JPS6314281A - Picture fetching method - Google Patents

Abstract

PURPOSE:To process the recognition of a recognition object at a high speed by setting visual fields of image sensors to overlap them so that the whole of the recognition object always exists in one image out of plural images taken in by image sensors. CONSTITUTION:The image of a label 3 which is stuck to the upper face of an object 2 on a convector line 1 and includes a character string is taken in by cameras 4 and 4' with quick shutters, and the number of the label is recog nized by a picture processing device 6. The signal based on the recognition result is sent to a controller 7, and the controller 7 controls destination guide bars 8 and 8' on a basis of this signal. An extent l of overlap between a visual field 20 of the camera 4 and a visual field 21 of the camera 4' is made longer than the diagonal line of the label at the time of taking in labels 30-34 as images. Thus, recognition of the recognition object is processed at a high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for capturing one image, and particularly to a method for setting the field of view of an image sensor suitable for capturing an image that facilitates recognition.

[Conventional technology]

Conventionally, the case where image capture is performed using a plurality of image sensors is discussed in Automation, Vol. 27, No. 2, p. 79.

[Problem that the invention seeks to solve]

The above-mentioned conventional technology does not take into account the setting positions of the fields of view of the plurality of cameras (image sensors).

For example, a recognition target such as a label or written character string affixed to an arbitrary position on the top surface of an object moving on a conveyor belt may be captured across the fields of view of multiple cameras (image sensors). There were problems such as a complicated recognition method and a long processing time.

An object of the present invention is to enable recognition of a recognition target to be processed easily and at high speed when it is necessary to continuously capture a plurality of images using a plurality of image sensors or one image sensor.

[Means for solving problems]

The above purpose is to combine the fields of view of multiple cameras (image sensors) or, when one image sensor is used to continuously capture images, the fields of view of the previous and current captured images of the same image sensor, respectively. This is achieved by setting the fields of view of the image sensors to overlap so that the entire recognition target is present in at least one of the images.

[Effect]

By setting the fields of view of multiple image sensors or the fields of view of multiple images of one image sensor so that the entire recognition target exists in at least one image, the fields of view of the image sensors are set to overlap. An image that includes the entire recognition target is extracted from the images captured in , and the recognition target is recognized from that image. This eliminates the need to combine and recognize recognition targets taken across multiple images, or to combine separate recognition results, which simplifies processing and reduces recognition errors.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG. 1st
The figure shows the overall configuration of the system for sorting objects. A label 3 containing a character string pasted on the top surface of an object 2 on a conveyor line 1 is captured by cameras 4 and 4' with high-speed shutters, the label number is recognized by an image processing device 6, and based on the recognition result. A signal is sent to the control device 7, and based on the signal, the control device 7 controls the destination guide bars 8, 8'.

Object 2.2' by destination guide bars 8, 8'.

2' are collected by destination via guide plates 9, 9'. Note that 5 is a lighting device. It is also assumed that the conveyor line 1 is moving in the direction of the arrow.

Hereinafter, a method for setting the field of view of an image sensor, which is the central part of the present invention, will be explained using FIGS. 2 to 5. 2
0 is the field of view of camera 4, and 21 is the field of view of camera 4'.

Assume that the object is moving from the direction of the arrow. 30～
34 is a label containing a character string. The amount of overlap Q between the fields of view 20 and 21 is set to be larger than the length of the diagonal line of the labels when the labels 30 to 34 are captured as images. If possible, it is desirable to set the length to be about 1 to 2 pixels longer than the length of the diagonal line. With this setting, the label will be in field of view 2.
It will fall within the field of view of either 0 or 21. For example, labels 30, 32, 33 are in the field of view 20;
31.34 shows the case where the entire image of each label enters the field of view 21 and is captured as an image, but when capturing images from multiple cameras (image sensors), the converter shown in Figure 3 is used. There is a method in which only one camera 36 is used, and a camera switch 37 is used to switch the camera and capture images into the image memory 39. FIG. 4 shows an example of setting the camera field of view in the case of three cameras or one camera. In addition, Fig. 4 (a) to (e)
The arrow shown in indicates the direction of movement of the object. Fig. 4(,) shows that when each camera has an A/D converter as shown in Fig. 3(a), and Fig. 4(b) shows an A/D converter as shown in Fig. 3(b). This shows the case where one D converter has a camera switch.

In 22, Q indicates the overlapping view of the camera field of view, and d indicates the amount of shift in the field of view due to a delay in the capture timing by the camera switch 37. In addition, ■ and ■ in the figure.

4(a) shows the order in which the cameras are activated.

(b) shows the case where multiple cameras are arranged perpendicular to the moving direction of the object due to the relationship between the recognition target and the resolution of the camera, but (c) of the same figure shows that the resolution of the camera is sufficient. However, since the moving speed of the conveyor line is high, this example shows a case in which the fields of view of multiple cameras are set to overlap in the direction in which the object is moving. The amount of overlap Q is as described above. Further, FIG. 4(d) shows the field of view 50 of the previous image capture when images are continuously captured and processed with one camera. The field of view 51 of the current image capture is shown. The amount of overlap m between both visual fields is
This is determined by the relationship between the speed of the conveyor line 1 and the recognition time, but it must be faster than the diagonal length of the label. FIG. 4(e) shows a combination of FIG. 4(a) and (d).

FIG. 4(f) shows a case where FIGS. 4(a) and (c) are combined.

As described above, there is at least one image that captures the entire label including the character string to be recognized.

Extract an image containing the entire label from multiple images,
Extract the label from the image, then extract the string,
Perform number recognition. Since these are already known, their explanation will be omitted here.

In the above explanation, we took as an example a case where a character string is written on the label and the label can be detected. However, in cases where the label is bad and the character string is written directly on the object, it is possible to detect the character string instead of the label. If the column itself can be detected directly,
The overlapping amount Q of the visual fields is determined using the length of the diagonal of the circumscribed rectangle of the character string instead of the length of the diagonal of the label.

In addition, in the fifth embodiment, the case where the target object is moving has been explained, but the case where the target object is stationary is also explained.
It goes without saying that similar ideas can be applied (particularly Fig. 4(a), Cc, > , and (f) are applicable. Fig. 4(b) is the same as Fig. 4(a) in the case of a stationary object. In this embodiment, a camera with a shutter is used as the image sensor, but the same applies to a normal ITV camera or a -dimensional sensor. However, in the case of a -dimensional sensor, only the fields of view in the direction perpendicular to the movement direction of the object or sensor overlap.

Incidentally, in view of the object to be recognized and the resolution of the cameras, the camera field of view when inputting images using multiple cameras is different from the present invention, and the overlap amount α is set to zero, as shown in FIG. This can be considered as one method. However, with this method, it is very difficult to set the camera overlap amount to zero, and maintenance is difficult. Furthermore, even if the amount of overlap could be reduced to zero, would it be possible to synthesize and process two images containing the i11 object input across the fields of view of the two cameras? ,
It is necessary to have an image memory that can combine and process two images and a function that can process it), or to process the two images separately and successfully combine the recognized results.

〔Effect of the invention〕

According to the present invention, there may be cases where it is necessary to recognize an object whose images have been captured from a plurality of image sensors.

When it is necessary to recognize a target by continuously capturing images with one image sensor, the entire recognition target can be captured in at least one image, which simplifies the recognition process and enables high-speed recognition. becomes possible. In addition, when setting the field of view of the camera, make sure that the range in which multiple image sensors can cover the appearance range of the recognition target and that the overlapping part of the field of view is the maximum length of the recognition target (for example, the diagonal length of a label affixed to an object). Since it is only necessary to set the field of view of the image sensor so that the field of view is larger than the size of the field of view, setting and maintenance of the field of view is easy. Furthermore, as described above, there is an effect that a large-scale image memory for combining and processing images from a plurality of image sensors and an image processing function for processing them are not required.

[Brief explanation of drawings]

Figure 1 shows the overall configuration of the system for sorting things, and Figure 2 shows the overall configuration of the system for sorting things.
An explanatory diagram of a method for setting the field of view of an image sensor, FIG. 3 is a diagram showing an example of how the image sensor is connected, FIG. 4 is a diagram showing an example of setting the field of view of the image sensor, and FIG. 5 is an example of a field of view setting example other than the present invention. FIG. 1...Conveyor, 2...Object, 3...Level, 4,
5 ・Camera, 6...Image processing device, 7...Control device, 8...1st grade, 2nd grade (α) Figure 4 (α) Showo map (C) (cl) 硼□ L J謬=+Figure (e) Plot 4-■ (Preliminary)

Claims (1)

[Claims] 1. In a device or system that recognizes an object by processing images captured by an image sensor, at least one image of a plurality of images captured by a plurality of image sensors or in one image sensor is used. The image capturing is performed by setting the field of view of the image sensor to overlap so that the entire recognition target is always present in at least one of the plurality of continuously captured images. Image import method.