JP2008146133A - Abnormality detector, program, and abnormality detection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology for hardly generating the erroneous decision of abnormality detection by solving the problem of a conventional abnormal detector that even when a person in a monitor area performs an operation such as physical exercises, the operation is determined to be abnormal. <P>SOLUTION: This abnormality detection device is provided with a periodic operation featured value calculation part 5 for calculating a periodic operation featured value for the purpose of determining an abnormal action or a normal action and a periodic operation determination part 6 for comparing the periodic operation featured value with a periodic operation featured determination threshold set by a threshold setting part 8 for determining the presence/absence of a periodic operation. Thus, it is possible to detect abnormality only with respect to such an abnormal operation as a riot without erroneously detecting such a periodic operation as physical exercises as an abnormal operation. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an anomaly detection technique for performing predetermined processing in response to an image captured from a photographing apparatus such as a camera.

  A conventional abnormality detection device calculates, for example, the direction and magnitude of movement of each point in an image from an image of a camera or the like, a variation amount of a person's movement, a person's abnormal behavior (for example, rampage) determination threshold, Are compared to determine the abnormal behavior of the person (see, for example, Patent Document 1).

JP 2006-276969 A (page 3-4)

  However, in the conventional abnormality detection device, the direction and magnitude of the movement of each point in the image is calculated from the image, and compared with a person's abnormal action determination threshold. Therefore, there is a problem that even if a person in the monitoring area performs an operation such as gymnastics, it is determined as an abnormal action.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a technique that is unlikely to cause erroneous determination of abnormality detection.

  In order to achieve the above object, the invention according to claim 1 is an abnormality detection device that performs predetermined processing in response to an image taken by a photographing device such as a camera, and is photographed by the photographing device. An image capturing unit that captures an image, a threshold setting unit that sets an operation amount determination threshold for identifying abnormal behavior and a periodic motion feature determination threshold, and two images captured at different times by the image capturing unit A motion amount calculation unit that calculates a motion amount based on the above, a motion amount determination unit that compares the motion amount with the motion amount determination threshold, and a periodic motion feature value that calculates a periodic motion feature value of the motion amount A calculation unit and a periodic motion determination unit that compares the periodic motion feature value with the periodic motion feature determination threshold are provided to detect an abnormality.

  The invention according to claim 2 is a program, wherein the computer is configured to input an image captured by an imaging device such as a camera. The computer includes an image capturing unit that captures the image, and abnormal behavior. Threshold value setting means for setting an action amount determination threshold value and a periodic action feature determination threshold value for identification, and an action amount calculation for calculating an action amount based on two images captured by the image capturing means at different shooting times Means, a motion amount determination means for comparing the motion amount with the motion amount determination threshold, a periodic motion feature value calculation means for calculating a periodic motion feature value of the motion amount, and the periodic motion feature value An abnormality is detected by functioning as a periodic motion determination means for comparing the periodic motion feature determination threshold.

  The invention according to claim 3 is an abnormality detection method for performing predetermined processing in response to an input of an image, the image capturing step for capturing the image, an operation amount determination threshold for identifying abnormal behavior, and a cycle A threshold setting step for setting a dynamic motion feature determination threshold, a motion amount calculating step for calculating a motion amount based on two images captured at the image capturing step and having different shooting times, and the motion amount and the motion amount A motion amount determination step for comparing a determination threshold value, a periodic motion feature value calculation step for calculating a periodic motion feature value of the motion amount, and the periodic motion feature value and a periodic motion feature determination threshold value are compared. It has a periodic operation determination step, and an abnormality is detected.

  According to the first to third aspects of the present invention, the state of the gymnastic exercise is determined by the imaging device that images the monitoring area and the periodic motion feature value calculated by the periodic motion calculation unit based on the image captured by the imaging device. Since the movement is determined to be a normal action and can be determined to be an abnormal action only when it is rampant, there is an effect that erroneous determination of abnormality detection is less likely to occur than conventional abnormality detection devices.

  Furthermore, since the present invention determines abnormal behavior based on the amount of change in luminance of the image in the monitoring area, the abnormal behavior can also be determined using a device that can measure the luminance of the monitoring area as an input device, for example, a luminance meter. There is also an effect that can be done.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the embodiment of the present invention, the monitoring area is a place where a background change is scarce and a person's movement is captured clearly, such as an elevator.

  FIG. 1 shows a block diagram of the abnormality detection device in the present embodiment. The abnormality detection device captures the image capturing device 1 such as a CCD camera for capturing a predetermined area to be monitored, the image capturing unit 2 for capturing data output from the image capturing device 1, and the image capturing unit 2. The operation amount calculation unit 3 that calculates the operation amount in the screen from the data, the operation amount and the threshold value stored in the operation amount determination threshold storage unit 9 described later are compared, and the operation amount and the operation amount determination threshold value are obtained. The movement amount determination unit 4 to be compared with a period so as not to erroneously determine a periodical movement as a physical exercise but a periodic movement when there is a possibility of an abnormal action within the shooting range. A periodic motion feature value calculation unit 5 that calculates a dynamic motion feature value, and compares the periodic motion feature value with a threshold value stored in a periodic motion feature determination threshold value storage unit 10 to be described later to determine abnormal behavior Periodic motion determination unit For example, an input device 7 for setting with a keyboard or a mouse, a threshold setting unit 8 for setting a motion amount determination threshold and a periodic motion feature determination threshold using the input device 7, and a motion amount determination threshold are stored. An action amount determination threshold value storage unit 9, a periodic operation feature determination threshold value storage unit 10 that stores a periodic operation feature determination threshold value, an action amount determination unit or a periodic operation determination unit, etc. The alarm output unit 11 sends a message to the monitor device and the like, and the monitor device 12 outputs the data of the image capturing unit 2 and the alarm output unit 11 to the screen.

  The image capturing unit 2, the motion amount calculation unit 3, the motion amount determination unit 4, the periodic motion feature value calculation unit 5, the periodic motion determination unit 6, the threshold setting unit 8, and the alarm output unit 11 are all, for example, predetermined calculations A ROM and a hard disk that store the processing program, a CPU that executes the arithmetic processing program, a RAM that temporarily stores the read image data, and the like. Further, the motion amount determination threshold value storage unit 9 and the periodic motion feature determination threshold value storage unit 10 are configured by a storage medium such as a hard disk, for example.

  In the block diagram of FIG. 1, only one photographing device 1 and one monitoring device 12 are provided, but a plurality of photographing devices 1 and monitoring devices 12 may be installed as necessary.

  1 shows an embodiment of an abnormality detection technique according to the present invention. In this embodiment, a camera is used as an imaging device in an elevator, and an abnormal behavior in a captured image is determined. The camera may be monochrome or a color camera. For example, the computer may be a known personal computer (PC) having a video capture circuit for capturing a moving image. In addition, a monitor, a keyboard, and a mouse are connected to the computer.

  Details of such operations will be described below. FIG. 2 is a flowchart showing the operation of the abnormality detection device. The operation of the abnormality detection apparatus will be described with reference to the block configuration shown in FIG. 1 and the flowchart shown in FIG. In step S1, an image is taken from the photographing apparatus 1, and the image is input.

  In the case of an elevator or the like shown in the present embodiment, it is possible to prevent erroneous detection of abnormal behavior and improve the accuracy of operation amount determination by performing door opening / closing determination in step S2. The door opening / closing determination is based on a method of determining a door opening / closing determination image near a door where a person in the image does not enter and determining that the door is open when the average brightness of the door opening / closing determination image changes. Further, instead of determining from the image, the determination may be made from a signal related to opening / closing from a door opening / closing motor, an opening / closing sensor or the like.

  Next, in the calculation of the movement amount in step S3, the movement amount calculation unit 3 calculates the movement amount. The operation amount is calculated by using two images whose photographing times are different by a predetermined time with respect to the image input in step S1. The amount of movement is, for example, a standard value within a certain continuous frame with respect to an average value of the difference image of two images whose shooting times differ by a predetermined time or an average value of the difference image of the two images. The deviation is obtained as the standard deviation within a certain continuous frame with respect to the correlation value between the two images and the correlation value between the two images. The two images that differ for a predetermined time can be two consecutive images depending on the environment to be used. However, by changing the interval of the images to be compared, an image with an interval of 4 frames or an interval of 16 frames It can also be an image.

  Next, in the determination of the operation amount in step S4, the operation amount determination unit 4 compares the operation amount determination threshold stored in the operation amount determination threshold storage unit 9 with the operation amount calculated in step S3, and the operation amount Is larger than the motion amount determination threshold, it is considered that there is a possibility of abnormal behavior, and data characterizing the periodic motion in step S5 is calculated.

  For example, the following method is used as the method of calculating data characterizing the periodic motion in the periodic motion feature value calculation unit 5 in step S5.

  The calculation method of the data characterizing the periodic motion is, for example, the correlation between images at the time when the motion amount value reaches the peak when the motion amount values of the two images calculated in step S1 are displayed on a graph. A method is used in which a value is obtained and the periodic motion feature value is large when the correlation value is high. FIG. 3 shows a luminance average value of a difference image between two images having a large motion but having a continuous periodic motion such as a gymnastic exercise. FIG. 4 shows an average luminance value of a difference image between two consecutive images during abnormal behavior. When the correlation is obtained for each image at the time indicated as “peak value” in FIG. 3 and FIG. 4, the correlation value becomes high in the case of the periodic operation of FIG. In the case of behavior or the like, the correlation is low compared to the periodic motion.

  In addition, a method for calculating data characterizing a periodic operation includes, for example, an interval for calculating a correlation value between two images in continuously captured images, a current image, an image one frame before, and a current image. A method is used in which the periodic motion feature value is high when the difference between the values calculated by shifting little by little between the current image and the previous 4 frame image, the current image and the previous 8 frame image is small. For example, in the case of a correlation value during a periodical movement such as gymnastics, although the movement is large, the position of the object is almost constant, so the correlation value remains constant even if the interval for correlating images is changed. However, in the case of abnormal behavior, the position at the time of operation is not constant, so the value changes greatly if the correlation interval is changed.

  In addition, a method for calculating data that characterizes periodic motion is, for example, that a spectrum is calculated for the motion amount values of the two images calculated in step S1, and data that increases in value for a certain frequency is obtained. A method is used in which the periodic motion feature value is high. The data from which the spectrum is calculated may be, for example, the luminance average value of the difference image between the two images or the correlation value between the two images. As a method for calculating the spectrum, for example, a well-known Fourier transform or the like is used. A certain frequency can be set arbitrarily according to the environment. FIG. 5 shows a spectrum calculated with respect to an average luminance value (FIG. 3) of a difference image between two images having continuous periodic operations. FIG. 6 shows a spectrum calculated for the average luminance value (FIG. 4) of the difference image between two images during abnormal behavior such as rampage. In the case of the periodic operation in FIG. 5, the spectrum value becomes large at the frequency component of 6 Hz, and in the case of the abnormal behavior such as the violence in FIG. 6, the spectrum value becomes small as compared with the periodic operation.

Finally, in the periodic motion determination in step S6, the periodic motion determination unit 6 uses the periodic motion feature determination threshold stored in the periodic motion feature determination threshold storage unit 10 and the periodic motion feature calculated in step S5. When the periodic motion feature value is smaller than the periodic motion feature determination threshold value, it is determined that the periodic motion is not performed and the behavior is determined to be abnormal.
The periodic motion feature value to be compared with the periodic motion feature determination threshold may be compared with a plurality of values calculated by the above-described method or the like according to the environment to be used, or may be compared with only one. . The processes from step S1 to step S6 are repeatedly executed.

  As described above, the operation amount in the image is calculated from the image capturing device 1 that captures the monitoring area, and two images captured by the image capturing device 1 are different from each other by a predetermined time, and the feature value of the periodic operation is calculated. By calculating, it is possible to determine whether the motion is periodic, and a detection technique can be obtained as long as a large movement such as gymnastics is not erroneously determined as abnormal behavior.

It is a block diagram which shows the structure of an abnormality detection apparatus. It is a flowchart which shows operation | movement of an abnormality detection apparatus. It is a graph of the luminance value at the time of periodic operation. It is a graph of the luminance value at the time of abnormal action of rampage action. It is a spectrum of the luminance value at the time of periodic operation. It is a spectrum of the luminance value at the time of abnormal action of rampage action.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Imaging device 2 Data acquisition part 3 Motion amount calculation part 4 Motion amount determination part 5 Periodic motion feature value calculation part 6 Periodic motion determination part 7 Input device 8 Threshold setting part 9 Motion amount determination threshold storage part 10 Periodic motion feature Determination threshold storage unit 11 Alarm output unit 12 Monitor device

Claims (3)

An anomaly detection device that performs a predetermined process in response to an image captured by an imaging device such as a camera,
An image capture unit;
A threshold setting unit for setting a motion amount determination threshold and a periodic motion feature determination threshold for identifying abnormal behavior;
An operation amount calculation unit that calculates an operation amount based on two images with different shooting times captured by the image capturing unit;
An operation amount determination unit that compares the operation amount with the operation amount determination threshold;
A periodic motion feature value calculator for calculating a periodic motion feature value of the motion amount;
An abnormality detection device comprising: a periodic motion determination unit that compares the periodic motion feature value with the periodic motion feature determination threshold value. In a computer configured to input an image taken by a photographing device such as a camera,
The computer,
Image capturing means;
Threshold setting means for setting a motion amount determination threshold and a periodic motion feature determination threshold for identifying abnormal behavior;
A motion amount calculating means for calculating a motion amount based on two images having different shooting times captured by the image capturing means;
An operation amount determination means for comparing the operation amount with the operation amount determination threshold;
Periodic motion feature value calculating means for calculating a periodic motion feature value of the motion amount;
A program that functions as a periodic motion determination unit that compares the periodic motion feature value with the periodic motion feature determination threshold value. An abnormality detection method that performs predetermined processing in response to image input,
An image capture step;
A threshold setting step for setting a motion amount determination threshold and a periodic motion feature determination threshold for identifying abnormal behavior;
An operation amount calculating step for calculating an operation amount based on two images having different shooting times captured by the image capturing step;
An operation amount determination step for comparing the operation amount with the operation amount determination threshold;
A periodic motion feature value calculating step of calculating a periodic motion feature value of the motion amount;
An abnormality detection method comprising: a periodic motion determination step of comparing the periodic motion feature value with a periodic motion feature determination threshold value in the previous period.

Images