JP3153718B2 - Intruder detection device and intruder detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an intruder detecting apparatus and an intruder detecting method for detecting an intruder such as an intruder who has entered a specific monitoring area.

[0002]

2. Description of the Related Art Conventionally, Japanese Patent Application Laid-Open No.
An intruder recognition system such as that shown in 3-226799 is known. The intruder recognition system includes a camera including a monitoring area in a field of view, and an image processing apparatus connected to the camera and performing image recognition by processing an image from the camera and invading the monitoring area. When recognizing the position and size of the intruding object, a scale bar arranged in the field of view of the camera, a monitoring area; an intruding object; the geometrical relative positional relationship between the camera and the scale bar; Is calculated.

In this intruding object recognition system, a scale bar is provided at a fixed distance from a camera, and an image of the scale bar and an image of a monitoring area are input together to obtain an intruder and a camera;
The position and size of the intruder can be known from the geometric relationship of the scale of the scale bar. That is, if the upper limit and the lower limit of the scale bar of the scale occupied by the found object (intruder or the like) are known, the position and size of the found object can be calculated by the calculating means.

[0004]

However, the conventional intruder recognition system described above has a problem that a scale bar must be provided at a fixed distance from the camera. That is, depending on the structure of the room in the building, it may not be possible to arrange the scale bar. Further, even when the scale bar can be arranged, there is a problem that the scale bar must be correctly arranged at a predetermined position in the monitoring area, and the appearance is impaired.

An object of the present invention is to provide an intruder detecting apparatus and an intruder detecting method capable of detecting reliable information on an intruder without separately installing a special tool such as a scale bar. The purpose is.

[0006]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides an image relating to detection of an intruder based on image data obtained by capturing an image in a predetermined viewing angle range by a predetermined image capturing means. When performing the processing operation, the image processing operation uses the viewing angle of the imaging unit and the installation condition of the imaging unit. This makes it possible to detect reliable information on an intruder (intruder) only by the viewing angle and the installation conditions of the imaging unit without separately installing a special tool such as a scale bar.

[0007]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of an embodiment of an intruder detection device according to the present invention. Referring to FIG. 1, an intruder detecting apparatus according to the present embodiment has an imaging unit 1 having a predetermined viewing angle and capturing an image in a range of the viewing angle, and an installation unit for inputting installation conditions of the imaging unit 1. A condition input unit 2, an operation unit 3 for performing an image processing operation relating to detection of an intruder based on image data captured by the imaging unit 1, and an output unit 4 for outputting a detection result of the intruder
And a power supply unit 5 for supplying power.

In the example of FIG. 1, the image pickup section 1 has an image pickup device (for example, a CCD sensor) 11 and optical means (for example, an optical lens) 12 having a predetermined viewing angle (angle of view). FIG. 2 shows a specific example (plan view) of the imaging unit 1 in which the optical lens 12 is incorporated in a CCD sensor as the imaging element 11.
It is shown. Note that, in FIG.
And the CCD sensor 11 are shown in an overlapping state. In FIG. 2, reference numeral 13 denotes an operation lamp,
Power is supplied from the power supply unit 5 to the CD sensor 11, and the CCD sensor 11
When the sensor 11 is operating, the operation lamp 13 is turned on.

FIG. 3 is a perspective view showing an example of a room in a building. As shown in FIG. 3, for example, the image pickup unit 1 includes a ceiling 102 or a wall 103 of a predetermined room 101 in the building.
The image picked up by the image sensor 11 includes a view angle (angle of view) of the optical lens 12 and an installation condition of the image pickup unit 1 (the floor 10 of the room 101).
4, the height h of the imaging unit 1 and the installation angle of the imaging unit 1
(Tilt angle) θ). Therefore, room 1
In order to monitor a desired monitoring area in the room 01, the optical lens 12 having a predetermined viewing angle is used.
It is necessary to previously install the imaging unit 1 at a predetermined angle θ at a predetermined position in the image sensor 01.

FIGS. 4A and 4B are cross-sectional views in the x-axis direction and y-axis direction of FIG. 3, respectively. FIGS. 4A and 4B show a desired monitoring area in the room 101. Like monitoring 105
The imaging unit 1 having a predetermined viewing angle θ ₀ (θ _x0 , θ _y0 ) is placed at a predetermined position at a predetermined tilt angle (tilt angle from the vertical line z) θ
The state of installation at (θ _x , θ _y ) is shown. Note that θ
_x, x-axis direction component of the theta _y each inclination angle theta, a y-axis direction component, theta _x0, x-axis direction component of each of theta _y0 viewing angle theta _0, the y-axis direction component.

FIG. 5 shows an image pickup unit 1 shown in FIG.
When the CCD sensor 11 is installed as shown in FIG.
Shows an example of a screen imaged by the above. In the example of FIG. 5, this screen is, for example, the number of pixels of the CCD sensor.
The number of pixels is divided into N and M (for example, 60, 60) pixels in the x-axis direction and the y-axis direction corresponding to (number of elements) N × M. That is, one pixel of the imaged screen is CC
This corresponds to one pixel of the D sensor.

The CCD sensor has a high resolution of about several hundred thousand pixels to a low resolution of about several tens of pixels, and any of these can be used. In this case, when a low-resolution CCD sensor is used, the output of each pixel of the CCD sensor can be directly used as image data to be processed. In this case, each pixel of the screen is And each pixel has a one-to-one correspondence. When a high-resolution CCD sensor is used, each pixel output of the CCD sensor can be used as image data to be processed as it is. However, in order to simplify the subsequent image processing, the image sensor 1 or In the arithmetic section 3, a plurality of pixel outputs of the CCD sensor may be combined (for example, by averaging the plurality of pixel outputs) to form one pixel as image data. For example, when the CCD sensor has 64 × 64 pixels, that is, 4096 pixels, four pixels are compressed into one pixel (for example, the average value of the levels of four pixels is taken and the level of one pixel is obtained. 16x1
Image data of 6 pixels, that is, 256 pixels can be obtained, and the time required for image processing can be reduced.
However, the resolution is reduced accordingly.

The computing unit 3 monitors the screen imaged in this manner, for example, periodically (at a constant time interval T). Basically, the image data of the current screen and the image data of the previous screen are basically displayed. And when there is a change between the image data of the current screen and the image data of the previous screen,
Processing for determining whether or not the changed portion is caused by an intruder is performed.

In the example shown in FIG. 1, the arithmetic unit 3 outputs analog image data from the image pickup unit 1 (that is, for example, the CCD sensor 1).
A / D converter 20 for performing analog-to-digital conversion on each pixel output), a processor (CPU) 21 for performing the above-described monitoring processing, determination processing, and the like based on digitally converted image data; It has a ROM 22 storing a processing program and the like, and a RAM 23 functioning as a work area of the processor 21. Image data of a current screen, image data of a previous screen, and the like used in the above processing are stored in the RAM 23. It has become so.

The A / D converter 20 has a function of converting analog image data into binary digital image data by a predetermined threshold value according to the processing capability of the processor 21 and the like. Alternatively, it may have a function of converting analog image data into multivalued digital image data using, for example, the luminance (gradation level) at the focal coordinate position as a threshold. That is, when the processor 21 has a processing capability for multi-valued digital image data, an A / D converter having a function of converting the data into multi-valued digital image data is used, so that the intrusion can be more accurately performed. An A / D having a function of converting into binary digital image data when the processor 21 has only a processing capability for binary digital image data, although it can detect an object (intruder). When a converter is used or an A / D converter having a function of converting to multi-value digital image data is used, it is necessary to further convert the multi-value digital image data to binary digital image data. is there.

In the following description, for convenience, the analog image data from the imaging unit 1 is finally converted into binary image data, and the processor 21 performs a predetermined image processing operation on the binary image data. Shall do. The comparison between the current image data and the previous image data for determining whether or not the image data has changed is performed by taking the difference between them and making the difference image data. In this case, since the image data is binarized, the value of all the pixels is “0” when there is no change in the difference image data, and only the portion where there is a change is other than “0”. , It is possible to immediately determine whether or not there is a changed portion.

By the way, in order to reliably determine whether or not the changed portion of the screen is caused by an intruder, the calculation unit 3 uses the installation condition of the imaging unit 1.
(The height h of the imaging unit 1 and the inclination angle θ (θ _x ,
θ _y )) is required. Therefore, in this embodiment, an installation condition input unit 2 for inputting the installation conditions of the imaging unit 1 is provided.

The installation condition input unit 2 includes, for example, a range finder for automatically measuring the height h (that is, the distance from the floor 104) of the imaging unit 1 and / or the inclination angle θ of the imaging unit 1 from the vertical direction z. It can be configured as an inclinometer that automatically measures (θ _x , θ _y ). In this case, these measuring instruments are connected to the imaging unit 1
And outputs the outputs from these measuring instruments
, The installation conditions can be automatically input. In this case, the device for measuring the height (distance) h, that is, the range finder, receives the ultrasonic wave or laser light reflected by the floor 104 from the time when the ultrasonic wave or laser light is emitted toward the floor 104. An apparatus or the like that measures the distance to the floor 104 by measuring the time until the time of the operation can be used. Further, as the inclinometer, for example, a low-cost, high-performance miniaturized inclination angle sensor manufactured by Lucas Corporation in the United States can be used.

Further, instead of such a distance meter and an inclinometer, the installation condition input unit 2 is used to input the height h and the installation angle θ (θ _x , θ _y ) of the imaging unit 1 as analog voltages. It can also be configured as an analog voltage regulator. In this case, the operator operates the analog voltage regulator
By adjusting the analog voltage corresponding to the height h and the installation angle θ (θ _x , θ _y ) of the imaging unit 1 (for example, by operating a volume), the height h and the installation angle θ (θ _x , θ
_y ) can be input to the arithmetic unit 3 as an analog voltage.

Alternatively, the installation condition input unit 2 is connected to the imaging unit 1
The height h and the installation angle θ (θ _x , θ _y ) can be configured as digital information, for example, as an 8-bit dip switch. In this case, the operator operates the dip switch to set the height h and the angle θ.
By setting (θ _x , θ _y ), the height h and the angle θ
(θ _x , θ _y ) can be input to the arithmetic unit 3 as digital information.

Based on the installation condition input from the installation condition input unit 2, the calculation unit 3 first determines where the position (x _C , y _C ) immediately below the imaging unit 1 on the screen as shown in FIG. Then, the object OBJ corresponding to the portion CH where the image data has changed from the installation position of the imaging unit 1 is calculated (see FIG. 1).
), The distance L _XY from the installation position of the imaging unit 1 to the object OBJ corresponding to the changed part of the image data, and the object OBJ corresponding to the changed part of the image data. Is calculated, and whether or not the target object OBJ corresponding to the changed portion of the image data is an intruder (intruder) is determined based on the calculation information and the like.

Here, the calculation of the position (x _C , y _C ) immediately below the imaging unit 1 on the screen is performed as follows. That is, as shown in FIGS. 4A and 4B, for example,
Is _assumed to have viewing angles θ _x0 and θ _y0 in the x-axis direction and the y-axis direction, respectively, and to be tilted at a tilt angle θ _{x in} the x-axis direction and a tilt angle θ _y in the y-axis direction. The angles dθ _x and dθ of one pixel (one section) on the x-axis and y-axis of the screen as shown in FIG.
_y is calculated by the following equations.

[0023]

## EQU1 ## dθ _x = θ _x0 / N dθ _y = θ _y0 / M

N and M are the number of pixels (the number of sections) in the x-axis direction and the y-axis direction, respectively. One pixel on x-axis and y-axis
When the angles dθ _x and dθ _{y of} (one section) are calculated by Expression 1, the position (x _C , y _C ) immediately below the imaging unit 1 on the screen is obtained by the following equation.

[0025]

X _C = N / 2−θ _x / dθ _x y _C = M / 2−θ _y / dθ _y

FIG. 5 shows the position (x _C , y _C ) immediately below the image pickup unit 1 on the screen thus obtained. As can be seen from FIG. 5, when the tilt angles θ _x and θ _y are 0, that is, when the imaging unit 1 is not tilted, the position (x _C , y _C ) immediately below the same is the center of the screen. , Tilt angle θ _x
Alternatively, (x _C , y _C ) is closer to the edge of the screen as θ _y is larger.

In this manner, the imaging unit 1 on the screen
When the position (x _C , y _C ) immediately below is calculated, the angle ψ (vertical line) from the installation position of the imaging unit 1 to the position of the object OBJ corresponding to the changed part of the image data is calculated based on this. z
(Angle with respect to). More specifically, for example, the object OBJ does not exist at the previous time point, and the screen at the previous time point is as shown in FIG. 6A, and the object OBJ appears at the current time point as shown in FIG. The screen at the moment is as shown in Fig. 6 (b),
When the changed part of the image data is detected as the code CH on the screen, that is, for example, the difference between the image data of FIG. 6B and the image data of FIG.
(c) is obtained, and a portion having a pixel value other than “0” in the difference image data (FIG. 6)
(In the example of (c), a black pixel portion) When CH is detected as a changed portion of the image data, information on the angle ψ to the actual object OBJ corresponding to the changed portion CH is obtained as follows. be able to.

That is, the changed part CH of the image data on the screen is (x ₁ , y ₁ ) as shown in FIG.
~ (X _2, y ₂₎ when in the range of, x-axis direction of the angle range ψ _x1 ~ψ _x2 to the actual object OBJ corresponding to the changed portion CH of the image data from the installation position of the imaging unit 1, The angle range 方向_{y1 to} ψ _{y2 in the} y-axis direction is obtained as follows.

[0029]

数_x1 = dθ _x × (x ₁ −x _C ) ψ _x2 = dθ _x × (x ₂ −x _C ) ψ _y1 = dθ _y × (y ₁ −y _C ) ψ _y2 = dθ _y × ( y ₂ -y _C )

Thus, the distance L _XY (the distance from the installation position of the imaging unit 1 to the object OBJ to the farthest part of the object OBJ) can be obtained by the following equation.

[0031]

[Number 4] _{L XY = (h / cosψ x2} ) × (1 / cosψ y2)

In addition, the actual size | X |, | Y | of the object OBJ in the x-axis direction and the y-axis direction and the actual size S of the object OBJ are calculated by the following equations. Can be calculated.

[0033]

| X | = L _XY × (ψ _x2 −ψ _x1 ) | Y | = L _XY × (ψ _y2 −ψ _y1 ) S = | X | × | Y |

Thus, S, | X |, | Y | can be obtained as information on the size of the object OBJ,
L _XY , ψ can be obtained as information on the position of the object OBJ. Further, as information on the schematic shape of the object OBJ, for example, a ratio | Y | of | X | and | Y |
/ | X | can be obtained.

The information of the object OBJ as described above is stored in the floor 104 to be the monitoring area 105 of the room 101.
The part is a flat plane, and the object OBJ is obtained on the assumption that the object is along the flat plane of the floor 104. The part of the floor 104 to be the monitoring area 105 has irregularities such as stairs. In such a case, it is necessary to perform a corresponding correction to each information of the object OBJ. In the following, for convenience of explanation, it is assumed that the floor 104 to be the monitoring area 105 is a flat plane.

Further, in the above example, the imaging unit 1, an arbitrary inclination angle θ (θ _x, θ _y) was to be able to be attached by, if necessary, the inclination angle θ (θ _x, θ _y ) Can be constrained. That is, the image pickup device 11 has a CCD
When an element is used, the CCD element is generally rectangular in shape, and one of the x-axis direction and the y-axis direction is longer than the other and has a larger number of pixels.
Now, for example, when the length in the x-axis direction is long, and when the tilt angle θ _{x in the} x-axis direction is not “0 °” (that is, when the x-axis of this CCD element is not parallel to the floor 104), The actual monitoring area in the x-axis direction may be insufficient, or the calculation of the characteristic parameters (such as the size) of the object OBJ may be complicated, resulting in an error in the characteristic parameters. When the object OBJ is displayed on the monitor, a problem such as difficulty in seeing the object OBJ may occur. Therefore, when a CCD element is used as the imaging element 11, the imaging unit 1 is attached so that the x-axis of the CCD element is parallel to the floor (that is, the inclination angle θ _x is “0 °”) in the above example. Is good. In other words, in the case of the above example, at least one of the x-axis direction component θ _x and the y-axis direction component θ _y of the inclination angle θ of the image sensor 1 from the vertical line z is set to “0 °”. Is good. At this time, the position of the x _C,
Alternatively, the position of y _C is the center on the x-axis of the imaging screen or the center on the y-axis.

As described above, the arithmetic unit 3 periodically collects the image data of the screen (for example, at fixed time intervals).
The change of the image data is monitored each time, and when there is a change in the image data, the information as described above is obtained for the object corresponding to the change, and the state of the change of the object is further examined to determine the object. It is determined whether an object is an intruder (for example, an intruder).

In this judgment processing, when the intruder is an intruder (human), the human image moves on the screen (image image) as shown in FIGS. 7A and 7B, for example. (Changes with time). FIG. 7A is a diagram for explaining a change in the screen when a human is moving substantially along the x-axis direction, and FIG.
FIG. 9 is a diagram for explaining a change in a screen when the image is moving along the axial direction (in a direction slightly oblique from the y-axis direction).

From FIGS. 7A and 7B, it can be seen that when a human enters and moves in the monitoring area, the change in the image has the following characteristics. That is, the shape and size of the changed portion of the image at a certain point during the movement are appropriate for humans (for example, the size (height HG) is about 1 m to 2 m, and the shape is If the distance L _XY from the imaging unit 1 is small (approaching to the imaging unit 1), the size of the changed part of the image becomes large, and the imaging unit 1 When the distance L _XY is increased with (moved away from the imaging unit 1), that the size of the changed portion of the image is reduced, also, FIG. 7
As can be seen from (a), the position (x _C , y
_C ), the image of a human (upright state) has verticality (tilt angle ξ = 0), and is located at the position (x _C , y _C ).
Human (upright) when positioned off the y-axis passing through
It is the image to be inclined (xi] ≠ 0), also moving speed, it is reasonable to distance L _XY of the imaging unit 1 (e.g., a range of 0 m / sec to 2 m / sec according to the distance L _XY , Etc.).

Focusing on this feature, the arithmetic unit 3 performs, for example,
X of the object OBJ corresponding to the changed part of the image data
| X |, | Y | in the axial and y-axis directions and the overall size S
And a, whether the shape and size of this portion is reasonable as a human (whether for example a height HG of the order of 1 m to 2 m), also the distance L _XY is increased in this portion size Becomes smaller (as the distance from the imaging unit 1 increases,
When the distance _LXY decreases, the size of this portion increases (the size increases when approaching the imaging unit 1). , The inclination angle 、, and whether or not the change speed is appropriate for the distance _LXY . The determination of the verticality can be made based on, for example, whether or not the image has human-like verticality (that is, whether or not the image has an unnatural verticality as a human upright image pattern). Also, the inclination angle ξ
Can be detected as the inclination angle の of the object OBJ around the coordinates (x _C , y _C ), as shown in FIG. 7B.

In the above example, the changed part of the image data is only one pixel connection area CH.
Although only one object OBJ has dealt with the case where the detected changed portion of the image data is a plurality of pixel connected region CH ₁ to CH _n as shown in FIG. 8, a plurality of objects OBJ ₁ ~ If the OBJ _n is detected, the pixel connected regions CH ₁ to CH _n, the arithmetic processing described above for each of the object OBJ ₁ ~OBJ _n is performed, the object OB
J ₁ ~OBJ _n the shape information like sizes are determined.

For obtaining information on the shape of the object corresponding to the changed part of the image data, instead of the above method (a method using the ratio between | X | and | Y |), or For example, a pattern recognition technique can be used together with the above technique. That is, for example,
From the difference image data resulting from the comparison between the current image data and the previous image data, a changed portion of the image data having a pixel value other than “0” is extracted, and a predetermined feature parameter is extracted for this portion. Thus, information on the shape of the object corresponding to the changed part of the image data can be obtained.

More specifically, when the changed part of the image data is, for example, as shown in FIG. 9A at a certain point in time (the example of FIG. 9A is a difference binary image). 9 (a), the white pixels are the unchanged portions and the black pixels are the changed portions), as shown in FIG. 9 (b),
In this difference binary image, a region where predetermined pixels are connected (a region where black pixels are connected in the example of FIG. 9A)
A labeling process for assigning the same label number to the pixels in CH ₁ and CH ₂ is performed, and then, for the pixel connected region CH ₁ assigned the same label number “1”, its area R ₁ (pixel connected region CH ₁ , The perimeter L ₁ is extracted as a feature parameter, and the area R ₂ (pixel connection area CH ₂ ) is similarly assigned to the pixel connection area CH ₂ with the same label number “2”. the total number of pixels in _2), extracts the perimeter L ₂ as a characteristic parameter.

Here, the area R ₁ , R ₂ of each pixel connection region CH ₁ , CH ₂ is such that the label number is “1” when the binarized image is sequentially scanned as shown in FIG. "Or" 2 "
Of each of the pixels is occurring, by stepping one has for example a counter provided corresponding to the pixel connected region CH ₁ or CH _2, it can be counted. That is, each time scanning each pixel of the label numbers in the pixel connected region CH ₁ is "1", the count value of the counter corresponding to the pixel connected region CH ₁ is one increment, when all the scanning has been completed count value of the counter becomes the area R ₁ of the connecting region CH ₁ of the. Similarly, the area R ₂ of the pixel connected region CH ₂ also, the counter corresponding to the connecting area CH _2, above all scan can be obtained when finished.

The perimeters L ₁ , L ₂ of the respective pixel connection areas CH ₁ , CH ₂ can be counted, for example, by the following method. That is, as one method, the binarized image is shown in FIG.
When sequentially scanning as shown in FIG. 10A, each pixel M _{11 to} M of a mask having a size of 3 × 3 as shown in FIG. ₃₃ satisfies the condition as a contour (for example, FIG.
(h) is checked each time, and when the condition as a contour is satisfied, for example, by incrementing a counter provided for the pixel connection area by one, The number of contour pixels C ₁ and C ₂ of each connected area CH ₁ and CH ₂ can be counted. For example, now to belong to the pixel M ₂₂ is pixel connected region CH ₁ are focused, each pixel in the mask M
_{When ij} is the state shown in any one shown in FIG. 11 (a) to (h), it is determined that the pixel M ₂₂ is a contour pixel, in this case, the count value of the counter corresponding to the pixel connected region CH ₁ There is one increment, the number of outline pixels count value of the connection area CH ₁ of the counter C ₁ when all the scan has been completed
Becomes Similarly, the number of contour pixels connecting region CH ₂ C
₂ also, the counter corresponding to the connecting area CH _2, above all scan can be obtained when finished. Note that in FIGS. 11A to 11H, the symbol “x” is replaced with “don '
t care ". That is, this part may be either a white pixel or a black pixel. With such scanning and counting of the number of contour pixels, in one connected area, for example, CH ₁ , Of the number of contour pixels C _1, the contour pixel of interest and the next contour pixel are directly connected (for example, as shown in FIG. 13A described below), and the number of contour pixels is C _1a. The pixel of interest and the next contour pixel are obliquely connected (for example, as shown in FIG. 13B described later), and the number of contour pixels is counted as (C ₁ -C _1a ). when the circumferential length L ₁ of the connecting region CH _{1 ^{is, [C 1a +2 1/2 · (}} C 1 -C 1a)]
・ It is obtained as α. Here, α is the distance between adjacent pixels.

In order to obtain the perimeter L, it is also possible to use a method of sequentially tracking the boundary of the pixel connection area. That is, as shown in FIG. 12A, the image is scanned to search for untracked boundary point pixels. When a boundary point pixel is detected, a 3 × 3 mask M as shown in FIG.
_{Using ij} , it is checked whether or not this boundary point pixel is an isolated point. That is, when eight pixels in the vicinity of the boundary point pixel M ₂₂ of interest now are all white pixels ( "0"), the boundary point pixel is determined to be an isolated point, end a trace I do. In contrast, when the boundary point pixel M ₂₂ of interest now is not an isolated point, as shown in FIG. 12 (b), 8 pieces of the vicinity of the boundary point pixel (boundary point pixels that are now focused) The pixels are sequentially examined (for example, counterclockwise), and a pixel that changes from “0” to “1” among the eight neighboring pixels is searched for as the next boundary point pixel. Incidentally, in FIG. 12 (b), FIG. 12 and the state of the mask M _ij is shown when the mask M _ij in the portion indicated by symbol F in (a) is set, in this state, attention now when examined are eight pixels in the vicinity of the boundary point pixel M ₂₂ in the order of _{M 13, M 12, M 11} , M 21, M 31, M 32, M 33, pixel M ₃₃ is "0" to " is detected as a pixel vary in 1 ", therefore, M ₃₃ is detected as the next boundary point pixel. Thus the next boundary point pixel in the is detected, so that the center M ₂₂ of the mask M _ij is set to the next boundary point pixels, in a way that shifts the mask M _ij, next boundary point pixel Are sequentially traced, and when the next boundary point pixel becomes the first boundary point pixel at which tracking is started, it means that one round of the boundary of the pixel connection area is completed, and the tracking is terminated. For example, the scanning as shown in FIG. 12 (a), when the first boundary point pixels D _a is detected, FIG. 1 from the border point pixel D _a
2 start tracking boundary points by using a mask M _ij of (b), and around the boundary, and terminates the track when returning to the boundary point pixel D _a.

In such tracking, the perimeter L is obtained by sequentially adding the distance between the currently focused boundary point pixel and the next boundary point pixel for each tracking. The distance between the currently focused boundary point pixel and the next boundary point pixel is determined by the relationship between the currently focused boundary point pixel and the next boundary point pixel, for example, as shown in FIG. In such a case, it is calculated by multiplying by 1. In addition, for example, in the case of such a case as shown in FIG. 13B (that is, when the connection of the contour lines (pixels) is in an oblique direction) Is calculated by multiplying by 2 ^1/2 .

When there are a plurality of pixel connection regions as shown in FIGS. 9A and 9B, for example, for one pixel connection region, the position of one round of the boundary point between the pixel connection regions is scans to the right to seek a pixel connected region, when found boundary points, performs tracking in the same manner for the next pixel connected areas, the perimeter L _1, L ₂ of each pixel connected regions CH _1, CH ₂ You can ask.

As described above, the pixel connection areas CH ₁ , C
When the area and perimeter are obtained for H ₂ , respectively,
The pixel connection areas CH ₁ , CH ₂ , that is, the object OBJ ₁ ,
Information on the shape of OBJ ₂ can be obtained by the following equation.

[0050]

E = 4π (area) / (perimeter) ²

That is, information e ₁ and e ₂ relating to the shapes of the objects OBJ ₁ and OBJ ₂ can be obtained by the following equation.

[0052]

E ₁ = 4πR ₁ / (L ₁ ) ² e ₂ = 4πR ₂ / (L ₂ ) ²

As can be seen from Expression 6, the value of e approaches "1.0" as the object OBJ is closer to a circle, and the value of e becomes "0.0" as the object OBJ becomes more complicated. Become closer to Therefore, information about the complexity of the shape of the object OBJ can be obtained from the value of e, and this information is used as information about the shape instead of the shape information (height information HG) obtained by the method described above. Used, or the shape information (height information H
G) can be further used.

In the above example, the image data is binarized. However, as described above, if the processor 21 is capable of processing multi-valued digital image data, the above-described processing is performed by multi-valued processing. In this case, the information on the shape can be obtained with higher accuracy.

The arithmetic unit 2 can perform the above-described monitoring and arithmetic processing from the time when the power of the power supply unit 5 (see FIG. 1) is turned on. For example, a sensor 6 (see FIG. 1) such as a passive sensor or a window entrance magnet sensor is further used, and usually, the sensor 6 such as a passive sensor or a window entrance magnet sensor performs intrusion detection (intrusion monitoring). 6, the intrusion may be detected, and when the detection information is notified to the arithmetic unit 2, the arithmetic unit 2 may be configured to perform the above-described monitoring and arithmetic processing by using this as a trigger.

In this case, normally, the processor 21 can be prevented from being exclusively used for image processing, and can use the idle time to perform other processing. Efficiency (throughput) can be improved.

In any case, in the present embodiment, the object OBJ corresponding to the portion where the image data has changed only by the installation condition of the imaging unit 1 without separately installing a special tool such as a scale bar. It is possible to obtain various information such as the size, shape, distance from the imaging unit 1, and angle of the object OBJ, and to determine whether or not the object OBJ is an intruder (intruder) based on the various information. Can be performed.

Next, an example of the processing operation of the intruder detecting device having such a configuration will be described with reference to the flowchart of FIG.
In the processing operation example of FIG. 14, normally, the sensor 6 such as a passive sensor or a window entrance magnet sensor detects intrusion, and when the sensor 6 detects intrusion, the detection information is used as a trigger to trigger the processor 21. Indicates that monitoring and arithmetic processing are to be started.

Referring to FIG. 14, the processor 21 includes:
First, an initialization process is performed. For example, the RAM 23 and the like functioning as a work area are initialized (step S
1). Next, it is determined whether or not intrusion detection information has been notified from the sensor 6 (step S2).
Then, a timer (not shown) built in is driven to measure the time from when the intrusion detection information is notified, and the image monitoring process is started (step S3).

That is, the processor 21 takes in the image data from the imaging unit 1 (more specifically, the digital image data from the A / D converter 20), for example, at predetermined time intervals T (step S3). For example, as shown in FIG.
The start time of the image processing is t _1, the time _{t 1, t 2, t 3} ,
_Are image data D _ij (t ₁ ), D _ij (t ₂ ), D _ij (t ₃ ),
.., The processor 21 sets the times t ₁ , t ₂ , t
_The image data D _ij (t ₁ ), D _ij (t ₂ ), D _ij (t ₃ ) _,.
.. Are taken and stored in the RAM 23.

[0061] At this time, such as to save the RAM23 capacity, basically, two times t _k temporally adjacent image data D _ij of _{t k + 1 (t k)} , D ij ( t _{k + 1} ) is RA
M23. Specifically, the processor 21 captures the image data D _ij (t ₁ ) at time t ₁ ,
And stores it as shown in FIG. 16 (a) to the area WK ₁ of RAM 23, then, when the time t ₂ captures image data D _ij (t _2), this in the area WK ₂ in RAM 23 16
Store as shown in (b). At this time, the RAM 23 stores the image data D _ij (t ₁ ), D 2 at two times t ₁ and t _2.
ij (t ₂ ) is held. Then, the image data D at time t _{3
When ij} (t ₃ ) is captured, the image data D _ij (t ₁ ) at time t ₁ held in the area WK ₁ of the RAM 23 is deleted, and
Instead of this, the image data D _ij (t ₃ ) at time t ₃ is
Store as shown in (c). Thus, storing the image data of the odd-numbered time the area WK ₁ of RAM 23,
The image data of the even-numbered time is stored in the area WK of the RAM 23.
Store in ₂ .

When the image data D _ij (t _k ) and D _ij (t _{k + 1} ) at the two times t _k and t _{k + 1} are thus obtained, the processor 21 sets the two image data D _ij (t _k ), D
_ij (t _{k + 1} ) is compared, and it is determined whether or not the image data D _ij (t _{k + 1} ) has changed from the image data D _ij (t _k ) (step S4). For example, two binary image data D _ij (t _{k + 1} ),
D _ij (t _k ), and the difference image data (D
_ij all pixel values _{(t k + 1) -D ij} (t k)) is determined that there is no change when the "0", when there is a portion having a pixel value other than "0" is changed Judge.

If the result of this determination is that there has been no change, it is checked whether a predetermined time has been measured by the timer since the intrusion detection information was notified (step S5). Then, the process returns to step S3, and the same processing is performed for the next time. The predetermined time is, for example, the time required for a person to reach the monitoring area 105 from the position of the sensor 6 when the sensor 6 is not installed in the monitoring area 105, or Is set in consideration of, for example, the time during which the stationary state can be maintained. Therefore, if there is no change in the image data even after the predetermined time has elapsed, it is determined that no human has entered, and the process returns to step S2 again to wait for the intrusion detection information from the sensor 6 again.

More specifically, in the example of FIG. 15, if the predetermined time is set to, for example, (t ₇ −t ₁ ), after starting the image processing, the images at times t ₁ and t ₂ Data D _ij
When there is no change between (t ₁ ) and D _ij (t ₂ ), time t ₂ ,
image data D _ij of t ₃ (t _2), compares the D _ij (t _3), the image data D _ij (t _2), when there is no change in the D _ij (t ₃₎ is
Time t _3, t ₄ of the image data D _ij (t _3), and so to compare the D _ij (t _4), 2 two times t _k, t _{k + 1} of the image data D _ij (t _k), D _ij (t _{k + 1} ) is compared for a predetermined time (t ₇ −t
₍₁ ) is repeated (steps S3, S
4, S5).

In such a repetitive processing, for example, the image data D _ij (t ₄ ) and D _ij (t ₅ ) at the times t ₄ and t ₅ are obtained, for example.
As a result of the comparison, if a change is recognized, the processor 21
Further determines whether the change in the image data is due to an intruder (intruder). That is, as described above, it is determined whether or not the shape and size of the object OBJ corresponding to the changed portion is appropriate as the shape and size of the intruder (intruder). O
It is determined whether the moving speed of the BJ is appropriate as the moving speed of the intruder (intruder).

To make this determination, the processor 21
Is, when a change is recognized, simply two times t ₄ and t ₅
Image data D _ij of the (t _4), not only D _ij (t ₅₎ only, subsequent time t _6, ..., the image data of _{t m D ij (t 6)} , ..., D
_ij performs processing using a (t _m).

That is, in determining whether or not an image is due to an intruder (intruder), the image data D _ij (t _{4) at} two times t ₄ and t ₅ at the time when a change is recognized. ), D
Using only _ij (t ₅ ) (by the difference image data),
Determine the shape and size of the object corresponding to the changed part, determine whether this is appropriate as the shape and size of the intruder (intruder), and determine whether it is an intruder (intruder) Can also be determined. However, only by comparing the two image data D _ij (t ₄ ) and D _ij (t ₅ ) (that is, by only comparing at one point in time), an intruder (intruder) detects this. There is a possibility that the shape and the size are determined to be inappropriate as an intruder (intruder) even though the target object is an intruder (intruder). Further, it is not possible to obtain information on the movement (movement speed) of the object only from the comparison result of the two image data D _ij (t ₄ ) and D _ij (t ₅ ).
Further, with only one piece of differential image data, an erroneous report may occur due to the action of light such as an instantaneous light. Therefore, in order to make a more accurate determination, step S
When it is determined in step 4 that there is a change, not only the comparison result of the image data D _ij (t ₄ ) and D _ij (t ₅ ) at times t ₄ and t ₅ , but also the subsequent times t ₆ ,. , _{Tm, and} the image data at a plurality of times including the image data D _ij (t ₅ ) at the time t ₅ . For example, the image data D at times t ₄ and t _5
ij (t ₄ ) and D _ij (t ₅ ), the comparison results of the image data D _ij (t ₅ ) and D _ij (t ₆ ) at the subsequent times t ₅ and t _{6 ,.} t _m−1 , t _m image data D _ij (t _m−1 ), D
_ij comparison result even with a (t _m), by monitoring the temporal change of the object is good object performs intruder whether (intruder) determination.

For this reason, in the example of FIG. 14, step S4
Is determined, the processor 21 further acquires image data at a subsequent time.
(Step S6). Then, the processor 21 monitors the temporal change of the target object, for example, as described above, and at least at one point in time of the temporal change of the target object, the shape and the size of the intruder (intruder) are changed. It is determined whether or not the size is appropriate (Step S7),
In addition, it is determined whether or not the movement (moving speed) of the object is not unnatural as an intruder from the temporal change of the position of the object (step S8). Here, the temporal change of the position of the object OBJ can be detected by monitoring the temporal change of the distance L _XY and the angle ま で to the object OBJ. For example, the object OBJ suddenly disappears from the screen, or , If it suddenly changes position on the screen, moves back and forth in the screen, or does not move at all,
It is determined that the movement of the target object is unnatural as an intruder (intruder), and otherwise, it is determined that the movement of the target object is natural as an intruder (intruder).

As a result of these determinations, if the shape and size of the intruder (intruder) are appropriate and the movement (moving speed) of the intruder (intruder) is appropriate, the An alarm is output (step S9). For example, a warning lamp is turned on or a warning sound is generated. Also, the shape and size of the intruder (intruder) are not appropriate, or the movement (moving speed) of the intruder (intruder)
If it is not appropriate, return to step S2 again,
Wait for the intrusion detection information from the sensor 6 again.

The processing after the change is confirmed in the image data, that is, the comparison processing of the image data D _ij (t ₄ ) and D _ij (t ₅ ),..., The image data D _ij (t _m-1 ) , D _ij (t _m ), two image data, for example, D _ij (t _m-1 ),
The shape, size, and movement of the changed portion can be monitored by taking a simple difference between D _ij (t _m ), but after a change is confirmed in the image data, Since there is an object, if a simple difference is calculated, the difference image data includes, as shown in FIG. 17, an object at the current time, in addition to the current object portion CH (t _k ). Part CH (t
_k-1 ) may appear and the processing may be complicated.

In order to avoid this problem, for example, the image data D _ij (t ₄ ) immediately before a change is recognized (time t ₄ ) is set as standard image data, and the image data D _ij (t ₅ ) after time t _{5 is used.} ,
_{D ij (t 6), ...} , when D _ij (t _m) are obtained sequentially, each image data _{D ij (t 5), D} ij (t 6) ..., D ij (t m) and standard The difference from the image data D _ij (t ₄ ) is calculated, and each difference image data
(D _ij (t ₅ ) −D _ij (t ₄ )), (D _ij (t ₆ ) −D _ij (t ₄ )),
, (D _ij (t _m ) −D _ij (t ₄ )), the shape, size, position, and the like of the changing part are calculated, and the object corresponding to this part is calculated from the lapse of time. It can be determined whether or not is an intruder (intruder).

Alternatively, it is also possible to determine whether or not an object is an intruder (intruder) by any desired image processing other than the above-described image processing.

In the above example, when a change is found in the image data, the object corresponding to the changed portion of the image data is appropriate as the shape and size of the intruder (intruder), and When the movement of an intruder (intruder) is appropriate, an intrusion alarm is output, but as a previous step, image data at each time after a time point at which a change in image data is recognized is displayed on a display or the like. You can also.
That is, the changed portion of the image data can be visually displayed on a display or the like as a moving image.

Further, in addition to determining whether or not an intruder (intruder), it is also possible to calculate the number of intruders (intruders) and output this information. For example, as shown in FIG. 8, when a plurality of objects OBJ _{1 to} OBJ _n are detected and each of the objects OBJ _{1 to} OBJ _n is determined to be valid as an intruder (intruder), an intruder is detected. It is also possible to output (for example, display) “n” as the number of (intruders).

As described above, according to this embodiment, the imaging unit 1 can be installed without special tools such as a scale bar.
Only with the installation condition, reliable information on an intruder (intruder) can be detected.

[0076]

As described above, according to the present invention, an image processing operation relating to detection of an intruder is performed based on image data obtained by capturing an image in a range of a predetermined viewing angle by a predetermined imaging unit. In this image processing operation, the viewing angle of the imaging means and the installation conditions of the imaging means are used. This makes it possible to detect reliable information on an intruder (intruder) only by the viewing angle and the installation conditions of the imaging unit without separately installing a special tool such as a scale bar.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an embodiment of an intruder detection device according to the present invention.

FIG. 2 is a diagram illustrating a specific example of an imaging unit.

FIG. 3 is a perspective view showing an example of a room in a building.

4 (a) and 4 (b) are cross-sectional views in the x-axis direction and y-axis direction of FIG. 3, respectively.

FIG. 5 is a diagram showing an example of an imaged screen.

FIGS. 6A to 6C are diagrams for explaining a method of detecting a changed portion of image data.

FIGS. 7A and 7B are diagrams for explaining how a human image moves. FIG.

FIG. 8 is a diagram showing a case where a changed part of image data is composed of a plurality of pixel connection areas.

FIGS. 9A and 9B are diagrams for explaining a process of extracting a pixel connection region and attaching a label thereto.

FIGS. 10A and 10B are diagrams for explaining a method of counting the area of a pixel connection region and the number of contour pixels.

FIGS. 11A to 11H are diagrams showing pixels in a mask satisfying a condition as an outline.

FIGS. 12A and 12B are diagrams for explaining how to count the perimeter of a pixel connection region; FIGS.

FIGS. 13A and 13B are diagrams for explaining how to count the perimeter of a pixel connection region; FIGS.

FIG. 14 is a flowchart illustrating an example of a processing operation of the intruding object detection device of FIG. 1;

FIG. 15 is a time chart showing the timing of capturing image data.

FIGS. 16A to 16C are diagrams for explaining a method of storing image data in a RAM.

FIG. 17 is a diagram illustrating an example of difference image data.

[Explanation of symbols]

REFERENCE SIGNS LIST 1 imaging unit 2 installation condition input unit 3 calculation unit 4 output unit 5 power supply unit 6 sensor 11 imaging element 12 optical unit 20 A / D converter 21 processor 22 ROM 23 RAM 101 room 102 ceiling 103 wall surface 104 floor 105 monitoring area h imaging Height θ (θ _x , θ _y ) Installation angle of the imaging unit (tilt angle)

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-63-70683 (JP, A) JP-A-6-258027 (JP, A) JP-A-6-174405 (JP, A) JP-A-6-174405 223187 (JP, A) JP-A-7-225828 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01C 3/00-3/32 G01B 11/00-11/30 102 G06T 7/60 150 G08B 13/189-13/196

Claims (7)

(57) [Claims] An image pickup means having a predetermined viewing angle, and for picking up an image in a range of the view angle, wherein the image pickup means is provided.
Setting the height and the installation angle of the imaging means
Installation condition input means for inputting as a condition, and arithmetic means for performing an image processing operation relating to detection of an intruder based on image data taken by the imaging means, wherein the arithmetic means comprises an image processing apparatus for detecting an intruder In the operation,
Using the viewing angle of the imaging unit and the installation condition input from the installation condition input unit , the imaging unit directly on the screen
Calculate the lower position, based on the position immediately below the imaging means,
The angle from the installation position of the imaging means to the position of the object
Information and the distance from the installation position of the imaging means to the object.
Information and information on the size of the object
An intruding object detection device characterized in that both devices calculate one piece of information . 2. The intruding object detection device according to claim 1 , wherein said installation condition input means has a range finder and an inclinometer attached to said image pickup means, and said distance measurement means detects the distance from the floor of said image pickup means. An intruding object detection apparatus, comprising: measuring a height, measuring an installation angle of the imaging unit with an inclinometer, and inputting the measurement angle to an arithmetic unit. 3. An intruder detection apparatus according to claim 1 , wherein said installation condition input means has a voltage adjusting means for inputting a height and an installation angle of said image pickup means as an analog voltage. Characteristic intruder detection device. 4. The intruder detection apparatus according to claim 1 , wherein said installation condition input means has digital input means for inputting a height and an installation angle of said imaging means as a code having a predetermined number of bits. An intruder detection device, characterized in that: 5. The intruder detection apparatus according to claim 1, wherein the image data is processed as one pixel of the imaging unit as necessary, regardless of the resolution of the imaging unit, or A plurality of pixels of the imaging means are collectively 1
An intruder detection device characterized by being processed as a pixel. 6. An intruder detection method for performing an image processing operation relating to detection of an intruder based on image data obtained by capturing an image in a range of a predetermined viewing angle by a predetermined imager, wherein the image relating to the detection of an intruder is provided. In the processing calculation, the viewing angle of the imaging unit, the height at which the imaging unit is installed, and the
Imaging on the screen using the installation angle of the imaging means
Calculate the position immediately below the means, and refer to the position immediately below the imaging means.
And the angle from the installation position of the imaging means to the position of the object
Information about the distance from the installation position of the imaging means to the object
Information about separation and information about the size of the object
A method for detecting an intruder, wherein at least one piece of information is calculated . 7. An intruder detection method according to claim 6 , wherein a sensor such as a passive sensor or a magnet sensor for window entrance is detected by an intrusion detection method. Performing an intruding object detection method.