JP2008224396A - Moving body detection device and moving body detection system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To minimize malfunctioning by limiting to large noncontiguous operation by a detecting object, and to accurately detect the movement of the moving body. <P>SOLUTION: This detection device for detecting the movement of the moving body comprises a means for dividing a picked up image of a monitoring camera 2 into a plurality of regions according to a monitoring region, an extraction section 321 of brightness values in the regions, a calculation section 322 of standard deviation (dispersion) of the brightness of each region divided from the picked up image, a calculation section 323 of time standard deviation in the regions, and a means for detecting the operation of the detecting object in the region based on the obtained standard deviation value of the variation. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a mobile body detection apparatus and a mobile body detection system used for a mobile body detection apparatus for a mobile body, for example, a bed removal detection apparatus that detects bed removal from a bed of a care recipient such as a hospital.

As a moving body detection apparatus, for example, various apparatuses have been developed for detecting a bed leaving a detection target person such as an inpatient. In these detection devices, for example, a mat with a sensor is used as a detection region, and this is laid on the side of the bed or on the bed to detect the motion of the detected person to be detected, or a clip is attached to the detected person. Thus, when the person to be detected leaves the bed, the clip is released and the bed is detected.
However, mat sensors that place mats on the bed or on the bed often have false detections, and it is necessary to take care such as affecting the sleep of the person being detected and preventing the caregiver from touching the sensor. . The method of attaching the clip also puts a burden on the person to be detected, so that the person to be detected may remove it.

Therefore, in order to detect getting out of bed without imposing a burden on the person to be detected or caregiver, an infrared pattern model of the body movement of the person to be detected is stored, and a pyroelectric sensor that detects infrared rays emitted from the person to be detected Has been proposed in which the body motion of the detected person is detected from the infrared pattern detected by the pyroelectric sensor (see Patent Document 1).
However, when detection is performed using this pyroelectric sensor, it is not always easy to determine the body movement only with infrared rays emitted from the person to be detected without noise, and there is a possibility of misperception.

  Therefore, detection by a monitoring camera is conceivable, but in order to detect a person leaving the bed with one monitoring camera, it is necessary to set the bed and other areas as monitoring areas. Therefore, for example, the area is set by placing a sheet of a predetermined color or pattern on the bed or the area outside the bed. However, since the captured image of the camera changes greatly depending on the surrounding environment, this method cannot reliably set the area, and as a result, it is difficult to accurately detect the bed leaving.

In view of this, an apparatus that divides an imaging screen including a bed into, for example, a bed area and an out-of-bed area and detects the movement of the human body in each area to detect getting out of bed is also known (see, for example, Patent Document 2). ). In this bed leaving detection device, a floor area other than the bed area and the area where the bed is placed and a boundary side representing the boundary between the bed area and the floor surface area are set in advance, and the boundary side is passed from the floor area. When a human body movement to the bed area is detected, this event is judged as an entrance event.When a human body movement from the bed area to the floor area is detected, this event is regarded as a bed leaving event. to decide. At this time, if it does not pass through the boundary, it is regarded as a movement in the bed space such as moving the limbs or turning over in the bed, and is not determined to be out of bed.
However, with this bed detection device, it is not possible to distinguish between the movement of a patient in the bed and the person in the vicinity of the bed in a place where multiple patients or caregivers go in and out, such as a hospital bed. There is also a possibility of detecting that the person is getting out of bed when sitting on the bed.

JP 2006-136666 A Method JP 2002-230533 A

  The present invention has been made to solve the above-described conventional problems, and its purpose is to change the operation to be detected by the detected object to a large non-continuous operation (the gradation of the captured image). By limiting the variance (standard deviation) to the variance (standard deviation) over which the variance (standard deviation) is greater than a predetermined threshold), the malfunction is minimized, and when this is used as a bed leaving detection means, bed leaving is more accurate. It is to be able to detect.

The invention according to claim 1 is a moving body detection apparatus, wherein the captured image of the monitoring camera is divided into a plurality of areas of the monitoring area, the gradation value in the divided area is extracted, and the extracted gradation value Means for obtaining the standard deviation value, means for obtaining the standard deviation of the temporal change amount for each obtained standard deviation value, and the detected object in the region based on the obtained standard deviation value of the temporal change amount And a means for detecting the operation of.
According to a second aspect of the present invention, in the moving body detection apparatus according to the first aspect, the means for detecting the operation of the detected object compares the standard deviation value of the temporal change amount with a predetermined threshold value, A pattern in which the standard deviation value of the change amount is binarized with the threshold value is created, and an operation of the detection target is detected based on a change with time of the pattern.
According to a third aspect of the present invention, in the moving body detection apparatus according to the first or second aspect, the means for dividing the captured image into a plurality of regions divides the captured image into a bed region and an out-of-bed region. The operation is characterized in that the person to be detected is getting out of bed.
The invention of claim 4 is a mobile body detection system, and an alarm is issued when an abnormality is detected by the mobile body detection device according to any one of claims 1 to 3, the imaging camera, and the mobile body detection device. It has the alarm means which generate | occur | produces, It is characterized by the above-mentioned.

(Function)
According to the present invention, the movement of the detected object is detected by a change in gradation for each divided area, and at that time, the amount of change over time in the standard deviation of the gradation for each area as a parameter for detection in particular. Standard deviation (here, referred to as temporal standard deviation) was adopted. In addition, a pattern in which the temporal standard deviation is binarized for each region based on a threshold value is formed, and the operation of the detected object is determined based on the temporal change of this pattern. Thereby, it is not influenced by the continuous small continuous movement of the detected object, and this is reliably detected only when a large intermittent movement of the detected object in the monitoring region occurs.

  According to the present invention, since the operation to be detected by the surveillance camera can be limited to only a large non-continuous operation, the erroneous detection of the moving object detection device can be minimized, and more detailed detection of the object to be detected. Is possible. Moreover, when this is used as a bed leaving detection means, it is to be able to detect bed leaving more accurately.

The embodiments of the present invention shown in the drawings will be described with reference to the drawings.
FIG. 1 is a block diagram schematically showing a configuration of an embodiment of a moving body detection system of the present invention.
The bed leaving detection system includes a moving body detection device 1, a monitoring camera 2, a memory 4, and an alarm notification unit 5 as shown in the figure.
The moving body detection apparatus 1 includes an image processing unit 3 that processes an image captured by a monitoring camera and a control unit 6 that controls the entire system.

The image processing unit 3 includes an intra-region change detection unit 32. The intra-region change detection unit 32 further includes an intra-region luminance value extraction unit 321, an intra-region variance (standard deviation) calculation unit 322, and an intra-region variance (standard deviation). ) Of time dispersion (standard deviation) calculation unit 323, and the change of the detected object in the set area obtained by dividing the imaging screen by the monitoring area setting unit 30, the gradation in each divided area (this embodiment) It is extracted as the standard deviation of the temporal variation of the standard deviation of the luminance in the form.
The memory 4 is an image memory that stores an image captured by the monitoring camera 2.
The alarm notification unit 5 generates a visible or audible alarm according to an instruction from the control unit 6 when the detected person leaves the floor, for example.
The control unit 6 is composed of a microcomputer including a CPU 61, a RAM 62, and a ROM 63. The control unit 6 instructs the surveillance camera 2 to take an image, and instructs the image processing unit 3 to process an image taken by the surveillance camera 2. Control the entire system.
The ROM 63 stores a program for executing the present bed detection system, and the RAM 62 temporarily stores data necessary for the processing of the CPU 61, or a change amount for each standard deviation value described later. The standard deviation threshold value and the standard deviation of the change amount of the standard deviation value due to the movement and movement of the detected object are binarized depending on whether or not the predetermined threshold value is exceeded, and a pattern is formed for each region. A table in which the time-series fluctuations are associated with the actual motion of the detected object is stored.

  In this embodiment, the monitoring camera 2 is installed on the upper part of the bed of a person to be detected such as a care recipient, and first, necessary processing is performed on an image obtained by imaging the bed and its surroundings and stored in the memory 4. To do.

  In the present embodiment, in accordance with the monitoring area set by the monitoring area setting unit 30, an image of the peripheral area including the bed imaged by the monitoring camera 2 is divided into a bed area and an outside bed area. A variance value of the luminance of the pixel is obtained, and a standard deviation is calculated from the variance value. Next, the standard deviation value (temporal standard deviation) of the temporal variation (time change) of the obtained standard deviation value of each region is obtained, and based on this temporal standard deviation value, the detected object in each region, Here, the movement or movement of the person to be detected in the bed or the area outside the bed is detected, and the bed leaving is monitored. When it is determined that the person is getting out of bed, a visual or audible alarm is issued from the alarm reporting unit 5.

Next, an algorithm for executing detection in each region of a captured image used in the moving object detection apparatus according to the present embodiment by the moving object detection apparatus will be described.
In the present embodiment, first, a luminance value indicating the brightness per pixel of each area is detected from the captured image according to the bed area and the out-of-bed area where the monitoring area is set.
Here, the luminance dispersion of each region and the temporal change of the dispersion value are obtained by the following equations 1 to 4.

A standard deviation σ is obtained from the variance value σ ² . Here, σ _t represents the standard deviation of time variation of the standard deviation of luminance of each area (referred to as temporal standard deviation), and this temporal standard deviation σ _t is calculated for each area (bed area, out-of-bed area). The temporal standard deviation of the bed area is σ _t1 , and the standard deviation of the non-bed area is σ _t2 .

In FIG. 2A, the vertical axis represents the standard deviation σ (or temporal standard deviation σ _t ) and the horizontal axis represents time, and the luminance standard deviation σ (or temporal standard deviation σ _t ) in a certain region changes with time. It is the graph which showed.
For example, if there is a continuous movement in the bed area, the standard deviation σ of that area constantly changes and increases as shown by the curve A in FIG. 2A.
However, when this change is seen by the temporal standard deviation σ _t , if the movement is continuous, if the movement in the bed area continues, no further change will occur. As shown by a curve B in FIG. 2A, it eventually converges to a constant value.
Note that when the bed area changes from a bright and stable state to a completely dark state (for example, when the light is turned off), the luminance in the area changes uniformly, so the standard deviation σ of the luminance does not increase.

FIG. 2B is a graph similar to FIG. 2A, but shows a state in which a large change has occurred in the movement of the person to be detected in the bed area. In this case, a large change appears in the luminance standard deviation σ, and the temporal standard deviation σ _t is also larger than a predetermined threshold (TH). Note that the threshold is appropriately determined, for example, by experimentally determining the actual movement of the person to be detected.

In this embodiment, when the temporal standard deviation σ _t is detected and exceeds a predetermined threshold value (TH), it is determined that a change to be detected has occurred in the region, and the threshold value is not exceeded. Determines that there is no change to be detected in the area.

Next, a description will be given of a case where the detection algorithm described above is actually detected for getting out of the person to be detected.
FIG. 3 is a view for explaining the monitoring area, FIG. 3A is a side view showing the relationship between the bed and the attachment position of the monitoring camera, and FIG. 3B is a plan view for explaining the monitoring area.
That is, a surveillance camera is installed on the head of a person lying on the bed, and the lens is directed obliquely downward so that the surrounding area including the bed enters the imaging range. As shown in FIG. 3B, a bed area and an outside-bed area are set as the imaging area, and one of the beds is in contact with the wall here.

  The monitoring camera 2 captures images at predetermined time intervals or according to a command from the control unit 6, and the captured images are sent to the image processing unit 3. The image processing unit 3 extracts the luminance value in the area from the sent image, obtains the luminance dispersion (standard deviation) in the area, and stores this in the memory. The intra-regional variance (standard deviation) calculation unit obtains the temporal standard deviation of the standard deviation in the region from the temporal change of the intra-area standard deviation (or variance) accumulated in the memory. This temporal standard deviation is compared with a predetermined threshold value for each region, and binarized depending on whether or not the threshold value is exceeded. Based on the binarized value of each area, a binarized pattern of all areas is formed, and the behavior of the detected object is discriminated from the temporal change of this pattern.

FIG. 4 is a diagram for explaining a procedure together with each state in the case of detecting the person leaving the bed actually lying on the bed that is the detection target based on the principle described above.
That is, when the temporal standard deviation σ _{t1 of the} bed area is larger than the threshold value (TH) and the temporal standard deviation σ _{t2 of the} non-bed area is smaller than the threshold value (TH), that is, the binarized pattern is σ _{If t1} = 1 and _σt2 = 0, it is clear that the motion of the person to be detected remains in the bed area, so that the temporal standard deviation pattern immediately before that is, for example, the temporal standard deviation σ of the bed area _{If t1} is smaller than the threshold (TH) and the temporal standard deviation σ _{t2 of the} out-of-bed area is smaller than the threshold (TH), that is, the binarized pattern is σ _t1 = 0, σ _t2 = If it is 0, the CPU 61 of the control unit 6 refers to, for example, a comparison table stored in the RAM 62, and the detected person changes from a sleeping state or lying on the bed to a change such as getting up or turning over. That it happened I refuse. This state is a change from state 1 to state 2 in the figure.

From this state (state 2), when the temporal standard deviation σ _{t2 of the} out-of-bed area is larger than the threshold value (TH) (state 3 in the figure), the binarized pattern of the temporal standard deviation is σ _{Since t1} = 1 and σ _t2 = 0 change to σ _t1 = 1 and σ _t2 = 1 (change from state 2 to 3 in the figure), the CPU 61 of the control unit 6 is similarly stored in the RAM 62. With reference to the comparison table, it is determined that the operation corresponds to the pattern change from the above state 2 to 3, that is, the detected person is sitting on the bed, for example.

When the temporal standard deviation σ _{t1 of the in-} bed area becomes smaller than the threshold value (TH) from the state 3 above (state 4 in the figure), the binarized pattern of the temporal standard deviation is σ _t1 = 1. , Σ _t2 = 1, σ _t1 = 0, and σ _t2 = 1 (in the figure, the change from state 3 to 4), the CPU 61 of the control unit 6 is similarly controlled in the RAM 62. Referring to the table, it is determined that the detected person has left the floor.

When the temporal standard deviation σ _{t1 of} the in-bed area is also smaller than the threshold value (TH) from the above state 2 (state 1 in the figure), the binarized pattern of the temporal standard deviation is σ _t1 = 1, Since σ _t2 = 0 changes to σ _t1 = 0 and σ _t2 = 0 (in the figure, the change is from state 2 to 1), the CPU 61 of the control unit 6 similarly stores the contrast table stored in the RAM 62. To determine that the detected person has gone to bed.

Next, when the temporal standard deviation σ _{t2 of the} out-of-bed area is larger than the threshold value (TH) from the above state 1, the binarized pattern of temporal standard deviations is σ _t1 = 0, σ _t2 = 0 to σ _{Since t1} = 0 and σ _t2 = 1 are changed (in the figure, the change is from state 1 to 5), the CPU 61 of the control unit 6 refers to the comparison table stored in the RAM 62 in the same manner. It is determined that someone has entered the bed around which the detector is sleeping, and the alarm notification unit 6 is activated as necessary.

When the temporal standard deviation σ _{t2 of the} out-of-bed area again becomes smaller than the threshold (TH) from the state of 5 above, the binarized pattern of temporal standard deviations is σ _t1 = 0, σ _t2 = 1 to σ _t1 = 0, σ _t2 = 0 (in the figure, it is a change from state 5 to 1), the CPU 61 of the control unit 6 refers to the comparison table stored in the RAM 62 and moves around the bed. It is judged that the intruder has passed through as it is, and the operation of the alarm reporting unit 6 is stopped.

When the temporal standard deviation σ _{t1 of} the in-bed area becomes larger than the threshold value (TH) from the state 5 above (state 6 in the figure), the binarized pattern of temporal standard deviations is σ _t1 = 0, σ _{Since t2} = 1 changes to [sigma] _t1 = 1 and [sigma] _t2 = 1 (in the figure, it is a change from state 5 to 6), the CPU 61 determines according to which state is to be shifted next. That is, when only the temporal standard deviation σ _{t1 of the in-} bed area becomes smaller than the threshold value (TH) (state 5 in the figure), the binarized pattern of temporal standard deviations is σ _t1 = 1, σ _t2 = 1 to σ _t1 = 0 and σ _t2 = 1 (in the figure, it is a change from state 6 to 5), the CPU 61 refers to the control table, and the invading person makes care such as bed-making. It is determined that

When the temporal standard deviation σ _{t2 of the} out-of-bed area again becomes smaller than the threshold (TH) from the state 6 above, the binarized pattern of temporal standard deviations is σ _t1 = 1, σ _t2 = 1 to σ _t1 = 1, σ _t2 = 0 (in the figure, it is a change from state 6 to 7), the CPU 61 determines that a person has entered the floor with reference to the comparison table from this change pattern.

  Similarly, if the state 3 is shifted from the state 3 (that is, the state of getting out of bed) to the state 2 described above, it can be seen that the person to be detected is getting up or turning over in the bed.

  According to the embodiment of the present invention, it is divided into a bed area and an out-of-bed area, and the standard deviation value of the time variation of the standard deviation of the luminance dispersion value for each pixel of each captured image, that is, the temporal standard deviation value. Attention is paid to detect the detected person's behavior based on the change in the magnitude of the change, so even if the detected person is constantly moving without being confused by turning off or turning on, the behavior changes greatly. Otherwise, it can be excluded from the change in the monitoring area, and the detection subject's bed can be detected more accurately.

  As described above, the case where the present movement detection apparatus is used as a person's bed detection apparatus has been described. However, the present invention is not limited to this and can be applied to other moving body detection. For example, the movement of a person in a specific area is monitored, or the above-mentioned area is set to 3 or more and the binarized pattern of temporal standard deviation of each area is used to detect more complicated movements of people and objects. It can be detected accurately. For example, it is possible to monitor the movement of a moving body having a predetermined movement pattern and to determine whether or not the movement is predetermined.

It is a block diagram which shows roughly the bed detection apparatus as one Embodiment of the mobile body detection apparatus of this invention. Taking the standard deviation σ (or simply the temporal standard deviation σ _t ) on the vertical axis and the time on the horizontal axis, the time variation of the luminance standard deviation σ (or the standard deviation temporal standard deviation σ _t ) in a certain region. It is the graph which showed. FIG. 3A is a side view showing the relationship between the bed and the attachment position of the surveillance camera, and FIG. 3B is a plan view for explaining the surveillance area. It is a figure for demonstrating the detection which concerns on this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Moving body detection apparatus, 2 ... Surveillance camera, 3 ... Image processing apparatus, 4 ... Memory, 5 ... Alarm alert part, 6 ... Control part, 30 ... Monitoring region setting unit, 32... Region change detection unit, 321... Region luminance value extraction unit, 322... Region dispersion (standard deviation) calculation unit, 323. ) Calculation unit.

Claims (4)

  Means for dividing the monitoring area of the captured image of the monitoring camera into a plurality of areas, means for extracting gradation values in the divided areas, means for obtaining standard deviation values of the extracted gradation values, and each acquired standard deviation A means for acquiring a standard deviation of a temporal change amount for each value; and a means for detecting an operation of the detected object in the region based on the obtained standard deviation value of the temporal change amount. A moving body detection device. In the moving body detection apparatus according to claim 1,
The means for detecting the operation of the detected object compares the standard deviation value of the temporal change amount with a predetermined threshold value, and creates a pattern in which the standard deviation value of the change amount is binarized by the threshold value, A moving body detection apparatus that detects an operation of an object to be detected based on a temporal change of the pattern. In the moving body detection apparatus according to claim 1 or 2,
The means for dividing the captured image into a plurality of areas is divided into a bed area and an out-of-bed area, and the movement of the detected object is a bed leaving the detected person.   A moving body detection device comprising: the moving body detection device according to any one of claims 1 to 3; an imaging camera; and an alarm unit that generates an alarm when an abnormality is detected by the moving body detection device. system.

Images