JP3252574B2 - Image processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for determining the quality of an object by processing an image obtained by imaging the object to be inspected, for example, in an object inspection system. In particular, the present invention relates to an image processing apparatus capable of performing appropriate discrimination processing even if the imaging position of an object fluctuates back and forth.

[0002]

2. Description of the Related Art FIG.
As shown in FIG. 0, the objects 1a and 1b to be inspected are transported in the direction of the arrow by the belt conveyor 2, and the objects 1a and 1b are sequentially imaged by the television camera 3 installed in the middle of the transport.
The quality of each of the objects 1a and 1b is determined by taking each image into the image processing device 4 and performing image processing. Characters 5 are printed on the sides of each of the objects 1a and 1b, and the image processing device 4 measures the area of the image portion of the characters, thereby inspecting the characters 4 for blurring, blurring, chipping, and the like. I have.

FIG. 11 shows another object inspection system. When a plurality of stacked objects 6 are sequentially taken out from the top and moved to the next process, the top-most object 6 is taken from directly above the television camera 3. And similarly inspects the blurring of the character 7.

[0004]

In the inspection system shown in FIG. 10, when each of the objects 1a and 1b is located at a fixed position in the width direction on the belt conveyor 2, the television camera 3 and each of the objects 1a and 1b are connected to each other. The distance between them becomes the same. The object 1a going ahead is located at a fixed position, and the object 1b following this
Is located behind the home position, the subsequent object 1b appears smaller on the television camera 3 than the preceding object 1a. As a result, the area of the image portion of the character becomes smaller than the original value, and the image processing device 4 determines that the target object 1b is defective although it is good.

In the inspection system shown in FIG. 11, each time a plurality of stacked objects 6 are sequentially taken out from above, the television camera 3
In this case, the distance from the object to the uppermost object 6a always fluctuates. In this case, the area of the image portion of the character fluctuates in accordance with the fluctuation of the distance, so that this type of inspection cannot be performed properly.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide an image processing apparatus capable of correctly determining the quality of an object even if the imaging position of the object fluctuates back and forth. I do.

[0007]

An image processing apparatus according to the present invention comprises: a changing unit for extracting a partial image in an image obtained by capturing an image of a model of an object at a reference position and changing the image size; A coincidence measurement unit that measures the degree of coincidence between the input image obtained by imaging the object, the partial image, and each image obtained by the change unit, and each of the coincidences measured by the coincidence measurement unit. Calculating means for calculating the positional relationship between the imaging position of the model and the imaging position of the object from the degree of coincidence of the images, wherein the degree of coincidence measuring means obtains an input image with the partial image and the changing means. After scanning each of the obtained images and measuring a normalized correlation value for each pixel position, the maximum value of the normalized correlation values is obtained as the degree of coincidence.

According to a second aspect of the present invention, the arithmetic means calculates a visual field ratio between a camera field of view of the model and a camera field of view of the object as a positional relationship between the image capturing position of the model and the image capturing position of the object. I have.

In the image processing apparatus according to a third aspect of the present invention, the calculating means calculates a visual field ratio between the camera visual field of the model and the visual field of the object as a positional relationship between the image capturing position of the model and the image capturing position of the object. Then, the distance to the imaging position of the object is calculated using the visual field ratio.

In the image processing apparatus according to a fourth aspect of the present invention, the calculating means calculates a visual field ratio between a camera visual field of the model and a camera visual field of the object as a positional relationship between the image capturing position of the model and the image capturing position of the object. Then, the distance to the imaging position of the object is calculated using the visual field ratio, and the calculated value is fed back to the focal length adjustment mechanism of the imaging device.

[0011]

The degree of coincidence between the input image of the object, the partial image of the model image, and each image whose image size has been changed is measured by measuring the normalized correlation value, and the model is determined from the degree of coincidence for each image. Since the positional relationship between the imaging position of the object and the imaging position of the object is calculated, for example, as a visual field ratio, even if the imaging position of the object fluctuates back and forth, it is possible to correctly determine the quality of the object based on the positional relationship.

In the third aspect, after calculating the visual field ratio,
Since the distance to the imaging position of the object is calculated using the visual field ratio, this image processing device can be applied to the field of distance measurement.

According to a fourth aspect of the present invention, the distance to the image pickup position of the object is calculated using the visual field ratio, and the calculated value is fed back to the focal length adjusting mechanism of the image pickup apparatus. Can be given.

[0014]

FIG. 1 shows an example of the configuration of an image processing apparatus 9 according to an embodiment of the present invention. TV camera 8 shown in FIG.
Captures an object 17 to be inspected and a model 16 of the object to generate a grayscale image. The object model 16 is imaged after being positioned at a reference position, for example, an intermediate position of a range S in which the position of the object 17 with respect to the television camera 8 varies.

The image processing device 9 includes a frame memory 1
0, a plurality of pairs of coincidence calculation units 11 and a model memory 1
2, a CPU 13, a ROM 14, and a RAM 15 as components, and the frame memory 10 stores a gray-scale image for one frame given from the television camera 8. The grayscale image of the model 16 stored in the frame memory 10 is taken into the CPU 13, and the CPU 13 extracts a partial image from the grayscale image, enlarges and reduces the partial image in a similar form, and performs a plurality of enlargement. Generate an image and a reduced image.

In this embodiment, a character M is obtained from an image obtained by imaging the model of the object 1a or 1b shown in FIG.
Is extracted as a partial image, and FIG.
From the partial image 20 of the character M, two enlarged images 21 and
The process of generating two and two reduced images 23 and 24 is shown. In this partial image 20, the vertical size of the image portion of the character (the area shown by hatching in the figure) is L, and this partial image 2
In the first enlarged image 21 obtained by enlarging 0, the vertical size of the character image portion is L + ΔL. The second enlarged image 22 is obtained by enlarging the first enlarged image 21, and the vertical size of the image portion of the character is L + 2ΔL. The first reduced image 23 obtained by reducing the partial image 20 has a vertical size of a character image portion of L-ΔL, and the second reduced image 23 obtained by reducing the first reduced image 23 is obtained. In the reduced image 24, the vertical size of the image portion of the character is L-2ΔL.

The above-described partial image 20, first and second enlarged images 21, 22 and first and second reduced images 23, 24
(Hereinafter, these are collectively referred to as “model images”).
13 is stored in each model memory 12 individually. Each of the coincidence calculation units 11 measures the degree of coincidence between a gray image (hereinafter referred to as an “input image”) of the inspection object 17 stored in the frame memory 10 and a model image stored in each model memory 12. .

FIG. 3 shows a specific example of the coincidence measuring method. In the figure, reference numeral 25 denotes an input image, and 26 denotes a model image. After the input image 25 is raster-scanned by the model image 26 and a normalized correlation value is measured for each pixel position,
The maximum value of the normalized correlation value is obtained as the degree of coincidence. Now, if the input image 25 is represented by f (i, j) and the model image 26 is represented by g (i, j), the normalized correlation value γ is generally expressed by the following equation, where N is the number of pixels in the area. Given.

[0019]

(Equation 1)

Assuming that the scanning position of the model image 26 is (I, J), a normalized correlation value γ _IJ is obtained by the following equation while changing the scanning position on the coordinates of the input image 26, and the maximum value is obtained. It will be obtained as the degree of coincidence.

[0021]

(Equation 2)

The coincidence calculated by each of the coincidence calculators 11 is taken into the CPU 13, and the CPU 13 uses the coincidence as a value representing the positional relationship between the imaging position of the object model 16 and the imaging position of the object 17. Then, the visual field ratio between the camera visual field of the model 16 and the camera visual field of the object 17 is calculated.

FIG. 4A shows the camera field of view (N ₁ × N ₁ ) for the model 16, and FIG. 4B shows the camera field of view (N ₂ × N ₂ ) for the object 17. In this case, the visual field ratio is N ₂ / N ₁ . In the illustrated example, the distance between the object 17 and the television camera 8 is the model 1
Since the distance between the camera 6 and the television camera 8 is larger, the field of view of the camera is widened, and the characters displayed on the object 17 are observed to be small.

FIG. 5 shows a method of calculating the visual field ratio. The figure plots the relationship between the vertical size of the image portion of the character in each model image and the degree of coincidence at points P _{1 to} P _{5. When} each point is approximated by Lagrangian polynomial and interpolated, Is obtained. In the case of the curve 30, the vertical size of the image portion of the character is L-2ΔL and L-ΔL.
When L-1.5ΔL is the middle point, the coincidence is maximized. Therefore, the vertical size L-1.5ΔL of the character image portion in the input image is set to the vertical size of the character image portion in the partial image 20 described above. The field ratio can be obtained by dividing by the size L.

FIG. 6 shows a processing procedure by the image processing apparatus 9 described above, and corresponds to step 1 in FIG.
1)), the grayscale image of one frame obtained by imaging the model 16 at the reference position is taken into the frame memory 10, and in the next step 2, the CPU 13 extracts the partial image 20 in the grayscale image and The partial image 20 is enlarged to generate enlarged images 21 and 22, and the reduced image is generated to generate reduced images 23 and 24. These partial images 20, enlarged images 21 and 22, and reduced images 23 and 24 are generated. It is individually stored in each model memory 12 as a model image.

In the next step 3, the obtained one-frame gray-scale image of the object 17, that is, the input image is fetched into the frame memory 10, and the next step 4
In each coincidence calculation unit 11, the frame memory 1
The degree of coincidence between the input image stored in 0 and the model image stored in each model memory 12 is measured. In the next step 4, the CPU 13 takes in the calculation results of the coincidence from the respective coincidence calculators 11, and in step 5, calculates the visual field ratio between the camera visual field of the model 16 and the camera visual field of the object 17 therefrom.

In the image processing apparatus 9 of the above embodiment, the measurement of the degree of coincidence between the input image and each model image is performed in parallel. However, the present invention is not limited to this. May be performed in series. FIG. 7 shows a configuration of an image processing apparatus 9 for performing such a serial processing method, and is composed of a frame memory 10, a CPU 13, a ROM 14, and a RAM 15. According to this embodiment, the processing time is longer than that of the above embodiment, but the circuit configuration is simplified.

FIG. 8 shows a processing procedure performed by the image processing apparatus 9 shown in FIG. 7, and a gray scale image for one frame obtained by imaging the model 16 at the reference position in step 1 of FIG. In the next step 2
U 13 extracts the partial image 20 in the gray image and transfers the image data to the RAM 15.

In the next step 3, one frame of the obtained gray-scale image of the object 17, ie, the input image, is loaded into the frame memory 10. Next, in steps 4 and 5, the CPU 13 enlarges or reduces the partial image in the RAM 15 to generate an enlarged image or a reduced image, and compares the enlarged image or the reduced image with the input image stored in the frame memory 10. Is measured. When the same procedure is repeatedly executed for the number of model images, the next step 6 becomes “YES” and step 7
Then, the CPU 13 calculates the visual field ratio between the camera visual field of the model 16 and the camera visual field of the object 17 from the calculation results of all the degrees of coincidence as described with reference to FIG.

Next, as shown in FIG. 9, assuming that the length of one side of the camera view of the object model 16 is S1 and the length of one side of the camera view of the object 17 is S2, the television camera 8 D1 between the camera and the model 16 and distance D2 between the television camera 8 and the object 17 are given by the following equations. In the equations (3) and (4), α and β are constants.

[0031]

(Equation 3)

[0032]

(Equation 4)

In equation (4), since the camera field of view S1 of the model 16 is known, the camera field of view S2 of the object 17 is known.
Is obtained by using the visual field ratio calculated by the image processing device 9, the distance D2 to the object 17 can be easily calculated by the equation (4), and can be applied to the field of distance measurement. If the distance D2 to the object 17 is fed back to the television camera 8 to control the focal length adjusting mechanism,
An automatic focus adjustment function can be provided.

[0034]

As described above, the present invention measures the degree of coincidence between an input image of an object, a partial image of a model image, and each image whose image size has been changed, and determines the model based on the degree of coincidence for each image. Since the positional relationship between the imaging position of the object and the imaging position of the object is calculated, even if the imaging position of the object fluctuates back and forth, the quality of the object can be correctly determined based on the positional relationship.

According to a fourth aspect of the present invention, the visual field ratio between the camera field of view of the model and the camera visual field of the object is calculated as the positional relationship between the image capturing position of the model and the image capturing position of the object. Since the distance to the imaging position of the object is calculated, this image processing device can be applied to the field of distance measurement. Further, in claim 5, the distance to the imaging position of the target object is calculated using the visual field ratio, and the calculated value is fed back to the focal length adjustment mechanism of the imaging device. There is an effect that can be.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a process of generating an enlarged image and a reduced image from a partial image.

FIG. 3 is an explanatory diagram showing a specific example of a matching degree measuring method.

FIG. 4 is an explanatory diagram showing a camera visual field of a model and a camera visual field of an object.

FIG. 5 is an explanatory diagram showing a method of calculating a visual field ratio.

FIG. 6 is a flowchart illustrating a processing procedure of the image processing apparatus.

FIG. 7 is a block diagram illustrating another embodiment of the image processing apparatus.

FIG. 8 is a flowchart showing a processing procedure of the embodiment of FIG. 7;

FIG. 9 is an explanatory diagram showing a distance measuring method.

FIG. 10 is a perspective view showing an outline of the object inspection system.

FIG. 11 is a perspective view showing an outline of another object inspection system.

[Explanation of symbols]

 Reference Signs List 8 TV camera 9 Image processing device 11 Matching degree calculation unit 12 Model memory 13 CPU

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06T ⁷ /00-7/60 G01B 11/00-11/30 G06T 1/00

Claims (4)

(57) [Claims] 1. A changing means for extracting a partial image in an image obtained by imaging a model of an object at a reference position and changing the image size, and an input image obtained by imaging the object. Measuring means for measuring a degree of coincidence between each of the images obtained by the partial image and the changing means, and an image pickup position of a model and an object based on the degree of coincidence of each image measured by the degree of coincidence measuring means. Calculating means for calculating a positional relationship with an imaging position of an object, wherein the coincidence measuring means scans the input image with each of the images obtained by the partial image and the changing means, and scans the input image for each pixel position. An image processing apparatus which measures a normalized correlation value, and then calculates a maximum value of the normalized correlation value as the degree of coincidence. 2. The apparatus according to claim 1, wherein the calculating means calculates a visual field ratio between a camera visual field of the model and a camera visual field of the object as a positional relationship between the image capturing position of the model and the image capturing position of the object. Image processing device. 3. The calculating means calculates a visual field ratio between a camera visual field of the model and a visual field of the object as a positional relationship between the image capturing position of the model and the image capturing position of the object, and then uses the visual field ratio. The image processing apparatus according to claim 1, wherein a distance to an imaging position of the object is calculated. 4. The calculating means calculates a visual field ratio between a camera visual field of the model and a camera visual field of the object as a positional relationship between the image capturing position of the model and the image capturing position of the object, and then uses the visual field ratio. Calculate the distance to the imaging position of the object,
2. The image processing apparatus according to claim 1, wherein the calculated value is fed back to a focal length adjusting mechanism of the imaging apparatus.