JP3497929B2 - Intruder monitoring device - 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 monitoring device for capturing an image of a predetermined monitoring area and detecting an intruding object from the captured image.

[0002]

2. Description of the Related Art In recent years, as one of the applications of image technology, a technique has been developed in which a predetermined surveillance area is imaged by a TV camera and the captured image is analyzed to automatically detect an intruding object in the surveillance area. For example, Japanese Patent Laid-Open No. 4-101594 discloses a technique of detecting an intruding object based on a difference in brightness between two images captured from one TV camera with a constant time difference.

However, when the brightness of the sun or the like changes significantly in the outdoor natural environment, or when trees or the like in the background are greatly shaken, these backgrounds are also falsely detected as intruders. In addition, since it is not possible to detect the exact position of the object in three-dimensional space, it is possible to detect when an unmonitored area behind (distant) or in front of the monitored object is reflected in the image. There is a problem that it is not possible to accurately determine whether or not the created object is within the monitoring target.

For this reason, in Japanese Patent Laid-Open No. 4-31996, two cameras for imaging the same monitoring area from different viewpoints are used, and after detecting a temporal change in the images captured by each camera, Detecting intruders and objects by determining the three-dimensional position of the changed part by performing stereo matching on the part where the change has occurred and determining whether or not the changed part (detection object) exists at an appropriate position. Discloses a technique capable of avoiding an erroneous determination even if a change in light or shadow occurs at a position where an intruder / object should not exist.

[0005]

However, in consideration of the place where an arbitrary object is present, especially when it is monitored outdoors, if the background brightness changes as long as the temporal change of the image is detected, the The entire background of is detected and the intruder is buried, making it difficult to accurately detect only the intruder.

Further, in order to detect the temporal change of the image, it is necessary for the cameras to be fixed with respect to the background, and when trying to cover a wide surveillance area, a large number of cameras are used for each surveillance area. Must be installed towards.

The present invention has been made in view of the above circumstances, and it is possible to detect an intruding object with high accuracy without being affected by the brightness of the background or an object existing in the background, and monitor over a wide range. It is an object of the present invention to provide an intruder monitoring device that can be performed.

[0008]

The invention according to claim 1 is
Stereo image processing means for processing a pair of stereo images obtained by capturing the same monitoring area from different viewpoints, and calculating a distance distribution over the entire image from the shift amount of the corresponding positions of the stereo image pairs, and the stereo image processing means. It determined the three-dimensional position of each part of the corresponding object to the distance distribution information, monitoring territory
The three-dimensional space of the region is divided into multiple cubic lattices, and each cubic case
Calculate the number of data of the above 3D position included in the child, and
Grid containing 3D position data of existing and existing objects
Is excluded as a non-detection target, and a cubic lattice to be monitored is added.
On the other hand, the number of data included in each cubic lattice is used as the frequency
Create a gram and set the histogram frequency to a preset threshold
A cubic lattice that exceeds the value is extracted as an object, and the
The cubic lattices that touch each other are grouped together and
Data for each group is projected on a two-dimensional image and each data is
Create a contour image by connecting with line segments and
And an intruding object detecting means for detecting an intruding object to be specified and detected .

According to a second aspect of the invention, in the first aspect of the invention, the intruder detecting means is a detection target.
When an intruder is detected, it is notified and
The position of the center of gravity of the detected object detected by
The position of the center of gravity of the detected object should be
When they match within the set threshold, it is judged as the same intruder
Then, the position of the center of gravity is tracked and the captured image is recorded.
If the center of gravity position and the previous centroid position do not match within the threshold,
It is characterized in that it is judged as a new intruder and the position of the center of gravity is recorded .

[0010]

[0011] That is, in the first aspect of the present invention, the
A pair of stereo images of one surveillance area taken from different viewpoints
Calculate the distance distribution over the entire image from the shift amount of the corresponding position of
To do. Then, each part of the subject corresponding to the distance distribution information
The 3D position of the
It is divided into square grids and the three-dimensional position
Calculate the number of data. And three of the background and existing objects
Exclude the cubic lattice containing the dimensional position data from detection
exclude. Then, for each cubic lattice to be monitored,
Histogram with the number of data included in the square grid as the frequency
Created and the histogram frequency exceeds a preset threshold
The existing cubic lattice is extracted as having an object. And things
For the cubic lattices that are determined to have a body,
The square grids are put together into one group, and the data for each group is
The data is projected on a two-dimensional image and each data is connected by a line segment.
By connecting, a contour image is created to identify the type of detected object,
Detects intruders as detection targets .

In this case, the invention according to claim 2 is the detection pair.
When an intruder that is an elephant is detected, it is notified and the intruder is entered.
The barycentric position of the detected object detected as the object is obtained. That
The center of gravity of the detected object obtained this time is the detected object obtained last time.
When the barycentric position matches the preset threshold, the same
It is determined that it is an intruder, the position of the center of gravity is tracked, and a captured image
Is recorded, and the current and previous barycentric positions do not match within the threshold.
Sometimes, it is judged as a new intruder and the position of the center of gravity is recorded.
It

[0013]

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. The drawings relate to an embodiment of the present invention, FIG. 1 is an overall configuration diagram of a monitoring device, FIG. 2 is a circuit configuration diagram of the monitoring device, FIG. 3 is a flowchart of deviation amount detection processing,
4 is a flowchart of the intruding object detection process, FIG. 5 is an explanatory diagram showing a coordinate system, FIG. 6 is an explanatory diagram showing a positional relationship between coordinate values on a distance image and a target object, and FIG. 7 is a distance image and a target object. FIG. 8 is an explanatory view showing the positional relationship of FIG. 3 by a vertical section, FIG. 8 is a top view showing the rotation angle of the camera in the left-right direction, and FIG.

In FIG. 1, reference numeral 1 is a monitoring device installed in, for example, an indoor or outdoor parking lot of a building or the like and monitoring an intruding object such as a person, an animal or a car invading a predetermined monitoring area, In the figure, an example is shown in which an image is taken of a monitoring area in which the plant 2, the table 3, etc. are present, such as the interior of an office or a corridor, and the intruder 4 is detected as an intruder from the captured image and monitored.

The above-mentioned monitoring apparatus 1 uses the stereo optical system 10 for imaging the same monitoring area from different viewpoints by a pair of cameras, and the stereo image pair captured by this stereo optical system 10 based on the amount of displacement of the corresponding position. The process of obtaining the distance information over the entire image by the principle of triangulation is performed,
A stereo image processing device (stereo image processing means) 20 for generating a three-dimensional distance image, a distance image from this stereo image processing device 20 is analyzed to obtain a three-dimensional position of each part of a subject, and the obtained three-dimensional position An intruder detection device (intruder detection means) 70 for determining the presence or absence of an object based on the above, and detecting the intruder in the monitoring area, and the position data of the intruder detected by this intruder detection device 70 is recorded. In addition, an image data recording device 80 for recording an image, and an alarm device 90 for outputting an alarm signal to a buzzer, a lamp, or the like when a detection signal of an intruding object from the intruding object detecting device 70 is input are provided. There is.

The stereo optical system 10 is mainly composed of two cameras using, for example, a charge coupled device (CCD), and the left and right cameras 11a and 11b are fixed to the stay 12. A camera rotation device 13 including a motor and the like is attached to a lower portion of the left camera 11a which is a main camera, and a camera position detection device 14 including an encoder and the like is attached to the camera rotation device 13.

That is, when the surveillance area is wide and cannot be covered by the visual fields of the cameras 11a and 11b at a fixed position, the present invention does not detect an intruder from the temporal change of the image at a fixed position, and will be described later. As described above, a stereo image pair captured at a predetermined position is processed to obtain a three-dimensional distance distribution over the entire image, and an intruder is detected based on the three-dimensional position. By rotating the entire optical system 10 to the left or right or up and down, the monitoring area can be covered over a wide range.

In order to detect the shift amount d between the left and right images picked up by the stereo optical system 10, it is necessary to find the image of the same object in the left and right images. In the present invention, in the stereo image processing device 20 described below, By dividing the image into small areas and comparing the brightness or color patterns in each small area on the left and right images to find a matching area, there is no effect from the brightness or color of the subject, and even outdoors. The distance distribution can be accurately obtained over the entire screen without being affected by a large change in the light conditions in.

The degree of coincidence between the left and right images is i of the right and left images.
The luminance of the th pixel (color may be used) is A
Letting i and Bi be, for example, the city block distance H shown in the following formula (1) can be evaluated, and when the minimum value of the city block distance H between the small regions of the left and right images satisfies a predetermined condition, It can be determined that the small areas correspond to each other.

H = Σ | Ai−Bi | (1) Therefore, the stereo image processing apparatus 20 has a circuit configuration as shown in FIG. 2, for example. In this circuit configuration example, the camera 11a is used. , 3b for converting the analog image picked up by the image pickup device 11b into a predetermined digital image.
0, for the image data from the image conversion unit 30, the city block distance calculation for sequentially calculating the city block distance H for determining the shift amount d of the corresponding small areas of the left and right images while shifting the pixels one by one A minimum value detecting / checking circuit for detecting the minimum value of the city block distance H, determining whether the obtained minimum value indicates a match between the left and right small areas, and determining the deviation amount d. 50, a distance image memory 60 for storing distance distribution information (distance image) based on the deviation amount d determined by the minimum value detection / check circuit 50.

The image conversion unit 30 includes an A / D converter 31 corresponding to the left and right image cameras 11a and 11b.
a, 31b are provided, and each A / D converter 31a, 3b
Look-up tables (LUTs) 32a and 32b each composed of a ROM are connected to 1b, and contrast correction for a low-luminance portion, correction of the intrinsic gain of the CCD amplifier, and the like are performed. Further, a left image memory 33a and a right image memory 33b are connected to the LUTs 32a and 32b, respectively.

The image data recorded in the left and right image memories 33a and 33b are repeatedly extracted by the city block distance calculation section 40 and processed. The city block distance calculation unit 40 uses the left image memory 33.
The left image shift register 43 is connected to the two sets of left image input buffer memories 41a and 41b to which the image data from a is inputted, and the right image memory 33b is also provided.
The right image shift register 44 is connected to the two sets of right image input buffer memories 42a and 42b to which the image data from the above are input, and each shift register 43 and 44 is connected to the city block distance calculation circuit 45. Has been done.

The city block distance calculation circuit 45 is
For example, it has a pipeline structure in which a high-speed arithmetic unit in which an input / output latch is connected to an adder / subtractor is combined and connected in a pyramid shape, and several pixels are input at the same time for calculation. The first stage of this pyramid structure is an absolute value calculator, the intermediate stages each constitute an adder, and the final stage is a summing adder.

Further, the minimum value detection / check circuit 50 is
The minimum value HMIN and the maximum value HMAX of the city block distance H are detected in synchronization with the output of the city block distance calculation circuit 45, and the minimum value HMIN of the obtained city block distance H is detected.
Indicates that the left and right small areas actually match each other, and only those satisfying the conditions correspond to the distance image memory 60 including the dual port memory or the like that interfaces with the intruder detection device 70. The shift amount d is output to the pixel position.

On the other hand, the intruding object detecting device 70 is mainly composed of a microprocessor 71 which reads out the deviation amount d written in the distance image memory 60 and performs various calculation processes, and mounts the cameras 11a and 11b. ROM 72 for storing parameters such as position and viewing angle, data such as position and range of monitoring area, and control program, RAM 73 for storing various data during calculation processing, output memory 74 for storing processing result, interface circuit It is composed of 75 etc.

The alarm device 90 is connected to the output port of the interface circuit 75, and the camera rotating device 13 is connected via a drive circuit 76.
The camera position detection device 14 is connected to the input port.

Further, the image data recording device 80 includes
When the intruder detection device 70 is connected via the output memory 74 and the main camera (left camera) 11a of the stereo optical system 10 is connected and an intruder is detected, position data of the intruder is detected. Etc. are recorded, and the image at that time is recorded on a video tape or the like.

The camera 11a of the stereo optical system 10,
The stereo image pair captured by 11b is matched with the left and right images by the stereo image processing device 20 to generate a distance image and stored in the distance image memory 60. In the intruder detection device 70, the distance image memory 6
Based on the three-dimensional position coordinates of the object based on the distance distribution information from 0, the intruder is recognized separately from the background, and the recognition result is output to the output memory 74.

The deviation amount detecting process by the stereo image processing device 20 and the intruding object detecting process by the intruding object detecting device 70 will be described below with reference to the flowcharts of FIGS. 3 and 4.

First, in the shift amount detecting process, in step S101 of FIG. 3, when the images captured by the left and right cameras 11a and 11b are input, the process proceeds to step S102, and the input analog image is input to the A / D converters 31a and 31b. It is converted into a digital image having a predetermined luminance gradation (for example, 256 gradation gray scale).

The digital images converted by the A / D converters 31a and 31b are LUTs 32a and 32, respectively.
b, contrast enhancement in low-brightness area, left and right cameras 1
The characteristics of 1a and 11b are compensated, and the image memory 33
a, 33b. The image memory 33a,
The image stored in 33b is C of the cameras 11a and 11b.
Of all the lines of the CD element, only the lines necessary for the subsequent processing are rewritten every predetermined time.

After that, the left and right image memories 33a, 33b
In step S103, the image data recorded in (1) is transferred to the input buffer memories 41a, 41b, 42a, 42b from each of the image memories 33a, 33b by several lines, and the left and right images are matched, that is, the matching degree is evaluated.

At this time, the input buffer memories 41a, 4b from the image memories 33a, 33b are read for the left and right images.
A read operation to 1b, 42a and 42b and a write operation to shift registers 43 and 44 are alternately performed. For example, while the image data is being read from the left image memory 33a into one of the input buffer memories 41a, the shift register 43 from the other input buffer memory 41b is read.
While the image data read in is written out and the image data is read from the right image memory 33b into one input buffer memory 42a, the image data read from the other input buffer memory 42b into the shift register 44. Is written out.

The shift registers 43 and 44 are
The contents of the even-numbered stages are simultaneously displayed in the city block distance calculation circuit 45.
When the calculation of the city block distance H is started, the data of the right image is held in the shift register 44, and the odd line data and the even line data are alternately output every clock.

On the other hand, the data of the left image is the shift register 4
The data of the odd lines and the even lines are alternately output, and are replaced by the data shifted to the right by one pixel every several clocks. This operation is repeated, and when the transfer to one small area is completed,
The content of the right image address counter (start address of the next small area) is set in the left image address counter in the address counter (not shown), and the processing of the next small area starts.

In the city block distance calculation circuit 45, data for several pixels is input to the absolute value calculator at the first stage of the pyramid structure, and the absolute value of the brightness difference between the left and right images is calculated. That is, if the brightness of the corresponding left pixel is subtracted from the brightness of the right pixel, and if the result is negative, the subtraction and subtraction are reversed by changing the operation instruction, and subtraction is performed again. Calculate the value. Therefore, the subtraction may be performed twice in the first stage.

Next, after passing through the first stage, two simultaneous input data are added one after another by each adder of the intermediate stage, and two continuous data are added by the total adder of the final stage to calculate the total sum. Then, the city block distance H for pixels forming the small area is output to the minimum value detection / check circuit 50 every several clocks.

Then, the process proceeds to step S104, and the above step
The maximum value HMAX of the city block distance H calculated in S103,
The minimum value HMIN is detected by the minimum value detection / check circuit 50. In the minimum value detection / check circuit 50, the city block distance H (deviation amount d = 0) that is output first
And the city block distance H (deviation amount d = 1) output at the next clock are compared, and the deviation amount d is stored only when the comparison result is small. Then, this comparison operation is repeated, and the maximum value, the minimum value during the calculation, and the deviation amount d at that time are stored, and the maximum value HMAX of the values so far,
The minimum value HMIN is updated, and the calculation is continued until the deviation amount d reaches the predetermined number of pixels.

When the calculation is completed (one clock after the last city block distance H is output), step S105
Then, it is checked whether or not the obtained minimum value HMIN really indicates the coincidence of the left and right small areas by the following three conditions.

(1) HMIN ≤ Ha (When the minimum value HMIN is larger than the threshold value Ha, the distance cannot be detected.) (2) HMAX-HMIN ≥ Hb (obtained minimum value HMIN is clearer than fluctuation due to noise) This is a condition for checking that the value has become low, and the difference between the maximum value HMAX and the minimum value HMIN is compared with the threshold value Hb, not the difference with the value in the vicinity of the minimum value HMIN. Due to
Distance detection can be performed even on an object such as a curved surface whose brightness changes gently. (3) Brightness difference between adjacent pixels in the horizontal direction in the small area of the right image> Hc (when the threshold value Hc is increased, edge detection is performed, but it is also possible to deal with a case where the brightness changes gently. As described above, the threshold value Hc is set to be much lower than the normal edge detection level.This condition is based on the basic principle that distance detection cannot be performed in a portion where there is no change in luminance, and thus in a small area. Therefore, only the pixels whose distances are actually detected are adopted in the small area, and a natural result is obtained.) That is, the shift amount that minimizes the city block distance H is The calculated displacement amount d is obtained, but if the above three check conditions are satisfied, the displacement amount d is output to the distance image memory 60 in step S106. If not satisfied, no data is adopted. "0" to the distance image And outputs it to the memory 60, the process exits.

The distance distribution information stored in the distance image memory 60 is extracted from a portion of each pixel of a stereo image pair in which a difference in brightness between adjacent pixels in the left-right direction is large (a portion having distance data). When the coordinate system on the range image is the origin, the horizontal direction is the i coordinate axis, and the vertical direction is the j coordinate axis, the coordinate values (i, j, d) is obtained as distance distribution information (unit is pixel).

The distance distribution information stored in the distance image memory 60 is read by the intruder detecting device 70. In the intruder detection process in this intruder detection device 70, in step S201 of FIG. 4, coordinate values (i,
(j, d), the process proceeds to step S202, and the main camera (left camera) detected by the camera position detection device 14 is detected.
The depression angle θf with respect to the horizontal line of the optical axis of 11a and the rotation angle θr in the left-right direction are read, and the coordinate value of the distance image is converted into the coordinate value of the real space.

That is, as shown in FIG.
As shown in, the origin is the floor surface directly below the left camera 11a, and from this origin to the right camera 11b side, the X axis (X> 0),
Above the left camera 11a, the Y-axis (Y> 0), the left camera 11
a Extend the Z-axis (Z> 0) forward and set the coordinate system so that the plane formed by the X-axis and the Z-axis coincides with the floor surface. From the coordinate values (i, j, d) of the distance image, Three-dimensional position of the subject (X,
Y, Z) is calculated.

In this case, for example, as shown in FIGS. 6 and 7, if a point on the head of the intruder 4 is P and the coordinate value of this point P on the distance image is (ip, jp, dp). , The distance Ddp to the plane including the point P and perpendicular to the optical axis of the left camera (main camera) 11a is r, the mounting interval between the left and right cameras 11a and 11b is r, and the viewing angle per pixel is PW,
The length Gp of the straight line Lp connecting the point P and the left camera 11a can be calculated by the following equation (2).
It can be calculated by equation (3).

Ddp = r / (PW × dp) (2) Gp = (Dip2 + Djp2 + Ddp2) (3) where Dip = Ddp × PW × ip Djp = Ddp × PW × jp Further, it is obtained from the camera position detecting device 14. Depression angle θf
(See FIG. 7) and the rotation angle θr in the left-right direction with respect to the Z axis.
(See FIG. 8), the depression angle θfp and the rotation angle θrp of the straight line Lp can be obtained by the following equations (4) and (5).

Θfp = θf + jp × PW (4) θrp = θr + ip × PW (5) Therefore, assuming that the mounting height of the left camera 11a is CH, the three-dimensional position of the point P in the X, Y, Z coordinate system. (Xp, Y
p, Zp) can be obtained by the following equations (6), (7), (8).

Xp = Gp × cos (θfp) × sin (θrp) (6) Yp = Gp × sin (θfp) + CH (7) Zp = Gp × cos (θfp) × cos (θrp) (8) As described above, when all the points on the range image are converted into the three-dimensional position in the real space, the steps from the above step S202
In S203, as shown in FIG. 9, the three-dimensional space to be monitored has a side length of lk (for example, 0.2 m to 0.5 m).
m) The cubic grid is divided, and the number of data points included in each cubic grid is obtained.

When setting the above-mentioned cubic lattice, the cubic lattice can be three-dimensionally set in an arbitrary shape by setting the cubic lattice, for example, at a distance of several cm from the floor or wall surface. It is possible to prevent the wall surface and the wall surface from being detected as an object again.

Next, in step S204, in order to distinguish the existing object originally existing in the monitoring area such as the plant 2 or the table 3 from the intruder 4, based on the known three-dimensional position of the existing object, Exclude the applicable cubic lattice from monitoring.

When the three-dimensional position of the existing object is not known due to the rearrangement or the like, the observer confirms the situation where there is no intruder, and the object is detected from the distance image obtained by imaging the situation. Exclude the cubic lattice in which the object was detected. Thereby, the existing object and the intruder can be easily distinguished from each other, and it can be accurately determined whether or not the intruder is within the monitoring area.

That is, even if arbitrary objects are present in the monitoring area, they can be easily excluded from the detection targets and can be monitored in various places such as the interior of an office or a corridor. Further, even if there is a tree swaying in the wind outdoors, by excluding the swaying range from being detected, it is possible to accurately detect an intruder without erroneously detecting the tree.

After that, the process proceeds to step S205, and a histogram is created for the cubic lattice to be monitored with the number of data included in each cubic lattice as the frequency.
Then, a cubic lattice in which the frequency of the histogram exceeds a preset threshold is extracted, and it is determined whether or not an object exists.

In this case, noise-like data that is erroneously detected is also present in the range image, and some data appears at positions where no object actually exists on the above histogram. It is recognized as a floating point in the empty space, and the number of data is not so large.

That is, since the frequency of the cubic lattice in which the object exists has a large value and the frequency generated by the incorrect distance data has a small value, it is determined that the object exists when the frequency of the histogram exceeds the threshold value. By extracting the corresponding cubic lattice and determining that there is no object when the frequency is less than or equal to the threshold value, the influence of noise is minimized even if the image data contains some noise. Object can be detected.

If it is determined in step S206 that there is no object, the process exits, and if it is determined that there is an object, the process proceeds to step S207, where a plurality of cubic lattices determined to have an object are adjacent to each other. The cubic lattices are grouped together into one group, and each group is labeled.

Next, in step S208, the data for each group is projected on a two-dimensional image, and each data is connected by a line segment to create a contour image, and the shape dimensions such as the height and width of the object are determined. After the calculation, in step S209, the type of detected object (person, animal, automobile, etc.) is determined, and the process proceeds to step S210 to determine whether or not to output an alarm.

For example, when it is determined that the detected object is a small animal other than human such as a cat, and there is no need to output an alarm, the process exits from step S210, and the detected object has the same shape and size. When it is determined that there is an intruder, the process proceeds from step S210 to step S211 to activate the alarm device 90, and an alarm is generated by a buzzer or a lamp to notify the surveillance staff or the like.

Then, from step S211 to step S2
When the position of the center of gravity of the detected object is calculated in step 12, step S2
The process proceeds to step 13 to compare with the position of the center of gravity in the previous processing, and it is checked whether or not they match within a preset threshold value.

As a result, when the barycentric position of the detected object in the current process matches the barycentric position of the previous process within the threshold value, it is determined that the intruder is the same, and the steps S213 to S214 are performed.
Then, the camera rotation device 13 is driven to track the position of the center of gravity of the camera, the image of the main camera (left camera) 11a is recorded, and the intruding object is tracked / recorded to continue monitoring over a wide range.

On the other hand, in step S213, when the barycentric positions of the current process and the previous process do not match within the threshold value, it is determined that the intruder is a new intruder, the process proceeds to step S215, and the barycentric position is stored and processed. Get out.

[0062]

As described above, according to the first aspect of the present invention, the same monitoring area is imaged from different viewpoints .
The distance distribution over the entire image is calculated from the shift amount of the corresponding positions of the stereo image pair . Then, the three-dimensional position of each part of the subject corresponding to the distance distribution information is obtained, and the three-dimensional position of the monitoring area is calculated.
Divide the space into multiple cubic grids,
Calculate the number of data at the 3D position. And the background and
A cubic grid containing the 3D position data of an existing object is detected.
Since it is excluded as an out-of-going target, any object in the monitoring area
Even if they exist, they are excluded as detection targets accurately.
Can be removed to detect intruders in various places with high accuracy
can do. And on the cubic lattice to be monitored
On the other hand, the number of data included in each cubic lattice is used as the frequency
Create a gram and set the histogram frequency to a preset threshold
Extract the cubic lattice exceeding the value as having an object
, The frequency of the cubic lattice where the object exists shows a large value,
The frequency generated by incorrect distance data is small.
Therefore, when the histogram frequency exceeds the threshold,
Judge that an object exists, extract the corresponding cubic lattice,
If the frequency is below the threshold, it is determined that there is no object.
As a result, the image data may contain some noise.
Even if there is an object, the effect of noise is minimized
Can be detected. And it is determined that there is an object
For multiple cubic lattices, combine adjacent cubic lattices
One group, and data for each group is displayed in a two-dimensional
Contour by projecting on the image and connecting each data with a line segment
Create an image to identify the type of detected object and
Detects incidents, ensuring only intruders to be detected
Can be detected and even if it is an intruder, it will not be detected.
If it is an object and there is no need to notify it,
It can be implemented without any action, and can be confirmed only when notification is required.
It will be possible to actually inform, and the reliability will be significantly improved.
You can

According to the invention of claim 2 , the object to be detected
When an intruder that detects
The center of gravity of the detected object detected as is determined. And
The center of gravity of the detected object obtained this time is the center of gravity of the detected object obtained last time
Identical intrusion when the position matches within a preset threshold
It is determined to be an object, the position of the center of gravity is tracked, and the captured image is recorded.
When the center of gravity position of this time and the previous time do not match within the threshold value
, The position of the center of gravity is determined as a new intruder.
In addition to the effect of the invention according to claim 1,
Can be tracked and recorded for a wide range of monitoring
Has the effect of being able to continue.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a monitoring device

FIG. 2 is a circuit configuration diagram of a monitoring device.

FIG. 3 is a flowchart of shift amount detection processing.

FIG. 4 is a flowchart of intruder detection processing.

FIG. 5 is an explanatory diagram showing a coordinate system.

FIG. 6 is an explanatory diagram showing a positional relationship between coordinate values on a distance image and a target object.

FIG. 7 is an explanatory diagram showing a positional relationship between a distance image and a target object in a vertical section.

FIG. 8 is a top view showing a left-right rotation angle of the camera.

FIG. 9 is an explanatory diagram showing division of a three-dimensional space.

[Explanation of symbols]

1 Monitoring device 10 Stereo optical system 20 Stereo image processing device (stereo image processing means) 70 Intruder detection device (intruder detection means)

─────────────────────────────────────────────────── ─── Continuation of the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T 1/00 315 G06T 1/00 340 H04N 7/18

Claims (2)

(57) [Claims] 1. Stereo image processing means for processing a pair of stereo images obtained by capturing the same monitoring area from different viewpoints, and calculating a distance distribution over the entire image from the shift amount of the corresponding positions of the stereo image pairs; The three-dimensional position of each part of the subject corresponding to the distance distribution information from the image processing means is obtained, and the three-dimensional position of the monitoring area is calculated.
The original space is divided into multiple cubic lattices and included in each cubic lattice
The number of data of the above-mentioned three-dimensional position is calculated, and the background and existing
Detects a cubic lattice containing three-dimensional position data of various objects
It is excluded as an outside of the elephant, and each cube is added to the cubic lattice to be monitored.
Histogram with the number of data included in the square grid as the frequency
Created and the histogram frequency exceeds a preset threshold
The existing cubic lattice is extracted as an object, and adjacent cubes are extracted.
The square grids are put together into one group, and the data for each group is
The data is projected on a two-dimensional image and each data is connected by a line segment.
By connecting, a contour image is created to identify the type of detected object,
Intruder monitoring device, characterized in that it comprises a intruder detection means for detecting an intruder to be detected. Wherein said intruder detection means, invasion to be detected
When an incident is detected, it is notified and the
The centroid position of the detected object detected by
The position of the center of gravity of the detected product is set in advance as the position of the center of gravity of the detected product
When they match within the specified threshold, it is judged as the same intruder,
Tracking the position of the center of gravity and recording the captured image,
If the previous barycentric position does not match within the threshold, a new
The intruder monitoring device according to claim 1, wherein the barycentric position is recorded as being an intruder.