JP2953497B2 - Object capturing method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for capturing an object, and more particularly to a method and apparatus for capturing an object at a predetermined position on a screen when capturing an image of an object to be monitored by a television camera or the like in a surveillance camera system. Regarding the device.

[0002]

2. Description of the Related Art Conventionally, a scene is monitored by a television camera or the like.
When a monitoring target is detected, a technique of capturing the monitoring target at the center of an imaging screen has been proposed. For example, JP-A-4-3322
No. 71 discloses a moving pixel, receives data P _x and P _y indicating distances in the x and y directions from the center of the screen, and calculates an average distance L _x ,
Seeking L _y, tracking control circuit is disclosed for driving the pan motor and a tilt motor.

[0003] A conventional method for obtaining the position of an object in a screen from an image and controlling the attitude of the television camera will be described. Referring to FIG. 2, parameters relating to the attitude of the television camera used in the following description will be described. In FIG. 2, the pan angle κ of the television camera 1 is controlled by the first motor 2, and the tilt angle ω is controlled by the second motor 3.
h is the height of the second motor 3, r is the distance between the rotation axis of the second motor 3 and the optical axis of the television camera 1, l is the rotation axis of the second motor 3 along this optical axis and the lens 4 shows the distance from the center of FIG. At this time, the world coordinate system (X, Y, Z)
Is the intersection of the rotation axis of the first motor 2 and the support (or the ground). The coordinate system of the image plane of the television camera 1 is represented by (x, y) with the center of the screen being (0, 0). x is the right direction of the screen, and y is the downward direction. With the optical center of the lens 4 as the origin, the two coordinate axes (X _P and Y _P ) are parallel to the coordinate system of the image plane, the forward direction of the television camera 1 is Z _P , and the scale is the world coordinate system. Positions in the same moving coordinate system are (X _P , Y _P , Z _P ).

In the above positional relationship, the world coordinates (X _O , Y _O , Z) of the origin (optical center of the lens 4) of the moving coordinate system
_O ) is given by Equation 1.

[0005]

(Equation 1)

When the position of a point in a three-dimensional space in a moving coordinate system is represented by (X _P , Y _P , Z _P ),
The coordinate axes, first rotates by pan angle κ respect Y _P axis, next time to match the world coordinate rotated by a tilt angle ω with respect to X _P axis, the coordinate transformation therebetween, the number 2.

[0007]

(Equation 2)

When an object at (X _P , Y _P , Z _P ) is imaged, the position (x, y) on the screen is given by f being the distance between the lens and the imaging surface (hereinafter referred to as a zoom parameter). , 3

[0009]

(Equation 3)

According to the above equations 1, 2 and 3, when an object at the position of world coordinates (X, Y, Z) is imaged in the attitude of the television camera 1 at the pan angle κ and the tilt angle ω, the image is displayed on the screen. (X, y) can be known. Based on this, the posture (κ, ω) of the television camera 1 is changed to the posture (κ ′, ω) so that the target currently imaged at the position (x, y) is imaged at the screen center (0, 0). ω ′), a new moving coordinate system (X _P ′, Y _P ′, Z
_P 'position in) _{is, X P' = Y P '} = 0 becomes, Z _P'
Is the distance from the optical center of the lens 4 to the object,
Equation 4 is obtained.

[0011]

(Equation 4)

Now, from the above relationship, the current position (x,
From y), world coordinates (X, Y, Z) are obtained as follows. That is, when solving Equation 2, the matrix is transposed to become Equation 5.

[0013]

(Equation 5)

Equation 6 is obtained by dividing the entire equation 5 by Z _P and obtaining a relational expression.

[0015]

(Equation 6)

X _P / Z _P and Y _P / Z _P are x / f and y / f, respectively, from Eq. (3), and are calculated from the current position (x, y) and attitude (κ, ω). The value on the left side of 6 is determined.

In the new moving coordinate system, X _P ′ = Y
_{Since P} ′ = 0, Equation 6 is simplified and becomes Equation 7.

[0018]

(Equation 7)

If the left side of Equation 7 is found, new pan angle κ ′ and tilt angle ω ′ are found by solving the equation. Here, since Z _P ′ has the relationship of Expression 4, the relationship of Expression 8 is obtained.

[0020]

(Equation 8)

[0021] However, X _O, Y _O, the number in the Z _O is generally 1
Since it is a function of the attitude (κ, ω) as shown in
X _O ′, Y _O ′, and Z _O ′ are functions of the new attitude (κ ′, ω ′), and Equation 7 cannot be solved analytically. However, if l = r = 0, that is, if the optical center of the lens 4 is arranged so as to coincide with the axial center of the second motor 3,
Since X _O ′ = Y _O ′ = 0 and Z _O ′ = h, X and Y on the left side of Expression 7 are obtained as shown in Expression 9, and X /
Since Z _P and Y / Z _P are known, the new pan angle κ ′ and tilt angle ω ′ are determined analytically.

[0022]

(Equation 9)

Conventionally, to solve Equation 7, for example, X _O '
= Y _O ′, an approximate value of the pan angle κ ′ and the tilt angle ω ′ is obtained under the approximation of 0, and by repeating the numerical calculation,
The television camera 1 was controlled in search of a more accurate solution.

[0024]

However, in order to obtain the pan angle κ ′ and the tilt angle ω ′ at a high speed by the above-mentioned numerical operation, a high-performance arithmetic device is required. It was difficult to control the movement of the target so as to capture it at the center of the screen. Further, when the television camera 1 is replaced, the mounting position changes or the distance l or r changes. Therefore, it is necessary to correctly measure this and reflect it in the above calculation.

Further, in Equation 3, there is a zoom parameter f as a parameter, and this value varies depending on the focus. Furthermore, when a zoom lens whose focal length can be changed is used as the lens 4, the value of the zoom parameter f greatly changes. In order to calculate the pan angle κ ′ and the tilt angle ω ′ based on the current value of the zoom parameter f, accurate measurement of the zoom parameter f is required. This is a serious problem in the operation of a real surveillance camera system.

In the above description, only the distances l and r are considered as structural parameters. However, actually, the optical axis of the television camera 1 does not intersect with the rotation axis of the first motor 2 and there is a deviation. And many other parameters to consider. In such cases, the problem is further complicated.

However, the purpose of the television camera 1 for monitoring a scene is to capture the object at the center of the screen or to zoom up to make it easier to grasp the situation regarding the monitored object at a remote place. Is not precisely controlled.

An object of the present invention is to provide a surveillance camera system capable of detecting an object and capturing it at a predetermined position on a screen even if parameters represented by the distances l and r and the zoom parameter f are not accurately known. And an apparatus for capturing an object.

[0029]

According to the object capturing method of the present invention, a target object is detected by image processing from a screen picked up by an image pickup means, and the detected position is shifted from a predetermined position on the screen (Δx, Δy), and the resulting deviation (Δ
x, Δy) and zoom parameter f
The control amount of the pan angle κ and the tilt angle ω is -Δκ = −Δx
/ F and -Δω = Δy / f.
The attitude of the imaging means is controlled by the control amounts -Δκ and -Δω , and the above process is repeated a plurality of times at a speed at which the movement of the target on the screen can be ignored, thereby capturing the target at a predetermined position on the screen. .

In the method of capturing an object of the present invention, a target object is detected by image processing from a screen imaged by the imaging means, and a deviation (Δx, Δy) of the detected position from a predetermined position on the screen is obtained. From the obtained deviation (Δx, Δy), the absolute error Δe = (Δx ² + Δ ) in the target object
y ² ) ^1/2 is obtained, and the absolute error Δe _{1 of the} previous time and the absolute
From the absolute error ratio _{r e} = Δe ₂ / Δe ₁ with relative error .DELTA.e ₂
The update amount Δf of the zoom parameter f is calculated to calculate the zoom
The parameter f is updated, and the deviations (Δx, Δy) and
From the zoom parameter f, the pan angle κ of the
And the control amounts of the tilt angle ω are -Δκ = −Δx / f and −
Approximately calculated as Δω = Δy / f, and obtained control amount−Δ
controlling the attitude of the imaging means by κ and -Δω ;
Capturing the object of interest at a predetermined position of the screen by repeating a plurality of times at a speed negligible movement of the in the target screen above process.

Furthermore, the target acquisition method of the present invention, prior to
Serial updating amount Delta] f, the absolute error ratio r _e is less than the threshold value th
If the set to 0, the absolute error ratio r _e is greater than the threshold th
Change of the sign of the deviation (Δx, Δy)
Examine, if the code is changed Δf = + k · _{r e} (only
And k is a positive constant), and if the sign does not change, Δf
= And -k · _{r e.}

Furthermore, the object capturing device of the present invention comprises:
An image pickup means for picking up an image of a scene, an image processing means for processing an image obtained by the image pickup means to detect a target, and obtaining a shift (Δx, Δy) from a predetermined position on the screen; , Δy) and zoom parameter f
The control amount of the pan angle κ and the tilt angle ω of the imaging means is −Δ
control parameter calculation means for performing approximate calculation as κ = −Δx / f and −Δω = Δy / f, and controlling the attitude of the imaging means by the control amounts −Δκ and −Δω. By posture control means, by repeatedly executing from the processing by the image processing means to the control by the posture control means as fast as negligible movement in the target screen,
The target object is captured at a predetermined position on the screen.

Also, the object capturing device of the present invention can be used to capture scenes.
Imaging means for imaging, and an image obtained by the imaging means
To detect the target object, and from the predetermined position on the screen
Image processing means for obtaining a shift (Δx, Δy);
From (Δx, Δy), the absolute error Δe =
(Δx ² + Δy ² ) ^1/2 is obtained, and the previous absolute error Δe
₁ and the absolute error ratio between the current absolute error Δe _{2 r e} = Δe ₂
Update of the zoom parameter f of the imaging means from / Δe ₁
A convergence determining means for obtaining the amount Δf;
y) and the zoom parameter f,
The control amounts of the pan angle κ and the tilt angle ω are given by −Δκ = −Δx /
f and -Δω = Δy / f
Meter calculation means and the control amounts -Δκ and -Δω.
And attitude control means for controlling the attitude of the imaging means.
From the processing by the image processing means to the attitude control means.
Movement in the screen until the control by the target can be ignored
By repeatedly executing at high speed every time, the target
Capture at a predetermined position on the screen .

According to the object capturing apparatus of the present invention, the convergence determination
It means the absolute error ratio r _e previous If it is less than the threshold value th
The serial update amount Δf is 0, the absolute error ratio r _e is the threshold value th
If greater than the sign of the deviation (Δx, Δy)
And if the sign has changed, the update amount Δf
Δf = + k · _{r e} (where, k is a positive constant), and the sign is
The updated amount Δf unless you change Δf = -k · _{r e}
And

[0035]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In the present invention, processing for detecting an object from an image is not only faster than moving the object, but also controls a pan angle κ and a tilt angle ω for controlling the attitude of the television camera 1. As the first motor 2 and the second motor 3, those having a higher speed than the movement of the object are used.
If the latest motor control technology is used, it is possible to execute several tens of degrees of rotation in a fraction of a second. In many surveillance camera systems, it is sufficient if the object is captured at the center of the screen in 1/2 second to 1 second and further zoomed in. Therefore, if such a control technique is used, it is not always necessary to perform one control. There is no need to capture the target at the center of the screen. As a high-speed image detection system that can be used in the present invention, for example, Okazaki et. al. , "A Compa
ct 30 GIPSReal-Time Visio
n System RVS-2 ", Proc. MVA '
94 IAPR Works on Machi
ne Vision Applications, p
p. 13-16, 1994. Is mentioned. As a high-speed camera control system, for example, Murray
et. al. , "Reactions to Per
ipheral Image Motion usin
ga Head / Eye Platform ”, Pr
oc. International Confere
nice on Computer Vision, p
p. 403-411, 1993.

Therefore, the parameters of the distances l and r shown in Equations 1 to 3 are set to, for example, 0, and Equation 7 is approximated to Equation 10, and control is performed by solving the pan angle κ ′ and the tilt angle ω ′. Do.

[0037]

(Equation 10)

If the deviation (x, y) from the center of the screen is large, the result is that the target cannot be accurately captured, but the control directions of the first motor 2 and the second motor 3 are almost correct. After the control, the target is detected near the center of the screen. At this position, the detection of the target is performed again, and control is performed by further obtaining the pan angle κ ′ and the tilt angle ω ′ from Expression (10) from the obtained shift (x, y). In this case, since the target has already been captured near the center of the screen, even if the parameters such as the distances l and r are not accurate, the approximate expression 10
Thus, the pan angle κ ′ and the tilt angle ω ′ can be obtained sufficiently correctly.

For the above reasons, the equations for determining the pan angle κ ′ and the tilt angle ω ′ can usually be approximated more coarsely. For example, in many surveillance camera systems,
Since the horizontal direction is monitored, ω ′ ≒ π / 2, | Z−Z ₀ |
<< Y. When the Y-axis direction is considered to be near the center of the screen, κ ′ 一般 0, Z _P ≒ Y, and X << Y can be handled while maintaining generality, so that Equation 11 is obtained from Equations 2 and 3. Can be

[0040]

[Equation 11]

Furthermore, by subjecting x and y to Taylor expansion to the first order with respect to the pan angle κ and the tilt angle ω under the above conditions, an approximate expression of the formula 12 can be obtained.

[0042]

(Equation 12)

This is because the pan angle and the tilt angle are respectively
This shows that Δκ and Δω are shifted.
To capture at the center of the screen, set the pan and tilt angles to
It is necessary to control -Δκ and -Δω, respectively. Therefore, when the target is detected at the position of the shift (Δx, Δy), the first motor 2 and the second motor 3 may be controlled so as to satisfy Expression 13.

[0044]

(Equation 13)

By repeating this control a few times faster than the movement of the object, Δx and Δy converge to 0, and the object is captured at the center of the screen.

Next, a method and an apparatus for capturing an object according to the present invention in consideration of a case where the zoom parameter f used for the above calculation is incorrect will be described. Since the above-described control is based on an approximation formula, a small error of the extent that the zoom parameter f is out of focus does not matter, and Δx and Δy converge to 0. But lens 4
Is a zoom lens, and when the zoom parameter f is set to be considerably larger than the true value, the control is carried out considerably more sparingly than the required amount by the formula (13). for that reason,
A large number of times is required until convergence, and capturing is possible when the moving speed within the target screen is low, but not capturing when the moving speed is high. Therefore, in the present invention, the zoom parameter f is updated as follows.

First, a deviation (Δx,
Δy), an absolute error Δe is defined as in Expression 14.

[0048]

[Equation 14]

[0049] Then, by comparing the absolute error .DELTA.e ₂ in absolute error .DELTA.e ₁ and the next target detection in the first object detection, than the absolute error ratio r _e = Δe ₂ / Δe ₁ threshold th (e.g., 1/2) When it is large, it is determined that the zoom parameter f is too large, and a negative value is set as the update amount Δf,
The setting of the zoom parameter f is updated as f → f + Δf.

On the other hand, when the zoom parameter f is set to be considerably smaller than the true value, the control becomes unnecessarily large according to Expression 13. Therefore, the sign of the deviation (Δx, Δy) from the center of the target screen is +, − for each repetition.
To fluctuate. Therefore, the absolute error ratio r _e is greater than the threshold value th, if there is a change of sign, set a positive value as the update amount Delta] f, similarly, updates the setting of the zoom parameter f as f → f + Delta] f .

The value of the update amount Delta] f is larger when the absolute error ratio r _e is large, since the in absolute error ratio r _e is small, small is desirable, | Delta] f | = variation as k · r _e Making it effective is also effective.

In the above description, κ 通常
Although the control method of the present invention has been described using an approximate expression near 0, ω ≒ π / 2, X ≒ 0, Z ≒ 0, the present invention provides an approximation that is easy to calculate as control converges as a result. , It is possible to use another approximation calculation according to need or situation. For example, when the tilt angle ω is not near π / 2, it is also effective to make Expression 13 slightly more precise and make Expression 13 as follows.

[0053]

(Equation 15)

[0054]

Next, the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of an object capturing apparatus to which an object capturing method according to one embodiment of the present invention is applied. The target capturing device includes a television camera 1 having a lens 4 as an imaging unit, a first motor 2 for controlling a pan angle κ, a second motor 3 for controlling a tilt angle ω, and the television camera 1. Processing the obtained image to detect the target object and shift from the center of the screen (Δx, Δy)
And a control parameter calculating means 6 for estimating the control parameters of the attitude of the television camera 1 by an approximate calculation from the deviation (Δx, Δy).
(x, Δy) from a temporal change in the convergence determining means 7 for obtaining the update amount Δf of the zoom parameter f of the lens 4 and the high-speed camera control means 8 for controlling the attitude (κ, ω) of the television camera 1 by the control parameters It is configured.

Various parameters between the television camera 1, the first motor 2, the second motor 3, and the lens 4 are as described above with reference to FIG. 2, and therefore, detailed description thereof will not be repeated.

Referring to FIG. 3, the method of capturing an object according to this embodiment includes an image processing object detection step S101, a shift calculation step S102, a zoom parameter update amount calculation step S103, and a control parameter estimation step S10.
4 and imaging means posture control step S105.

Next, the operation of the object capturing apparatus of the present embodiment thus configured will be described together with the object capturing method.

The attitude (κ, ω) of the television camera 1 is such that the pan angle κ is controlled by the first motor 2 and the tilt angle ω is controlled by the second motor 3. An image is captured by the television camera 1, and the position (x, y) of the target in the screen is detected by the high-speed image processing means 5 (step S101). This position (x, y) is slightly shifted from the center (Δx, Δ
y) (step S102), and the control parameter calculation means 6 calculates the current pan angle κ and tilt angle ω as κ−Δκ and ω−
Δω is set (step S104). In this case, the value of the zoom parameter f in Expression 13 is set to a standard value in advance, and may be different from a correct value. Note that when Expression 13 is used, Expression 16 is obtained.

[0060]

(Equation 16)

The value of the deviation (Δx, Δy) is also sent to the convergence judging means 7. At the time of the first target detection, the convergence determining means 7 does not output.

The high-speed camera control means 8 obtains the control amounts -.DELTA..kappa. And -.DELTA..omega. Output from the control parameter calculation means 6, and controls the first motor 2 and the second motor 3 at high speed (step S104). ). However, as described above, since the control amounts −Δκ and −Δω are approximate values, the target is not captured at the center of the screen.

The processing of the high-speed image processing means 5 and the control of the high-speed camera control means 8 are sufficiently fast as compared with the movement of the object on the screen. , The target is detected at the second time point from the next captured image (step S10).
1) The deviation (Δx, Δy) from the center of the screen is calculated again (step S102).

The convergence determining means 7 calculates the update amount Δf of the zoom parameter f (step S103). Specifically, the previous deviation (Δx, Δy) and the current deviation (Δx, Δ
y) and the error Δe is calculated from equation (14). And
If the absolute error ratio r _e = Δe ₂ / Δe ₁ the threshold th is smaller than the corrected output of the zoom parameter f is not issued.
However, if the absolute error ratio r _e is greater than the threshold value th, the zoom parameter f which is used in the calculation is not appropriate, examine the change in the (positive and negative) sign of Δx and [Delta] y. When the code has not changed, since the zoom parameter f is set to be larger than the correct value, when the k is positive constant, and outputs for example the updating amount Δf = -k · r _e to the control parameter calculator 6. Conversely, Δx and Δy
Of when the (positive or negative) sign is changed, since the zoom parameter f is set smaller than the correct value, and outputs for example the updating amount Δf = + k · r _e to the control parameter calculator 6.

The control parameter calculation means 6 adds this update amount Δf to the current zoom parameter f, and calculates f → f + Δ
and f, by using this new zoom parameter f, number 1
6 , the control amounts -Δκ and -Δω are calculated and sent to the high-speed camera control means 8.

By repeating the above processing and control several times, the object is captured at the center of the screen, and even when the lens 4 of the television camera 1 is a zoom lens and the zoom parameter f is deviated from the standard value, the object is detected. Can be captured in the center.

In the above embodiment, the convergence judging means 7 is included as a configuration considering that the lens 4 is a zoom lens. However, in a surveillance camera system in which the zoom parameter f does not change or is small, FIG. As shown, the convergence determination means 7 can be omitted. At this time,
It goes without saying that the processing of the method of capturing the target is a processing in which the zoom parameter update amount calculation step S103 is omitted as shown in FIG.

In the above embodiment, the pan angle κ is consistently
Has been described on the assumption that a support exists on the first motor 2 for controlling the tilt angle and the second motor 3 for controlling the tilt angle ω exists on the support. The configuration of the imaging means to which the capturing method of (1) is applied is not limited to such a configuration, and the dependency of the control motor on the pan angle κ and the tilt angle ω is reversed, or a different control mechanism is used. In this case, the content of the approximation calculation in the control parameter calculation means 6 only needs to be performed in accordance therewith, and it goes without saying that the present invention can be similarly applied.

Furthermore, in the above-described embodiment, the description has been given of capturing the target at the center of the screen. However, this is because it is normal to capture the target in an application such as a surveillance camera system. Just writing,
The present invention can be applied without any change even when the target is captured at a predetermined position other than the center of the screen.

[0070]

As described above, according to the method and the apparatus for capturing an object of the present invention, the surveillance camera system detects the object without knowing its mounting position and zoom parameters accurately and displays the image on the screen. Control can be performed so as to capture at a predetermined position, and the effect of monitoring can be achieved.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a target capturing apparatus to which a target capturing method according to an embodiment of the present invention is applied.

FIG. 2 is an explanatory diagram of a coordinate system for describing control parameters of the television camera.

FIG. 3 is a flowchart illustrating a process of an object capturing method according to the present embodiment.

FIG. 4 is a block diagram showing a configuration of a target capturing apparatus to which a target capturing method according to another embodiment of the present invention is applied.

FIG. 5 is a flowchart illustrating a process of an object capturing method according to the present embodiment.

[Explanation of symbols]

 Reference Signs List 1 TV camera 2 First motor 3 Second motor 4 Lens 5 High-speed image processing unit 6 Control parameter calculation unit 7 Convergence determination unit 8 High-speed camera control unit S101 Target detection step S102 Shift calculation step S103 Zoom parameter update amount calculation step S104 Control parameter estimation step S105 Image pickup unit attitude control step

Claims (6)

(57) [Claims] 1. A target object is detected from a screen imaged by an imaging means by image processing, a deviation (Δx, Δy) of the detected position from a predetermined position on the screen is obtained, and the obtained deviation (Δx, Δy) is obtained. ) And the zoom parameter f
The control amount of the pan angle κ and the tilt angle ω of the image means is given by −Δκ =
Approximate calculation as −Δx / f and −Δω = Δy / f,
The attitude of the imaging unit is controlled by the obtained control amounts -Δκ and -Δω , and the above process is repeated a plurality of times at a speed at which the movement of the target in the screen can be ignored. A method for capturing an object, comprising capturing. 2. A target object is detected by image processing from a screen imaged by an imaging means, and a deviation (Δx, Δy) of the detected position from a predetermined position on the screen is obtained, and the obtained deviation (Δx, Δy) is obtained. ) To the absolute error Δe in the target object
= (Δx ² + Δy ² ) ^1/2 is obtained, and the previous absolute error Δ
absolute error ratio r _e = Δe between e ₁ and the current absolute error Δe ₂
From ₂ / .DELTA.e ₁ calculates the update amount Δf zoom parameter f
Then, the zoom parameter f is updated, and the deviation (Δ
x, Δy) and the zoom parameter f
The control amount of the pan angle κ and the tilt angle ω of the means is −Δκ = −
Approximate calculation as Δx / f and −Δω = Δy / f
The attitude of the imaging means is controlled by the control amounts −Δκ and −Δω , and the above process is repeated a plurality of times at a speed at which the movement of the target on the screen can be ignored, thereby capturing the target at a predetermined position on the screen. A method for capturing an object, comprising: Wherein said update quantity Delta] f, the absolute error ratio r _e
There is a 0 if less than the threshold value th, the absolute error ratio r _e
If the difference (Δx, Δy) is larger than the threshold th,
The change of the negative sign is checked, and if the sign is changed, Δf = +
k · _{r e} (where, k is a positive constant), and the sign is changed
Target according to claim 2, unless Δf = -k · _{r e} have
Capture method. 4. Image capturing means for capturing an image of a scene, image processing means for processing an image obtained by the image capturing means, detecting a target, and calculating a deviation (Δx, Δy) from a predetermined position on a screen. From the displacement (Δx, Δy) and the zoom parameter f
The control amount of the pan angle κ and the tilt angle ω of the imaging unit is-
Approximators as Δκ = −Δx / f and −Δω = Δy / f
A control parameter calculating means calculate for the control amount -Δκ and consists of a posture control means for controlling the attitude of the imaging means by -Derutaomega, from processing by the image processing means subjects the up control by the attitude control means Screen An object capturing apparatus characterized in that a target object is captured at a predetermined position on a screen by repeatedly executing the object at a high speed such that the movement in the area is negligible. 5. An image pickup means for picking up an image of a scene, and processing an image obtained by said image pickup means to obtain a target object.
Is detected, and a deviation (Δx, Δy) from a predetermined position on the screen is determined.
The absolute error in the target object is determined from the image processing means to be obtained and the deviation (Δx, Δy).
The difference Δe = (Δx ² + Δy ² ) ^1/2 is obtained, and the previous absolute ^value is calculated.
The absolute error ratio r _e between the error Δe ₁ and the current absolute error Δe ₂
= Δe ₂ / Δe ₁ from the zoom parameter of the imaging means
convergence determining means for obtaining an update amount Δf of f, the shift (Δx, Δy) and the zoom parameter f
From the control amount of the pan angle κ and the tilt angle ω of the imaging means
As -Δκ = −Δx / f and −Δω = Δy / f
A control parameter calculating means for performing a similar calculation, and an image of the imaging means based on the control amounts -Δκ and -Δω.
And posture control means for controlling the force .
Control so that movement on the target screen can be ignored
The target can be displayed on the screen by repeatedly executing it at high speed.
A target capturing device, which captures at a predetermined position . 6. The convergence determining means according to claim 1 , wherein said absolute error ratio r
_{If e} is less than the threshold th, the update amount Δf is set to 0,
The deviation when serial absolute error ratio r _e is larger than the threshold value th
Examine the change of the sign of (Δx, Δy) and change the sign
And the updated amount Δf if Δf = + k · _{r e} (only
And k is a positive constant), and if the sign does not change,
The update amount Δf Δf = -k · r _e to claim 5 pairs <br/> elephant capture device according.