JP2001092979A - Unsuitable image detecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for exactly judging whether an image from a camera is a suitable image or unsuitable image. SOLUTION: A vertical or horizontal edge component is extracted and projected while using output images from cameras by using at least two cameras and the unsuitable image of blurring is detected by comparing the width of the projected edge component exceeding a threshold value. Besides, the unsuitable image of bleeding is detected by comparing the number of projected edge component exceeding the threshold value. Further, the unsuitable image of bleeding is detected by comparing the direction of an edge in the edge image provided from the extracted vertical or horizontal edge component. When the unsuitable image is detected, this unsuitable image is corrected and the corrected image is used for vehicle control. When the unsuitable image can not be corrected, however, system control based on the image is stopped and abnormality is reported to a driver.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an inappropriate image detecting method for detecting whether or not an object cannot be recognized because an inappropriate image has been input.

[0002]

2. Description of the Related Art A camera is mounted on a vehicle, and an output image of the vehicle is used to monitor a preceding vehicle and its surroundings, detect a danger, and notify a driver of the danger, or control the vehicle to avoid the danger. That is being done. For this purpose, an image processing apparatus is used. However, when an object such as a preceding vehicle cannot be recognized, the object cannot be recognized because the object does not exist, or the object cannot be clearly recognized due to rain, dust, or the like. It is important to distinguish whether an appropriate image has been input and cannot be recognized.

Conventionally, a single camera is used to monitor a preceding vehicle and its surroundings using an output image thereof, detect a danger, and notify a driver of the danger. For this reason, an inappropriate image, that is, whether the image is blurred or a carrot, is detected by performing an inappropriate image determination process separately from objective recognition using the output image of the one camera.

[0004]

When an inappropriate image is detected by a single camera as described above, the determination of whether the image is blurred or carrot is made by determining the width of the edge of the image,
Use numbers, directions, etc. However, in the case of a camera provided in a vehicle, the background greatly changes in accordance with the environment such as the weather and the movement of the vehicle, and accordingly, the width, number, direction, and the like of the edges of the image greatly change. It is difficult to detect whether the output image is an unsuitable image such as a blurred image or a bleeding image using only the output image.

Accordingly, an object of the present invention is to provide a method for accurately determining whether an image from a camera is an appropriate image or an inappropriate image.

[0006]

Accordingly, in the present invention, at least two cameras are used, and a vertical edge or a horizontal edge is extracted and projected by using an output image from each camera. The inappropriate image of blur is detected by comparing the widths of the obtained edges. Further, a vertical edge or a horizontal edge is extracted and projected using the output images from the respective cameras, and the number of peaks of the projected edge exceeding a threshold value is compared to detect a bleeding inappropriate image. Further, a vertical edge or a horizontal edge is extracted from the output image of each camera, and the direction of the edge is compared in the edge image obtained from the extracted edge to detect an unsuitable image of bleeding.

When an inappropriate image is detected, the inappropriate image is corrected, and the corrected image is used for a vehicle control system. However, when the improper image cannot be corrected, the vehicle control system is stopped and the driver is notified of the abnormality.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS How to detect a blurred image or a blurred image using an output image of a camera will be described below. FIG. 1 is a diagram showing a sample of the output image of the camera. In FIG. 1, a hatched portion A is a sample image, the horizontal width is the length between x1 and x2, and the height is a on the left and b on the right. First, a vertical edge component is extracted from this sample image. Extraction of a vertical edge extracts only a vertical component from an edge of an image, that is, a contour line. FIG. 2 shows an extracted vertical component of the outline of the sample image A of FIG. 1, that is, a vertical edge component.

FIG. 2 is an enlarged view of FIG. 1 showing vertical edge components in pixel units. As can be seen from FIG. 2, the number of vertical edge components of the left and right vertical contours is large, and the number of vertical edge components of the horizontal contour is small. FIG. 3 shows a diagram in which vertical edge components are extracted and projected. The horizontal axis is the same as FIG. 1, and the vertical axis represents the number of vertical edge components.
In FIG. 3, the peak a represents the left edge component a of the sample image A, and the peak b represents the right edge component b of the sample image A. And
The number of vertical edge components of peak a is a1 and peak b
Is the number of vertical edge components b1. Peaks a and b have a certain width. For example, peak a has a width D, and D is referred to as "edge width". In FIG.
th is a threshold value set to a predetermined value, and the number of peaks (a and b in this case) exceeding the threshold value is referred to as “the number of edges”.

FIG. 4 is an edge image obtained by extracting vertical edge components of the image shown in FIG. 1 and extracting the extracted edges. As a result, a left edge a, a right edge b, and an upper edge c of the sample image are displayed. As shown by arrows in this figure, the directions of a and b are vertical directions, and c
Is oblique. These directions are called "edge directions"
Say

When detecting a blurred image or a blurred image using an output image of a camera, the "edge width"
“Edge number” and “edge direction” are used. Next, FIG.
In the case where the sample image shown in FIG. 5 is blurred, FIG. 5 shows a vertical edge component extracted and projected as in FIG. When the image is blurred, the width of the peak becomes wide. For example, the width of the edge of the peak a is D + α, and the width is increased by α.

FIG. 6 shows that when the output image of the camera is not a clear image as shown in FIG. 1 but a blurred image, vertical edges of the blurred image are extracted and obtained from the extracted edges. FIG. As shown in the figure, a left edge a, a right edge b, and an upper edge c of the sample image are displayed. In this case, since the image is a blurred image, the width of the edge is wider than that in FIG.

Next, when the output image shown in FIG. 1 is made of, for example, nidine due to rain, it becomes distorted as shown at A 'in FIG. FIG. 8 shows an image obtained by extracting and projecting the vertical edge component of this carrot image in the same manner as in FIG.
When the image is made of nidine, the image is distorted, and the edge is distorted as shown in FIG. 8, so that the peaks a and b become low as a1-β1 and b1-β2, respectively. As a result, as shown in FIG. 8, only the peak a exceeds the threshold th and the peak b becomes smaller than the threshold th.

4 based on the output image shown in FIG.
When an edge image is obtained in the same manner as in the above, an edge image as shown in FIG. 9 is obtained. As can be seen from FIG. 9, the directions of the edges a and b are not vertical but slightly inclined. Next, in the present invention, how to detect an inappropriate image such as the blurred image or the blurred image using the output image of the camera will be described.

FIG. 10 is a diagram showing a configuration of an image processing apparatus used for detecting an inappropriate image according to the present invention. In FIG. 10, reference numeral 1 denotes an image processing apparatus, which includes an inappropriate image detection unit 11, a vehicle recognition and distance measurement unit 12, and an image correction unit 13. The inappropriate image detecting unit 11 receives output images from the first camera 21 and the second camera 22, and determines whether the image is appropriate or inappropriate. If it is determined that the image is appropriate, the image is sent to the vehicle recognition and ranging unit 12 as an appropriate image, and measurement is performed based on this image. For example, when the image is a preceding vehicle, an inter-vehicle distance with the preceding vehicle is measured, and the data is sent to the vehicle control device 31. Based on the data, inter-vehicle distance control, collision prevention control, and the like are performed. Reference numeral 32 denotes a system stop unit that receives a correction impossible signal described later and stops the vehicle control system based on the image. When it is detected that the vehicle is too close to the preceding vehicle, an alarm is issued from the alarm 33.

On the other hand, if it is determined that the image is inappropriate, the image is sent to the image correction unit 13 as an inappropriate image. Here, the image determined to be an inappropriate image is corrected, and if the image can be corrected, the corrected image is sent to the vehicle recognition distance measuring unit 12. However, when the correction is impossible, the correction impossible signal is sent to the system stop unit 32 of the vehicle control device to stop the vehicle control system. At the same time, the driver notification unit 3
This signal is sent to No. 4 to notify the system abnormality.

Although the above image processing apparatus uses two cameras, it suffices that two or more cameras be used as long as they can be compared with each other. Next, how an inappropriate image is detected by the image detection device shown in FIG. 10 will be described. FIG. 11 is a flowchart showing how to control inappropriate image detection. The control shown in this flowchart is performed by a microcomputer (not shown) provided in the image processing apparatus of FIG. Also, the description will be made assuming that two cameras are used.

When the control for detecting an inappropriate image is started, first, the output image of the first camera 21 is taken into the inappropriate image detecting section 11 (S1), and then the output image of the second camera 22 is converted to the inappropriate image. The data is captured by the detection unit 11 (S2). Next, edges are extracted and projected from the image of the first camera as shown in FIG. 3 (S3), and subsequently, edges are extracted and projected from the image of the second camera (S4). In this case, either the vertical edge component or the horizontal edge component may be extracted, but which one to extract can be selected according to the shape of the object. Next, the edges extracted and projected from each of the two extracted cameras are compared. First, it is determined whether the width of the edge is normal (S
5). For this purpose, the widths of the edges exceeding the projected threshold value of each camera are compared, and if the width of one edge is clearly wider than the other, it is determined that the width of the edge is not normal (No. ), Which is detected as a blurred image generated in one of the cameras, and is determined as an inappropriate image by the image correction unit 1
3 and the image is corrected (S8).

If both edge widths are normal (Yes), it is determined whether the number and / or orientation of the edges is normal (S6). To do so, the number of edges that exceed the projected threshold of each camera is compared. If the camera image has blurring, the number of edges exceeding a certain threshold value is reduced. Therefore, if the number of one edge is clearly smaller than the other, it is determined that the edge is not normal (N
o), it is detected as a blurred image generated only in one camera, and sent to the image correcting unit 13 as an inappropriate image to correct the image (S8). In addition, the direction of the edge is compared between the edge images shown in FIGS. 4 and 9. If the directions of the edges do not match, it is determined that it is not normal (N
o), sent to the image correction unit 13 as an inappropriate image.

If both the number of edges and the direction of the edges are normal (Yes), normal processing is performed using the detected image. For example, the image is sent to the vehicle recognition and ranging unit 12 as an appropriate image, and measurement is performed based on this image. For example, if the image is a preceding vehicle,
The inter-vehicle distance from the vehicle is measured, and the data is sent to the vehicle control device 31, and the inter-vehicle distance control or the like is performed based on the data.

On the other hand, the inappropriate image sent to the image correcting section 13 is corrected to an appropriate image here (S8), but it is determined whether the image can be corrected (S9), and the image can be corrected. In the case (Yes), the correction is sent to the vehicle recognition and distance measurement unit 12. The correction is performed based on the other appropriate image. However, if the correction is not possible due to large distortion or the like and the degree of inappropriateness is large (No), a correction impossible signal is sent to the system stop unit 32 and the driver notification unit 34, and the vehicle control system is stopped (S10). Then, the driver is notified of the fact (S11).

In the case of the above embodiment, the detection of an inappropriate image is performed on the image of the entire screen. However, instead of the entire screen, an object that always appears on the screen, such as a white line on the road or the hood of the own vehicle, is used. An edge is extracted and projected, and an inappropriate image such as a blurred image can be detected based on this. Hereinafter, such a case will be described. FIG. 12 is an image viewed from the windshield of the own vehicle during traveling. In the figure, B is the hood of the own vehicle, and C is the rear part of the preceding vehicle. W is a white line representing a lane. In this figure, peaks p and q are vertical edge components of a white line, and r is a horizontal edge component of hood B extracted and projected. Since the vertical edge component of the white line is not so large, the height of the peak is low. Since the vertical edge component is small in the case of the hood, the horizontal edge component is extracted. As described above, it is not always necessary to use the vertical edge component, and the horizontal edge component may be extracted according to the shape.

As in the case described above, a vertical edge component or a horizontal edge component of an object which always appears on a screen such as a white line or a bonnet is extracted from images obtained by two cameras and projected. By obtaining an edge image and comparing the edge width, the number of edges, and the direction of the edge, an unsuitable image such as a blurred image or a blurred image can be detected.

[0024]

As described above, according to the present invention, at least two cameras can be used to detect an unsuitable image such as a blurred image or a blurred image, which was difficult to detect with one camera. Further, the usability of the system is improved by correcting an inappropriate image such as a blurred image or a blurred image.

Further, by accurately detecting an inappropriate image, the safety of the vehicle control system is increased.

[Brief description of the drawings]

FIG. 1 is a diagram showing a sample of an output image of a camera.

FIG. 2 is an enlarged view of FIG. 1 showing vertical edge components in pixel units;

FIG. 3 is a diagram in which a vertical edge component is extracted and projected.

FIG. 4 is a diagram showing an edge image of the sample image of FIG. 1;

FIG. 5 is a diagram in which a vertical edge component of a blurred image is extracted and projected.

FIG. 6 is a diagram showing an edge image of a blurred image.

FIG. 7 is a diagram showing a sample of a carrot output image.

FIG. 8 is a diagram in which a vertical edge component of a carrot output image is extracted and projected.

FIG. 9 is a diagram showing an edge image of a blurred image.

FIG. 10 is a diagram showing a configuration of an image recognition device used for detecting an inappropriate image according to the present invention.

FIG. 11 is a flowchart showing how to control inappropriate image detection according to the present invention.

FIG. 12 is a diagram in which an image viewed from a running windshield and an extracted edge component are projected.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Image processing apparatus 11 ... Unsuitable image detection part 12 ... Vehicle recognition distance measurement part 13 ... Image correction part 21 ... First camera 22 ... Second camera 31 ... Vehicle control device 32 ... System stop part 33 ... Warning 34 ... Driver Information department

Claims (9)

[Claims] 1. An inappropriate image is detected by extracting and projecting a vertical edge or a horizontal edge from an output image of a camera using at least two cameras and comparing a width of the projected edge exceeding a predetermined threshold. how to. 2. An inappropriate image by extracting and projecting a vertical edge or a horizontal edge from an output image of each camera using at least two cameras, and comparing the number of projected edges exceeding a predetermined threshold. How to detect. 3. Using at least two cameras, extracting a vertical edge or a horizontal edge from an output image of each camera, and in an edge image obtained from the extracted edges,
A method of detecting inappropriate images by comparing the directions of edges. 4. The method for detecting an inappropriate image according to claim 1, wherein a vertical edge or a horizontal edge is extracted and projected from the entire output image of the camera. 5. The method for detecting an unsuitable image according to claim 1, wherein a vertical edge or a horizontal edge is extracted from a target that always appears on a screen in an output image of the camera and projected. 6. The method according to claim 3, wherein a vertical edge or a horizontal edge is extracted from the entire output image of the camera to obtain an edge image. 7. The method for detecting an unsuitable image according to claim 3, wherein a vertical edge or a horizontal edge is extracted from a target that always appears on a screen in an output image of the camera to obtain an edge image. 8. The method for detecting an inappropriate image according to claim 1, further comprising the step of correcting the detected inappropriate image into an appropriate image based on the appropriate image. 9. The vehicle control system using the output image of the camera is stopped if the image cannot be corrected to an appropriate image in the step of correcting the image to the appropriate image.
Method for detecting inappropriate images described in.