JP2913882B2 - Road object detection device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for detecting a road object, and more particularly to a device for detecting a road object such as an obstacle from an image obtained by a surveillance camera installed on the road.

[0002]

2. Description of the Related Art Conventionally, a system has been developed which monitors an object such as an obstacle located on a road in order to monitor and control the flow of traffic on an expressway or the like, and notifies the vehicle driver or the like of this as timely road information. Have been. In such a monitoring system, how to accurately and reliably detect a road object has become a major issue, and various methods have been developed.

For example, in the traffic flow measurement system disclosed in Japanese Patent Application Laid-Open No. 61-2144100, in order to improve the followability to environmental changes, a luminance value which is a reference independently for each measurement cycle is set to an input image. 3 shows a configuration for recognizing a road object while learning by an exponential smoothing method using a coefficient changed according to the difference value of.

That is, if the background image at time t = 0 is Y _i-1 and the input image at time t = T is X _i , the background image Y _i at time t = T is Y _i = (1-a ) X _i + aY _{i -1 is} sequentially updated.

As described above, by sequentially updating the reference background image, even when a vehicle that is not an obstacle is captured in the input image at time t = T, the weight of the input image can be reduced. Thus, it is possible to eliminate the influence of the vehicle and obtain a background image that matches the surrounding environment.

[0006]

As described above, in the related art, when an object is detected by comparing a background image with an input image to be processed, the background image is subjected to environmental changes such as changes in weather and movement of shadows. Although the on-road object is detected while being updated by the exponential smoothing method so as to follow up, there is a problem that it is possible to detect a moving object or a stationary object on the road, but it is not possible to detect that the stationary object has been removed.

That is, for example, when a cargo falls from a cargo bed such as a truck on the road, it can be detected by comparing it with a background image immediately after the occurrence of the cargo. However, the cargo remains at the same position on the road for a while. In this case, the image including the cargo, which is a stationary object, is registered as a background image, so that there is no difference between the background image and the input image, and the image cannot be detected. Further, when the stationary object that has been present at that position is removed, a difference occurs again between the background image including the stationary object and the input image not including the stationary object. This is erroneously detected as having occurred.

The present invention has been made in view of the above-mentioned conventional problems, and an object of the present invention is to provide a road object detecting apparatus capable of reliably detecting the occurrence or disappearance of a stationary object.

[0009]

In order to achieve the above object, a road object detecting apparatus according to the present invention comprises: a difference calculating means for calculating a difference between a photographed image and a reference background image; Storage means for storing the obtained positive or negative code, and determination means for determining occurrence or disappearance of a road object based on the obtained code and a change in the previous code stored in the storage means. And

[0010]

The road object detecting apparatus according to the present invention has such a structure, and does not detect a road object based on only the difference value between the photographed image and the reference background image as in the related art. Focusing on the code, it detects the occurrence and disappearance of a road object.

That is, for example, when a white stationary object is generated on the road as compared with the road surface, the luminance level of the input photographed image is higher than that of the reference background image, and the difference (reference background image-input image) is obtained. Takes a negative sign.
On the other hand, when the white still object is removed or disappears from the road, the sign of the difference between the input captured image and the updated reference background image is positive.

Therefore, the code obtained from the difference is stored in the storage means, the obtained code is compared with the previous code, and the occurrence or disappearance of a stationary object is also detected by paying attention to the code change. Becomes possible.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of a road object detecting apparatus according to the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the overall configuration of this embodiment, and FIG. 2 is a conceptual diagram of a road monitoring system in which the road object detection device is a part.
In FIG. 1, a road image captured by a monitoring TV camera 10 installed on a roadside such as a highway is an image processing device 1.
2 and a predetermined image processing is performed. The image processing device 12 includes an 8-bit A / D conversion unit 12a that converts an analog image signal into a digital image signal, a background image creation unit 12b that creates a background image from the digital image signal by an exponential smoothing method, a background image and input photographing. A difference calculation unit 12c for calculating a difference from an image, and a binarization process 12 for the difference data
d, a measuring unit 12e for measuring the number and position of road objects from the difference value and the binarized difference data, and a determining unit 12f for determining a moving object or a stationary object, and information on the detected road object. On the road control device 1 as shown in FIG.
4 and further transmits the information to a beacon 16 installed on the road side to notify the vehicle driver of information on the road object.

Hereinafter, image processing performed in the image processing apparatus 12 will be described in detail with reference to the flowchart of FIG.

First, in S101, an image to be processed is input. In this embodiment, as described above, the TV camera 10 is provided on the road side, and is approximately 10 m × on the road surface.
An area of 30 m is input as a processing target image. And
In step S102, a reference background image is created based on the captured image. This background image is created by the exponential smoothing method as in the related art. That is, if the background image at time t = 0 is Y _i-1 and the input image at time t = T, which is the next processing time, is X _i , the background image Y _{i at} time t = T is Y _i = ( 1-a) the X _i + aY _i-1. In this embodiment, a = 29/30. The image updated in this manner is registered as a reference background image, and the difference from the input image at time t = 2T is calculated in next S103. This difference calculation is performed by calculating a difference between each pixel luminance of the digital input image at time t = 2T and each pixel luminance of the background image created in S102.

After the difference between the input image and the reference background image is calculated in S103, the difference image data is binarized in S104. As the threshold value at the time of performing the binarization processing, for example, an average value of all pixels of the differential image data may be used, or an intermediate value between the maximum level and the minimum level of luminance may be used. After the binarization process is performed in S104, the position of the road object and its difference value are measured in the next S105.

Here, considering a case where no road object exists in the reference background image and a road object is included in the input image, the luminance of a pixel corresponding to the road object image in the input image is calculated based on the luminance of the reference background image. Since the pixel brightness is different from the corresponding pixel brightness, the difference value is a finite value other than 0, is determined to be white in the binarization process, and the position of the road object is measured by the pixel position in the entire image. Further, for example, when the detected road object is higher than the luminance of the road, the sign of the difference value is negative. Then, information on the code is stored in the memory 12g in the image processing apparatus 12.

After the position of the road object and the difference value have been measured, the flow shifts to the next step S106, where the judgment unit 12f judges whether or not an object at the same position exists. This determination is made by comparing the position and the difference value of the previously detected object. If it is determined that there is no object at the same position, the data of the position and the difference value of the detected object is stored in S107. I do.

On the other hand, if the position data match and it is determined that an object exists at the same position, the flow shifts to S108.

In S108, it is determined whether the detected stationary object has occurred or has disappeared. Therefore, using the sign of the difference value measured in S105,
It is determined whether the code is the same as the previous code stored in the memory 12g. If it is determined that the codes are the same, it means that a stationary object has just been generated and no new object has been generated, and data of the position and the difference value of the stationary object is created in S109.

On the other hand, if it is determined in step S108 that the code is a different code, it is determined that the stationary object has disappeared or been removed, and in step S110, data on the position and the difference value is created.

The process of detecting the occurrence and disappearance of a road object will be described more specifically with reference to FIGS.

FIG. 4 shows a temporal change of a background image, an input image and a difference value with the occurrence and disappearance of a stationary object. FIG. 4A shows a stationary object having a higher luminance than a road surface.
FIG. 4B shows a case where a stationary object having a lower luminance than the road surface is generated and disappears. First, in FIG. 4 (a), when a stationary object having a higher luminance than the road surface is considered, the luminance of the background image is updated by the exponential smoothing method due to the generation of the stationary object, and the luminance of the background image is increased by the luminance of the stationary object. Rises, and its luminance becomes constant in due course. On the other hand, the luminance of the input image discontinuously increases at the time when the stationary object occurs, and the difference value between the background image and the input image becomes a negative value. When this value exceeds a predetermined determination threshold Δd, It is determined that a stationary object has occurred.

When the stationary object continues to exist at the position, the stationary image is included in the background image, so that there is no difference from the input image, and the difference comparison value becomes zero. However, the sign has not changed,
Therefore, even if the difference value is 0, it is not determined that there is no stationary object, but it is determined that there is a stationary object as described in S109, and the position and the data of the difference value are stored.

When the stationary object disappears or is removed, the luminance of the background image gradually decreases by the exponential smoothing method with the disappearance of the luminance of the stationary object. Brightness level.

On the other hand, the luminance of the input image discontinuously decreases with the disappearance of the stationary object, so that the difference value shows a positive value, and this value exceeds a predetermined judgment threshold value Δd. In this case, since the sign has changed from negative to positive, it is determined that the stationary object has disappeared.

The case where a stationary object having a lower luminance than the road surface shown in FIG. 4 (b) is generated and disappears is the same as the above case, except that the sign polarity changes from positive to negative. It is. That is, when a stationary object occurs, the difference value indicates a positive value, and when the stationary object disappears, the difference value indicates a negative value.

FIG. 5 shows a background image, an input image and a temporal change of a difference value when a moving object passes. FIG. 5A shows a moving object having a higher luminance than the road surface. FIG. 5B shows a change when a moving object having lower luminance than the road surface passes. First, in FIG. 5A, when a moving object passes, the luminance level of the input image changes on a pulse as shown in the figure, while the luminance level of the background image gradually increases by exponential smoothing. Rises, the difference value has a slope-shaped pulse waveform at the peak portion, and when this pulse value exceeds a predetermined threshold value Δd for a short time, it is detected as a moving object.

Similarly, in the case of FIG. 5B, when a moving object having a low luminance passes, a positive pulse waveform having a slope-like peak appears in the difference value, and the moving object is detected. The Rukoto.

As described above, in the present embodiment, when comparing the input image with the reference image, attention is paid to the sign of the difference value, and the stationary object that was present when the sign is different is eliminated. Or to accurately determine the occurrence and disappearance of a stationary object, and to inform vehicle drivers etc. of accurate information on the road without erroneously detecting that a stationary object that does not exist originally exists. Becomes possible.

[0032]

As described above, according to the road object detecting apparatus of the present invention, the occurrence and disappearance of a road object can be reliably detected. This has the effect of significantly improving the performance.

[Brief description of the drawings]

FIG. 1 is a configuration block diagram of an embodiment of a road object detection apparatus according to the present invention.

FIG. 2 is a conceptual explanatory diagram of a road monitoring system using the road object detection device of the embodiment.

FIG. 3 is a flowchart in the embodiment.

FIG. 4 is an explanatory diagram showing occurrence and disappearance of a stationary object in the embodiment.

FIG. 5 is an explanatory diagram showing occurrence and disappearance of a moving object in the embodiment.

[Explanation of symbols]

 Reference Signs List 10 TV camera 12 Image processing device 14 Road control device 16 Beacon

Claims (1)

(57) [Claims] 1. A road object detecting apparatus for detecting a road object included in a captured image by comparing a sequentially updated reference background image with a captured image, a difference calculation for calculating a difference between the captured image and the reference background image. Means, and storage means for storing the positive or negative sign obtained by the difference calculation means,
A road object detecting apparatus comprising: a determination unit configured to determine occurrence or disappearance of a road object based on the obtained code and a previous change in the code stored in the storage unit.