JPH04195477A - Pattern recognition device for circular body - Google Patents

Abstract

PURPOSE:To recognize the pattern of the circular body by cutting and processing data of a preset radial window part from pattern data which are inputted. CONSTITUTION:An input image signal is converted by an A/D conversion part 2 into a multi-valued digital image signal, which is stored in an image memory part 4. Then a center calculation part 5 finds the center position of the circle by processing the image signal and stores the found center point of the image signal in the memory in the center calculation part 5, and also calculates the diameter of the circle simultaneously with the center of the circle and stores it in the memory in the center calculation part 5. Then a radial window setting part 6 sets windows radially from the center of the circle. A data processing part 7 cuts data in the respective windows and concentrates the data in the respective windows. Consequently, the circular body is identified with small quantities and the data processing is speeded up.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a pattern recognition device for circular objects, which is capable of recognizing and sorting patterns on the surface of circular objects such as coins.

(Prior Art) Conventional devices that recognize patterns on circular objects such as coins use a feature extraction method or a pattern matching method.

However, according to such feature extraction methods or pattern matching methods, it is necessary to process the entire range of imported pattern data, which increases the amount of data to be processed and reduces the time required for pattern recognition. The problem was that it was long.

Therefore, the present applicant has proposed a discrimination method and a pattern recognition device in which a ring-shaped window is set for the captured pattern data of a circular object, and the circular object is discriminated by processing the image data within this window. A patent application was filed first.

This is the invention disclosed in Japanese Patent Application No. 1-177686 and Japanese Patent Application No. 2-91905.

(, 5! Problems to be solved by Ming) According to the invention of the above application, only the data within the set ring-shaped window needs to be processed, so the amount of data to be processed is small and pattern recognition can be performed in a short time. It has the advantage of being able to

However, depending on the pattern on the surface of a circular object, the pattern data within the set ring-shaped window may not have a characteristic portion, and in such a case, pattern recognition cannot be performed. Therefore, in order to obtain a predetermined recognition rate, it is necessary to set a large number of ring-shaped windows or to widen the window width, which has the disadvantage that the amount of data to be processed increases and the processing time increases. . Of course, there is no need to process the entire captured image data, so
The processing time is shorter than conventional image processing methods.

The present invention has been made in view of this problem, and an object of the present invention is to provide a pattern recognition device for circular objects that can further shorten processing time and improve the recognition rate.

(Means for Solving the Problems) The present invention includes an image input section that inputs pattern data of a circular object, an image memory section that stores the input pattern data of the circular object, and an image input section that inputs pattern data of a circular object. A center calculation unit that calculates the center of a pattern of a circular object, a window setting unit that sets windows radially from the calculated center, and data that imports data of the set radial window from the image memory unit and processes this data. The present invention is characterized in that it includes a processing section and an identification section that identifies a pattern of a circular object based on data of the processed radial window.

(Operation) Data of a preset radial window portion is fetched from the input pattern data and processed. If the radial window is set so that the captured data includes the characteristic parts of the pattern, pattern recognition of circular objects is possible.

(Example) In FIG. 1, an image input unit 1 is composed of a television camera or the like, and receives an image signal obtained by imaging a pattern of a circular object such as a coin. The image input section 1 uses a one-dimensional sensor or a two-dimensional sensor. In any case, it should be possible to process it as a two-dimensional image. Therefore, in the case of a -dimensional sensor, one screen is input by moving relative to the object. In the hardware configuration shown in FIG. 2, the imaging lens 11 and the television camera 12 correspond to the image input section 1 described above. Since the object is circular, camera 1
Set the camera so that the object looks circular when viewed from 2. It is also assumed that the object is illuminated in a manner suitable for image capture by the camera 12.

The analog image signal taken in by the image input section 1 is
The D converter 2 converts the signal into a multivalued digital image signal, for example, an 8-bit, 256-gradation signal, and inputs it to the image processor 3. The image processing section 3 includes an image memory section 4 and a center calculation section 5.
, a radial window setting section 6 , a data processing section 7 , and an identification section 8 .

The image memory section 4 stores the multivalued digital image signal. The center calculation unit 5 calculates the center of the pattern of the circular object stored in the image memory unit 4. A conventionally known method may be used to calculate the center. For example, since the object is circular, the center can be found by finding its center of gravity, or it can also be found by finding a histogram on the X and Y coordinates and finding the center positions of each. I can do it. The center can also be determined by taking a plurality of points on the edge of the image signal and using the method of least squares from the plurality of points.

In any case, find the center as accurately as possible.

The radial window setting section 6 sets one or more windows radially from the center calculated by the center calculation section 5.

The data processing section 7 takes in data of the radial window set by the window setting section 6 from the image memory section 4 and processes this data.

The data processing method is to aggregate the data for each window and apply FFT (fast Fourier transform) to it.
Use an appropriate method such as pattern matching.

The identification unit 8 recognizes the pattern of a circular object based on the radial window data processed by the data processing unit 7. When using FFT, recognition is performed based on characteristic frequency components, and when using pattern matching, recognition is performed while comparing with a pre-registered pattern. It goes without saying that whether using FFT or pattern matching, only data within a preset radial window is processed and recognized.

The center calculation section 5, radial window setting section 6, data processing section 7, and identification section 8 are included in the processing device 13 in the hardware configuration shown in FIG. The processing device 13 includes a CPU, R
It can be configured with a computer circuit having AM, ROM, etc., and if necessary, speeding up can be achieved by using a processor dedicated to signal processing such as a digital signal processor (DSP).

Next, the details of the pattern recognition operation according to the above embodiment will be explained in the third section.
This will be explained with reference to FIGS. 6 through 6.

First, an image signal of a circular object such as a coin is input from the image input section 1 . FIG. 4(a) shows a simplified example of the image 14 of a coin. The input image signal is converted into a multivalued digital image signal by the A/D converter 2 and stored in the image memory 4.

Next, the center calculation section 5 calculates the image signal to find the center position of the circle. The center point of the obtained image signal is determined by the center calculation unit 5.
stored in internal memory. At the same time as calculating the center of the circle, the diameter of the circle is also calculated and stored in the memory within the center calculation unit 5.

Next, the radial window setting section 6 sets windows radially from the center of the mouth.

Depending on the diameter of the target circular object and various other conditions, the length, width, number, etc. of windows are set so that the amount of data to be processed is as small as possible and a sufficient recognition rate is ensured. FIG. 4(b) shows a state in which a plurality of radial windows 15 are set for the image 14 of the coin. Also,
The data in each window 15 is taken in by the data processing section 7 and stored in the memory within the data processing section 7.

FIG. 4(c) shows each window Wl, W2. The data in W3°-.wn is shown, and this data is stored in the memory. (a), (b), and (c) in FIG. 4 correspond to (a), (b), and (c) in FIG.

Next, the data processing unit 7 aggregates the data of each window. Data aggregation methods include: - simple addition of the values of all pixels within the window, - sum of the absolute values of the differences between adjacent pixel values, - simple sum of the absolute values of the differences with the average value, - moving average. There is the absolute value sum of the difference between the values, and so on. (2) corresponds to the sum of differential values, and is effective when the pattern changes small. (2) is effective when the pattern changes greatly. (2) is effective when there is uneven lighting.

Figure 5 shows the pattern data of the 500 yen coin currently in use in Japan, and Figure 6 shows the pattern data of the Korean 500 won coin.The upper half of each figure shows each window in the radial direction. This is a compilation of data.

Here, the sum of the absolute values of the data differences of each pixel with respect to the average value within each window was determined using the method described in (2) above.

Now, when the pixel configuration in one window W1 is 1×10 pixels, the value of each pixel in that window W1 is W
ll, W12. W13. ...If it is WIIO, the aggregate value W1 of window W1 is as follows.

Next, process the aggregated data for all windows. Here, the periodic component was extracted by applying FFT to the aggregated data. The waveforms in the lower half of Figures 5 and 6 are FFT
Show the output. In the case of 500 yen shown in FIG. 5, the maximum peak is at the 4th cycle, excluding the 0 cycle, and in the case of 500 won, shown in FIG. 6, the maximum peak is at the 2nd cycle, excluding the 0 cycle. Therefore, by identifying this difference using the identification unit 8, it is possible to identify whether it is 500 yen, 500 won, or another circular object.

In this way, according to the above embodiment, the window 15 is set radially from the center of the circular object 14, and the window 15 is set radially from the center of the circular object 14.
Since the circular object 14 can be identified simply by processing the data in 5, the circular object 14 can be identified with a small amount of data, speeding up data processing, and improving the identification rate. It is also possible to aim for

Here, the advantages of recognizing circular objects by setting a radial window as in the present invention will be explained in comparison with the case of recognizing circular objects by setting a ring-shaped window as in the previously filed application. .

Now, consider the case where there is a vertically symmetrical and horizontally symmetrical pattern 17 on the surface of the circular object 16, as shown in FIG. If only one ring-shaped window is set on the concentric circle of the circular object 16, the window may cover only the upper and lower patterns, or only the left and right patterns.
A circular object 2 with patterns 21 only on the left and right sides as shown in the figure
0 and a circular object 22 having patterns 23 only on the top and bottom as shown in FIG. In such a case, it is necessary to set up multiple ring-shaped windows so that the windows span the top, bottom, left, and right patterns. FIG. 7 shows a state in which two ring-shaped windows 18 and 19 are set. As shown in FIG. 11, a symmetrical pattern 25
When the pattern 25 is formed to be wide, a plurality of ring-shaped windows may overlap the pattern 25. Therefore, the phases of the plurality of ring-shaped windows must be matched and data processing becomes troublesome. In this case, identification can be made by increasing the width of the ring-shaped window, but doing so increases the amount of data to be processed and increases the time required for data processing.

In this regard, according to the embodiment of the present invention, as shown in FIG. 8, by setting an appropriate number of radial windows 15, any window 15 will overlap the pattern 17 necessary for identification. , it is possible to extract the characteristics of a pattern relatively easily, and the recognition rate can be improved with a small amount of data.

If the pattern of a circular object to be recognized is limited to some extent, recognition is possible even if only one radial window is set. For example, as shown in FIG. 12, if the circular object 26 is a 500 yen coin, there are many parallel lines 27 carved near the center of the coin, so one radial window 28 is set across the parallel lines 27. Then, since the data obtained from this window 28 includes characteristic data due to the existence of the parallel line 27, it is possible to identify that the amount is 500 yen.

The window 28 is set to be located on a line slightly offset from the center of the coin 26 in order to intersect the parallel line 27 carved on the coin 26 at an angle as close to a right angle as possible. In this way, the radial window does not necessarily need to be set on a line passing through the center of the circular object; in short, it is sufficient that it is set radially. Furthermore, a plurality of windows 28 may be set as shown in FIG. 13. In FIG. 13, a concentric circle 29 is set near the center of a circular object, and a plurality of windows 28 are set in the tangential direction of this concentric circle 29.

In this way, even though radial windows are set, various forms can be considered depending on their length, width, number, direction, etc., so it is necessary to select the most suitable window form depending on the pattern of the circular object to be recognized. to be set.

When multiple radial windows are set, data may be processed for each individual window and used for pattern recognition, or data processed for each individual window may be aggregated and processed as in the above embodiment. .

The recognition target of the present invention is not limited to coins. For example, bottle crowns and various other circular objects are eligible.

(Effects of the Invention) According to the present invention, a circular object can be identified by simply setting a window radially from the center of the circular object and processing the data within this window. This makes it possible to speed up data processing and improve the identification rate.

[BRIEF DESCRIPTION OF THE DRAWINGS] Fig. 1 is a functional block diagram showing an embodiment of the pattern recognition device for circular objects according to the present invention, Fig. 2 is a block diagram showing the hardware configuration of the same embodiment, and Fig. 3 is the above-described block diagram. A flowchart showing the signal processing operation of the embodiment, FIG. 4 is a conceptual diagram showing the signal processing operation process of the above, '! Figure J5 is a waveform diagram showing aggregated data and its processed data in the above signal processing operation.
The figure is a waveform diagram showing aggregated data and its processed data for different objects, FIG. 7 is a conceptual diagram showing the conventional ring-shaped window setting, and FIG. 8 is the radial window setting according to the present invention. FIG. 9 is a conceptual diagram showing another example of the pattern of the object; FIG. 10 is a conceptual diagram showing yet another example of the pattern of the object;
The figure is a conceptual diagram showing yet another example of an object pattern, Figure 12 is a conceptual diagram showing an example of an object with a specific pattern and a window suitable for this, and Figure 13 is another example of a radial window. FIG. 1. Image input section, 4.. Image memory section, 5..
Center calculation section, 6. Window setting section. 7...Data processing unit, 8...Identification unit, 14...Pattern, 15...Radial window, 26...Circular object. Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Number of laps - 1◆ Number of laps Figure 7 Figure 8 Figure 9 Figure 10 Figure 11

Claims (1)

[Scope of Claims] An image input unit that inputs pattern data of a circular object; an image memory unit that stores the input pattern data of the circular object; and a center of the pattern of the circular object that is determined from the data stored in the image memory unit. A window setting section that sets windows radially from the calculated center; A data processing section that reads the data of the set radial window from the image memory section and processes this data; An identification unit for identifying a pattern of a circular object based on data of a radial window.