JPH0426514B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for counting the number of objects and holes present in an image from a video signal from a television camera or the like.

Recently, image processing technology has been used for automatic product inspection and robot vision devices, and one of the important image processing technologies is to quickly count the number of objects and holes.

For example, in order to automatically inspect products that are mass-produced at high speed, such as press-cut parts, by image processing, it is necessary to display a large number of products on one screen and count the number of products at high speed. At the same time, the total number of through holes in one screen is counted in order to detect defects such as holes that should be punched with a press being incomplete or adjacent holes being connected. Since the number of through holes that should exist in one product is known, the total number of through holes that should exist can be calculated from the number of products on the screen, and if this value does not match the hole count result, it is determined that there is a defective product. do. In such a case, it is necessary to count the number of holes at the same time as counting the number of objects.

The function of counting the number of objects and the number of holes can be realized by computer software, but there is a problem that the processing speed is slow. In addition, hardware-based counting methods have been proposed to increase processing speed, but there is a problem in that if there is a hole in an object, it will result in incorrect counting. For example, as shown in Fig. 1, take the Laplacian L of four neighbors at each pixel, and if the Laplacian L is smaller than zero, -
1. When the number is zero, it is set as 0, and when it is larger than zero, it is set as +1, and the total sum divided by 4 is used as the number of objects, but as shown in Figure 1b, there is one hole in the object. In this case, the above result will be zero and will not match the solid number of the object. Furthermore, if the number of holes is greater than the number of objects, the above result has a negative value. That is, the above conventional method does not count the number of holes, but counts {(number of holes) - (number of holes)}.

In addition, since holes can lead to erroneous counts, a method has been proposed to fill in the holes as preprocessing, but even with hardware processing, this cannot be achieved with one scan per screen, and multiple scans are required. However, there is a problem in that it cannot be applied in terms of processing speed.

It is therefore an object of the present invention to eliminate the above-mentioned disadvantages and to
To provide a number/hole counting method capable of counting the number of objects and holes in an image at high speed with very simple processing.

Hereinafter, the present invention will be explained in detail with reference to the drawings.

FIG. 2 is an example of an image signal waveform for explaining the principle of the present invention. In the figure, a is a waveform obtained by binarizing an input image containing three objects, one of which has a hole, b is a waveform obtained by ANDing the current scanning line and the previous scanning line, and c indicates the run located after when one run, which is a series of "1"s in the previous scanning line, is divided into two in the current scanning line, and is the waveform to be stored in the memory circuit, and d is the waveform detected at c. The waveform used to read out the information of the previous scanning line to determine whether the runs are connected in the next scanning, e is the waveform used when one run in the previous scanning line is divided into two in the current scanning line. A waveform that is reproduced one pixel earlier to be stored in the memory circuit as information about the currently located run, f is a waveform similar to d, and g is a waveform that is reproduced one pixel before in order to be stored in the memory circuit as information about the currently located run.
The waveforms outputted when the current is reached are the waveforms obtained by performing the logical product of the d waveform, the f waveform, and the g waveform, respectively. Note that the broken line in the figure is drawn for comparison with the binarized waveform, and the signal is "0".

3 and 4 are circuit diagrams of one embodiment of the present invention.

In FIG. 3, 1 is a camera that inputs images;
2 is a binarization circuit that sets the background to "0" and the object to "1"; 3 is a shift register for one scanning line; 4a,
4b is a 2-bit shift register, 5 is an AND circuit that takes the AND of each input, 6 is a hole detection circuit, which will be explained in detail in FIG. 8 is an arithmetic circuit that performs the calculation (x-y+z) when each input is x, y, z, and 9 separates the synchronization signal from the input image. A synchronization signal separation circuit 10 indicates a pallet conveying device for sequentially conveying a pallet on which a plurality of objects having through holes are placed via a position below the camera.

Figure 4 shows an example of the hole detection circuit shown in 6 in Figure 3. 61 and 71 are terminals i1 and i2 which are "1".
Outputs “1” when the terminal changes from “1” to “0”, and always outputs “0” when the i1 terminal is “0”.
Flip-flop circuits 62a, 62 that output
b, 72, 73 are AND circuits that take the AND of each input; 63a, 63b are OR circuits that take the OR of each input; 64a, 64b, 64c are not circuits that invert the input; 65a, 65b, 65c are One-shot circuit 66a, 66 that outputs a pulse for a certain period of time from the time the input changes from "0" to "1"
When the s terminal is "0", the input of the terminal i1 is output to the terminal o1, the input of the terminal i2 is output to the terminal o2, and when the s terminal is "1", the input of the terminal i1 is output to the terminal o2, and the terminal A select circuit in which the input of i2 is output to terminal o1, 67 has an input from "0" to "1"
68a and 68b are binary counter circuits whose output is inverted when the input changes from "0" to "1", counting circuits that count up and output 10 bits when the input changes from "0" to "1", and 69a and 69b are counter circuits whose s terminal is "0" When , the input of the io terminal is output to the o terminal and the i terminal becomes high and impedance, and when the s terminal is "1", the input of the i terminal is output to the io terminal and the o terminal becomes high and impedance. The input/output switching circuits 70a and 70b respectively represent storage circuits capable of storing information for one scanning line. The circuit configuration of 6-2 is exactly the same as that of 6-1. However, the input shown in 6-1
The circuit in which I1, I2, and I3 are each connected to the (2) side is 6-2.

With this configuration, as described below, the number of objects and the number of holes can be counted immediately after the input of one screen is completed.

In FIG. 3, an image of one pallet of the pallet conveying device 10 is captured by the camera 1. At this time, a plurality of objects are placed on the pallet, including objects with through holes. The input signal from the camera 1 is converted into a logic “0” by the binarization circuit 2.
Then, the object is binarized so as to have a logic "1", a 2x2 local area image is extracted using the shift registers 3, 4a, and 4b, and one pixel on the current scanning line of this 2x2 local area image is (C pixel in the figure) is input to the counting circuit 7a and the number of runs in one screen is counted by counting the points of change from "0" to "1". In the example of FIG. 2a, the count value x is 31. Also, the AND circuit 5 calculates the AND of the previous scanning line (pixel A in FIG. 3) and the current scanning line (pixel C in FIG. 3), inputs it to the counting circuit 7b, and similarly outputs "0".
Count the points of change from to "1". FIG. 2b shows the input waveform to the counter 7b; in this case, the count value y of the counter 7b is 29. The difference between the counted values x and y (31-29=2) is {(number of objects) - (number of holes)}, which can also be obtained using conventional technology, but this value can also be calculated using hole detection, which will be explained in detail below. By adding the output of the circuit 6 to the value counted by the counting circuit 7c, that is, the number of holes, the number of objects can be calculated. That is, the arithmetic circuit 8
The output N1 of is the number of objects, and the output N2 of the counting circuit 7c is the number of holes.

Below, a method for detecting holes will be explained.

Figure 5 shows the principle of hole detection. That is, as shown in section a in the figure, one run in the previous scanning line captures the position where the current scanning line is divided into two runs, and in the next scan, each run is divided into two runs as shown in sections b and c in the figure. Confirm that they are connected, and when the two connected runs become one run as shown in section d in the figure, the two connected runs
Since the area of logic "0" sandwiched between two runs can no longer be connected to the background, it is recognized as a hole and one pulse is output. An example of a circuit for realizing this is shown in FIG. 4, where circuit 6-1 is a circuit that captures part a and confirms part b shown in FIG. 5, and circuit 6-2
is a circuit that captures part a and confirms part c shown in Figure 5, circuit 6-3 is a circuit that captures part d shown in Figure 5, and circuit 6-4 is an AND circuit that takes the AND of the above three outputs. It is a circuit. The number of output pulses of this AND circuit matches the number of holes.

The operation will be explained in more detail below with reference to FIG.

Due to the configuration of the flip/flop 61 and the AND circuit 62a, when one run of the previous scanning line is divided into two or more runs in the current scanning line, the output becomes "1". In the example of FIG. 2, pulses are output as shown in the d scanning line of c in the same figure. This information is OR circuit 6
3b, and when the current scanning line is an odd line, it passes through the input/output switching circuit 69a and is stored in the storage circuit 70a, and when the current scanning line is an even numbered line, it passes through the input/output switching circuit 69b and is stored in the storage circuit 70b. . At this time, the address is determined by the not circuit 64b and the select circuit 6 based on the information of the C pixel of the current scanning line.
6b, the counting circuit 68b or 68a, and the write signal is determined from the information of the D pixel of the current scanning line by the not circuit 64a and the one shot circuit 65.
a, the OR circuit 63a, and the select circuit 66a output one pulse when the pixel is at the end of the run.
That is, in the example of the d scanning line in FIG. 2c, addresses 1,
“0” is written to addresses 2 and 4, and addresses 3 and 5
“1” is written to. This stored information is read out at the appropriate location during the next scan.
That is, when the line is an odd number, it is read out from the memory circuit 70b, and when it is an even line, it is read out from the memory circuit 70a, and the input/output switching circuit 69b or 6
9a and becomes an input to the AND circuit 62b. The read address is determined by the select circuit 66b and the counting circuit 68a (or 68b) based on the information of the A pixel of the previous scanning line, and the read signal is determined by the one shot circuit 65b, based on the information of the A pixel of the previous scanning line. The select circuit 66a outputs one pulse at the start pixel of the run of the previous scanning line. The state is shown in Figure 2 d. In the 1st, 2nd, and 4th runs in the scanning line e, there is no output because it is "0", and in the 3rd and 5th runs it is "1", so there is no output "1" is output from the start of a line run, and "1" is maintained until no run exists in both the current and previous scan lines. This output is one input of the AND circuit 62b, and the other input is information on the C pixel of the current scanning line. Therefore, the output of the AND circuit 62b becomes "1" when the previously captured runs are connected in the current scanning line, and becomes "0" when they are not connected. This information passes through the OR circuit 63b and the input/output switching circuit 69a or 69b again and is stored in the storage circuit 70a or 70b. By repeating the same procedure, f in Fig. 2 d
As shown in the scanning line below, it becomes "1" if it is connected to the run extracted in c of the figure, and becomes "0" if the connection is broken. This makes it possible to detect the connection of runs located at the rear, starting from the hole.

The next step is to detect the connection of runs located in front of the hole starting from the hole, which is basically the same as the above explanation. The only major difference is that, as shown in Figure 2e, the run splits into two and the hole begins when the rear run appears, whereas the front run appears. I don't know when it exists.
In this state, information indicating that the hole is the beginning must be written to the previous address. To solve this problem, as shown in circuit 6-2, the inputs of I1 and I2 are set to pixels B and D, which are reproduced one pixel ahead of the input of circuit 6-1, and the beginning of the hole is set one pixel ahead of the input of circuit 6-1. The information is extracted and stored before the address of the capture memory circuit changes. I3 is an input for generating a write pulse at this time. In this case, the information of the third run is stored at address 2 and the information of the fifth run is stored at address 4, as shown in scan line d in FIG. 2e. That is, addresses 1, 3, and 5 are “0”,
Addresses 2 and 4 become "1". In the next scan, the method of extracting runs connected to the second and fourth runs is exactly the same as described above, and the output waveform is as shown in FIG. 2f. This makes it possible to detect the connection of runs located in front of the hole.

Next, in order to detect the end of the hole, it can be realized by inputting the C and A pixels to the circuit 6-3 and using the flip-flop 71 and the AND circuit 72. This output waveform is as shown in Fig. 2g.

By inputting the three outputs of the circuits 6-1, 6-2, and 6-3 to the circuit 6-4 (AND circuit 73), the output of the circuit 6-4 becomes as shown in FIG. 2h.
When there is one hole in the object, one pulse is output. If there are multiple holes, the same number of pulses will be output.

This allows the number of holes to be counted, and as described above, the number of objects can also be counted accurately.

As described in detail above, according to the present invention, it is possible to count the number of objects and holes within a screen by simple processing, to perform counting at high speed, and to improve the hardware configuration of the counting device. Remarkable effects such as miniaturization can be obtained.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram for explaining the conventional counting method, Fig. 2 is an explanatory diagram for explaining the processing of the present invention, Fig. 3 is an overall circuit diagram of one embodiment of the present invention, and Fig. 4 is an explanatory diagram for explaining the conventional counting method. is a circuit diagram of one embodiment of the hole detection circuit of the present invention,
FIG. 5 is an explanatory diagram for explaining the hole detection method. 1...Camera, 2...Binarization circuit, 3, 4a, 4b
...Shift register, 5...AND circuit, 6...hole detection circuit, 7a, 7b, 7c...counting circuit, 8...arithmetic circuit, 9...synchronous signal separation circuit, 10...pallet transport device, 61, 71...flip, flop circuit ,6
2a, 62b, 72, 73...AND circuit, 63
a, 63b...OR circuit, 64a, 64b, 64c
...Knot circuit, 65a, 65b, 65c...One shot circuit, 66a, 66b...Select circuit, 6
7... Binary counter circuit, 68a, 68b... Counting circuit, 69a, 96b... Input/output switching circuit, 70a,
70b...Memory circuit.

Claims (1)

[Claims] 1. Binarize the image signal obtained by scanning the image, count the number of runs that are connections of "1" on the scanning line for one screen, and at the same time calculate the logical product of the signals of the current scanning line and the previous scanning line. The number of resulting runs is counted, and the position where one run in the previous scan line is divided into two runs in the current scan line is extracted, and the two divided runs are connected in the next scan. When the two connected runs become one run, it is recognized as a hole and counted, and the number of objects and holes is calculated based on the above three values. Features a number of pieces/holes counting method.