JPH10153407A - Measuring and processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To greatly improve the precision in measuring the position of an object. SOLUTION: A plurality of model picture images M1 to M3 which indicate different features of respective models are set on picture images obtained by photographing models which are objects to be measured and are registered in a model memory. At the time of measurement and processing, a control part scans each model picture image M1 to M3 on an input picture image sequentially and specifies a scanning position at which the highest correlation value has been obtained as a matching position. Furthermore, the control part calculates a coordinate of a central point O of the object by using the matching result of the model picture image M3 by which the correlation value becomes the maximum value among the matching results of each model picture image M1 to M3 .

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for inputting an image obtained by imaging an object and measuring the position of the object on the input image.

[0002]

2. Description of the Related Art In a device for inspecting the quality of a printed matter such as characters and figures by image processing, the position of a printed matter to be observed is first measured on an image, and then the measured position is measured. Detailed observations are made by setting a window for the inspection.

In order to perform the above-described position measurement, it is necessary to extract image data indicating the characteristics of the model of the object in advance from an image obtained by imaging the model of the object, and register the image data in the memory as a model image. . At the time of the measurement process, by scanning the model image on the input image and obtaining a correlation value with the model image for each scanning position, an image region most similar to the model image on the input image is included in this model image. It is extracted as the corresponding area. Thereafter, based on the extracted position data of the image area and the setting conditions of the model image, the coordinates of a predetermined reference point of the object such as the center point are obtained and output as a measurement result.

FIG. 6 shows an example of registration of the model image.
The illustrated example is an image obtained by positioning the model of the target object at a predetermined position in the observation area and capturing an image.
The operator specifies an image region R indicating the feature of the target object on an image displayed on a monitor or the like.

[0005] A control unit inside the apparatus extracts image data in the designated image area R and registers it as a model image M in a memory. At the same time, the control unit sets a predetermined point of the image area R (upper left vertex MP of the area R in the illustrated example) as a reference point for indicating the set position of the model image, and sets the center point MO of the model as the reference position of the object. And the coordinates of these points MP and MO (x _MP , y _MP ),
(X _MO , y _MO ) is stored in the memory.

FIG. 7 shows a position where the correlation value with the model image becomes highest as a result of scanning the model image M on the image of the observation object (hereinafter, this position is referred to as a “matching position”).
Is shown. The control unit extracts, at the matching position, the coordinates (x _P , y _P ) of the position of the point P overlapping with the reference point MP of the model image M, and extracts the extraction result and the reference point MP, center at the time of setting the model image. The following equations (1) and (2) are executed using the registered data of the point MO, and the coordinates (x _o , y _o ) of the center point O of the object are calculated.

[0007]

(Equation 1)

[0008]

(Equation 2)

When an object having a curvature such as a circle is to be measured, a model image M 'including a part of the contour of the object is usually set as shown in FIG. When matching is performed using the model image M ′, as shown in FIG. 8 (2), when the scanning position of the model image M ′ is shifted in the radial direction a from the original matching position (indicated by a chain line in the figure), Since comparison with image data which is clearly different from the model image is performed, it can be clearly determined from the obtained correlation value that the scanning position is deviated from the matching position. On the other hand, if the scanning position of the model image M ′ is displaced in the circumferential direction b from the original matching position, the image data at this scanning position is not so different from the image data at the correct matching position, so that it is difficult to specify the correct matching position. Become.

As described above, since the matching accuracy in the circumferential direction is lower than the matching accuracy in the radial direction, for this type of object, image data at a plurality of positions along the contour of the model are respectively And the position is measured by integrating the matching results for each model image.

[0011]

The measurement processing shown in FIGS. 6 and 7 is based on the premise that the highest correlation value is obtained when the scanning position of the model image overlaps the corresponding portion of the object. However, if the image data in the corresponding image area fluctuates, the correlation value between this image area and the model image becomes low, making it difficult to extract as a matching position. The problem of erroneous recognition arises.

FIGS. 9 (1) to 9 (3) show an example of a change in image data at a position corresponding to a model image.
(1) shows an example in which deformation such as chipping has occurred in the characteristic portion 20 of the target object, and FIG. 9 (2) shows that the characteristic portion 20 is an image 2 of another object.
9 (3) shows an example in which an image 22 such as dust is included in the background portion. In any case, since the image area different from the model image is included in the image area corresponding to the model image, the correlation value is reduced, and the position of the characteristic portion 20 may not be able to be extracted as the matching position.

[0013] The above problem is the same when the measurement is performed using a plurality of model images. If the image data in the image area corresponding to any one of the model images fluctuates, the final measurement is performed. The result is also affected by the fluctuation, and a problem arises that the measurement result does not accurately reflect the actual position of the object.

The present invention has been made in view of the above-mentioned problems. For a model of a measurement object, a plurality of image data including different features are registered in advance, and an input image and a plurality of registered models are registered. By measuring the position of the object using only the correct result of the collation results,
A technical task is to significantly improve the position measurement accuracy.

[0015]

According to a first aspect of the present invention, there is provided a measurement processing apparatus, comprising: a storage unit for storing, as a model image of an object, a plurality of image data each including a different characteristic for a model of the object to be measured; Extraction means for extracting an image area corresponding to each model image on the image and the similarity between the area and the model image, and among the extraction results by the extraction means, a similarity of a predetermined value or more was obtained. And position measuring means for measuring the position of the object on the input image using the position information of the image area.

According to a second aspect of the present invention, there is provided a measurement processing apparatus including the same storage means and extraction means as described above, and using the position information of the image area having the maximum similarity among the extraction results by the extraction means. There is provided a position measuring means for measuring the position of the object on the input image.

[0017]

A plurality of model images each having different characteristics are registered for a model of a measurement object, and these model images are scanned on an input image, and an image region corresponding to each model image and the region and the model image Is extracted. Thereafter, in the invention of claim 1, the position information of the image region having the similarity equal to or more than the predetermined value is used, and in the invention of claim 2, the position information of the image region having the maximum similarity is used. Thus, since the position of the target on the input image is detected, the position of the target can be accurately measured by using the position information of the image area corresponding to any one of the model images.

[0018]

FIG. 1 shows the configuration of a measurement processing apparatus according to an embodiment of the present invention. This measurement processing device is for processing an image obtained by imaging a predetermined observation region and recognizing where in the observation region the object is located,
TV camera 1 (hereinafter simply referred to as “camera 1”), A /
It comprises a D conversion unit 2, an image memory 3, a D / A conversion unit 4, a monitor 5, an input unit 6, a control unit 7, a model memory 8, a matching processing unit 9, a position measurement unit 10, and the like.

The camera 1 is positioned and fixed so that its imaging direction is directed to the observation area. The analog image data from the camera 1 is converted into a digital amount by the A / D converter 2. , Are stored in the image memory 3. The digital amount of image data is supplied to the control unit 7.
And read by the matching processing unit 9 and D
The signal is decoded into an analog amount by the / A converter 4 and output to the monitor 5.

This measurement processing device recognizes the position of an object on an image by sequentially capturing a plurality of objects by the camera 1 and comparing the input image with a predetermined model image. In order to prevent measurement values from being affected by missing feature parts on the input image or changes in the background part image, a plurality of model images containing different features are set for the model. The coordinates of the center point are calculated using the result of obtaining the highest correlation value among the matching results for the images.

Prior to the measurement processing, the set model image data, the position of the reference point at the time of the setting, and the coordinates of the center point of the model indicating the reference position of the object are registered in the model memory 8. You. The matching processing unit 9 sequentially scans the model images registered in the model memory 8 on the input image stored in the image memory 3 and executes a matching process using a known normalized cross-correlation operation. ,
An image region having a maximum correlation value with each model image is extracted as a region including a feature corresponding to the model image. The position measurement unit 10 calculates the center point of the object on the input image using the extraction result and the position data registered in the model memory 8, and outputs the calculation result as the position measurement result of the object. .

FIG. 2 shows an example of setting a model image. The illustrated example shows a display screen of an image obtained by capturing an image of a model of a target object. An image region R _{1 to} R _{3 in} which an operator includes a plurality of characteristic portions of the model on this display screen.
Is set, the control section 7 sets the set image area R
The image data in the ₁ to R ₃ and extracted separately, stored in the model memory 8 as a model image M ₁ ~M ₃ respectively.

At this time, the control unit 7 sets each model image M
₁ image region R ₁ ~ as a reference point indicating the set position of ~M ₃
And identifies each upper left corner MP ₁ to MP ₃ of R _3, these points MP ₁ to MP ₃ coordinates _{(x MP1, y MP1),} (x
_MP2, y _MP2), and registers the (x _MP3, y _MP3) in correspondence to the model image M ₁ ~M ₃ respectively in the model memory 8. Further, the control unit 7 determines the coordinates (x _MO , y _MO ) of the center point MO of the model based on input data relating to the positioning of the model, and similarly registers the coordinates in the model memory 8.

[0024] FIG. 3 shows an example of an input image during the observation process, the model of the image 11 of the observed object M ₁
Characteristic portion 11a corresponding is included other well that is no longer able to identify hidden image 12 of the object, an image 13 of background dust characterizing part near corresponding to the model M ₂ in.

FIG. 4 shows the result of matching processing of the input image of FIG. 3 with each of the model images M _{1 to} M ₃ . Note that, here, for simplicity of description, each model image M _{1 to} M
₃ as well as deployment superimposed on each matching point in the vicinity of the matching position, is represented a specific calculation value of the calculated correlation value r _l1 ~r _l3.

[0026] figure P ₁ to P ₃ are at each matching position of each model image M ₁ ~M _3, be a point overlapping the reference point MP ₁ to MP ₃ of the model image, these points P ₁ ~ P ₃ coordinates _{(x P1, y P1),} (x P2, y P2),
(X _P3 , y _P3 ) and each of the correlation values r _{L1 to} r _L3 are associated as a matching result and output to the position measurement unit 10.

In the case of the illustrated example, the first and second model images M
_1, because it contains different image data to the model image in an image region corresponding to correctly M _2, matching the position of these model images are misaligned from the original position corresponding, yet the resulting correlation The values r _L1 and r _L2 are extremely low.

On the other hand, in the image area corresponding to the third model image M ₃ , the image data similar to the model image is held in both the characteristic part and the background part, so that matching is obtained at the position of this image area. And the correlation value r
_L3 is a value close to 100%. Therefore, in this case, the registered coordinates (x _MP3 , y _MP3 ) of the reference point MP ₃ of the third model image M3 and the coordinates (x _P3 , y _P3 ) of the corresponding point P ₃ and the center point MO of the model The registered coordinate data (x _MO , y _MO ) of (1) and (2)
By applying the equation, it is possible to obtain coordinate data that most accurately reflects the position of the center point O of the object.

As described above, a plurality of model images each including a different feature are registered in the model memory 8, and among the matching results for each model image, the matching result with the largest correlation value is adopted. Thus, a highly accurate measurement value can be obtained using the position data of the image area in which the image data closest to the model image is held. The number of matching results to be employed is not limited to one. For example, the above-described equations (1) and (2) are executed for matching results having correlation values equal to or greater than a predetermined threshold value, and the average value of each calculation result is obtained. May be output as the final measurement result. In particular, when measuring the position of an object having a curvature, a plurality of model images are set in the radial direction, a matching process is executed, and a measurement process is performed using only a highly reliable one of the matching results. May be implemented.

FIG. 5 shows an example of setting a model image when a circular object is measured, and the center point mo of the model is shown.
, Two image regions R ′ ₁ , R ′ ₃ in the x-axis direction
Two image regions R ′ ₂ and R ′ ₄ are also set in the y-axis direction. Any of these image regions R ′ _{1 to} R ′
Reference numeral ₄ also includes the contour of the model and is set at a position separated by an equal distance from the center point mo. The image data in these regions R ′ _{1 to} R ′ ₄ are model images m _{1 to} m _4, respectively. Is registered in the model memory 8.

When the matching process is sequentially performed on each of the model images m _{1 to} m ₄ , if the image data in the image area corresponding to any one of the model images fluctuates, the matching position of the model image becomes As in the embodiment of FIG. 4, the position is shifted from the original corresponding position, and the obtained correlation value is extremely low.

Therefore, the position measurement unit 10 compares the correlation value obtained by each matching process with a predetermined threshold value, and uses only the matching result having a correlation value equal to or larger than the threshold value to determine the position of the target object. It is configured to determine a center point. As described with reference to FIG. 8, in a target having such a shape, the matching position in the radial direction is more accurately detected than the matching position in the contour direction of the target. The x coordinate is set to be calculated by using the matching result of the model m ₁ and the model m ₃ , and the y coordinate is set to be calculated by using the matching result of the model m ₂ and the model m ₄ .

Now, each image region R ′ _i (i
= 1, 2, 3, 4), the coordinates of the reference point mp _i are (x _mpi ,
y _mpi ) and the coordinates of the center point mo of the model are (x _mo , y _mo )
And the reference point m at each matching position during the measurement process
The coordinates of the corresponding points p _i to _{p i (x pi, y pi} ) When,
The x coordinate x _o of the center point o on the input image is given by one of the equations (3) to (5), and the y coordinate y _o is given by the following (6) to
Each is calculated by one of the equations (8). In the preceding conditions of each equation, r _{1 to} r ₄ indicate the correlation values obtained at the matching positions to the model images m _{1 to} m ₄ , respectively, and th indicates the threshold value.

[0034]

(Equation 3)

[0035]

(Equation 4)

[0036]

(Equation 5)

[0037]

(Equation 6)

[0038]

(Equation 7)

[0039]

(Equation 8)

By adopting only the highly reliable matching result, the measurement accuracy is greatly improved.
Thereafter, even at the time of detailed observation of the object, it is possible to accurately set an observation window and perform a highly accurate observation process. Moreover, since there is no need to significantly change the configuration of the conventional apparatus, the cost for introducing the apparatus is small.

[0041]

As described above, according to the present invention, a plurality of image data including different features are stored as model images of a target object model, and the model images are associated with each other at the time of measurement processing. Among the extracted image regions, the similarity with the model image is equal to or more than a predetermined value, or using the position information of the image region having the maximum value,
The position of the object on the input image is measured. As a result, the position of the target object can be accurately measured even if a part of the input image data fluctuates, and the measurement accuracy can be greatly improved without making a significant change in the device configuration.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a measurement processing device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a setting example of a model image.

FIG. 3 is an explanatory diagram showing a specific example of an input image.

FIG. 4 is an explanatory diagram showing a result of a matching process on the input image of FIG. 3;

FIG. 5 is an explanatory diagram showing another example of setting a model image.

FIG. 6 is an explanatory diagram showing a setting example of a conventional model image.

FIG. 7 is an explanatory diagram showing a matching process using the model image of FIG. 6;

FIG. 8 is an explanatory diagram showing a matching process for an object having a curvature.

FIG. 9 is an explanatory diagram illustrating an example of a change in image data at a position corresponding to a model image.

[Explanation of symbols]

7 control unit 8 model memory 9 matching unit 10 position measuring unit _{M 1 ~M 3, m 1 ~m} 4 model image

Claims (2)

[Claims] 1. An apparatus for inputting an image obtained by capturing an image of an object and measuring the position of the object on the input image, wherein a plurality of models including different characteristics are provided for the model of the object to be measured. Storage means for storing image data as a model image of the object; extraction means for extracting an image area corresponding to each model image on the input image and a similarity between the area and the model image; and the extraction means And a position measuring means for measuring the position of the object on the input image using the position information of the image area in which a similarity equal to or more than a predetermined value is obtained from the extraction results of the above. 2. An apparatus for inputting an image obtained by capturing an image of an object and measuring a position of the object on the input image, wherein a plurality of models each including different features are provided for the model of the object to be measured. A storage unit that stores image data as a model image of the object; an extraction unit that extracts an image region corresponding to each model image on the input image and a similarity between the region and the model image; A measurement processing apparatus comprising: a position measurement unit that measures a position of an object on the input image using position information of an image region having a maximum similarity among the extraction results.