e) an AND device or an AND gate device (6, -6 ″) for gating the clock pulse signal or the clock pulse series (SC)) by means of each of the gate signals (öi-e,); and

f) a device (7, -7 ") for counting the clock pulse signals or rows (α _Γ ε ^) ^from the AND device or the AND gate device (O ₁ -O _n ) and for generating a series of numbers, which represents the captured object;

characterized by a device (9) for frequency modulating the clock pulse signal (SC).

The invention relates to a pattern information processing method, which comprises the following process steps:

a) Recording an object by means of an image sensor for generating a video signal which the object is equivalent to;

b) Splitting or subdividing the video signal by means of a Plurality of threshold levels that are different

are from each other;

c) gating clock pulse trains with the subdivided video signals for generation clock pulse trains corresponding to the plurality of threshold levels; and

d) Counting the clock pulse series to generate a series of numbers, which is a pattern of the recorded Object represents.

The invention also relates to a pattern information processing device to carry out the above procedure comprising:

a) an image sensor which picks up an object so that it generates a video signal of that object;

b) means for generating a plurality of threshold levels which differ from one another;

c) means for dividing the video signal by the plurality of threshold levels, so that gate signals are generated which are divided by the plurality of threshold levels or divided video signals;

d) a device for generating a clock pulse signal or a clock pulse series;

e) an AND device or an AND gate device for gating the clock pulse signal or the Clock pulse train by means of each of the gate signals; and

f) means for counting the clock pulse signals or rows from the AND device or the AND gate device and for generating a series of numbers representing the recorded object.

According to the prior art are in the field of testing, investigating or inspecting Various types of pattern information processing methods and devices are proposed for objects been. According to a prior art method, the pattern information becomes a Object to be examined, checked, inspected or recognized in two values of black and White, serving as data representing the shape, area, etc. of the object, and these data are compared with the data of a reference body which has been previously determined so that in this way it is decided whether the object is good or not.

However, there are actually few such practical objects that can be clearly divided into two values of black and can subdivide white or characterize it by two values of black and white. Generally Objects have a certain amount of light and shadow in addition to black and white. As a result is the prior art method is disadvantageous in that it is difficult to locate an object that a complicated pattern has to examine, investigate or inspect.

In a case where pattern information processing methods In the prior art, when a television camera or the like is used, the screen of this television camera is used in, for example 500 x 500 picture elements, the degree of light and darkness of an image that affects the picture elements is projected is through an analog-to-digital converter is converted into digital values that serve as data, and this data is converted by means of a Computer or the like. Processed to the characteristics of the

to extract the object captured by the TV camera. In this case, the amount or the Amount of data massive or very large and it takes a long time to create such a massive one Amount or such a large amount of data to process.

There are also methods for inspecting, investigating or checking errors, defects or the like. an article in which the screen of a television camera is not divided into a plurality of picture elements is, but a certain value of a threshold level is set, and then as an example only the part of the video signal from the television camera that exceeds the threshold level into two values converted, and these values are counted to represent the characteristic of the object or a To identify defects or deficiencies in the object.

The above prior art method should be trimmed with reference to FIG. In this case, a single threshold value S _{1 is} set and the inspection of an object is carried out in such a manner that it is decided whether a video signal S ₂ coming from a television camera shooting the object to be inspected exceeds the threshold value S _x or not. This method is effective only for inspecting objects or defects that are simple in pattern, but it is not effective for inspecting, examining, checking objects with complicated patterns.

A sample information processing method and a sample information processing device of the type mentioned are known from the article "Video signal analysis: The quantitative microscope" by Gerhard Kamin in the magazine "radio mentor electronic" 1969, pages 156 to 157 and have the disadvantage that there are a number of from cases of pattern recognition to wrong decisions about the equality or difference of patterns, as is explained in more detail below with reference to FIGS. 2A to 2D with reference to the video signals SP and SP 'shown in FIG. 2A.

In contrast, the object of the invention is to provide a pattern information method and a pattern information processing device of the type mentioned to make available with which the possibilities the decision-making error about the equality or difference of a pattern is significantly reduced will.

According to the invention, this object is achieved by the method step of frequency modulating Clock pulses.

The pattern information processing device according to the invention, with which this object is achieved, is characterized by a device for frequency modulating the clock pulse signal.

In this way it is achieved that different patterns, which differ for example in the way as explained with reference to the video signals SP and SP ' , which according to the prior art are incorrectly recognized as the same despite their differences, by means of the pattern information processing method according to the invention and correctly recognized as different by means of the pattern information processing device according to the invention.

A preferred embodiment of the invention will be described below with reference to FIGS. 2A to 3 of the drawing explained in more detail; it shows

Fig. 1 is a waveform diagram used for explaining an information processing method according to is state of the art;

Figs. 2A, 2B, 2C and 2D are waveform diagrams, respectively. those for explaining an embodiment of the pattern information processing method according to FIG serve the invention; and

Fig. 3 is a schematic block diagram showing an embodiment of a pattern information processing device according to the invention, which illustrates the pattern information processing method of the invention executes.

First, referring to Figs. 2A, 2B, 2C, and 2D, the pattern information processing method without applying frequency modulation, as described in principle from the above Article of the magazine "radio mentor electronic" is known for the quantitative evaluation of original images, to show how this pattern information processing method corresponds to the above-mentioned Wrong decisions about the sameness or diversity of patterns can result.

In Fig. 2A, SP denotes a video signal from an image sensor, such as a television camera, which picks up an object or goods that are to be recognized or inspected, examined or checked. In this case, the video signal SP corresponds to a horizontal scanning interval of the television camera. A plurality of threshold levels are set, namely five threshold levels a, b, c, d and e in the illustrated example, these threshold levels are set so that they differ from one another and lie within the level range of the video signal SP , as in FIG Figure 2A is shown. The number of threshold levels and the interval (level difference) between adjacent threshold values are not limited to a specific number and a specific value, but can be selected as desired.

When the parts of the video signal SP which intersect the respective threshold levels α to e are extracted, gate signals a _u b _u c>, d, and e _{x of} the type shown in Fig. 2C are obtained.

If a clock pulse train SC with a constant frequency, as shown in Fig. 2B, and the gate signals α, to e, shown in Fig. 2C are each given together to the inputs of an AND gate, then the Clock pulse trains a ₂ , b _2j C ₂ , d ₂ and e _{2 of} the type shown in Fig. 2D are obtained.

The number of clock pulses SC in each of the clock pulse series a ₂ to e ₂ is counted to form a series of numbers from the numbers of the clock pulses SC in the respective clock pulse series a ₂ to e ₂ , and the series of numbers is used to create a pattern or feature or to represent a property of the video signal SP and therefore of the object which corresponds to the video signal SP.

If the threshold levels are set to α to e and the frequency of the clock pulse series SC is chosen to be constant, as shown in Fig. 2, the numbers of the clock pulses in the respective clock pulse series a ₂ to e _{2 become} 4 + 3 = 7, 14, 3 + 17 = 20, 2 + 22 = 24 and 34. Therefore, the series of numbers corresponding to the video signal Sf shown in Fig. 2A and representing this video signal is expressed in the following form: {7, 14, 20, 24, 34 }.

If objects or articles, objects,

Goods or the like recorded by a television camera are different from each other the video signals coming from the TV camera differ accordingly. It means that then, when video signals, the different articles (this term is used collectively for the Terms objects, goods, articles or the like used) or objects correspond by the above Processes are processed in order to generate respective series of numbers, the respective objects or Articles are represented by their own series of numbers. Although it would have to be assumed that the series of numbers obtained by processing the video signals as explained above correspond accordingly the objects or articles are different, there are a number of cases in which different Video signals from different articles or objects lead to the same series of numbers, which leads to wrong decisions in pattern recognition. Such a case is explained below.

The video signal SP ' in Fig. 2A differs from the video signal SP in the way, as is indicated by the associated dashed lines in Fig. 2A, ie it has a different course in the area of the dashed lines than the video signal SP and has in the remaining areas in which there are no dashed lines have the same course as the video signal SP. Due to the illustrated difference between the video signals SP and SP ' , the interval in which the respective video signal intersects the threshold level e is different for the video signals SP and SP'. This means that the video signal SP ' crosses the threshold level e in its rising part sooner than the video signal SP, namely by the time period T1; and in a similar manner the video signal SP ' in the descending part crosses the threshold level e sooner than the video signal SP by the same time period Tl. As a result, in this case, as indicated in Fig. 2C by the corresponding dashed lines, the length (Time duration) of the gate signal e obtained from the video signal SP 'is not changed in relation to the length of the gate signal e _t obtained from the video signal SP , because the portion of the length which is added to the gate signal e _t on the left in FIG. 2C is to the right of taken away from him, so that the total length of the gate signal e _t remains unchanged. As a result of these relationships, as indicated by the dashed lines in Fig. 2D, the number of clock pulses _{contained in the clock pulse series e 2} in the video signal SP ' remains unchanged from the video signal SP , and as a result, the series of numbers is which represents the video signal SP ' , the same as the series of numbers representing the video signal SP. In this way, although the video signal SP 'is different from the video signal SP , a decision is made that both video signals are the same, which means that the article or object corresponding to the video signal SP' is different from that Differentiates article or object which corresponds to the video signal SP is erroneously recognized as being the same as the latter.

To avoid such an error, a clock pulse series SC is used, which is frequency-modulated, that is, does not have a constant frequency. The number of frequency-modulated clock pulses is not the same in a unit time interval, but changes over time. The number of frequency-modulated clock pulses in the first time interval 71 therefore differs from that in the later time interval 71 (see FIG. 2A), the duration of which is equal to the first time interval T1 . As a result, the series of numbers generated by counting the frequency-modulated clock pulses and representing the video signal SP ' becomes different from the series of numbers representing the video signal SP . Therefore, the above erroneous detection can be avoided by using the frequency-modulated clock pulse train.

It is readily apparent from the above explanation that the use of a frequency-modulated Clock pulse series has the same effect for the other threshold levels.

Next, let us refer to an embodiment of a pattern information processing device 3, which carries out the above method.

In Fig. 3, reference numeral 7 ", an image sensor referred to, which may for example be a television camera. For example, if a reference item or object (not shown) is received by the image sensor T, then this generates a reference video signal of the way it at SP in Fig. 2A The video signal SP from the image sensor T is inputted through an amplifier 1 such as an operational amplifier to a low-pass filter 2 which serves to remove undesirable noise components present in the amplified video signal SP and a Video signal SP _1. The video signal SP obtained at the output of the low-pass filter 2 is applied to the input terminal of each of the level comparators 3 1, 3 ₂ , 3 _{3 ... 3 "" provided as a means for dividing the video signal} the voltage B of a reference voltage source is applied to potentiometers P ₁ , P ₂ , P ₃ ... P _n , which is a device for generating thresholds levels THj, TH ₂ , TH ₃ . . . Form TH _n . These mutually different threshold levels are selected so that they are each within the level range of the video signal SP ₁ from the low-pass filter 2. The threshold levels TH ₁ to TH _n (which correspond, for example, to the threshold levels α to e shown in FIG. 2A) are applied to the other input terminals of the level comparators 3] to 3 ", respectively. Therefore, the respective level comparators 3, 3 to 3 "generate output signals during the time intervals during which the video signal SP ₁ from the low-pass filter 2

so the respective threshold levels TH ₁ to TH _n , which have been set by the respective potentiometers P ₁ to P _n , exceeds. The output signals from the level comparators 3] to 3 "are sent via Schmitt trigger circuits 4", 42, 43 ... 4 ", which form the rising part of the waveform and generate gate signals which, for example, correspond to the gate signals a _x to e shown in FIG. 2C , correspond to one input terminal of AND circuits 6 _U 6 ₂ , 6 ₃ . . . 6 ″, which form an AND gate device and whose other input connections are given a frequency-modulated clock pulse series. The frequency-modulated clock pulse series SC, which is shown only schematically in Fig. 2B, that is, without the frequency modulation being expressed in the illustration, is generated by means of a device 5 for generating a clock pulse signal or a clock pulse series, which is generated by a device 9 for frequency modulating the Clock pulse signal is controlled. As a result, give

the AND circuits 6, to 6 ″ clock pulse series which correspond, for example, to the clock pulse series a ₂ to e ₂ , which are shown in FIG. 2D. These clock pulse series from the AND circuits 6, to 6, are each given to counters 7 |, 7 ₂ , 7, ... 7 ", which form a device for counting the clock pulse signals of the clock pulse series a ₂ to e ₂ . The counting results produced by the counters 7 to 7 ″ are input to a computer 8, which then generates reference pattern information, ie reference number series, and stores it.

Then an object to be examined, checked, inspected or recognized is picked up by the image sensor T , and the video signal which the latter supplies at its output is processed in a similar manner to that set out above, so that a series of numbers can be Pattern represents, is generated, and this series of numbers of the object to be inspected, checked, examined or recognized is compared with the series of reference numbers in the computer 8 in order to decide whether the object corresponds to the reference article or object. As a result, the object can be precisely inspected, examined, checked or recognized with the described pattern information processing device which correctly executes the pattern information processing method explained above.

It is readily apparent that the pattern information processing method described is not only applicable to a pattern recognition par excellence, but can also be used to check whether an object or Item is good or not, whether a flaw or deficiency, for example a crack, a tear, a bubble or the like., is or is not present in goods, as well as for determining the shape of an object and for measuring a Area, area or the like of an object. In such a case, the numbers and sizes the threshold levels according to the item to be examined, tested or inspected changed so that one gets the required series of numbers that are appropriate to the desired purpose to reach.

With this pattern information processing method as explained above, the articles even if they are complicated in their pattern, course or the like, recognized or carefully examined, be checked or inspected, since, as already mentioned, the number etc. of the threshold levels accordingly the patterns of articles that are to be recognized or checked can be changed to the corresponding to get series of numbers representing patterns.

For this purpose 2 sheets of drawings

55

60

65

Claims (1)

Patent claims: 1. Pattern information processing method that includes the following process steps: a) Recording an object by means of an image sensor to generate a video signal corresponding to the Object corresponds to; b) dividing the video signal (SP, SP ') to by means of a plurality of threshold levels (ae; TH _{1 -TH n} ) which are different from one another; c) gating clock pulse trains (SC) with the subdivided or divided video signals to generate clock pulse trains (az-e ₂ ), ^we ' ^{cne correspond to} the plurality of threshold levels (ae; TH _x -TH _n ) ; and d) counting the clock pulse series (azei) to generate a series of numbers which represents a pattern of the recorded object;