JP2995803B2 - Parking detection device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an apparatus for detecting a vehicle parked and stopped on a traffic path of a vehicle such as a road by image processing.

2. Description of the Related Art Conventional technology and its problems By capturing an image of a vehicle such as a road from above or diagonally from above, image data including vehicles parked and stopped and running is obtained, and this image data is processed. 2. Description of the Related Art As a device for detecting a parked vehicle, there is a device that uses a frame difference method and a spatial difference method. A parking / stop detection device using this image processing method
A frame difference is applied to a plurality of input images captured with a time difference to extract a running vehicle, and thereby a spatial difference is applied to a region other than the region where the running vehicle exists to emphasize the lines and contours of the parked and stopped vehicles. Then, parked and stopped vehicles are extracted from the image.

However, such a conventional device requires sufficient brightness to extract the lines and contours of the vehicle, and requires a large-scale lighting device to detect the lines and contours of the vehicle at night. In addition, since the traveling vehicle is extracted based on the frame difference, the traveling vehicle must change (move) to some extent between the two images. Therefore, when the vehicle is traveling at a low speed such as a traffic jam, the amount of change due to the frame difference is small, and it is difficult to determine whether the vehicle is a parked or stopped vehicle or a traveling vehicle.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a device capable of correctly detecting only parked and stopped vehicles by excluding a vehicle traveling at a low speed without being affected by brightness or the like.

Configuration, operation, and effect of the present invention A parking / parking detection device according to the present invention captures an image of a vehicle such as a road at regular intervals, and includes images of vehicles parked and stopped on the captured road and running. Imaging means for outputting a video signal representing an image for each imaging, means for converting the video signal output from the imaging means to digital image data, and adding the converted digital image data between corresponding pixels for each scene imaging Means for storing the added cumulative image data, and, after the cumulative addition of the image data of the predetermined number of scenes is completed, the information on the parked vehicle based on the cumulative image data stored in the storage means. Is provided with an image processing means for extracting

According to the device of the present invention, the taken images are cumulatively added. Therefore, in the daytime, by adding a large number of images, the parked and parked vehicles appear floating because the moving vehicle is smoothed as a noise component against the background and the parked vehicles that do not move. Become. After the noise component is removed from such an accumulated image, a parked or stopped vehicle can be detected by extracting a contour line or the like which is a feature of the vehicle by using a spatial differentiation method or the like.

Also, at night, by adding the light trails of the headlights and tail lights, the light trails can be clearly identified, and when there are parked and stopped vehicles that obstruct traffic flow, These parked and stopped vehicles can be indirectly detected by changes in the light trace of the light.

As described above, according to the present invention, since the light trace of the vehicle light can be used at night, no illumination device is required at night. Further, even if the brightness of the screen changes due to a sudden change in the weather or the like during the cumulative addition of the image data, it can be treated as noise and the influence is hard to appear. Further, even when the moving vehicle is at a low speed due to traffic congestion or the like, a sufficient number of images are cumulatively added, so that the image is less affected. Since there is no need to continuously perform high-speed image processing, it is possible to perform processing in another place by image transmission, and in this case, a low-capacity communication line or the like can be used.

FIG. 1 shows an arrangement of an embodiment for detecting vehicles parked and stopped on a road.

A television (video) camera 9 is installed on an upper part of a column erected on the side of the road L. The camera 9 looks down on the road L, captures an image of a predetermined range of the road L, and outputs a video signal representing the captured image. Road L, Road L by camera 9
Vehicles that are parked and stopped on the road L and vehicles running on the road L are imaged. The video signal output from the camera 9 is input to the image processing device 10. It is also possible to stand uprights on both sides of the road, pass a horizontal line on top of these columns, and install the camera 9 on this horizontal line. Further, the camera 9 can be attached to an overpass or other structure to take an image of a road from above.

FIG. 2 shows an electrical configuration of the image processing apparatus 10.

A video signal for one scene (one frame or one field) output from the camera 9 is converted into digital image data by an A / D converter 11 and, if necessary, a bit conversion circuit 12.
To the adder 13 through. The adder 13 is an image memory
The accumulated image data stored in 14 and the image data provided from the A / D converter 11 are added for each pixel. The added image data is stored in the image memory 14. The image data of the first scene is stored in the image memory 14 without being added. In this manner, the image data of the predetermined number of scenes is added by the adder 11, and the result of the addition is stored in the image memory 14 as the accumulated image data. The number of accumulated scenes can be determined according to the gradation (number of bits) of the image data and the capacity of the image memory (number of data bits), and according to the parked and stopped vehicle extraction processing. For example, if the image memory is 16 bits and the gradation of the image data is 8 bits, cumulative addition of 8 bits, that is, 256 scenes is possible. If more accumulated additions are required for the process of extracting parked and stopped vehicles due to extreme congestion, etc., the conversion circuit 12 converts the 8-bit image data into 7-bit image data.
Cumulative addition of 512 scenes becomes possible.

The CPU 15 stores image data of the required number of scenes in the image memory 14.
After that, the information on the parked / stopped vehicle is extracted according to a predetermined image processing algorithm using the image data obtained by the cumulative addition.

FIG. 3 shows an example of an image processing algorithm for extracting parked and stopped vehicles that can be suitably used in the daytime.

One scene consists of vertical and horizontal 125 × 512 pixels, each pixel is 8
Image data represented by bits (256 gradations) is cumulatively added for 256 scenes. Assuming that it takes 1/15 second per scene, the cumulative added image data (512
× 512 × 16 bits). An image of a running (moving) vehicle is stored as if the camera was shaken, and when accumulated for a long time, becomes an noise component. On the other hand, the background and the image of the parked vehicle reinforce each other by addition and clearly appear.

In order to remove noise components such as the accumulated data of the image of the traveling vehicle, the image data of 16 bits per pixel is converted into 8 bits.
Shift, discard lower 8 bits, and leave only upper 8 bits. The image data is 512 × 512 × 8 bits.

An operation is performed using the measurement area mask to remove the background. As a result, only the image of the parked vehicle on the road remains. Further, the image data obtained in this manner is subjected to a spatial differentiation (spatial difference) calculation or the like to extract a feature amount such as a contour of the vehicle, and a parked or stopped vehicle is detected based on the feature amount. .

FIG. 4 is a diagram for explaining an image processing algorithm for extracting parked and stopped vehicles that can be suitably used at night.

 The outline of this nighttime algorithm is as follows.

Headlights and taillights can be used to identify vehicles at night. By accumulating the images of these light spots, the progress of the traveling vehicle is stored as light trails. By performing cumulative addition over a long period of time, it is possible to grasp the traffic flow situation as a group. That is, when there is no parked or stopped vehicle that hinders the traffic flow, a light trace is obtained along each lane, and the accuracy of the image is increased in that part. On the other hand, if there are obstacles (such as parked and stopped vehicles) that hinder the traffic flow, the traffic flow changes in a curve so as to avoid those obstacles, and the accuracy of the image near the obstacles increases. , The parked / stopped vehicle can be indirectly detected by detecting this.

Referring to FIG. 4, 16-bit accumulated image data is obtained by cumulatively adding 256 scenes of 8-bit image data. More scenes can be added by converting the image data of each scene into fewer bits by performing bit conversion. For example, by extracting a tail light (or a head light) from image data of each scene, it is possible to express two bits per scene. 16 in this case
Addition of 65536 scenes is possible in order to obtain the cumulative addition image data of bits. If it takes 1/15 second to measure one scene, it takes about 17 seconds and 655 seconds to add 256 scenes and obtain 16-bit cumulative image data from 8-bit image data.
It takes about 73 minutes to obtain 16-bit accumulated image data from 2-bit image data by adding 36 scenes. By appropriately selecting the bit conversion, the number of scenes to be added can be set optimally in consideration of the total measurement time.

Next, a right shift process of a predetermined number of bits is performed on the accumulated image data in accordance with the number of accumulated scenes or the time to obtain 8-bit image data. In this image data, light trails from the tail light or headlight of the vehicle clearly appear.

Subsequently, the background is removed using the measurement area mask.
Finally, a small area is set on the road along the lane (shown by a solid line), and the luminance in each small area is checked.
Where the parked and parked vehicles are present, the light trace curves and turns, so a small area set at that portion (indicated by A)
It can be seen that there are parked and stopped vehicles in this small area A that hinder traffic flow.

[Brief description of the drawings]

FIG. 1 is a diagram showing an arrangement configuration in a case where the present invention is applied to detection of parking and stopping on a road. FIG. 2 is a block diagram showing an electrical configuration of the image processing apparatus. FIG. 3 is a diagram for explaining an algorithm suitable for detecting parking and stopping at daytime, and FIG. 4 is a diagram for explaining an algorithm suitable for detecting parking and stopping at night. 9: TV (video) camera, 10: Image processing device, 11: A / D converter, 13: Adder, 14: Image memory, 15: CPU.

Claims (1)

(57) [Claims] An image of a vehicle such as a road is taken at regular time intervals, and a video signal representing an image including an image of a vehicle parked and stopped on the taken road and a running vehicle is output for each image taking. Imaging means, means for converting a video signal output from the imaging means into digital image data, adding the converted digital image data between corresponding pixels for each scene, and storing the added cumulative image data And after the cumulative addition of the image data of the predetermined number of scenes is completed,
An image processing means for extracting information on the parked vehicle from the accumulated image data stored in the storage means.