JP2002259953A - Congestion status determining method and its device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a congestion status determining method and its device capable of improving determination accuracy of a vacant state to almost 100%. SOLUTION: The device is provided with a camera 1 shooting an overview of an object area, an image memory 3 storing the shot image as an luminance image, an edge image transforming part 4 transforming the stored image into an edge image, a template image storing part 5 stored beforehand with template images of each congestion status such as vacancy, congested, and full, a correlation value calculating part 6 calculating correlation values with each template image by determining cross-correlations between the template images and the image of the object area converted into the edge image, and a determining part 7 comparing the calculated correlation values and determining the congestion status of the object area from their magnitudes. As the vacancy template image, a pseudo vacancy template image is used provided by applying edge processing to a pseudo image with white lines such as parking grids drawn in respective widths in accordance with the actual object area and then subjecting it to gradation processing.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a congestion state determination method and apparatus for determining a congestion state using a cross-correlation between a captured image and a template image.

[0002]

2. Description of the Related Art In a conventional parking lot system using image processing, the presence or absence of a vehicle is determined for each parking square by using a luminance change, edge information, and the like, and the results of individual detections are accumulated to obtain a parking lot system. A method of estimating and determining the overall congestion state is known. This method is based on the vehicle determination accuracy for each parking space, and the more information on the details of the vehicle, the higher the determination accuracy. Therefore, it is necessary to project the vehicle larger on the screen in order to increase the determination accuracy. However, the larger the vehicle is projected on the screen, the more it becomes impossible for one camera to detect a wide range at a time, and a large number of cameras are required to photograph the entire parking lot.

In order to solve the above-mentioned problem, the present inventor has previously proposed a method of determining the congestion state based on a cross-correlation between a captured image of a target area and a template image (Japanese Patent Application No. 2000-098083). This congestion status determination method takes a cross-correlation between a template image of each congestion status such as empty, congested, full, etc. prepared in advance and a target image captured from a camera, and determines the relative magnitude of the obtained correlation value. The congestion situation closer to the template image is found from the relationship. For this reason, it is possible to determine a wide-ranging parking situation only from the entire situation without paying attention to each parked vehicle as in the conventional parking lot system using image processing.

[0004]

However, when the congestion status determination method of the prior application was actually tested and operated, it was found that among the congestion statuses such as empty, congested, and full, the determination accuracy particularly when the vehicle is empty is low. It turned out that there was a problem of falling. The present invention has been made in order to solve such a problem, and an object of the present invention is to provide a congestion state determination method and apparatus capable of improving the determination accuracy of an empty state to almost 100%. is there.

[0005]

The present invention employs the following means in order to solve the above problems. In other words, the congestion state determination method of the present invention performs bird's-eye view photographing of the target area for which the congestion state is to be determined, and interoperates the photographed target image with a template image representing each congestion state such as empty, congested, or full, prepared in advance. In the congestion situation determination method in which a correlation value between each template image is calculated by taking a correlation and the congestion situation is determined from the magnitude of the calculated correlation value, the empty vehicle template image is artificially used. It is characterized in that the created pseudo empty vehicle template image is used.

Further, the congestion state determination device of the present invention includes a photographing means for bird's-eye view photographing of a target area where the congestion state is to be determined,
Image storage means for storing a captured image of the target area captured by the image capturing means as a luminance image, an edge image conversion unit for converting the luminance image of the target area stored and stored in the image storage means into an edge image, A template image storage unit that stores in advance a template image representing each congestion state such as congestion and fullness, an image of a target area converted into an edge image by the edge image conversion unit, and the template image storage unit. A cross-correlation with all the template images that are present and calculating a correlation value with each template image, and comparing the correlation values calculated by the correlation value calculation unit, and comparing the correlation value with the correlation value to determine whether the target area is congested. A determination unit for determining a situation, wherein a pseudo empty vehicle template created artificially as an empty vehicle template image stored in the template image storage unit. It is characterized in that using the image.

[0007] As a pseudo empty vehicle template image used in the congestion state determination method and apparatus, edge processing is performed on a pseudo image in which a white line such as a parking space is drawn at each thickness in accordance with an actual target area. After subjecting the blurred image or a pseudo image in which a partition line such as a parking space is drawn with a single thin white line in accordance with the actual target area to an expansion process, edge processing is performed, and the edge image is further blurred. This is the one using the processed image.

The present inventor has repeatedly conducted various experiments and studies to solve the problem that the accuracy of determination in an empty state is reduced in the congestion state determination method of the prior application. First, FIG. 7 shows the state of congestion determination when the vehicle is empty according to the conventional method, and Table 1 shows the accuracy of the empty vehicle determination at that time. According to this conventional technique, an edge image (FIG. 7A) of a representative empty state image actually obtained by photographing as an empty vehicle template image and a representative image obtained by actually photographing an empty vehicle template image are used. 7 (b)), and a cross-correlation was obtained between these template images and the target image in an empty state (FIG. 7 (c)). Things.

[0009]

[Table 1]

As is clear from Table 1, even in the case of the conventional method, a high empty vehicle judgment accuracy of 97.6% is obtained, but it is still not 100%. This is because, in the cross-correlation between empty vehicle images, a white line such as a parking space greatly affects the correlation value. However, since an apparatus for determining the congestion state is generally installed after the parking lot is put into service, the empty vehicle template is used. When an actual photographed image is used as an image, some parked vehicles are always present in a plurality of empty vehicle template images prepared in advance, and this parked vehicle hides a white division line such as a parking space. Was found to be one of the causes. That is, in the congestion situation determination method of the prior application, the determination accuracy decreases when the vehicle is empty, which is largely due to the fact that an actual photographed image is used as the empty vehicle template image.

The present inventors have conducted various experiments from the above viewpoint, and as a result, as a template image of an empty vehicle,
It was confirmed that the determination accuracy was significantly improved by using an ideal empty vehicle image in which no vehicle was parked. FIG. 3 shows a determination state when the empty vehicle determination accuracy by the cross-correlation is performed using an ideal actual photographed image in which no vehicle is parked as a template image as an empty vehicle template image, and the empty vehicle determination accuracy at that time. Are shown in Table 2.

[0012]

[Table 2]

As is apparent from Table 2, when an ideal actual photographed image in which no vehicle is parked is used as a template image, 100%
Can be obtained. That is, if an ideal template image in which no vehicle is parked is used as an empty template image and the relative comparison of the correlation values by the cross-correlation is performed, the complete empty detection can be performed.

From the above results, it can be seen that it is ideally desirable to use an actual photographed image in which no vehicle is parked as an empty vehicle template image. However, it is often difficult to actually obtain such an image in a parking lot in service. The inventor of the present invention has taken several images of an empty vehicle state, cuts out regions where no vehicles exist from these captured images, and connects these to form an empty vehicle template in which no parked vehicle exists. When an image was created and cross-correlation was performed, it was found that an empty vehicle judgment rate of 100% could be obtained in the same manner.

However, when an empty vehicle template image is created by cutting and pasting a plurality of empty vehicle images actually photographed as described above, the amount of work is enormous, it takes too much time, and it is difficult to easily adopt the image. I also found out. Therefore, the present inventor further repeated experiments and studies, and drawn a parking section or zebra line with the thickness of each white line on the computer screen using the coordinates of the parking square obtained from the actual image, and the white line and the zebra A pseudo empty vehicle image in which none of the parked vehicles is present is created by artificially creating an image of a drawn empty vehicle drawn in such as an image, and sequentially performing edge processing and blurring processing on this image. I tried to determine the empty car using.

FIG. 4 shows a determination state when the empty vehicle determination accuracy is determined by the cross-correlation using the pseudo empty vehicle template image. It has been found that even if the pseudo empty vehicle template image created artificially in this way is used, 100% empty vehicle determination accuracy can be obtained. That is, a pseudo empty vehicle template image created artificially can obtain the same accuracy as an ideal empty vehicle template image, and can be used as an empty vehicle template image. Furthermore, there is an advantage that it can be created more easily than an ideal empty template image in which no actually parked vehicle exists.

[0017]

[Table 3]

The present invention has been made based on the above findings, and uses a pseudo empty vehicle template image artificially created as an empty vehicle template image. Then, as an artificially created pseudo empty vehicle template image, specifically, after performing edge processing on a pseudo image in which a lane marking such as a parking space is drawn with a white line in accordance with the actual target area,
After subjecting the blurred image or the proceeding image in which a partition line such as a parking space is drawn with one thin white line in accordance with the actual target area to an inflating process, edge processing is performed, and the edge image is further processed. This is an image adopting a blurred image. In this way, when an artificially created pseudo empty vehicle template image is used for the empty vehicle template image,
The determination accuracy of the empty state can be improved to almost 100%.

[0019]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an embodiment of a congestion state determination device according to the present invention, wherein 1 is a camera (photographing means) for taking a bird's-eye view of a parking lot, and 2 is a digital image signal output from an analog image signal output from the camera 1. An AD converter 3 for converting the image into an image memory 3 (an image memory for storing an image captured by the camera 1 until a series of determination processes is completed); Is an edge image conversion unit that converts the image into an edge image, 5 is a template image storage unit, 6 is an image of the target area converted into the edge image by the edge image conversion unit 4, and all the images stored in the template image storage unit 5. A correlation value calculator 7 calculates a cross-correlation with the template image and calculates the correlation value. The correlation value calculator 7 compares the correlation values calculated by the correlation value calculator 6 to determine the congestion state of the target area based on the magnitude. It is a determination section for.

The template image storage unit 5 previously stores template images representing congestion states such as empty, congested, and full. In the present invention, among these template images, an empty car template image is used as an empty car template image. An artificial pseudo empty vehicle template image created by the method shown in FIG. 5 was used.

The method of creating the pseudo template image shown in FIG. 5 will be described. First, as shown in FIG. 5A, a white line is drawn on a computer screen or the like using the coordinates of an actual target area such as a parking space. Make a pseudo image subtracted by each thickness. Then, the pseudo image is subjected to edge processing as shown in FIG. 5 (b), and then subjected to blur processing as shown in FIG. 5 (c), and the image subjected to this blur processing is referred to as a pseudo empty vehicle template image. It was done.

The blurred image shown in FIG. 5C is obtained by an image blur processing means such as a Gaussian filter. However, as for the blurring process, the blurring process is performed with a relatively small number of pixels in order to smooth the data of the edge portion so that the actual white line approaches the edge image when the edge processing is performed.

Next, the determination operation of the congestion state determination device will be described with reference to the flowchart of FIG. When the camera 1 takes a bird's-eye view of the parking lot to be determined (step S1), the image signal is converted to a digital signal by the AD converter 2, sent to the image memory 3, and stored as a luminance image.

The target image composed of the luminance image stored and stored in the image memory 3 is sent to an edge image conversion unit 4, where only an edge component is extracted from the image by an edge extraction filter such as a Roberts filter or a Laplacian filter.
It is converted to an edge image (step S2). Then, the target image converted into the edge image is sent to the correlation value calculation unit 6.

The correlation value calculating section 6 represents the target image converted into the edge image by the edge image converting section 4 and each congestion state such as empty, congested, full or the like stored in the template image storing section 5. A cross-correlation with the template image is obtained, and a correlation value with each template image is calculated (steps S3, S4, S5), and the result is determined by the determination unit 7.
Send to

The determination unit 7 compares the correlation values sent from the correlation value calculation unit 6 and selects the correlation value having the largest value, and determines the congestion status represented by the template image corresponding to the selected correlation value. The congestion state of the parking lot at the time is determined (step S6).

In this manner, the congestion state can be determined based on the state of the entire parking lot without paying attention to each parked vehicle. In addition, as described above, as an empty vehicle template image, a pseudo image in which a white line such as a parking space is drawn at each thickness in accordance with an actual target area is subjected to edge processing, and then an image subjected to blur processing is used. Therefore, the determination accuracy of the empty vehicle state can be improved to almost 100%.

In addition, instead of drawing a line with the thickness of each white line according to the actual target area, a white line is drawn with one thin line, and the image is subjected to expansion processing to make the white line thick. The same determination accuracy can be obtained even if a pseudo empty vehicle template image is used. FIG. 6 shows a method of creating the pseudo empty vehicle template image.

The method of creating the pseudo template image shown in FIG. 6 will be described. First, as shown in FIG. A pseudo image drawn with a single thin line is created. Then, after performing a dilation process as shown in FIG. 6 (b) on the pseudo image composed of one thin white line, as shown in FIGS. 6 (c) and 6 (d), the edge processing and the blurring described in FIG. After performing the processing, the image subjected to the blur processing is used as a pseudo empty vehicle template image.

[0030]

As described above, according to the method and apparatus for judging congestion status according to the present invention, the accuracy of judging the empty state which has been a problem in the congestion status judgment method of the prior application is almost 100%.
In addition to this, there is an effect that the inspection time can be shortened because comparison with a plurality of empty vehicle template images is not required.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a congestion state determination device according to the present invention.

FIG. 2 is a flowchart for explaining a determination operation of the block diagram of the above.

FIG. 3 is an explanatory diagram of a determination state when an ideal actual photographed image is used as an empty vehicle template image,
(A) is an ideal empty vehicle template image, (b) is a full template image, and (c) is a target image.

4A and 4B are explanatory diagrams of a determination state when an artificially created pseudo empty vehicle template image is used, wherein FIG. 4A is a pseudo empty vehicle template image, FIG. 4B is a full template image, and FIG. It is an image.

5A and 5B show an example of a method for creating a pseudo empty vehicle template image. FIG. 5A shows a pseudo image (original image) in which white lines such as a parking space are drawn at respective thicknesses in accordance with an actual target area, and FIG. b) is an image after edge processing, and (c) is an image after blur processing.

FIG. 6 shows another method of creating a pseudo empty vehicle template image, in which (a) is a pseudo image (original image) in which a demarcation line such as a parking space is drawn with one thin white line in accordance with an actual target area. , (B) the image after the dilation processing, (c) the image after the edge processing, and (d) the image after the blurring processing.

FIG. 7 is an explanatory diagram of a congestion state determination method of the prior application,
(A) is an empty vehicle template image including an actual photographed image, (b) is a full template image, and (c) is a target image.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Camera (photographing means) 2 AD converter 3 Image memory (image storage means) 4 Edge image conversion part 5 Template image storage part 6 Correlation value calculation part 7 Judgment part

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Tetsuya Tanizaki 29-1 Otonoku Shinoda, Miwa-cho, Kaifu-gun, Aichi Prefecture F-term in Nagoya Electric Industries Co., Ltd. 5B057 AA16 BA02 CA08 CA12 CA16 CE04 DA06 DC16 DC34 5H180 AA01 CC04 KK06 5L096 BA04 CA02 EA02 FA06 HA09 JA03 JA09

Claims (6)

[Claims] 1. A bird's-eye view of a target area for which a congestion state is to be determined, and the taken target image is used for an empty vehicle, a congestion,
A congestion state in which a correlation value between each template image is calculated by calculating a cross-correlation with a template image representing each congestion state such as fullness, and the congestion state is determined from the magnitude of the calculated correlation value. In the determination method, an artificially created pseudo empty vehicle template image is used as the empty vehicle template image. 2. The pseudo empty vehicle template image is an image obtained by performing an edge process on a pseudo image in which a white line such as a parking space is drawn at each thickness in accordance with an actual target area and then performing a blur process. The congestion status determination method according to claim 1, wherein: 3. The pseudo empty vehicle template image is subjected to an expansion process on a pseudo image in which a partition line such as a parking space is drawn with one thin white line in accordance with an actual target area, and then subjected to edge processing. 2. The congestion state determination method according to claim 1, wherein the edge image is an image obtained by performing a blurring process. 4. A photographing means for bird's-eye view photographing of a target area for which a congestion state is to be determined, an image storing means for storing a photographed image of the target area photographed by the photographing means as a luminance image, and storing the image in the image storing means. An edge image conversion unit that converts the stored luminance image of the target area into an edge image; a template image storage unit that stores in advance a template image representing each congestion state such as empty, congested, and full, and the edge image conversion unit A correlation value calculating unit that calculates a cross-correlation between the image of the target area converted into the edge image and all template images stored in the template image storage unit, and calculates a correlation value with each template image; A determination unit that compares the correlation values calculated by the correlation value calculation unit and determines the congestion state of the target area based on the magnitude of the correlation value; As unladen template image to pay, congestion status determining apparatus characterized by using a pseudo unladen template images created artificially. 5. The pseudo empty vehicle template image is an image obtained by performing an edge process on a pseudo image in which a white line such as a parking space is drawn at each thickness in accordance with an actual target area and then performing a blur process. The congestion status determination device according to claim 4, wherein: 6. The pseudo empty vehicle template image is subjected to an expansion process on a pseudo image in which a partition line such as a parking space is drawn with one thin white line in accordance with an actual target area, and then subjected to edge processing. 5. The congestion state determination device according to claim 4, wherein the edge image is an image obtained by performing a blurring process.