JPH01199216A - Discrimination method for presence/absence of driving mark - Google Patents

Abstract

PURPOSE:To avoid such a trouble where the drive of an unmanned carrier is stopped due to the absence of a driving mark by changing the processing areas for inclusion of a driving mark in case said mark is not included in the processing area due to such a trouble that is naturally recovered. CONSTITUTION:A CPU 11 carries out in order an image pickup processing, a recognition processing, an arithmetic processing, and a drive control processing respectively. Then the CPU 11 reads out the lightness data on a group of picture elements contained in an area e1 including no mark to obtain the picture element of the highest lightness, i.e., the maximum value MaxE1 in case an ambiguous mark or the presence of the mark is decided in the recognition process. At the same time, the picture element of the highest lightness, i.e., the maximum value MaxE2 is obtained from the lightness data on a group of picture elements contained in an area e2 including the mark. The absolute value of the difference between both maximum values is compared with the reference value DELTAQ1. When said absolute value larger than or equal to the value DELTAQ1, the inclusion of a mark 6 is decided for a process area (e). Then the drive control is performed.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a method for determining the presence or absence of a driving mark for determining the driving route of an unmanned vehicle. , images are taken at least multiple times with the R imaging device, the marks in the images taken at each time are recognized, the driving route of the unmanned vehicle is determined and sequentially updated, and the unmanned vehicle is driven along the updated driving route. In particular, the present invention relates to a method for determining the presence or absence of a driving mark, which determines the presence or absence of a mark in an image that is captured from time to time.

[Prior Art] The present applicant predicts the position of a driving mark on a road surface in an image captured by an imaging device mounted on an image-based unmanned vehicle from time to time, and based on the predicted position, the same image is By setting an area that includes the mark inside and recognizing the mark only within that limited area, the image processing time is shortened, the influence of noise other than the driving mark is reduced, and accuracy and high viscosity are achieved. We have proposed an image processing method that recognizes driving marks.

[Problems to be Solved by the Invention] In this image processing method, it is possible to immediately perform the recognition process for the driving mark and calculate a new driving route when the processing area is set. First, it is necessary to determine whether a mark exists within that area.

In this case, if it is determined that there is an abnormality based on the result of determining the presence or absence of a mark only once, there may be a large loss if measures are taken immediately, such as stopping the unmanned vehicle.

For example, a part of the mark may slightly deviate from the processing area due to a problem that would naturally repair itself, such as a sudden disturbance, and if the processing area is large, it will definitely be included, or the second image capture may cause the mark to be reliably included. Immediately stopping an unmanned vehicle in a case where the vehicle is affected by the accident would be a huge loss considering the enormous amount of effort and time required for recovery work.

The purpose of the present invention is to change the processing area so that the mark is included when there is a problem such as a sudden disturbance that naturally recovers and the mark is not included in the processing area so that the mark does not exist. It is an object of the present invention to provide a method for determining the presence or absence of a driving mark, which can prevent unnecessary stopping of an unmanned vehicle based on the above.

[Means for Solving the Problems] In order to achieve the above object, a first invention provides a first coordinate system whose origin is an old point at which a driving mark placed on a road surface was first imaged. is set, the position of the mark is determined in the first coordinate system, a traveling route is determined based on the previously imaged mark, and the pupil image fold is determined in which the next imaging operation is performed for the same mark on the traveling route. Find the point, then set a second coordinate system with the new point of the R image as the origin, transform the mark position in the first coordinate system to the position in the second coordinate system, and I based on the transformation position and mark size
In an image captured at a new imaging point, a processing area for image recognition that is smaller than the same image and includes a mark in the image is set, and then it is determined whether or not the mark is included in the processing area. A method for determining the presence or absence of a running mark, in which, if the mark is not included, the processing area is enlarged, and it is again determined in the enlarged processing area whether or not the mark is included.

The second invention sets a first coordinate system whose origin is the old point at which the driving mark placed on the road surface was first imaged, and the position of the mark is determined in the first coordinate system. At the same time, a traveling route is determined based on the previously imaged mark, an imaging corner point at which the next imaging operation of the same mark is performed on the traveling route, and then a second imaging point with the new imaging point as the origin is determined. A coordinate system is set, the position of the mark in the first coordinate system is transformed into a position in the second coordinate system, and an image is captured at a new point of the i-image based on the transformed position and the size of the mark. In the image, a processing area for image recognition that is smaller than the same image and includes the mark in the image is set, and then it is determined whether or not the mark is included in the processing area, and if the mark is not included. In this case, the travel route is used as it is, and then the new point of the i-image at which the image movement is performed is determined, a new second coordinate system is set, and the mark in the first coordinate system is The position is transformed into a new position in the second coordinate system, and based on the transformed position and the size of the mark, a new image that is smaller than the same image and includes the mark in the image is recognized. A method for determining the presence or absence of a running mark, in which a new processing area is set for a running mark, and whether or not the mark is included in the new processing area is determined again.

[Operation] In the first invention, the imaging corner point at which the next imaging operation is performed is determined by a function of the travel route determined based on the previously imaged mark. Then, the mark position of the first coordinate system whose origin is the old imaging point taken previously is coordinate-transformed into a second coordinate system whose origin is the obtained imaging corner point.

Next, using the size of the mark and the conversion position, a processing area is set in the image that is smaller than the image and includes the mark in the image, and it is determined whether the mark is included in the processing area. .

If a part of the mark is slightly out of the processing area for some reason, the processing area is enlarged. the result,
The mark can be completely included within the expanded processing area, and the mark recognition processing operation can be performed only within the expanded processing area.

Next, in the second invention, the processing area is set as in the first invention, and if it is determined that a part of the mark is slightly outside the processing area, the current driving route is used as it is. Next, the imaging corner point where the flu image operation is performed and a new second coordinate system are set for the mark, and a new processing area is set in the same manner as the previous processing area.

Therefore, even if it is determined that the mark is out of the processing area due to a repairable problem that occurred for some reason, the mark will be completely included in the newly set processing area in the next new imaging operation. This makes it possible to perform mark recognition processing operations only within the same processing area.

[First Embodiment] An embodiment embodying the method for determining the presence or absence of a running mark according to the present invention will be described below with reference to the drawings.

FIG. 2 shows a side view of the image-based unmanned vehicle 1, and a C0D (Charae c
A camera 3 is provided. The CCD camera 3 is mounted on the support frame 2 so as to photograph a preset area 4a on the road surface 4 in front of the unmanned vehicle 1. The CCD camera 3 has captured the area 4a shown in FIG. 3 as an image 5 shown in FIG. and,
In this embodiment, the image 5 is composed of 256×256 pixel data.

As shown in FIGS. 2 and 3, on the road surface 4, marks 6 made of a white color different from the color of the road surface 4 for determining the traveling route of the unmanned vehicle 1 are arranged at predetermined intervals. In this embodiment, as shown in FIG. 1, the shape is a circle with a radius r1, and a pair of pointed corners 6a extend from opposite sides of the circle. The knot connecting the pair of corners 6a indicates the direction of movement of the unmanned vehicle 1 when passing the same mark 6, and also indicates the direction of the next mark 6 in this embodiment.

Further, the length from the center (center of gravity) of the mark 6 to the cusp of the corner is defined as r2 (>rl). Then, while running this constant and discretely arranged marks 6, the C
The OD cameras 3 will dance one after another.

In this embodiment, the CCD camera 3 employs a camera in which the pixel signal of the white mark 6 portion has a high level, and the pixel signal of the dark road surface 4 portion has a low level.

Next, the electrical configuration of the travel control device mounted on the unmanned vehicle 1 will be explained with reference to FIG.

The microcomputer 10 includes a central processing unit (hereinafter simply referred to as CPU) 11 and a program memory 1 consisting of a read-only memory (ROM) that stores control programs.
2, a working memory 13 consisting of a readable and replaceable memory (RAM) in which arithmetic processing results etc. of the CPU 11 are temporarily stored, a timer 14, and the like. Then, the CPLJll operates the CCD camera 3 according to the control program stored in the program memory 12, determines the traveling route of the unmanned vehicle 1 based on the captured image information, and performs various calculations for steering control. It is designed to perform processing operations.

The CPLJII connects the input/output interface 15 and the A/
The CCD camera 3 is scanned and controlled via the D converter 16, and the pixel signals from the CCD camera 3 are converted into pixel data and brightness data via the A/D converter 16 and the bypass capacitor 1-○-ra 17. and stored in the working memory 13. Based on the control signal, the A/D converter 16 converts the pixel signal of a value (analog value) proportional to the signal strength of the pixel signal from the CCD camera 3 (the brightness of the road surface 4 imaged by Ifi) into 256 levels according to the brightness. The divided brightness data is digitally converted and stored in the working memory 13 via the bus controller 17. In addition, the A/D converter 16 is connected to the C
When converting pixel signals from the CD camera 3 from analog values to digital values, it is determined whether each pixel signal is greater than a predetermined set value, and if the pixel signal is greater than the set value, a white mark is displayed. 6 as a pixel, and conversely, if the pixel signal is less than r, as a pixel in a dark road 4 part.
Each pixel signal that is sequentially input as OJ is binarized and stored as pixel data in the working memory 13 via the bus controller 17.

Therefore, the working memory 13 stores the image 5 captured by the CCD camera 3 as 256×256 pixel data.

In this embodiment, for convenience of explanation, a scanning method is adopted in which the CCD camera 3 scans in the horizontal direction and the scanning moves from the top to the bottom of the image 5, but other scanning methods may also be used. Of course it's a good thing.

The binarization level controller 18 is adapted to output setting value data for the A/D converter 16 to perform binarization based on a control signal from the CPU 11 to the A/D converter 16.

The drive controller 19 controls the driving motor 20 and the steering mechanism 21 based on the same control signal from the CPU 11. The rotational speed of the traveling motor 20 is controlled based on the control signal, and the steering angle O3 of the steering mechanism 21 is controlled based on the control signal. In this embodiment, the unmanned vehicle 1 is run at a constant speed V except for starting and stopping, and the CPU 11 controls the rotational speed of the running motor 20 to rotate at a constant speed. are doing.

Furthermore, in this embodiment, the CCD camera 3 is controlled to perform an imaging operation at predetermined intervals T, and as shown in FIG. It is now possible to take images by changing the Therefore, as the unmanned vehicle 1 approaches the mark 6, the mark 6 in the image 5 captured by the CCD camera 3 gradually approaches the front and appears larger as shown in FIG. Further, in this embodiment, when the mark 6 is imaged three times, the mark 6 moves out of the area 4a and the next new mark 6 enters the area 4a when the next image is taken.

Next, the processing operation of the CPLJ 11 will be explained.

First, to explain the basic operation of the CPU 11, (, CD
An imaging processing operation that activates the camera 3, a recognition processing operation that performs image processing to recognize the mark 6 provided on the road surface 4 based on the image taken by the camera 3, and an unmanned vehicle driving based on the recognition result. There is a calculation processing operation for calculating a route, and a travel control operation for controlling the steering mechanism 21 based on the calculation result to cause the unmanned vehicle 1 to travel. The CPLlll performs the following steps in the order of imaging processing operation→recognition processing operation→arithmetic processing operation→driving control operation, and the operating times T1 to T4 of each operation are set in advance. Then, the unmanned vehicle 1 is made to travel by repeating each of these operations in sequence.

Note that the period from the end of the imaging processing operation to the arithmetic processing operation and the start of the driving processing operation (=TI +T2 +T3
), the CPU 11 controls whether the unmanned vehicle 1 is controlled by a travel control operation based on the travel route obtained in the previous arithmetic processing operation.
The driving is controlled.

To briefly explain in detail, as shown in FIG.
When reaching , the CPU 11 causes the CCD camera 3 to take an image, starts the Wi image processing operation, performs the recognition processing operation and the arithmetic processing operation, and determines the next new travel route L1.The time Ta (=TI +T2 +T3 ) later,
That is, the unmanned vehicle 1 travels from point PO to point P1 based on the travel route LO. Then, CPtJll starts a travel control operation based on the travel route L1 from this point P1, and controls the unmanned vehicle 1 to travel along the travel route L1.

When the unmanned vehicle 1 has traveled to point P2 after a predetermined time T4 has elapsed since the start of the travel control operation based on the travel route L1, C
The PU 11 starts the next imaging processing operation. And C.P.
U11 determines the driving route L2 based on the image information captured at the point P2 after the elapse of time Ta (when the unmanned vehicle 1 travels along the driving route L1 to the point P3), and starts the determined new driving route from the point P3. The unmanned vehicle 1 is controlled to travel along the route L2. Thereafter, by repeating the same operation, the CPU 11 calculates each travel route LO, Ll, L2, etc. for the unmanned vehicle 1 at each time.

The vehicle is designed to control travel based on the following. Therefore, C.P.
U11 is time T (=T1 +T2 +T3 +T4
), the vehicle is traveling while updating a new route every time.

Next, details of each processing operation will be explained.

In FIG. 9, the travel route LO for the unmanned vehicle 1 is calculated first.
When reaching the point PO in a state where the travel is controlled according to the following, the CCD
When the camera 3 is scan-controlled, the CCD camera 3 scans the road surface 4.
Since the image is not taken perpendicularly to the object but tilted at a certain angle, the area 4a in front is imaged as an image 5 as shown in FIG.

The pixel signal corresponding to each pixel from the CCD camera 3 is digitally converted by the A/D converter 16 according to the signal strength, and is stored in the working memory 13 as brightness data. In addition, each pixel signal is compared with a preset value in the A/D converter 16, and the pixel signal of the mark 6 part and the pixel signal of the road surface 4 part are "1", and the pixel signal of the rOJ part is "2".
The pixel data is determined as a value and stored in a predetermined storage area of the working memory 13 as pixel data.

For convenience of explanation, it is assumed that the mark 6 imaged by VA at the point PO is the mark 6 imaged for the first time, and is imaged from the farthest position.

CPLJll performs image recognition of the mark 6 based on all pixel data stored in the working memory 13.

The CPU 11 selects mark 6 in this picture @5 using a known method.
The center position q of the portion determined to have the shape, that is, the position G on the actual road surface 4 at which the center of gravity of the mark 6 is located is determined and stored in the working memory 13.

As shown in FIG.
Find points a, b, c, and d. For these four points, the 128th vertical pixel column counting from the left of each pixel forming image 5 is defined as the y-axis, and the 128th horizontal pixel column counting from the top side is defined as the y-axis. It is expressed in X and y coordinates. In this case, since this mark 6 is the first imaged mark 6, the mark 6 is recognized by identifying the mark 6 in the image using a known method using all pixel data of the image 5. 4 points a, b, c, d
has been decided.

Next, the CPU 11 performs projective transformation on these four points a to d.

Projective transformation is carried out to determine which position on the actual area 4a shown in FIG. 9 (hereinafter referred to as base point>A-
D, that is, in this embodiment, the base point Δ~D is located at the center position of the unmanned center 1 (more precisely, the CCD camera 3) as shown in FIG.
The coordinate positions of base points A to D in the X, Y coordinate system are taken as the origin PH, the direction of movement of the unmanned vehicle 1 is the X axis, and the direction orthogonal to the X axis is the Y axis. Perform calculation processing to determine.

a (xa, ya)−+△(Xa, Ya)
b (xb, yb) → B (Xb, Yb)
c (xc, VC) → c (Xc, YC)d
(xd, yd) → D (Xd, Yd) As mentioned above, this is because the CCD camera 3 does not image the road surface 4 vertically, so the mark 6 in the image 5 is different from the mark 6 in the actual area 4a. This is the process of matching the

Note that this projective transformation processing operation is performed using a preset CCD.
Projective transformation, ie, coordinate transformation, is performed based on standards such as the focal length and magnification of the camera 3, and installation conditions such as tilt and height. The general formula for this projective transformation is as follows.

The position coordinates of each point are X, y, and the position coordinates of the base point are X.

Let Y, the height of the camera 3 be H1, the tilt of the camera 3 be θ, the focal length of the camera 3 be F, and the magnification constant be k.

Note that p is the distance between the center of the vehicle and the COD camera installation point.

After projective transformation, CPLJll calculates the coordinates (XIII, Yll) of the intersection of the line g connecting base points A and C and the line connecting base points B and D from the coordinates of base points △ to D as the center position of mark 6 @G. At the same time, the same mark 6 is calculated from the coordinates (Xa, Ya), (XC, YC) of base points A and C.
The slope Φm (which is the slope of the line 1 connecting the pair of corners 6a and indicates the direction in which the unmanned vehicle 1 moves) is determined.

In this embodiment, the center position G is defined by four points a-d in the image.
However, the midpoint of the line 1 connecting the two corner cusp points a and C is taken as the center position q of the mark 6 in the image, and the center position Q is transformed into the 9A shadow and set as the center position G. Good too. Alternatively, the center of gravity may be determined from the mark in the image, the center of gravity may be set as the center of the mark, and the center position G may be determined by projective transformation.

Next, a point P where travel control is started according to the travel route L1 determined based on the image information captured at the point PO.
1 and the coordinates (Xpl, YpI) of the same point P1 are determined. This calculation is a cubic function f (X
) can be easily determined using

Both positions G <XQ, Y(] ), Pi (XI
)1. Ypl) and its position G, slope Φm at Pi
, Φp1, the CPU 11 determines a travel route L1 that passes through both positions and is expressed by the following cubic function f (X).

f(X)=αX +βX2+γX+δ The coefficients α, β, γ, and δ can be easily obtained by solving the following simultaneous quartic equations. (below,
Margin) That is, YQ = αXg3 + βXg2 + γxg + δYp1 = αX
p13+βxp12+γXp1+δjan(Dllll
-3αX92-1-2βX(1+7'tanΦ1)1-
3αXI) 12+2βXrii+7CPLJ11 is this 3
The next function f (X) is determined as the traveling route L1 of the unmanned vehicle 1 passing from the point P1 to the mark center position G at an attitude angle Φm.

Then, the unmanned vehicle 1 follows the previous travel route LO to point P1.
When this is reached, the CPU 11 shifts to a travel control operation and controls the steering m mechanism 21 based on this function f (X). This control is performed from point P1 to function f (X>
In the control operation for moving the unmanned medium 1 along the curve of
, and unmanned vehicle 1 calculates its attitude angle (X
) The steering angle θS is determined so that the steering gear 4i
21 is operated and controlled.

Then, the arctangent of the differential value of the cubic function f (X) is the attitude angle (X) (=jan-1(f-(X))), and a certain point (Xn) on the traveling route L1.

f (Xn) ) to the next point (Xn+1, f (
When moving up to (Xn+1)), the attitude angle (X>
It can be seen that it is sufficient to satisfy the condition that Φ(Xn) to Φ(Xn+1'').

In this embodiment, a driving control method for satisfying this condition is implemented in steady turning circle driving.

That is, as shown in FIG. 12, steady turning circle running is running with a constant radius R when the steering angle θS is held constant, and the amount of change in attitude angle Φ(X) after ΔT seconds is defined as Δ
When Φ is assumed, the following formula holds true.

ΔΦ=V・θS・ΔT/W R=W/es ■ is the traveling speed, and W is the wheel base.

From both formulas, in order to change the attitude angle by ΔΦ while moving by V・ΔT, the radius R (=V・ΔT) is required for every 6 guns.
T/ΔΦ), and from its radius R the steering angle θ
s (=W/R=W・ΔΦ/■・6 pieces should be calculated.

By calculating the angle es based on the above formula and controlling the operation of the steering mechanism 21, the unmanned vehicle 1 can be driven along the travel route L1.

The unmanned vehicle 1 travels along the traveling route L1, and the unmanned vehicle 1 travels along the traveling route L1.
When T4 hours have elapsed since the start of the travel control operation based on (the point at that time is P2), the CPU 11 executes the next processing operation of the m-image device and operates the COD camera 3 in the same manner as above to display the area at that time. The mark 6 in 4a is imaged. Then, the processing operation to obtain the three engine numbers f (X>) of the next new travel route L2 is performed until reaching point P3.

In this case, unlike the above, all pixel data and all brightness data of the image 5 captured by the COD camera 3 are stored in the working memory 1.
3, the CPU II executes calculations to set a processing area e that completely includes the mark 6 in the image 5, as shown in FIG.

This setting calculation uses the point P2 as the origin from the X, Y coordinate system shown in FIG. Coordinate transformation to XI, Ym coordinate system as axes,
That is, the coordinate (X
IIJ, Yg) woXm.

Convert to coordinates (XmL Ym(II) in the ym coordinate system.

Therefore, in this case, the imaging concavity point is at point POStti, the image new point is at point P2, and the first coordinate system is at X.

The Y coordinate system and the second coordinate system are Xm and YI11 coordinate systems.

This transformation is performed using affine transformation, and the coordinate position (XmO, Ym(J)) of the mark center position Gm in the Xm, YI11 coordinate system is determined by the following calculation formula.

Xmg= (XIIJ -XI)2) CO5(-0m
)-(Yg-Yp2) sin (-0m)YmC
l = (xc+ -xpz) sin (-0m
) + (YO-'v'pz> COS (-0m) In addition, X
l)2. Yl)2 is the coordinate of point P2 in the X, Y coordinate system, and can be determined from the cubic function of the travel route L1.

Also, em is xm with respect to the Y axis of the X, Y coordinate system.

The rotation angle of the Ym axis of the Ym coordinate system, that is, the amount of change in the attitude angle of the unmanned vehicle 1 at point P2 with respect to the attitude angle of the unmanned vehicle 1 at point PO, which is determined from a cubic function of the traveling route L1. You can ask for it.

The coordinate position f2i of the mark center position Gm on the road surface 4 in the xIll, YIII coordinate system is
When m(1) is imaged at point P2, calculation processing is performed to determine which position in image 5, that is, which position in image 5 the mark center position q is located.

Gm (XIII(1, Y@(J) →Q (XO,
MO>This is a transformation process (
This can be obtained by performing a reverse projective transformation process. The general formula for this inverse 9A shadow transformation formula is as follows.

(Hereinafter, blank space) In other words, when the position 1xc+, ya) obtained by the inverse transformation is determined, the CPU 11 determines that the mark 6 in the image 5 is at the position Q(X(11
, yg), and then a calculation is performed to set the processing area e.

The calculation for setting the processing area e is based on the xm calculated above.

It is calculated based on the coordinate position (XmO, YIIlg) of the mark center position @Gm on the road surface 4 in the Yllll coordinate system.

This is because the size of the mark 6 on the road surface 4 (the length r2 from the center to the peak point of the corner) is stored in advance as data, and the shape of a rectangular frame that completely includes the mark 6 on the road surface 4 is A region Z is set in advance. The size of area Z is stored in the program memory 12 in advance so that it is a regular rectangle whose center of gravity is the center point of mark 6, and the - side is slightly longer than 2·r2, as shown in FIG. Based on that data, the CPU 11 executes Xm as shown below.
, Point Z at each corner of area Z on road surface 4 in Ym coordinate system
Each coordinate of a, Zb, ZC, and Zd is 0 (mza.

Ymza) (Xmzb, Ymzb) (
Xmzc, ``y'mzc) (>(mzd, Y
Find mzd).

Xmza =Xmg -r2-α1 Ymza =ymgr2 +CX1 Xmzb =Xmg -r2-cl Ymzb =Ymg -r2-(21
=Ymg -r2-α1 Furthermore, the larger the value of α1 is, the higher the safety factor is that the mark 6 is included, but on the other hand, the image processing time, which will be described later, becomes longer.

Next, the CPU II calculates the coordinates (X Za.

yza) (xzb, yzb) (xzc, y
zc) (xzd.

yzd) is obtained by inverse projective transformation in the same manner as above. The range connecting these points becomes the processing area e as shown in FIG. 16, and the area e is a horizontally long rectangular area that reliably includes the mark 6 to be imaged. Furthermore, this area e
becomes longer laterally because the COD camera 3 images the road surface 4 obliquely, causing distortion in the X-axis direction.

As mentioned above, when setting the processing area e, it can be set without finding the coordinates of the center position Q of the mark 6 in the image. No action is required.

When the processing area e is set, the CPU II executes mark presence/absence determination processing.

First, as shown in FIGS. 18 to 20, the CPU 11 sets an outer peripheral frame-shaped non-existence area e1 consisting of a group of pixels surrounding the outer periphery of the processing area e, and at a predetermined interval within the processing area e. An intersecting frame-shaped existence region e2 is set, which is composed of three vertically extending pixel groups and a horizontally extending pixel group perpendicular to the three pixel groups.

That is, since the non-existent area e1 is a part surrounding the outer periphery of the processing area e, the CPU II accurately recognizes the image, accurately calculates the travel route L1, and the unmanned vehicle 1 travels along the travel route L1. Once the COD camera 3 has reliably performed the imaging operation, the area becomes a region in which there are no pixels having brightness data corresponding to the mark 6. Therefore, the brightness data of the pixel group in the non-existent area e1 must all have data corresponding to the road surface 4.

As a result, the pixel group of the non-existent area e1 is included in the range of the mountain on the left side indicating the brightness of the road surface in the histogram of each signal strength (brightness) of all pixels of the image 5 including the mark 6 shown in FIG. Brightness data with strong brightness (hereinafter referred to as “maximum value M
axE1J), it is up to the right end of the mountain on the left, and the lightness data with the weakest brightness (hereinafter referred to as "minimum value MinEIJ") is up to the right end of the left mountain.
) But it is up to the left end of the mountain on the left.

On the other hand, since the pixel group in the existence area e2 is vertically traversal within the processing area e, there is always a pixel with brightness data corresponding to mark 6, and there is always a pixel with brightness data corresponding to road surface 4. exist. As a result, the existence area e
Pixel group 2 is included in the entire range in the histogram shown in FIG.
The maximum value MaxE2J) is at the right end of the right-hand mountain indicating the brightness of mark 6, and the lightness data with the weakest brightness (hereinafter referred to as the minimum value M inE 2 J) is up to the left end of the left-hand mountain. .

The CPU 11 reads the brightness data of the pixel group included in the non-existent area e1, and selects the pixel with the strongest brightness, that is, the maximum value Ma.
Find XE1. Similarly, the CPU 11 is located in the existence area e2.
The pixel with the strongest brightness, that is, the maximum value MaXE2, is determined from the brightness data of the pixel group included in the pixel group.

Subsequently, the CPU 11 changes the maximum value Ma from the maximum value MaXE1.
The absolute value of the value obtained by subtracting XE2 and the preset reference value ΔQ
Compare with 1. Then, it is determined that I MaxE 1-MaxE 21≧ΔQ1 is completely included. That is, the mark 6 is completely included in the processing area e, as shown in FIG. Maximum value MaXE2
is at least to the right of the left end of the right-hand mountain, so it can be seen that the absolute value of the difference is a large value, that is, greater than the reference value ΔQ1 determined in advance through tests or theoretically. Also, l
MaxE 1-MaxE 21 <ΔQ2(kuΔQ1)
At this time, it is determined that the mark 6 is not completely contained within the processing area e as shown in FIGS. 23 and 24. That is, as shown in FIG. 23, when the mark 6 is completely outside the processing area e, the maximum value M of the non-existing area e1
Both aXE1 and the maximum value MaXE2 of the existence area are on the left mountain, and if part of the mark 6 exists in the processing area e as shown in FIG. 24, the maximum value MaXE1 of the non-existence area e1 and the existence area Since both the maximum values MaXE2 are located on the right side of the mountain, it can be seen that the absolute values of the differences are both small values, that is, less than the reference value Δ02 determined in advance through tests or theoretically.

Therefore, the non-existent area and the existing area e1. It can be determined whether or not the mark 6 is included in the processing area e using only the brightness data of e2, and the time required for the determination process is extremely short.

In addition, in this embodiment, the non-existent ri area and the existing area e1゜e2
was set to determine the presence or absence of a mark, but this was done in processing order 10. The center of gravity of mark 6 is determined based on the pixel data corresponding to mark 6 in e, and the presence or absence of mark 6 is determined from the deviation between the center of gravity and the mark center position q determined by the above-mentioned inverse projective transformation. In short, any method may be used as long as it can determine whether or not the mark 6 is included in the processing area e, such as recognizing the shape of the mark 6 in the area e and determining the presence or absence of the mark based on the recognized shape pattern.

If mark 6 is not completely included in processing area e for some reason, CPtJll is
The enlarged new processing order 1f1 is shown by the two-dot chain line in FIG.
．． Set a. In this processing order 1ry, the setting of the processing area e larger than e is performed at each point Za to Zd of the processing order vAe.
Coordinates of (xza, yza) (xzb.

yzb) (xzc, yzc) (xzd, y
Each point zla- to determine the enlargement processing area e using
This is done by determining z1d as follows. Note that ζ is the amount of expansion.

Za (XZa, yZa) → Z1a (XZa-ζ
, yza ten ζ) zb (xzb, yzb) → zlb
(xzb-ζ, yzb-ζ)zc (xzc,
yzc) →zlc (xzc ten ζ, yzc-ζ)zd
(xzd, yzd) −+z1d (xzd×ζ, yzd+ζ) Then, when the expanded processing area e is set, the CPU 11 sets the non-existing area e1 and the existing area e2 that match this processing area e in the same manner as above. Then, the presence or absence of the mark 6 is determined based on the brightness data of both ends '1jAe'l and e2. If the mark 6 is not included, a new processing area e@ stage is set in the same manner as described above.

The number of times the processing order IIAe is expanded is not particularly limited in this embodiment, but the number of times is not particularly limited as long as the next new travel route L2, which will be described later, is devised at least before the travel control operation start time is reached.

Note that in this embodiment, the processing area e is expanded in all directions, but if a direction deviating from the processing area e can be determined during mark discrimination, only the direction in which it deviates can be enlarged. good.

When the macro is not included in the processing order vie even if the mark presence/absence determination process is performed multiple times, the CPU 11 determines that the travel route L2 cannot be calculated and immediately activates the travel motor 20 to stop the unmanned vehicle 1. Control the stop.

Therefore, even if part of the mark 6 deviates from the processing area e for some reason, it will not be immediately determined that there is an abnormality and the unmanned vehicle 1 will not be stopped based on the abnormality. As a result, when it is determined that the mark 6 is included in the enlarged processing area e, there is no need to perform unnecessary recovery work to prevent the unmanned vehicle 1 from traveling again.

In this embodiment, the processing area e is expanded while driving, but if it is determined that the mark 6 is not recognized, the unmanned vehicle may be temporarily stopped while the mark 6 is still included. In this case, the number of times of enlargement may be increased until the processing area e has the same human features as the image 5.

On the other hand, if it is determined that the mark 6 is completely included in a certain processing order Vie, the CPU 11
3, only the pixel data group in the processing area e is read out, and image recognition of the mark 6 that should be in the processing area e is executed. The CPU 11 determines the shape of the mark 4 within the area e using a known method in the same manner as described above, and determines the shape of the mark 4 at four points a including the cusp points a and C of the corner 6a of the mark 6.
Find the position of ~d in image 5.

Therefore, as in the past, the mark 6 is
This means that the processing time is significantly reduced compared to performing image recognition.

When the four points a-d in the image are determined, the CPU II changes the projection in the same way and calculates the unmanned vehicle 1 at each point G, A to D, and point G.
Find the slope of 0m. Further, cPUll determines the position of point P3 and the slope Φp3 at that time based on the cubic function f (X) of travel route L1.

Then, based on these eight values obtained, the next new travel route L2 formed by a cubic function is determined in the same manner as when the travel route L1 was determined.

At this time, CPLlll calculates the function f ( The running of the unmanned vehicle 1 is controlled based on the function f (X).

Then, when the unmanned vehicle 1 reaches the point P2, the CPU 11
controls the unmanned vehicle 1 to travel along the travel route L2 formed by the obtained new cubic function f (X).

Then, when the CCD camera 3 performs an imaging operation to obtain the next new travel route while traveling on the travel route L2, the CPU II performs the same processing operation as the method used to determine the travel route L2. Therefore, the CPU 11 first performs calculations to set the processing area e, and then determines the presence or absence of the mark 6 in the newly set processing area e and performs mark recognition. In this case, when setting the area vA setting e, we first set the area Z that includes the mark 6 on the actual road surface 4 and obtained it by projective transformation, so that the mark 6 is Even if the image gets larger as it approaches, the 17th
As shown in the figure, the processing area e is also enlarged to completely include the mark 6, and the presence or absence of the mark 6 can be detected and recognized with certainty. Therefore, in this case, once it is determined that the mark 6 is out of the processing area e, the processing order lii! of the size shown in FIG. 17! It will be expanded based on e.

Incidentally, in the case of this embodiment, when a traveling route is determined based on the mark 6 shown in FIG. 17 and the next image is taken while traveling on the same route,
The mark 6 will be removed from the imaging area 4a of the CCD camera 3, and the next new mark 6 will enter the area 4a.

Since it is difficult to predict at what position the new mark 6 will be imaged, the CPU 11 recognizes the image of the mark 6 using all pixel data only when the new mark 6 is imaged for the first time. Execute.

Thereafter, CPtJll repeats the same operation to sequentially update the traveling route and control the unmanned vehicle 1 according to the mark 6 captured in the image.

In this way, in this embodiment, even if a part of the mark 6 deviates from the processing area e for some reason, it is not immediately determined to be an abnormal state and the unmanned vehicle 1 is not stopped based on the abnormality. Since the processing area e is enlarged and the presence or absence of the mark 6 is determined again, when it is determined that the mark 6 is included in the enlarged processing area e, there is no need to make the unmanned vehicle 1 run again. It is possible to prevent unnecessary recovery work.

Furthermore, in this embodiment, recognition of the mark 6 is made possible by using only the pixel data group of the processing area e that completely includes the previously predicted mark 6 in the image taken by the COD camera 3.
Image processing time can be shortened, and unmanned vehicles1
The speed can be increased. Moreover, since the processing area vAe is set and mark 6 is recognized, even if noise that does not impede driving is included in the pixel data group, as long as the noise is not included in the processing area e. Since it is not affected by the influence, highly accurate image recognition is possible, and in turn, highly accurate driving routes can be determined.

[Second Embodiment] FIGS. 25 and 26 are explanatory diagrams for explaining the second embodiment. In the embodiment, the processing area e is enlarged on the same two images 5 multiple times. Although the presence or absence of the mark 6 was determined, this embodiment differs in that the presence or absence of the mark is determined using a plurality of different images instead of determining the presence or absence of the mark multiple times on the same image. Therefore, only the different parts will be explained in detail,
The other parts are substantially the same as the first embodiment, so detailed explanation will be omitted.

That is, as in the first embodiment, when it is determined that the mark 6 imaged wL at point P2 is not included in the processing area e, the CPU 11 determines that the image cannot be captured due to some reason, such as a sudden disturbance. It is determined that the unmanned vehicle 1 is unable to do so, and the unmanned vehicle 1 is driven to the next imaging point P4 using the previous travel route L1. Until the point P4, the CPU 11 calculates the center position Qm of the mark 6 on the road surface 4 in the new Xm and Ym coordinates with the point P4 as the origin by performing coordinate conversion in the same manner as described above, and Set. In this case, as in the embodiment described above, the mark 6 in the image 5 is photographed larger because the mark 6 on the road surface is closer to the unmanned vehicle 1, and the processing area e that includes the mark 6 is also inevitably However, it may be expanded further.

If the mark 6 is included in the set processing area e, the CPU 11 calculates a new travel route L2 in the same manner as described above, and causes the unmanned vehicle 1 to travel along the new travel route L2. Therefore, even if it is determined that mark 6 is not included due to a repairable trouble 1 to 6 that has occurred for some reason, the mark will be completely included in the newly set processing area in the next new @image operation. This makes it possible to prevent unnecessary restoration work to make the unmanned vehicle 1 run again, as in the embodiment described above.

On the other hand, when it is determined that the vehicle is not included, it is determined that an abnormal state exists and the unmanned vehicle 1 is immediately stopped.

In this embodiment, the processing area e was set twice and stopped when the mark 6 was not included both times, but if the same mark is imaged many times and the traveling route is updated, it may be necessary to The number of times the presence/absence determination is made may be determined. or,
This embodiment may be carried out independently, but it may also be carried out in combination with the first embodiment.

Incidentally, in each of the above embodiments, a rectangular processing area e that includes the mark 6 is set, but as long as the shape includes the mark 6, it may be circular, oval, triangular, etc. It can be a single shape without increasing the size, and of course,
Of course, the shape of the mark 6 may be changed.

Further, in the embodiment, the center position Gm of the mark 6 on the road surface 4
The center position Gm is set as the center of gravity and the area Z@: is set in advance, and the processing area e is set by inverse projective transformation. 6 may be set. In this case, when setting the processing area e in the image, the image processing time can be further reduced by appropriately selecting the size of the area e including the margin α1. Note that as the mark 6 on the road surface approaches the unmanned vehicle 1, the mark 6 in the image becomes less crowded, so it is necessary to enlarge the processing area e according to the X component value of the mark center position q of the image 5.

Roughly speaking, in each embodiment, when setting the area Z or directly setting the processing area e in the image, a margin α1 is set to ensure safety, but this is set at the center position of the mark in the image 5. It may also be implemented by changing it depending on the X component value of q, that is, depending on the distance of the mark 6 on the road surface from the unmanned vehicle 1.

Further, in the embodiment described above, when one mark 6 is imaged three times, the next new mark 6 is set to enter the area 4a at the next imaging, but the present invention is not limited to this. It may be implemented in such a way that a new mark 6 is imaged after a state in which the mark 6 is not imaged continues.

Furthermore, in the above embodiment, the processing area e was set immediately before image recognition, but for a bond for which one running route has been determined, the next imaging position is set before starting running based on the running route. The processing area may be set in advance by determining the following.

Further, in the embodiment described above, the travel route is determined using a cubic function, but the present invention is not limited to this, and the travel route may be determined using other functions.

Furthermore, although the COD camera 3 was used as the imaging device in the above embodiment, it may be implemented using other @imaging devices, and in the above embodiment, the pixel configuration (resolution) of the image in the COD camera 3 Although the number of pixels is 256×256, the number of pixels is not limited to this, and the number of pixels may be changed as appropriate, such as 512×512 pixels, 1024×1024 pixels, etc.

Furthermore, in the above embodiment, the imaging processing operation→recognition processing operation→
The driving route of the unmanned vehicle 1 is determined and the driving is controlled in the order of arithmetic processing operation → driving control operation, but the order may be changed as appropriate.In short, the driving route is determined after capturing an image of the mark. Any order of operations may be used as long as the travel route is determined and the travel route is determined in consideration of the time difference between when travel control based on the travel route is started.

Furthermore, in the embodiment described above, the running speed V during unmanned mode 1 was kept constant, but it may be changed from time to time. In this case, a vehicle speed detector may be provided to measure the vehicle speed of the unmanned vehicle, and the vehicle speed, travel distance, etc. may be calculated based on the detector.

[Effects of the Invention] As detailed above, according to the present invention, when a mark is not included in the processing area due to a 1-rubble that naturally recovers due to sudden disturbance, etc., the mark is included. By changing the processing area, it is possible to prevent an unmanned vehicle from stopping unnecessarily due to the absence of a mark, and this method has an excellent effect as a method for determining the presence or absence of a driving mark.

[Brief explanation of the drawing]

Fig. 1 is an electric block circuit diagram of a travel control device embodying the first invention, Fig. 2 is a side view of an unmanned vehicle in motion, and Fig. 3 is a side view of a running unmanned vehicle.
The figure is also a plan view, Figure 1 is a front view of the mark, Figure 5 is an explanatory diagram for explaining the image captured by the COD camera,
Figure 6 is a diagram showing the relationship between the number of images taken and the mark, Figure 7 is a diagram showing changes in the mark in the image, and Figure 8 is a diagram showing the relationship between the operating order of the travel control device and the travel position of the unmanned vehicle. Figure, No. 9
The figure shows the area imaged by the COD camera while driving.
Fig. 10 is a diagram showing an image descending into the area, Fig. 11 is a diagram showing a travel route determined by the travel control device, Fig. 12 is an explanatory diagram for explaining steady turning circle travel, and Fig. 13 is a diagram showing a travel route determined by the travel control device. A diagram showing the relationship between attitude angle and radius, Figure 14 is a diagram to explain coordinate transformation, Figure 15 is a diagram to explain the area on the road surface, and Figures 16 and 17 are inside the image. Figure 18 is a diagram to explain the processing area, non-existence area, and existence area, Figure 19 is a diagram to explain the non-existence area, and Figure 20 is the existence area. 21 is a diagram showing a brightness histogram of a pixel signal, FIG. 22 is an explanatory diagram illustrating a state where a mark is included in a processing area, and FIG. 23 is an illustration showing a state in which a mark is included in a processing area. FIG. 24 is an explanatory diagram illustrating a state in which a mark does not exist; FIG.
Figure 5 is a diagram showing the relationship between the operation order of the travel control device and the travel (1 position) of the unmanned vehicle in an embodiment embodying the second invention, and Figure 26 is a diagram showing the relationship between the mark and each processing area. It is an explanatory diagram. In the diagram, 1 is an image-based unmanned vehicle, 3 is a CCD camera, 4 is a road surface, 4a is an area, 5 is an image, 6 is a mark, 6a is a corner,
10 is a microcomputer, 11 is a CPU, 12 is a program memory, 16 is an A/D converter, 18 is a binary level controller, 19 is a drive controller, 20
11. is a driving motor, 21 is a steering mechanism, L, Lo, 11.
12 is the travel route, G is the mark center position, Φm, Φp1.
Φp3 is the attitude angle, PO is the point as the R image concave point, P2 is the point as the imaging concave point and break point, Z is the area, e is the processing area, el is the absent area, e2 is the existing area, and Q is the image The middle mark center position, Gm is the mark center position on the road surface, r2
is the length. ＼ Drawing No. 6 Signal strength
Th color name 彊degree

Claims (1)

[Claims] 1. A first coordinate system whose origin is the old point at which the driving mark placed on the road surface was first imaged is set, and the position of the mark is determined in the first coordinate system. At the same time, a travel route is determined based on the previously imaged mark, and a new imaging point at which the next imaging operation of the same mark is performed on the travel route is determined. A second coordinate system is set, the position of the mark in the first coordinate system is transformed into a position in the second coordinate system, and an image is captured at a new imaging point based on the transformed position and the size of the mark. In the image, a processing area for image recognition that is smaller than the same image and includes the mark in the image is set, and then it is determined whether or not the mark is included in the processing area, and if the mark is not included. If so, the processing area is enlarged, and it is again determined whether the mark is included in the enlarged processing area. 2. Set a first coordinate system whose origin is the old point at which the driving mark placed on the road surface was first imaged, find the position of the mark in the first coordinate system, and Find a travel route based on the imaged mark, find a new imaging point on the travel route where the next imaging operation for the same mark will be performed, and then set a second coordinate system with the new imaging point as the origin. Then, the position of the mark in the first coordinate system is transformed into the position in the second coordinate system, and based on the transformed position and the size of the mark, the position of the mark in the first coordinate system is transformed from the same image in the image taken at the new imaging point. A processing area for image recognition that is small and includes the mark in the image is set, and then it is determined whether or not the mark is included in the processing area, and if the mark is not included, the travel route is is used as it is to find a new imaging point at which the imaging operation is performed using the mark, a new second coordinate system is set, and the position of the mark in the first coordinate system is changed to the new second coordinate. The coordinates are transformed to the position in the system, and based on the transformed position and the size of the mark, a new processing area for image recognition is set in the newly captured image that is smaller than the same image and includes the mark in the image. , a method for determining the presence or absence of a running mark, which determines whether or not the mark is included again in the new processing area.