JPS61173391A - Method for setting the field of view for image sensor - Google Patents

Abstract

PURPOSE:To perform superposing processing of objects of regular pattern promptly and accurately without special illumination by setting the image sensor during superposing processing of plura objects so that the field of view of image sensor does not include the contour part of the objects. CONSTITUTION:An image sensor output for the contour line of a pocket 1 setting the field of view of the image sensor 4 is not produced when the field of view of the image sensor 4 is et inside the contour line of a pocket 2 superposed upon a shirt 1. In this state, two-dimensional shift amount of stripe patterns are calculated from binary coded image signals of the shirt 1 and the pocket 2 without obstruction by the contour line and the superposing state is adjusted by a robot 6. By this, special illumination blurring the contour line is dispensed with so that registration of the objects of regular pattern can be effected promptly and accurately during sewing.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a method of capturing images, and in particular, a method of setting the field of view of an image sensor to obtain optimal image input when superimposing objects having regular patterns. Regarding.

[Background of the invention]

Conventionally, in apparel manufacturers, patterns have been matched by human judgment in laminated cutting of patterned fabrics and sewing such as attaching pockets.

However, in recent years, attempts have been made to automatically perform such sewing using image processing technology.

Therefore, when using this image processing technology to sew a pocket onto a shirt, for example, it is necessary to place the pocket on the shirt, process the placed state into image processing, and correct the misalignment of the pocket. . This means that the geometrical features of the pocket are captured and the position of the pocket is recognized relative to the shirt, but the contour of the pocket may interfere with this recognition. It often happens.

[Purpose of the invention]

The purpose of the present invention is to shorten the time required for image recognition processing without requiring any special lighting measures in the superposition processing of two or more objects having the same regular pattern. The purpose of this invention is to provide a method for inputting images.

[Summary of the invention]

In order to achieve such an object, the present invention provides two images having the same regular pattern using an image sensor for inputting an image and an image processing device for processing the image captured by the sensor. In the process of overlapping two or more objects, the field of view of the image sensor is set so as not to include the outline of the objects.

[Embodiments of the invention]

Hereinafter, a method for setting the field of view of an image sensor according to the present invention will be explained using examples.

FIGS. 1 to 5 are diagrams showing an embodiment in which a pocket having the same pattern as the regular pattern is superimposed on a dress shirt having a regular pattern. First, FIG. 1 shows a case where a pocket 2 is attached to a dress shirt 1 having a striped pattern (knitted pattern). Images of the shirt 1 and the pocket 2 of the shirt 1 illuminated by the illumination device 3 are captured by the image sensor 4 (ITV camera, etc.), and these images are processed by the image processing device 5, so that both the shirt 1 and the pocket 20 are captured. The amount of inclination (Δθ) and the amount of deviation (ΔX, Δy) of the handle are determined, the amount of inclination and the amount of deviation are transmitted as control signals to the material handling robot 6.6', and the robot 6.6' is controlled. The pocket 2 is rotated and moved based on the signal. After the patterns of both (shirt and pocket) match, x-y-e
The table 7 is moved and the automatic machine 8 performs sewing. Note that when the pocket 2 is operated by the material handling robot 6.61, the shirt 1 is fixed on the X-X-θ table 7 by, for example, temporary adhesion or vacuum suction.

Here, the image sensor 4 recognizes the shirt 1 and the pocket 2 separately based on the patterns attached to them, but when recognizing the pocket in particular, it is performed as described below. . That is, as shown in FIG. 2, when matching the plaid pattern 10, consider the case where the upper object 12, which is a pocket, is superimposed on the lower object 11, which is a dress shirt. Since the set position of the upper object 12 is previously instructed to the material handling robot 6, it will not deviate greatly from the target position, and the pattern will only be misplaced. Therefore, the field of view of the easy sensor 4 is set to be inside the contour line 14 of the upper object 12, as shown by the broken line 13. Note that this setting can be performed by a human.

In addition, suppose that the setting of the field of view of the image sensor 4 is shown by the dashed line 15, the outline 14 of the upper object i2,
Furthermore, if the settings are made to include the outline of the lower object 11, the lighting method must be devised so that the outline does not appear as an image, or the outline may be removed using some image processing method. This requires processing to remove contour lines or to recognize contour lines and distinguish them from patterns.

The necessity of distinguishing contour lines from patterns will be explained below.

FIG. 3 shows an image of the shirt body 1 when the field of view of the image sensor 4, which is the main point of the present invention, is set inside the contour line 14 and an image of the shirt body 1 when the pocket 2 is placed on the body 1. The image is the result of performing an inter-image operation (logical OR operation) on two images after binarization processing. From this image, a histogram process or the like is performed to determine the amount of deviation between the patterns of the two objects (indicated by ΔX, ΔX/, Δy, and Δy' in the figure).

) is easy to find. Here, the smaller of ΔX and ΔX' is considered to be the amount of deviation in the X-axis direction, and the smaller of Δy and Δy/ is considered to be the amount of deviation in the y-axis direction. This is because if a larger amount of displacement is selected, the problem arises that the pocket will be displaced from its original position.

In this way, the outline of the pocket 2 appearing on the image plane will not be easily confused with the dress shirt and pocket pattern, and the pocket can be positioned with respect to the dress shirt only by the pattern.

FIG. 4 shows an image of the shirt 1 and the pocket 2 placed on top of the shirt 1 when the field of view of the image sensor 4 is set to include the outside or part of the outline 14 of the pocket 2 as in the conventional method. The image is the result of performing inter-image operation (logical sum operation) after binarizing the two images when placed (this image also includes the pattern of the shirt 1 shown by the dotted line within the outline 14). ). For simplicity, if we treat an image obtained by binarizing only the image when pocket 2 is placed on shirt 1, the pattern portion indicated by the dotted line will not appear as an image. In either image, as shown in the figure, the contour line 14 becomes an obstacle, and the deviation amounts ΔX, Δ
xl and Δy.

Determining each of Δy/ will be a difficult process.

FIG. 5 is a block diagram showing an embodiment of the above-described image processing. In the figure, there are an image sensor 21 that recognizes the pattern of a dress shirt, and an image sensor 22 that recognizes the pattern of a pocket, and their outputs are input to binarization processing devices 23 and 24, respectively. It is stored in storage devices 25 and 26. The stored data is stored in the storage device 28 via the OR circuit 27. This stored data in the storage device 28 is C
It is output to the RT 29 and also input to the arithmetic circuit 30. This calculation circuit 30 calculates Δx1 and Δy shown in FIG. 3, and drives the arm 31 of the robot based on the calculated values.

〔Effect of the invention〕

According to the present invention, (1) lighting costs are reduced because special lighting measures are not required as described above; (2)
This eliminates the need for processing to remove or ignore the outline of the object, which has the effect of speeding up image recognition processing.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of the method for setting the field of view of an image sensor according to the present invention, and is a block diagram showing an image processing system to which the method is applied, and FIG. 2 is a diagram showing the method for setting the field of view of an image sensor according to the present invention. An explanatory diagram showing one embodiment; FIG. 3 is an explanatory diagram showing a method for correcting the displacement of an object in the image processing system; FIG. 4 is an explanatory diagram showing an example of a method for setting the field of view of a conventional image sensor. FIG. 5 is a circuit diagram showing an embodiment of the processing circuit in the image processing system. 1... Dress shirt, 2... Pocket, 3... Lighting device, 4... Image sensor, 5... Image processing device, 6... Material handling robot, 7... x-y-e Table, 8... Automatic machine (automatic sewing machine), 10... Checkered pattern, 11... Lower object, 12... Upper object, 13... Field of view of image sensor, 14...・The contour line of the upper object. Patent applicant Todoroki Director of the Agency of Industrial Science and Technology

Claims (1)

[Claims] 1. In superposition processing of two or more objects having the same regular pattern, which is performed using an image sensor for inputting images and an image processing device for processing images captured by the sensor, A method for setting a field of view for an image sensor, characterized in that the field of view of the image sensor is set so as not to include an outline of an object.