JP2641237B2 - Circular object recognition method - Google Patents

Description

The present invention relates to a method for recognizing a circular object using a two-dimensional wave equation. The object of the present invention is to provide a method for recognizing a circular object in which a plurality of frame memories are provided to increase the calculation speed. Alternatively, an image of an object similar to this is input to the arithmetic unit as a two-dimensional gray value, and a two-dimensional wave equation with the gray value as an initial value is solved, and the center of the circular object is determined by the extreme value of the amplitude in the solution of the equation. In the circular object recognition method for confirming a position, at least two frame memories, that is, a first frame memory and a second frame memory are provided, and the first frame memory stores two images of the image after a predetermined time.
A two-dimensional gray value is stored in a second frame memory, and a time differential value of the two-dimensional gray value is stored in a second frame memory. Substituting 0 into the memory, obtaining the Laplacian of the image held in the first frame memory in step 2, adding the result to the content of the second frame memory, writing the result in the second frame memory, and writing the result in step 3 The contents of the two-frame memory are multiplied by a predetermined number less than 1, and the result of adding this value to the contents of the first frame memory is written to the first frame memory. The center position of the circular object is obtained from the x and y addresses of the first frame memory, and the radius of the circular object is obtained from the number of repetitions.

TECHNICAL FIELD The present invention relates to a circular object recognition method using a two-dimensional wave equation.

In the machine industry, circular objects occupy a large proportion as objects for visual recognition of objects. Most of the work of lathes and milling machines is the formation of circles, and bolts and rivets are fastened to round holes. In addition, the terminals of printed boards widely used for mounting electronic components are often ring-shaped. The position of each terminal part is known because it is formed by printed wiring, but it is unavoidable that a slight displacement occurs, so the center position of the terminal part is determined at the time of automatic bonding, and accurate bonding is performed It is desired to do so. In production facilities that have introduced robots, robots take parts one by one that are created and transported in the upstream process, mount them on processing machines, or move them to a conveyor for downstream processes. However, also in this case, it is necessary to correctly recognize the position and shape of the component.
When recognizing a circular object, the quantities to be measured are the center position coordinates and the radius of the circle.

2. Description of the Related Art Various methods for recognizing a circular object have been proposed. However, many of the conventional methods extract a contour and calculate a center of gravity and a perimeter, so that when a contour line is interrupted, it cannot be traced as a closed loop. And the center of gravity and circumference cannot be calculated. If the break is very small, a technique for filling the gap is known, but it is often difficult to remove a large portion.

This drawback is particularly problematic when processing grayscale images using reflected light.

Therefore, it has been pointed out that uniform processing and the effectiveness of a macroscopic method are effective as a recognition method that is not disturbed by background, surface patterns, and noise. A method of extracting a skeleton line by generating a wave from the periphery of an image is one example.

A similar idea is used to compare the outer periphery of a circular object to a wave on the water surface, and utilizing the fact that waves propagated from the outer periphery gather at the center after a certain time to form a peak, and a figure recognition method for determining the center position is disclosed in No. 51387.

In the figure recognition method disclosed in this publication,
A wave equation is used for recognizing a circular object, and the wave equation is used from the following viewpoints. For example, watering a circular basin and impacting the rim will cause the waves generated from the surroundings to propagate toward the center and peak at a certain time later. Therefore, if the circumference of the circular object is regarded as a portion surrounded by a circle on the water surface and the movement of the wave is simulated inside the calculator, a peak should similarly occur. If the position of the peak is known, the center position of the circle can be known, and the radius can be obtained from the time until the peak is reached.

The strength of this recognition method is that some circles may be missing or multiple circles may overlap. The reason why the peak occurs is that all waves generated simultaneously on the circumference reinforce in phase with each other at the center point, but if the circle is partially missing, the remaining arc has a sufficient length. This is because there is no difference in constructing waves generated therefrom in the same phase. As for the patterns and noises on the surface, the phases of the generated waves do not randomly generate large peaks unless they are circular, and do not interfere independently of the waves originally generated from the circumference.

The graphic recognition method described in the above-mentioned publication has an advantage that the center position and the radius of a circular object can be accurately obtained without being interfered by a pattern or noise on the surface. However, there is a problem that the calculation takes time.

The present invention has been made in view of such a point,
An object of the present invention is to provide a circular object recognition method in which a plurality of frame memories are provided to increase the calculation speed.

Means for Solving the Problems FIG. 1 shows a principle flowchart of the present invention.

In the present invention, at least two frame memories are used. That is, a first frame memory A and a second frame memory V are provided, the first frame memory A holds a two-dimensional gray value of an image after a predetermined time, and the second frame memory V stores a time differential value of the two-dimensional gray value. Hold.

In step 1, the two-dimensional gray value A (x, y, 0) of the original image is written into the first frame memory A, and 0 is substituted into the second frame memory V. Further, the counter n indicating the number of approximations is set to 1. . In step 2, the Laplacian ΔA of the image held in the first frame memory A is obtained, and the result is added to the contents of the second frame memory V, and the result is written to the second frame memory V. Next, in step 3, the second
A predetermined number less than 1 such as 1/8 in the contents of the frame memory V
And the result of adding this value to the contents of the first frame memory A is written to the first frame memory A.

In step 4, the values of the addresses x, y, and n of the first frame memory which are equal to or larger than the predetermined threshold value S are detected, and these are written in the peak table. In step 5, the counter n
Is incremented by one, and in step 6, n is It is determined whether the number is equal to or more than the pixel. Where R is the maximum radius of the circle to be found and in one approximate cycle the wave is It is assumed that the display advances by a pixel. In the case of a negative determination in step 6, steps 2 to 5 are repeated, and in the case of an affirmative determination, this routine ends.

According to this processing routine, the center position of the circular object is obtained from the x and y addresses of the first frame memory A having the predetermined threshold value S or more, and the radius of the circular object is obtained from the number of repetitions.

In the present invention, the brightness of the grayscale image is calculated based on the coordinates in the horizontal and vertical directions.
It is regarded as a three-variable function of x, y and time t. This is A (x, y,
t) and write, of 1 frame input to the time t _o image A (x,
y, t _o ) as two-dimensional wave equation However, a is to solve the velocity of the wave. The boundary condition is always A (x, y, t) = 0 on the periphery (square) of the image frame. Physically, this is the same as the equation describing the state of the elastic membrane in which the periphery of the square is fixed.

Since the equation (1) is converted to a difference equation and solved sequentially, a new variable Is introduced as the first-order simultaneous equations.

That is, the following equation is obtained by rewriting the equation (1) using V (x, y, t).

When equation (2) is replaced with a difference approximation equation, the following is obtained.

V (x, y, t + δt / 2) V (x, y, t−δt / 2) + a ² δtΔA (x, y, t) A (x, y, t + δt) A (x, y, t) + δtV ( x, y, t + δt / 2) (3) The algorithm for solving equation (3) is given below.

Step 1: V = 0, A = A (0) Step 2: V ← V + a ² δtΔA Step 3: A ← A + δtV Step 4: Return to Step 2 With the above algorithm, a ² δ = 1, δt = 1/8 FIG. 1 is a flow chart showing the principle of the present invention. This operation is efficiently executed by using a general-purpose local parallel image processor as disclosed in Japanese Patent Application Laid-Open No. 62-137669. can do.

In the principle flowchart of FIG. 1, steps 2 and 3 mean that calculations are performed on the entire contents of each frame memory, and the time required for each step is one frame period, that is, about 33 ms. When the repetition of step 2 and step 3 is called one approximation cycle, the first frame memory A holds A (x, y, t + nΔt) at the time when the nth approximation cycle is advanced. In steps 2 and 3, since it is necessary to perform calculations on the entire contents of the frame memory within one frame period, a high-speed image processor is generally required to execute this processing routine. Is as shown in FIG. 2, and the coordinates x, y of the center of the circle and the radius thereof can be easily obtained from the peak table.

EXAMPLE Although the flow chart of FIG. 1 illustrates the principle of the present invention, a certain image processor may not be able to execute the processing flow of FIG. For example, the data read from the second frame memory V in step 2 is added to the Laplacian ΔA and written into the same frame memory V, but it is generally difficult to do this within one frame period. This is because it is necessary to switch between reading and writing of the memory element within the sampling period. If each frame memory is always in the read state or the write state within one frame period, control becomes easier and the circuit becomes simpler.

Also, depending on the processor, it may not be possible to add Laplacian to one frame memory and add it to another frame memory as in step 2 in FIG. In other words, it may have only a simple frame addition function. In order to execute the present invention on a processor having only such a simple function, the number of working memories may be increased and the processing flow shown in FIG. 3 may be used.

That is, in the present embodiment, two work memories W ₁ and W ₂ are added in addition to the frame memory A and the frame memory V. In step 11, the two-dimensional gray value of the original image is written into the frame memory A, and the frame memory V
To 0. Further, a counter n indicating the number of approximations is set to 1
Set to. Then writes the Laplacian image ΔA proceeds to step 12 in the working frame memory W _1, writes the working frame memory W ₂ by adding the contents of the frame memory V in the data read from the working frame memory W _1, the working frame memory W ₂ The read data is written to the frame memory V. In step 12, a time of three frame periods is required because the above-described operations are sequentially performed. Next, at step 13, the frame memory V
Multiplied by 1/8 to read data from a write to a working frame memory W _1, writes the working frame memory W ₂ by adding the data read from the data and the frame memory A read from the working frame memory W _1, working and writes the data of the frame memory W ₂ in the frame memory a.
Also in step 13, each operation described above requires one frame period, so that a total of three frame periods is required. Step 12 performs an operation equivalent to step 2 of the flowchart of FIG. 1, and step 13 performs an operation equivalent to step 3 of FIG.

Next, the routine proceeds to step 14, where the values of the addresses x, y and n of the frame memory A which are equal to or larger than the predetermined threshold value S are written in a peak table as shown in FIG. In step 15, the counter n is incremented by one, and in step 16, n is incremented. Judge whether it is more than the pixel or not.
Steps 12 to 15 are repeated, and in the case of an affirmative determination, this processing routine ends. As described above, in this embodiment, four frame memories are used, and the time required for one approximation is six.
This is the frame period.

By using more working memory, the operation can be executed in a shorter time than in the embodiment of FIG. FIG. 4 shows an embodiment of the present invention using five frame memories.

In FIG. 4, first, in step 21, the frame memory A
substituting to zero the frame memory V _{the even} writes the two-dimensional gray value of the original image to _even. Further, a counter n indicating the number of approximations is set to one. In step 22, the frame Laplacian image .DELTA.A _{the even} memory A _{the even} writes into the frame memory W, written into the frame memory V _odd added and read data from the read data and the frame memory V _{the even} frame memory W. In step 23, the data read from the frame memory V _odd
Writes in the frame memory V _{the even} multiplied by 1/8, written in the frame memory A _odd adding data read from the frame memory V _{the even} the data read from the frame memory A _{the even.} In Steps 22 and 23, one frame period is required to perform each operation. Therefore, to execute Steps 22 and 23, two frame periods are required. Step 22
Performs the same operation as in step 2 of the flowchart of FIG. 1, and step 23 performs the same operation as in first step 3.

Then, the process proceeds to a step 24, wherein the values of the addresses x, y and n at which the frame memory A _odd is equal to or larger than the predetermined threshold value S are written in a peak table as shown in FIG. Step 22-
In step 24, the odd-numbered approximation cycle is being executed.

In step 25, the counter n is incremented by one,
Proceed to steps 26 to 28 to execute even-numbered approximation cycles. Step 26-Step 28 is Step 22-
In step 24, even is changed to odd, and the same process is executed by changing odd to even.

After incrementing the value of the counter n by one in step 29, the process proceeds to step 30 where n is Judge whether it is more than the pixel or not.
The processing from 22 is repeatedly executed, and in the case of an affirmative determination, this processing routine ends. In the present embodiment, since five frame memories are used, the calculation can be performed in a period of four frame periods.

FIG. 5 is an example of the configuration of an image processor suitable for executing the calculation of the present invention, and is particularly suitable for the calculation processing of the embodiment shown in FIG. In the figure, 10 _{0 -} 10 ₄ a frame memory is provided a frame memory 5. Each frame memory is connected to an 8-bit video bus 14 via a bidirectional bus buffer 12. Reference numeral 16 denotes a spatial filter, which includes a line buffer 18, a Laplacian processing unit 20, and a scaling circuit 22. Reference numeral 24 denotes an adder circuit, and the adder circuit 24 and the spatial filter 16 are pipeline-coupled by an internal bus as shown by a broken line. 26
Is a conversion circuit that performs A / D conversion and D / A conversion.

Advantageous Effects of the Invention The present invention is configured as described above in detail, and thus has an effect that recognition of a circular object can be performed accurately and at high speed.

[Brief description of the drawings]

FIG. 1 is a flowchart of the principle of the present invention, FIG. 2 is a schematic diagram of a peak table, FIG. 3 is a flowchart of an embodiment of the present invention, FIG. 4 is a flowchart of another embodiment of the present invention, and FIG. FIG. 1 is a schematic configuration diagram of an image processor applicable to the present invention. 10 ₀ to 10 ₄ … Frame memory, 12… Bidirectional bus buffer, 14… Video bus, 16… Spatial filter, 24… Addition circuit.

Claims (1)

(57) [Claims] 1. An image of a circular object or an object similar thereto is input as a two-dimensional gray value to an arithmetic unit, and a two-dimensional wave equation with the gray value as an initial value is solved. In a circular object recognition method for confirming a center position of a circular object by an extreme value, at least two frame memories, that is, a first frame memory and a second frame memory are provided, and a two-dimensional shading of an image input to the first frame memory is provided. The second frame memory is configured to hold the time differential value of the two-dimensional gray value. In step 1, the two-dimensional gray value of the original image is written to the first frame memory and the second frame memory is stored in the second frame memory. In step 2, the Laplacian of the image held in the first frame memory is obtained, and the Laplacian and the contents of the second frame memory are obtained. Is written in the second frame memory. In step 3, the content of the second frame memory is multiplied by a predetermined number less than 1, and the sum of the value obtained by the multiplication and the content of the first frame memory is stored in the first frame memory. Writing to the frame memory, repeating the above steps 2 and 3 to obtain the center position of the circular object from the x and y addresses of the first frame memory equal to or greater than the predetermined threshold value, and calculate the radius of the circular object from the number of repetitions. A method for recognizing a circular object, comprising: