JPH0388079A - Method for recognizing object - Google Patents

Abstract

PURPOSE:To improve the accuracy of measurement by converting the camera output of each plural feature points of an object to be observed into binary image data by two binarizing reference levels and executing lens focus correction for proper focusing based upon the difference image between two image data. CONSTITUTION:Plural feature points of the object to be observed are specified as an observing area, the output of a video camera 6 in each measuring area is supplied to two binarizing circiuts 10, 11 respectively different reference levels Q, Q-DELTAq and the inputs of the circuits 10, 11 are converted into binary image data. Exclusive OR between the output signals of both the circuits 10, 11 is found out by an arithmetic circuit 12, the number of differential picture elements is counted up by a counter 13 and the focus correcting value m of the camera 6 is found out by the number of differential picture elements.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is an object that images an observed object having a complicated external shape using a video camera and determines the presence or absence of the object, the presence or absence of defects in the object, etc. This is related to the recognition method.

[Prior art] This type of object recognition device that has been used in the past involves experimentally capturing an image of an object as a standard sample using a camera, converting the obtained camera output into a binary value, and then using a binary image as a judgment standard. Obtain a digitization level, binarize the output of the camera obtained by imaging the inspection object, and compare it with the binarization level obtained in advance as a standard for determining non-defective products. Good or defective products are determined based on the number of pixels in an image and their distribution. For objects with complex external shapes, several feature points are set and the above-mentioned determination is performed for each region to determine the overall shape.

However, it is only in extremely special cases that the distinction between good and defective products is clearly expressed as a binary level.
It is usually at an ambiguous level, and it is difficult to uniformly determine the binarization level as a standard for determining non-defective products. Again, I decided this.

Even if a good product is determined to be defective, it may be difficult to judge a good product as a defective product because there are changes in the test environment, such as the brightness of the room depending on the weather on the day of the test or slight differences in the surface condition of the test object. Sometimes I end up doing it.

Especially for objects with complex shapes or depth, one of the multiple feature points may be out of focus with the camera, making it impossible to obtain images with good reproducibility and reducing reliability. descend. Therefore, it is conceivable to use a plurality of cameras to make separate determinations for each feature point, but this operation would be extremely complicated.

Furthermore, for example, in the method disclosed in Japanese Patent Application Laid-Open No. 62-47788, a video signal corresponding to an image with blurred edges is obtained using a camera, and standard data obtained from a reference object and the observed This is done by comparing the measurement data obtained from the object.

[Problems to be Solved] However, even in this conventional technology, standard data must be constantly determined in response to changes in the environment, and this determination is extremely difficult.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide an object recognition method that makes it possible to easily observe a plurality of feature points and to easily obtain a suitable focused image for each feature point, even if the object has a complex shape. It is in.

[Means for Solving the Problems] In order to achieve the above object, the object recognition method of the present invention specifies a plurality of feature points as an observation area for an observed object having a complex shape, and Two reference levels with different binarization reference levels are set for each observation area, images are captured by a video camera for each observation area, and the output of the video camera for each observation area is used as the two standards for the observation area. Convert to binarized image data via a level-based binarization circuit,
A lens focus correction amount for proper focus is calculated from the difference image of the image data, and the focus of the video camera is corrected by the lens focus correction amount to obtain a proper focus image.

[Example] FIG. 1 is a perspective view showing the overall configuration of an apparatus for implementing the object recognition method of the present invention, in which an observed object 1 is positioned and fixed on a carrier 2, While the object is being conveyed from the pre-process 3 to the post-process 4, recognition of the object according to the present invention is performed. The sensor 5 detects that the carrier 2 has reached this recognition position. A video camera 6 equipped with a focusing mechanism 6a is arranged directly above this recognition position so as to be able to image the object 1. The video camera 6 and the sensor 5 are controlled by a control circuit built into the recognition device main body 7. The final judgment after recognition is sent to the subsequent process 4 as an output signal from the terminal 8. The device main body 7 is equipped with a display 7a, which displays a binarized image based on binarized image data obtained by imaging.

The video camera 6 is scanned along line 1-i in FIG.
The object 1 is imaged and the output voltage after photoelectric conversion is output to the control circuit. This object 1 has a complicated shape, and has planes 1a, ib, lc located at different distances from the video camera 6.
4 holes 1e in ld.

If, Ig, lh are opened. Therefore, Figure 2 (
b) As shown in hole 1 e, 1 f + 1
g rlh is the feature point, and the area centered on this hole is the observation area jl, j2. j3. It is specified as j4.

On the line 1-i-, there are observation areas jl, including holes 1e and lf,
j2 exists, and each observation area is observed sequentially according to the following observation method.

FIG. 3 focuses on the holes 1e and lf, and shows the cross section 1-i- as the output voltage for one line.

As mentioned above, the holes 1e and lf are opened on the surfaces 1a and l.
b are at different distances from the video camera 6, so -
The camera cannot be focused at the same time. In particular, when close-up images are taken in detail, the depth of focus of the lens system becomes shallow, and this tendency becomes more pronounced. So now page 1a
When focused on, the hole 1e can be imaged sharply, but the hole 1f becomes blurred. For this reason, when trying to binarize these images and measure the hole diameter, the observation area j
The diameter of the hole 1e can be accurately determined from the two reference levels Q1 and Ql-Δq of t, but in the observation region j2, setting the reference level Q2 itself is difficult, and it is difficult to accurately determine the diameter of the hole 1f. I can't.

For the hole 1f in such a state, the binarization reference level Q2 and 2 when Q2Δq differs by the amount of change Δq.
The valued image is shown in FIG. 4(a).

Shown in (b). FIG. 4(C) shows a difference image obtained by calculating the exclusive OR of these two binarized images. Note that when the focal points match as in the hole 1e, the binarized image hardly changes with respect to the amount of change Δq, and therefore a difference image cannot be obtained. In the case of a blur like the hole 1f, a considerable change occurs in the binarized image with respect to the amount of change Δq, and a difference image is obtained for this change as shown in Fig. 4 (C). .

Letting the number of pixels of this difference image be K, the following relationship exists.

K-fl (Δq) (1) Therefore, when considering the lens system focus correction amount Δm with respect to the destination prime number K, the following relationship exists.

Δm=f2 (K) (2) Therefore, for each observation area, two levels, the reference level Q and Q-Δq, are set, and the lens system focus correction amount Δm is determined by the destination prime number obtained from these. can be found. By instructing this Δm to the focusing mechanism 6a of the video camera 6, an appropriate focal length can be obtained. In this way, an appropriate focus can be set for each observation region of the observed object 1, and an accurate binarized image of feature points can be obtained.

FIG. 5 shows a basic configuration diagram for calculating the lens focus correction amount Δm using such an object recognition method. The output of the video camera 6 for each observation area is supplied to a binarization circuit 10.11 using two different reference levels of binarization, Q and Q-Δq, and is converted into binarized image data. be done. This image data is output as binary signals a and b of "1" or "0". The signals a and b are subjected to an exclusive OR in the logic operation circuit 12, and the destination prime number K is counted by the counter 13.

Prior to observation, the focal length to each observation area is determined in advance with respect to a standard observation object. Further, when correcting the focal length in the direction of increasing or decreasing the focal length, the focal length is set shorter or longer than the appropriate focal length so that a slight blurring occurs. This makes it possible to control the focal length correction in only one direction.

Generally, the functions fl and f2 in equations (1) and (2) are nonlinear, but when considering the minute change Δq, they may be considered linear as in equation (3).

Δm=α·K· (3) Here, α is a proportional coefficient, and formula (3) is used in a table.

Using the destination prime number K counted by the counter 13 as described above, the focus correction amount Δm is obtained from the tabulated equation (3), and this is supplied to the focusing mechanism 6a to increase or decrease the focal length by Δm. and a new value of K is determined. In the normal case, new K<K. Therefore, as a convergence condition for the correction, find Gosho K in the difference Δ,
If Δk is less than or equal to the reference value KO, it is determined that proper focus has been reached, and an image is captured at this focal length to obtain a properly focused image.

After properly focusing and obtaining proper binarized images for all observation areas j1 to j4 as described above, the focus is reset to the first initial focal length, and the next object to be observed is transported. Wait for.

[Effects] By the method of the present invention as explained above, even for objects with complex shapes, it is possible to
In the l region, the focal length can be automatically corrected, a properly focused image for each region can be automatically and efficiently obtained, and overall accurate observation results can be obtained, improving reliability.

[Brief explanation of drawings]

The drawings are explanatory diagrams for explaining the recognition method of the present invention, and FIG. 1 is a perspective view showing the overall configuration of an apparatus for implementing the method of the present invention, and FIGS. 2(a) and (b) are observation areas. Fig. 3 is a diagram of the output voltage waveform obtained by imaging two observation areas with cameras with the same focal length, and Fig. 4 (a) shows the reference level Q for binarization when out of focus. The binarized waveform diagram of FIG. 4(b) is the reference level Q- in the same as above.
Figure 4(c) is the binary output waveform diagram of the exclusive OR of the above two binary outputs. Figure 5 is the basic configuration for calculating the lens focus correction amount. It is a diagram. 1. Observed object, 6#◆ Video camera, 10.11 Binarization circuit, Q, Q-Δq... Reference level, jl - j4 Observation area, Δm... Lens focus correction amount. Applicant Seiko Co., Ltd. Synthesis Attorney Kazuko Matsuda J1~JImaichi...

Claims (1)

[Claims] For an observed object having a complex shape, a plurality of feature points are designated as observation areas, and two different reference levels for binarization are set for each observation area. set, image each observation area with a video camera, and convert the output of the video camera for each observation area into binarized image data via a binarization circuit based on the two reference levels of the observation area. and calculating a lens focus correction amount for proper focus from the difference image of the image data, and correcting the focus of the video camera using the lens focus correction amount to obtain a proper focus image. Method.