JP3384033B2 - How to monitor parking space - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention
Large-scale parking lots attached to a ket, department store, amusement park, etc.
How to accurately and quickly detect free space in a car park
About. The parking lot has one car lined up in rows and columns.
A set of unit spaces that can be occupied. This simple
The place space is framed on the ground and visually
It has a great spread. One car in this space
If occupied, this is called the occupied space and
If no vehicle is parked, this is called empty space.
And [0002] When entering a new car, empty space
And have to guide the car here.
In the case of a small parking lot, open space by visual inspection
Easy to find. Empty space even if large
If there is a lot of pace, visually detect empty space
it can. However, the parking lot is large and the number of parking lots is large.
In some cases, it is difficult to find empty space visually
You. Because the car you ’re in will block your vision.
You. In this case, where is the empty space left?
Not easily understood. If you do not know, guide the car properly
Can not do it. In large parking lots, the detection of empty spaces
It must be done quickly. Building a tall watchtower or window
It would be possible to constantly monitor the space.
But relying on the human eye is less reliable and lowers labor costs.
It will be expensive because it is expensive. Not the desired way. There
And free by mechanical, electrical or optical means
Space and occupied space.
You. [0005] A method of detecting an empty space and an occupied space
For example, the following can be considered. (1) Detection by mechanical switch This means that a gate is provided for each space,
Open and close the gate to enter the unit space
It is. A mechanical switch that turns on and off by opening and closing the gate
A switch is provided. Depending on the direction of opening and closing the gate
Can be distinguished. Or if the car is taking up space
Mechanical switches that are turned on as far as
It may be h. However, in this method, it is the number of unit spaces.
Gates and switches must be provided. Moreover
Many to transmit these gate signals to the central monitoring office
Need transmission lines. With these shortcomings,
Nevertheless, it has never been applied as a parking guard. (2) For optical sensors (laser light, infrared light)
A light emitter and a receiver are installed on both sides for each detection unit space.
You. When the car enters, the light is blocked.
Absent. This makes it possible to accurately determine whether a car is parked.
You can know. But this also has many unit spaces
Each time an optical sensor must be installed. Also many
Several signal lines are installed from the unit space to the monitoring office.
Must be placed. A lot of capital investment for this reason
Is required. This is also useful as a parking lot monitoring device.
Never used. (3) A transmitter and a receiver for a detected radio wave by electromagnetic waves (such as electromagnetic induction) are installed in each unit space.
It is a thing. When transmitting radio waves such as microwaves,
Because of its short wavelength, it can go straight. By this
Can recognize the presence or absence of a car. But this too
It requires a number of sensor groups and transmission lines. Present in terms of cost
Not practical. (4) Detection by an ultrasonic sensor This is performed by connecting an ultrasonic oscillator and a sound meter for each unit space.
It is to know the existence of a car. This is (1) ~
As in the case of (3), the cost is high and is not realistic. As described above, the existence of a car for each unit space is
The non-detection method has never been used.
Such a sensor for each unit space of a vast parking lot
Installation is not very difficult, but there are many transmission lines
Must be laid. In addition, sensors and transmission lines
It also hinders the running of moving vehicles. Also, if the number of sensors is large,
The probability of disability increases and management is troublesome. (5) Detection by Image Processing It is desirable that the free space can be monitored collectively. This
One or more TV cameras in the parking lot
To monitor the status of each unit space.
available. Such a comprehensive method is
Even if the number increases, the equipment cost does not increase so much. Therefore
Is the only realistic way to monitor parking lots
Conceivable. However, optically, the existence and non-existence of automobiles
Even if you monitor the presence, even if you take a picture with a TV camera
Why do you handle it? This is the problem. TV set
When the observer looks at the image and observes the state of each space
So it ’s not reliable, and it ’s good in terms of labor cost.
Absent. After applying some image processing, each unit
I want to know the pace situation. (A) For example, Japanese Utility Model Laid-Open No. 1-175400
Is equipped with a TV camera that captures the entire parking lot.
Analyzed the image and painted it on the ground in the unit space with paint etc.
I try to see if the numbers are visible. Unit spec
Write the numbers 1,2 ‥‥ in large numbers with paint.
Keep it. When the car is parked, the numbers are hidden so you can see
Absent. If no car is present, you can see the numbers. This difference
Space and occupied space. [0014] This is named as a free space recognition method.
I can do it. Paint on the ground instead of focusing on cars
The focus is on the numbers that have been set. This is a unit space
Because the number that should be for each is determined, this
And if this number is visible, it can be determined that it is empty. this
Can be identified by only one TV camera in principle
Therefore, there is an advantage that the installation of the device is easy. [0015] However, this has the following disadvantages.
is there. If you have one or two TV cameras,
A tall car is in the Nth space
And the adjacent space in the opposite direction to the TV camera (Mth
Is not clearly visible. This is the ground
Must be observed. Mth is empty
I don't know if it's space or occupied space. Also, one camera can save a lot of space.
In order to be able to monitor, the camera must be installed at a high position
No. In addition, disturbance signals due to light reflection
This is sometimes mistaken for numbers. Also, from the vehicle
If the brightness and color information of the reflected light is similar to the ground,
And the ground. Also, there is no height on the ground
Possibility of misrecognition as a vehicle when there is a strange object (such as paper)
There is. [0017] Due to these disadvantages, free space
The recognition method is considered to be unreliable. Actually parking
It has never been applied as a surveillance method. This is the reflected light
This is because image processing is performed based on color information and luminance. Earth
It is not unlikely that surfaces and automobiles cannot be clearly distinguished
It is possible. (B) There is also a method of recognizing the presence of a car
sell. This is a pre-determined sample point (s
Brightness, color information and car reflected from one point in the pace)
Compare the brightness and color information known as the reflected light from both models.
It is to compare. In other words, cars have a finite shape.
There are finite colors, but all of these
Brightness and color information of the reflected light
Let me know. Is this a sample point in unit space?
The brightness and color of the reflected light are compared. This maximizes the use of TV camera information.
One for each detection system
All you need is a TV camera. Easy installation of equipment
You. The benefits of such a comprehensive system are expected.
You. However, on the other hand, let's use the information of reflected light to a high degree
From the optical signal containing a lot of noise
May extract incorrect information. Even if cars of the same color, the position of the sun
The reflected light intensity,
Different wavelengths. Also depending on the position of the sun
Intense reflected light may enter the camera. Many
The reflected light information for all colors and models
Since it must be entered in the base,
It takes a long time to make the source itself. Also, this is
The brightness and color information of the emitted light must be matched
No, but there is a lot of data, so it takes time to match
You. Also, if there is a vehicle with a high vehicle height, it will be close to this.
It is possible that the bordering space may not be visible.
For this reason, a car that detects the car itself
Neither recognition method is actually used. [0022] A large number of unit spaces are required.
Because it is installed in a large parking lot with
For example, if you install a detector for each source,
Cost and transmission line laying cost
Not. By optical means that can observe the entire space
Is influential, but image processing from image information
The method of discriminating the car itself and the ground itself may cause malfunction
High performance. These take advantage of the colors and brightness contained in the light information.
Because it uses. The color, brightness or lightness is light
And the most important attribute. But this is
Varies significantly depending on the weather and time. The object itself and the reflected light
It is different. Sky that does not malfunction due to weather, time, etc.
A space detection method is desired. Also large paper on the ground
Is not mistaken for a car even if it has fallen
Must. Higher car next to it
Even if you are present, if it is not a method that can detect free space
No. In other words, equipment costs are low and reliable
It is an object of the present invention to provide a pace detection monitoring method.
You. According to the present invention, an empty parking lot space according to the present invention is provided.
-The surveillance method uses two TV cameras,
Image the area with the left and right or top and bottom TV cameras, and
Data and divide it into predetermined unit spaces.
I do. This divided area (sampling area is called
), One or more feature points are extracted. The feature point is
It is defined as the point where the luminance and color information of the image data changes.
You. This converts the luminance or color information into a numerical value and
Extracted by first or second derivative
You. In the image data of the two cameras, the same
Match feature points. This is a small unit space
Therefore, it can be easily performed. Now you can handle feature points
Then, the difference between the coordinates of the feature points in the two image data
Call Since the distance between cameras and the focal length are known, three
The distance to the feature point can be calculated from the principle of angle measurement. image
Since the coordinates of the feature point at are known, this and the distance
Can be used to calculate the three-dimensional coordinate components of the feature points. In the three-dimensional coordinates, a threshold at which a height component exists
It is determined whether the value is equal to or less than the value. Height component
Is greater than or equal to that threshold, there is a car here
judge. If below threshold, this is free space
Is determined. Thus, the method of the present invention has two (two or more)
May be used), and feature points are extracted from the feature points.
Obtain the distance and calculate the three-dimensional position, and set the height Z to a certain threshold.
The presence or absence of a car at a certain position by comparing
Is what you want. FIG. 1 shows a schematic configuration of the system of the present invention. Big
Large parking lots have many vertical and horizontal unit spaces.
I do. The arrangement of unit spaces is regular. Also parking
Multiple TV cameras in a position overlooking the whole
Is installed. If there is a unit space, leave it empty
Others have cars parked. Also, the unit
In addition to the pace, there are other roads that cars must pass through. Exchange
The difference point may be provided with a guide plate or a vehicle detector. Te
The input image from the Levi camera is processed by the image processing device.
The image is processed. In addition, an empty space that displays available space
A source display device is also provided. FIG. 2 is a diagram showing a processing flow of the present invention.
You. The image input unit 1 has a plurality of television cameras. this
Are installed at two or three different points. Again
The direction is also fixed. Therefore, the position of the screen and the
The position of the place space corresponds. In the image processing section 2, a plurality of television image data
Process data. Feature point extraction and image data of this feature point
The operation to find the coordinates above is performed. The control unit 3
Find the distance from the coordinates on the image data to the feature point,
Calculates the three-dimensional coordinates of feature points from coordinates on the image data
And whether the height Z is less than or equal to a certain threshold Q
Judge. If the height Z is smaller than Q, this means that the car
It does not exist in that space. Z is Q
Larger means that there is a car. Empty space
The presence of the pace and the location of the empty space
This is displayed on the space guidance display section. The control unit simply seeks an empty space
The vehicle can be automatically vacated here.
I try to guide to the pace. The vehicle guidance section is "turn right
Please, please go straight, please turn left.
It is a sign consisting of any wording, but this is at the turning point
It is. At the same junction, whether a car has passed
An optical sensor is provided to check for a failure. this is
Light emitting element and light receiving element are installed on poles on both sides of the road
Things. If the car passes, it will be shaded so
You. Such an optical sensor is provided for each unit space.
It is not necessary to provide it only for important branches. The same calculation is repeated even if there are three or more cameras.
Only in the case of two cameras.
The processing at each stage will be described in more detail below. This is shown in FIG.
Corresponds to the flow. Two cameras may be left and right
However, it is better to put it up and down to find the height Z clearly
No. [Initial setting] Cameras are set up and down, and
Adjust the optical axes of the lower two cameras. XY flat parking area
Let X = 0 and Y = 0 the point where the camera is located.
That is, two cameras are installed on the Z axis. The height of the ground is Z
= 0. The height of the camera is given by Z = U, Z = D
You. This is a constant. Camera looks at the entire parking lot
Position. If it is too large, a similar
Multiple cameras are installed to view the area divided for each camera.
Make sense. The camera is not rotated. Diagonally downward
Fix it. In this way all unit spaces in the parking lot
Shall enter the camera. Because of this, the image on the screen
The position of the element and the position of the unit space of the parking lot almost correspond
Will be. The unit space is a narrow surface of several meters x several meters.
You. On the camera image, this is [several pixels] x [several
Element]. Unit space image
The area inside is called a sampling area. This
That this is a sufficiently small area makes the method of the present invention
Make it highly accurate. [Image processing and transfer] From the imaging units of the two televisions,
The whole view is input. Sampling the input image data
The image data is divided into image data for each garea. Sampling
The image data for each rear is sequentially transferred to the processing device.
You. [Processing in Each Processing Apparatus] Find feature points from luminance and color information in the sampling area
Extract. A feature point means that brightness or color information is spatially
Defined as a changing point. This is luminance or color
Report by first or second derivative
You. Extract feature points in image processing and represent the points
Means for finding a position as a point are already widely used.
However, in that case, there are two images and the same feature between them
Matching points is often difficult. It
Is generally the area where feature points should exist between the two images.
There is no limitation. If there are multiple feature points,
It can be wrong. However, in the case of the present invention, the sampling area
Since the correspondence between them is determined in detail,
Error does not occur. Number the unit space of the parking lot
Add this to 1, 2, １，２i, ‥‥ N. N is a number
Tens to hundreds. The screen is also sampled N times
Divided into areas. The i-th sump on the TV screen above
Let the area of the ring area be Hi. That is, the image is H ₁ ,
H _Two , ‥‥ H _N Is permanently divided in advance like this
You. Even on the TV screen of F, L ₁ , L _Two , ‥‥ L _N That
The screen is divided like this. Feature point extraction for each sampler
One or two are obtained in the calling area. Part of narrow range
It is extremely easy to associate the feature points of the divided regions Hi and Li.
No mistake. However, there is something to be aware of. On the parking lot plane
Sampling area is strictly three-dimensional
Division. Sampling area diagonally from camera
If the height Z is different, the same sampling
Even if the point is in the rear, the position may change slightly on the camera image.
Come over. That is, it belongs to the i-th sampling area
Hi on the screen is an adjacent area
That is, they have a degree of overlap. This is the number
It is different from the normal academic division (in this case,
There is no intersection between the divided areas). The part whose height from the parking scene is up to E is adopted.
Assuming that the horizontal distance from the camera is R,
The difference from the height of LA (Z = U, D) is from UE, DE
Each sampling area includes the area from U to D
Therefore, the angle in the vertical direction is: tan ^-1 (U / R) -tan ^-1 {(U−E) / R} (1) The following difference occurs. In this way, the vertical expansion
The image area is completely divided.
And the adjacent regions Hi and Hj partially overlap. 2. On the two television images, the ith (i =
(1,2, ‥‥ N) corresponding to the feature points of the divided areas Hi and Li
Let it. The divided area is small enough and one or two
Since only the feature points of the degree are required,
It is easy to find only the mark. One person for each divided area
When finding only one feature point, there is no need to support feature points.
No. Find multiple feature points for each sampling area
In these cases, these feature points must be matched on two images.
I have to. The means for coping is the brightness of two images.
By the correlation value of LDH within the sampling area
(Frequency of local parallax), geometry using redundant system
Either method, such as a correspondence by a strategic arrangement, may be used. 3. Feature points are paired in each sampling area.
Then, the distance to each feature point is calculated. This is a triangle
The principle of surveying may be used. Distance between cameras, on screen
Uses feature point coordinates, camera direction, magnification, focal length, etc.
Can be calculated. Set the center point of the camera baseline to M
And the distance between the cameras is S,
The angle between the straight line and the base line is φ ₁ , Φ _Two Then, the feature points and
The horizontal distance R from the base line is as follows: R = S (cotφ ₁ -Cotφ _Two ) ^-1 (2). 4. On the screen, the horizontal distance R to the feature point
And the three-dimensional coordinates (X,
Y, Z). On the images of the two cameras, the divided areas Hi and L
i corresponds. One or a few feature points
The coordinates on the image are (previous in Hi, Li)
It is required in 2 above. This is used in the calculation of 3.
You. Since the camera is set up and down, the horizontal position on the screen
Direction coordinates should be almost the same,
different. Since the light goes straight,
The angle φ between the coordinate and the base line of the feature point is 1: 1
Should be supported. After all, from the values of the vertical coordinates,
The horizontal distance R to the feature point can be obtained. Ma
The horizontal coordinates on the image correspond to the angle φ around the baseline.
Therefore, the X and Y coordinates of the feature point can be obtained by the following equation: X = Rcosφ (3) Y = Rsinφ (4) Originally this feature
Is known to be in a certain sampling area
Therefore, X and Y are included in the i region. X and Y coordinates
The value is not very important. i-th sampling area
If we write Ai, it is certain that (X, Y) ∈Ai
You. What is important is the Z coordinate. Lower camera coordinates
Is (0,0, D) and the coordinates of the upper camera are (0,0, U).
(UD = S). And connect the feature points with the camera
The line is aligned with the baseline (the line connecting the cameras, in this case, coincident with the Z axis).
Φ ₁ , Φ _Two Z = D− (Scotφ _Two ) (Cotφ ₁ -Cotφ _Two ) ^-1 (5) or Z = (Dcotφ ₁ -Ucotφ _Two ) (Cotφ ₁ -Cotφ _Two ) ^-1 (6) or Z = (Dtanφ _Two -Utanφ ₁ ) (Tanφ _Two -Tanφ ₁ ) ^-1 (7) It can be obtained by: This is also
Of course, it is smaller than the sampling height E (Z <E). 5. This Z coordinate is predetermined
It is determined whether it is larger or smaller than a threshold value Q. Q is a car
Lower than the height of the ground, higher than paper or cloth laid on the ground
Set to a value. The value is about 0.5 to 1 m. And Z
If <Q, a car is located in this sampling area Ai.
It is determined that it does not exist. If Z> Q, sampling
It is determined that a car exists in the rear Ai. Thus empty
The space and the occupied space are determined. About one sampling area Ai
If multiple feature points are to be taken, threshold
Examine the magnitude relationship with Q. If Z <Q for all
It is determined that there is no car and that there is an empty space. either
If Z> Q for one, this is the occupied space
judge. 6. This operation is performed for all sampling
We perform sequentially for area. The data obtained in this way is
Transfer to the data integration part. 7. Finally, in the data integration section,
The availability of all parking spaces is integrated from the results determined in
Combine and record. It can also be displayed on a display device. What has been described above, two cameras are arranged vertically
Met. Height Z is a problem
Arranging them is the most accurate. But high resolution, high resolution
If you have two cameras in the horizontal direction, the same calculation
It can be carried out. Also, you may have three instead of two
No. In this case, the calculation amount is tripled, but the accuracy is
improves. The present invention uses a system as shown in FIG.
Find the free space by processing like 2.
This is displayed. For each sampling area
To extract feature points on two or more camera images and perform triangulation
The coordinates are obtained according to the principle described above. And the Z coordinate
Of empty space by whether or not is larger than threshold Q
I do. Divide camera image by sampling area
{Hi}, {Li}, but this division is partially
There is an overlap. Claim that this is a special split
Revealed. In addition, the sampling area where the feature point extraction is narrow
It is explained that it is very easy because it is divided into
Was. In addition, the direct subtraction from the vertical coordinate of the image to the feature point
It is also possible to know the oblique angle of the line with respect to the base line (Z axis).
Revealed. In a sampling area with this bevel, its value
Because it is almost decided, by computer
In processing, the range of the oblique angle can be limited, so the processing time
Can be reduced. It is assumed that an empty space can be identified in this way.
You. The present invention is the one before, but the empty space is
Automatically guide car to empty space after finding out
be able to. FIG. 4 is a diagram showing a processing flow of the vehicle guidance unit.
It is. 1. First, as an initial setting,
Receive parking status. This is 1,2, ‥‥ N
Each unit space in the eye is occupied or empty
It represents the difference. Empty space on signboard
Displaying is also effective. 2. When a new vehicle enters the parking lot,
Work. Is this from the guidance device at the entrance of the parking lot
You. Photoelectric switches are provided facing each other, and the car
The switch opens and closes when you pass, so you can understand the entrance of the car
You. 3. Optimal parking space for entering vehicles
To determine This is of course one of the free spaces
Yes, but there are usually many empty spaces. Most
The appropriate parking space depends on certain criteria.
Is selected from among the available spaces. For example,
Give priority to the one closest to the doorway, or
Is a criterion of preferentially selecting distant objects. 4. Once the optimal parking space has been determined,
Determine the route from the mouth to the space. 5. For the driver of the approaching vehicle, the location and number of the optimal parking space
By displaying the number and the route to it on the guidance display
Let me know. 6. The optimal parking space is determined and the car
Is going to be there
And the same as being occupied. There parking lot use letter
Update status here. 7. The display of the guidance signboard at the entrance is also changed. 8. The car is exactly as instructed
Must be monitored for access to For this reason
The progress of the vehicle in the parking lot
Detect using a sensor. This is an infrared
Line sensors and cameras. 9. Installed at each intersection in the hall according to the progress of the vehicle
A sign indicating the direction of travel to the vehicle
You. 10. Vehicle enters optimal parking space
And the space occupied by multiple TV cameras
It becomes clear that it is. This allows you to
Check the changed parking lot usage status and change of the guidance display board. 11. The vehicle is not in the designated optimal parking space and other
It may be possible to stay in an empty space. in this case
This can be understood from the image of the TV camera,
The pace is the occupied space. As the optimal parking space
Parking lot to return designated space to empty space
Correct the usage status and the signboard display. The present invention optically monitors the status of a parking lot.
However, all the color and brightness information in the image data is
Not used in all aspects. Varies depending on the weather, season, and time
If you use a remarkable product, it malfunctions and it can not be used
Absent. In the present invention, the entire parking lot is used by using two or more cameras.
Capture the scene. This image corresponds to the unit space in advance
Divided into sampling areas,
Extract feature points in a. The same feature points on the two images are made to correspond,
From the pixel value (vertical and horizontal numbers), the three-dimensional position of the feature point
To determine whether the height Z is larger or smaller than a certain threshold Q
I do. Feature point correspondence is precise regardless of weather, season, and time
Can be done. These are the image information itself (color
(Chroma, brightness, brightness)
It does not depend much on the sun or the sun altitude. The height of the feature point in the unit space is calculated.
Even if paper, cloth, etc. fall on the ground,
It cannot be mistaken for the existence of a moving vehicle. Also the car itself
When trying to recognize
Have to store a lot of data
It takes time to create data and the processing time becomes longer. The present invention
Does not recognize the car itself, but the height of the feature points
Therefore, there is no need for preliminary data storage. There is little possibility of malfunction,
It is an excellent invention that can monitor the source quickly.
You.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram of a parking lot monitoring system of the present invention using a three-dimensional distance measurement method. FIG. 2 is a basic configuration block diagram showing a processing flow of the present invention. FIG. 3 is a processing flowchart of an image processing unit. FIG. 4 is a processing flowchart of a control unit, a display guide unit, and a vehicle guidance unit.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G08G 1/14 G06T 1/00 G08G 1/01 G08G 1/04 H04N 7/18

Claims (1)

(57) [Claim 1] In a parking lot having a large number of unit spaces divided vertically and horizontally so as to be able to park one car at a time, an occupied space where the car is parked, A method for determining an empty space that has not been completed, comprising: installing two or more television cameras at different positions that can overlook the inside of a parking lot, and capturing all unit spaces of the parking lot with the television cameras. Each of the images is divided into unit spaces to form a sampling area, and one or a plurality of feature points are obtained for each sampling area by performing spatial differentiation of luminance and color information, and are included in images obtained from different television cameras. The same feature points of the same subject are associated with each other, and the three-dimensional coordinates of the feature points are obtained from the pixel coordinates in different images of the same feature points. Compared to a threshold, determines that the empty space if the threshold value or less, parking free space monitoring method being characterized in that so as to determine that the space occupied by equal to or greater than the threshold value.