JPH08297092A - Method and apparatus for image recognition - Google Patents

Abstract

PURPOSE: To obtain an apparatus whose constitution is compact and by which an object to be recognized is recognized in a short recognition time and with a high accuracy even when the object is large. CONSTITUTION: The apparatus is provided with a one-dimensional line sensor camera 1 as the one-dimensional image input part, a θ-rotation motor 2 as the drive part used to turn the camera, an encoder 3 which generates the signal of its angle of rotation and an image processing part 4 which fetches a one- dimensional image from the one-dimensional line sensor camera 1 in synchronization with the signal from the encoder 3 and which forms a two-dimensional image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a recognition method and apparatus which can be suitably used when an object is recognized in an assembly process or the like.

[0002]

2. Description of the Related Art In recent years, as an image input means in an image recognition method, a device to which a CCD element which responds to visible light, infrared light or the like is applied has become mainstream. Among them, two-dimensional C
The method using a CD is general, but the pixel is 512 ×
When used as a wide range image input means for recognizing a large object as small as about 480 using a wide-angle lens, 1
The image area per pixel becomes large, and the image accuracy is degraded. Therefore, 1000 with more pixels
The utilization of one-dimensional line sensors using CCDs with more than pixels has become common.

A conventional image recognition method will be described below with reference to FIGS. 9 to 11. In FIG. 9, 21 is a one-dimensional line sensor camera, and 22a and 22b are linear motion robots arranged in directions orthogonal to each other. Reference numeral 23 is an encoder, and 24 is an image processing unit.

In FIG. 10, reference numeral 25 denotes a recognition object whose one side is slightly less than twice the image input dimension length of the one-dimensional line sensor camera 21. The sensor camera 21 is moved by the linear motion robots 22a and 22b along the loci indicated by arrows 26, 27, and 28. Reference numeral 29 is an input range of a one-dimensional image captured by the one-dimensional line sensor camera 21.

The operation of the image recognition apparatus shown in FIG. 9 will be described in detail. The one-dimensional line sensor camera 21 is linearly moved by the linear robot 22a. The encoder 23 outputs a signal each time the one-dimensional line sensor camera 21 moves by one pixel of the one-dimensional line sensor, and the image processing unit 24 captures the one-dimensional image from the one-dimensional line sensor camera 21 in synchronization with the signal. The one-dimensional line sensor camera 21 linearly moves on the recognition target object 25 along the loci shown by arrows 26, 27, and 28 in FIG. 10. In the movement of the arrow 26, the image processing unit 24 takes in the first one-dimensional image 31 of the recognition target object 25 in synchronization with the signal from the encoder 23, as shown in FIG. Next, the linear motion robot 22
The one-dimensional line sensor camera 21 detects the object 2
5, the image processing unit 24 captures the second one-dimensional image 32 of the recognition target object 25 in synchronization with the signal of the encoder 23, as shown in FIG. Next, the linear robot 22a causes the one-dimensional line sensor camera 21 to move on the recognition target object 25, and the image processing unit 24 synchronizes with the signal of the encoder 23 and, as shown in FIG.
The third one-dimensional image 33 of 5 is captured. Thereafter, similarly, the image processing unit 24 passes through the locus indicated by the arrow 26.
Captures a one-dimensional image of the recognition object 25 from the one-dimensional line sensor camera 21. Next, in order to capture the remaining image of the recognition object 25, the linear motion robot 22b is used to move the linear motion robot 22a, and the one-dimensional line sensor camera 21 is moved along the locus shown by the arrow 27.
Next, in moving the arrow 28, the image processing unit 24 causes the arrow 2 to move.
In the same manner as when moving at 6, the one-dimensional image of the recognition object 25 is captured in synchronization with the signal from the encoder 23. The image processing unit 2 receives the one-dimensional image captured as described above.
Then, the two-dimensional image 34 of the recognition target object as shown in FIG.

[0006]

However, in the above configuration, since the one-dimensional line sensor camera 21 is moved by using the linear motion robots 22a and 22b, the one-dimensional line sensor camera 21 is driven. There is a problem in that the apparatus space becomes large due to a large space, and particularly the large object to be recognized 25 requires the two-axis linear motion robots 22a and 22b, resulting in a larger size and higher cost. In addition, when the recognition target object 25 is large, there is a problem that a wasteful time that is not related to image capturing, such as moving the one-dimensional line sensor camera 21, is required as shown by an arrow 27, and recognition takes time. Further, since the linear motion robot 22a is operated in the opposite direction as shown by the arrows 26 and 28, an error such as mechanical backlash is included and the image accuracy is deteriorated.

In view of the above-mentioned conventional problems, the present invention provides an image recognition method and an apparatus therefor capable of achieving a compact structure, a short recognition time even when a recognition target is large, and high-accuracy recognition. It is intended to be provided.

[0008]

The image recognition method of the present invention comprises the steps of rotating a one-dimensional image input section to input a one-dimensional image, and capturing the one-dimensional image to form a two-dimensional image. It is characterized by having.

Further, the image recognition apparatus of the present invention, the image input means for rotating the one-dimensional image input section to input the one-dimensional image, and the image for capturing the one-dimensional image from the image input means to form a two-dimensional image. And a processing means.

Preferably, a drive unit for rotating the one-dimensional image input unit, an encoder for generating a signal of a rotation angle of the drive unit, and a one-dimensional image from the one-dimensional image input unit in synchronization with a signal from the encoder. And an image processing unit that forms a captured two-dimensional image. Further, the one-dimensional image input unit is arranged in the radial direction with respect to the rotation axis of the drive unit and connected by the connecting unit, and the connecting unit has a variable length.

[0011]

According to the image recognition method and apparatus of the present invention,
Since the one-dimensional image input unit is rotated to input the one-dimensional image, the driving unit of the one-dimensional image input unit can be made compact and simple, and the size of the drive unit can be twice the length of the one-dimensional image input unit. Even when recognizing an object to be recognized, it can be recognized by one-axis drive, and when a large object is to be recognized, the moving time of the one-dimensional image input unit can be shortened to shorten the recognition time, and the object can rotate in one direction. Since the recognition is performed by being performed, it is hardly affected by a mechanical error such as a backlash and the recognition can be performed with high accuracy.

Further, the rotation angle of the one-dimensional image input section is detected by the encoder, and the one-dimensional image is input from the one-dimensional image input section in synchronism with the signal from the encoder in the image processing section. An image can be easily formed.

Further, when the one-dimensional image input section is radially separated from the rotation axis of the drive section and connected by the connecting means, it is necessary to recognize the shape of the outer peripheral portion of the recognition target when the primary Even in the case of recognizing a large-sized recognition target having a length equal to or more than twice the length of the original image input section, it can be recognized in a short time by the uniaxial drive, and by using the connecting means of variable length, an arbitrary size can be obtained. The recognition target can be recognized.

[0014]

【Example】

(Embodiment 1) Hereinafter, a first embodiment of the present invention will be described with reference to FIG.
~ It demonstrates with reference to FIG.

In FIG. 1, 1 is a one-dimensional line sensor camera as a one-dimensional image input unit, and 2 is a drive unit θ.
The rotary motor 3, 3 is an encoder that detects the rotation angle of the θ rotary motor 2 and outputs a signal, and 4 is an image processing unit. Figure 2
The recognition object 5 is shown in FIG. This recognition object 5 has one side slightly less than twice the length of the one-dimensional line sensor in the one-dimensional line sensor camera 1.

Next, the recognition operation of the recognition object 5 of FIG. 2 by the recognition device of FIG. 1 will be described with reference to FIG. The one-dimensional line sensor camera 1 is arranged so that an arbitrary end of the one-dimensional line sensor is located on the rotation axis of the θ rotation motor 2, and the recognition target is on the axis of the θ rotation motor 2. The recognition target object 5 is positioned so that it is located substantially at the center of the object 5, and the θ rotation motor 2 is rotated. Then, the one-dimensional line sensor camera 1 rotates around the center of the recognition target 5 and the outermost CCD of the one-dimensional line sensor 1 moves to 1
A signal is output from the encoder 3 each time it rotates by an angle moved by the number of pixels.

The image processing section 4 captures a one-dimensional image 6 of the recognition object 5 from the one-dimensional line sensor camera 1 as shown in FIG. 3A in synchronization with the signal from the encoder 3.
Next, when the one-dimensional line sensor camera 1 is rotated by one pixel by the θ rotation motor 2 and the signal from the encoder 3 is input, the one-dimensional line sensor camera 1 is synchronized with the signal as shown in FIG. The next one-dimensional image 7 of the recognition object 5 is captured from the sensor camera 1.

Next, the one-dimensional line sensor camera 1 is further rotated by one pixel by the θ rotation motor 2, and the encoder 3
When the signal from is input, it synchronizes with the signal and
As shown in (c), the next one-dimensional image 8 of the recognition object 5 is captured from the one-dimensional line sensor camera 1. Similarly, the one-dimensional line sensor camera 1 is rotated by the θ rotation motor 2 and the image processing unit 4 is synchronized with the signal from the encoder 3.
Sequentially takes in one-dimensional images of the recognition object 5. In this way, the one-dimensional line sensor camera 1 moves over the recognition object 5
When rotated, all the one-dimensional images of the recognition target object 5 are captured and image-processed by the image processing unit 4, and the image is processed as shown in FIG.
A two-dimensional image 9 of the recognition object 5 is formed as shown in FIG.

As described above, one such as the θ rotation motor 2
By moving the one-dimensional line sensor camera 1 by a driving unit having a compact structure having only axes, a two-dimensional image of the recognition target object 5 having a size twice as long as that of the one-dimensional line sensor as shown in FIG. 2 is formed. Can be recognized. Further, in forming the two-dimensional image 9, since there is no movement of the drive unit that does not take in the one-dimensional image unlike the conventional example, the two-dimensional image forming time can be shortened.

(Second Embodiment) A second embodiment of the present invention will be described below with reference to FIGS.

In FIG. 4, the one-dimensional line sensor camera 1, the θ rotation motor 2, the encoder 3, and the image processing unit 4 are
The components are the same as those in FIG. Reference numeral 11 denotes a variable-length connecting member that connects the output shaft of the θ rotation motor 2 and the one-dimensional line sensor camera 1. As shown in FIG. 5, the recognition target of this embodiment is a disc-shaped tray 12 on which a plurality of (six in the illustrated example) recognition targets 13a to 13f are arranged.

Next, the recognition operation of the recognition objects 13a to 13f of FIG. 5 by the recognition device of FIG. 4 will be described with reference to FIG. First, the tray 12 is arranged such that the center of the tray 12 is located on the rotation axis of the θ rotation motor 2, and
The one-dimensional line sensor camera 1 uses the recognition objects 13a to 13a.
The connecting member 11 is manually adjusted so that it is on top of f.

When the θ rotation motor 2 is rotated, the one-dimensional line sensor camera 1 causes the recognition target objects 13a to 13f.
Pass over. Hereinafter, as shown in the first embodiment, the one-dimensional image 14 is captured from the one-dimensional line sensor camera 1 as shown in FIG. 6A in synchronization with the signal from the encoder 3. Next, when the one-dimensional line sensor camera 1 is rotated by one pixel by the θ rotation motor 2 and the signal from the encoder 3 is input, the one-dimensional line sensor camera 1 is synchronized with the signal as shown in FIG. Next one-dimensional image 15 from the sensor camera 1
Take in. Similarly, the one-dimensional line sensor camera 1 is rotated by the θ rotation motor 2, and the image processing unit 4 sequentially captures one-dimensional images of the recognition objects 13a to 13f in synchronization with the signal from the encoder 3. Then, when the one-dimensional line sensor camera 1 makes one rotation, all the one-dimensional images are captured, image-processed by the image processing unit 4, and the two-dimensional images 16 of the recognition objects 13a to 13f are obtained as shown in FIG. 6C. Is formed.

As described above, according to the present embodiment, it is possible to form a two-dimensional image of the plurality of recognition objects 13a to 13f arranged on the circumference, and the parts arranged on the circumference can be collectively collected. It can be used for recognition.

(Embodiment 3) A third embodiment of the present invention will be described below with reference to FIGS. 4, 7, and 8.

The structure of the recognition device is the same as that of the second embodiment shown in FIG. The recognition object 17 of this embodiment is
As shown in FIG. 7, it is an electronic component called QFP in which leads are provided in four directions, and in order to detect the number, bending, position, etc. of the leads, only the shape near the tips of the leads on the outer periphery of the component is recognized. It is a target recognition target object.

Next, the recognition operation of the recognition object 17 of FIG. 7 by the recognition device of FIG. 4 will be described with reference to FIG. First,
The recognition target object 17 is arranged such that the center of the recognition target object 17 is located on the rotation axis of the θ rotation motor 2, and the one-dimensional line sensor camera 1 passes over the lead on the outer peripheral portion of the recognition target object 17. The connecting member 11 is manually adjusted.

When the θ rotation motor 2 is rotated, the one-dimensional line sensor camera 1 passes over the lead of the recognition object 17. Hereinafter, as shown in the first embodiment, the encoder 3
The one-dimensional image 18 is captured from the one-dimensional line sensor camera 1 as shown in FIG.
Next, when the one-dimensional line sensor camera 1 is rotated by one pixel by the θ rotation motor 2 and the signal from the encoder 3 is input, the one-dimensional line sensor camera 1 is synchronized with the signal as shown in FIG. The next one-dimensional image 19 is captured from the sensor camera 1. Similarly, the one-dimensional line sensor camera 1 is rotated by the θ rotation motor 2, and the one-dimensional image of the recognition object 17 is sequentially captured by the image processing unit 4 in synchronization with the signal from the encoder 3. And the one-dimensional line sensor camera 1
When the image is rotated, the one-dimensional images of all the leads of the recognition object 17 are captured and image-processed by the image processing unit 4.
As shown in (c), the two-dimensional images 20 of all the leads of the recognition object 17 are formed.

The two-dimensional image formed as described above is Q
Since only the lead image necessary for FP recognition is formed and the central image unnecessary for QFP recognition is not formed, the time required for forming a two-dimensional image is reduced.

[0030]

According to the image recognition method of the present invention, as is clear from the above description, since the one-dimensional image is input by rotating the one-dimensional image input unit, the drive unit of the one-dimensional image input unit is The device can be made compact and simple, and the size and cost of the device can be reduced. In addition, even when recognizing an object that is twice the length of the one-dimensional image input unit, it can be driven by a single axis. It is possible to shorten the movement time of the one-dimensional image input unit to shorten the recognition time in the case of a large recognition target because it can be recognized by rotating it in one direction, and it is possible to detect mechanical problems such as backlash. Highly accurate recognition can be performed without being easily affected by errors.

Further, according to the image recognition device of the present invention, the above recognition method can be implemented by the image input means and the image processing means.

Further, the rotation angle of the one-dimensional image input section is detected by the encoder, and the one-dimensional image is taken in from the one-dimensional image input section in synchronization with the signal from the encoder in the image processing section. An image can be easily formed.

Further, when the one-dimensional image input section is radially separated from the rotation axis of the drive section and connected by the connecting means, the shape of the outer peripheral portion of the recognition target needs to be recognized when the primary shape is recognized. Even in the case of recognizing an object to be recognized which is larger than the length of the original image input section, it can be recognized in a short time by uniaxial driving.

Further, by using the variable length connecting means, it is possible to recognize an object to be recognized of any size.

[Brief description of drawings]

FIG. 1 is a perspective view of a recognition device according to a first embodiment of the present invention.

FIG. 2 is a plan view of an object to be recognized according to the first embodiment.

FIG. 3 is an explanatory diagram of a recognition operation according to the first embodiment.

FIG. 4 is a perspective view of a recognition device according to second and third embodiments of the present invention.

FIG. 5 is a plan view of an object to be recognized according to the second embodiment.

FIG. 6 is an explanatory diagram of a recognition operation according to the second embodiment.

FIG. 7 is a plan view of an object to be recognized according to a third embodiment.

FIG. 8 is an explanatory diagram of a recognition operation according to the third embodiment.

FIG. 9 is a perspective view of a recognition device in a conventional example.

FIG. 10 is an explanatory diagram of a movement locus during an image capturing operation of a conventional example.

FIG. 11 is an explanatory diagram of a recognition operation of a conventional example.

[Explanation of symbols]

 1 One-dimensional line sensor camera (one-dimensional image input unit) 2 θ rotation motor (driving unit) 3 Encoder 4 Image processing unit 11 Connecting member

Claims (5)

[Claims] 1. An image recognition method comprising: a step of rotating a one-dimensional image input section to input a one-dimensional image; and a step of capturing a one-dimensional image and forming a two-dimensional image. 2. An image input means for rotating a one-dimensional image input section to input a one-dimensional image, and an image processing means for taking a one-dimensional image from the image input means and forming a two-dimensional image. Characteristic image recognition device. 3. A drive unit that rotates the one-dimensional image input unit, an encoder that generates a signal of a rotation angle of the drive unit, and a one-dimensional image is captured from the one-dimensional image input unit in synchronization with the signal from the encoder. The image recognition apparatus according to claim 1, further comprising an image processing unit that forms a three-dimensional image. 4. The image recognition apparatus according to claim 3, wherein the one-dimensional image input unit is arranged in a radial direction with respect to the rotation axis of the drive unit and is connected by a connecting means. . 5. The image recognition apparatus according to claim 4, wherein the connecting means has a variable length.