JP2006276969A - Violence detecting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a simple violence detecting device hardly causing incorrect determination with respect to installation conditions of a camera, change of an environment, and entry and exit actions of a person following movement of a door in various places such as a closed space as in an elevator car. <P>SOLUTION: A movement calculating means is provided for calculating a direction and magnitude of movement of each point in an image from two images wherein photographing times are different by a predetermined time, and calculating a dispersion amount of a person's movement of at least one of a direction or magnitude of a person's movement from the direction and the magnitude of movement of each point. A violence determining means is provided for comparing the dispersion amount of the person's movement with a violence determination threshold to determine violence of the person. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a ramp detection device that automatically detects the occurrence of a rampant motion of a person from a camera image in an elevator car or the like.

In the conventional rampage detection device, a part of the body such as the head, arms, and legs is extracted from the image, and the rampage is detected from the positional relationship and movement (for example, Non-Patent Document 1).
Also, in terms of crime detection, it detects passengers in the elevator car and analyzes the number of passengers, passenger gender, passenger movement speed, identity of the floor on which they were boarded, relative distance of passengers, and passenger orientation. Then, a plurality of abnormal sign coefficients are obtained to determine the possibility of crime (for example, Patent Document 1).
Also, if a person's movement in the surveillance area continues, does not fit indefinitely, and a word such as “help” that reminds a crime is detected by voice recognition, a crime has occurred in the area. (For example, Patent Document 2).

Japanese Patent Publication No. 6-74114 (page 3-4, Fig. 1) Japanese Patent Laid-Open No. 2003-248882 (page 4, FIG. 1) Ankur Datata, Mubarak Shah and Niels Da Vitoria Lobo, “Person-on-Person Violence Detection in Video Conferencing, International Conferencing”. 433-438, 2002

In the conventional rampage detection device, a specific part of the body must be extracted by photographing the person from the side and away from the person. However, in a small elevator car or the like, it is difficult to install a camera so that a specific part of the body can always be seen.
In addition, the conventional rampage detection device detects the possibility of crime by analyzing the number of passengers, the gender of the passengers, the moving speed of the passengers, the identity of the boarded floor, the relative distance of the passengers, and the direction of the passengers Therefore, it is impossible to detect the occurrence of a human rampage. Furthermore, a misdetection of a rampage detection due to a change in camera installation conditions and environment is likely to occur, and a misdetection of a rampage detection also occurs for a person's entrance / exit operation associated with the movement of a door. In addition, for the individual detection of passengers necessary for the detection of rampage, when the camera is taking a bird's-eye view close to the passengers in the elevator car, the passengers in the image overlap, so advanced processing is performed. As a result, the apparatus becomes large and the processing time becomes long.
Also, the conventional rampage detection device detects crimes and only focuses on the continuation of voice and human movement, so it cannot distinguish between normal person movement and rampage movement, and voice recognition means As a result, the device becomes complicated.

  The present invention has been made to solve the above-described problems, and in a variety of places such as a closed space such as in an elevator car, a false detection of a detection of a detection is unlikely to occur. To get.

  In the rampage detection apparatus according to the present invention, the direction and size of the movement of each point in the image is calculated from the camera that captures the monitoring area and the two images that are captured at different times by the camera. Motion calculation means for calculating the amount of variation in the movement of at least one of the people from the direction and size of the movement of each point, the amount of variation in the movement of the person, and a person's rampage determination threshold And a rampage judging means for judging the rampage of the person.

  Further, in the rampage detection apparatus according to the present invention, a camera that captures the monitoring area, and a motion amount calculating means that calculates the magnitude of the person's motion from two images that are captured by the camera at different shooting times for a predetermined time; And a rampage amount judging means for comparing the magnitude of the person's movement and the rampage amount judgment threshold of the person to judge the rampage of the person.

  The rampage detection device according to the present invention calculates the direction and magnitude of movement of each point in an image from a camera that captures a surveillance area and two images that are captured by the camera at different shooting times for a predetermined time. A motion calculation means for calculating a variation amount of the movement of at least one of the direction and magnitude of the movement of the person from the direction and magnitude of the movement of the point, a variation amount of the movement of the person, and a person's rampage determination threshold; It is equipped with a rampage judgment means that compares people with each other to determine whether a person is rampant.Therefore, in various places such as closed spaces such as in an elevator car, the camera installation conditions, environmental changes, and the entrance of people due to door movements A false detection of a rampage detection hardly occurs with respect to the leaving operation, and a simple rampage detection device can be obtained.

  In addition, the rampage detection apparatus according to the present invention includes a camera that captures a monitoring area, a motion amount calculating unit that calculates the magnitude of a person's motion from two images that are captured by the camera at different shooting times, and Since the amount of movement of the person is compared with a person's rampage amount determination threshold and the rampage amount judging means for judging the rampage of the person is provided, the processing speed of the rampage detection can be increased.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a violence detection apparatus showing Embodiment 1 for carrying out the present invention. The camera 1 captures a predetermined monitoring area to be monitored and outputs the captured image to the motion calculation unit 2. The motion calculation unit 2 calculates the direction and magnitude of the movement of each point in the image from two images taken by the camera 1 at different shooting times, and the person's motion is calculated from the direction and magnitude of the movement of each point. It is a motion calculation means for calculating the amount of variation in the motion of at least one of the motion direction or magnitude. The two images having different shooting times for a predetermined time are the two images before and after consecutive shooting. For example, when a video camera is used to capture an image using the NTSC system, the image is captured every 1/30 seconds. In this case, the two images that are different from each other for a predetermined time can be two consecutive images. However, the images can be changed every 1/10 second by thinning out the intermediate images or changing the shooting interval. Or every second.

  In addition, the motion calculation unit 2 includes a storage area and can appropriately store the received image. The variation amount of the person's motion calculated by the motion calculation unit 2 is output to the correction unit 3. In the present embodiment, the monitoring area is a closed space, and the door open / close determining unit 6 is a door open / close determining means for determining an open / closed state of a door that communicates with the inside and outside of the monitoring area that is a closed space. The door open / close determining unit 6 determines the door open / closed state from the image taken by the camera 1 and outputs the result to the motion calculating unit 2 and the correcting unit 3. Note that the motion calculation unit 2 calculates the amount of variation in the movement of the person excluding the person entry / exit data into the monitoring area.

  The correction unit 3 is based on a statistical model composed of a database of an average value and a variance value of person movement variations classified for each door open / closed state, and a person movement variation amount or person Correction means for correcting the rampage determination threshold. The correction unit 3 corrects the amount of variation in the movement of the person using the statistical model stored in the statistical model DB (database) 4 and outputs the result to the rampage determination unit 5. The rampage determination unit 5 is rampage determination means for comparing the amount of variation in the movement of the person and the rampage determination threshold value of the person to determine the rampage of the person. The rampage determination unit 5 performs threshold processing on the corrected variation in the movement of the person to determine whether the person is rampant.

  The content of the database constituting the statistical model may be an average value and a variance value of the amount of variation in human movement classified for each type of closed space. The types of closed space include a case where the wall is glazed, a window is attached to the wall, the color of the wall is different, or the wall is different. The database contents may be a database of average values and variance values of the amount of variation in the movement of a person classified for each angle of the camera 1. Based on the statistical model configured from such a database, the correction unit 3 corrects the variation amount of the person's movement or the person's rampage determination threshold value, and outputs it to the rampage determination unit 5.

  FIG. 2 is a flowchart showing the operation of the ramp detection apparatus in the first embodiment. In step ST10, an image is captured by the camera 1 and an image is input. In step ST11, the door open / close determining unit 6 determines the open / closed state of the door. A method for determining the open / closed state of the door will be described later. Next, in step ST12, the motion calculation unit 2 calculates an optical flow from two images whose shooting times are different by a predetermined time. The optical flow will be described later. Subsequently, in step ST13, the motion calculation unit 2 calculates the amount of variation in human motion from the optical flow. The amount of variation in the movement of the person is, for example, the variation in the direction of the movement of the person or the variation in the size of the movement of the person, the movement direction obtained at each point of the image, the variance value of the movement magnitude, or the standard deviation As required.

  By the way, when the door is open, for example, the environmental conditions in the monitoring area such as the lighting conditions greatly change. In addition, movements associated with the opening and closing of the doors, movements of people in and out of the door, movements from the door to the outside, and the like are observed, which cause misdetection of rampage detection. Therefore, when the door is open, the movement of the person is not used without using the information on each point in the image of the predetermined area and the information on each point in the image having the predetermined moving direction. Calculate the variation.

Further, in step ST14, the correction unit 3 corrects the variation amount of the person's movement using the statistical model stored in the statistical model DB 4, and corrects the value to reduce the dependence on the installation conditions and environment of the camera 1. To do. The statistical model is prepared in advance for each monitoring area and each door open / closed state. For example, after the camera 1 is installed, a variation amount of the person's movement is experimentally calculated with respect to a normal person's movement, and the temporal average value M and the variance value σ are stored as a statistical model. . Assuming that the variation amount of the movement of the person calculated in step ST13 is V, in step ST14, the correction value V ′ is calculated from equation (1).
V ′ = (VM) ÷ σ (1)

  Finally, in step ST15, the rampage determination unit 5 performs threshold processing on the correction value V 'to determine whether or not the person is rampant. Here, when the correction value V ′ is equal to or greater than the threshold value, it is determined that there is a rampage. The above processes from steps ST10 to ST15 are repeatedly executed at predetermined time intervals.

  FIG. 3 is an explanatory diagram of a determination window which is a method for determining the open / closed state of the door in the first embodiment. For example, when a door is reflected in the image, a determination window is installed near the upper end of the door where no person in the image enters, and it is determined that the door is open when the average brightness in the window changes. . Further, the open / closed state of the door may be determined from the optical flow. On the other hand, when the door is not reflected in the image, attention should be paid to the change in average brightness around the image where no person enters. Instead of judging from the image, it may be judged from a signal relating to opening / closing of the door from a door opening / closing motor, an opening / closing sensor or the like.

  FIG. 4 is an explanatory diagram of the optical flow in the first embodiment. The optical flow is an apparent velocity distribution representing the direction of movement and the magnitude of movement at each point of the image. In FIG. 4, the direction of movement and the magnitude of movement calculated at each point of the image are displayed by line segments.

  In this way, paying attention to the amount of variation in the movement of the person that correlates with the rampage movement, it is judged that there is a rampage when the variation is large. Rampage detection can be realized. Further, the present invention does not individually detect a person or a specific part of the person, and further, the movement amount of the person excluding the movement generated in a specific area or the movement in a specific direction according to the door opening / closing state is detected. In addition to the calculation, the variation amount or threshold value of the movement of the person is appropriately corrected for each monitoring area and each door open / closed state. For this reason, it is difficult to depend on changes in the installation conditions or environment of the camera 1, and there are few false detections of rampage detection for the movement of the door and the person entering and leaving the room, including in the elevator car, train, bus, office, etc. It can be applied in various environments. Since no voice recognition is required, the rampage can be detected with a minimum apparatus configuration.

  As described above, the direction and size of the movement of each point in the image are calculated from the camera 1 that captures the monitoring area and the two images that are captured by the camera 1 at different shooting times. The amount of movement variation of at least one of the person's movement direction or magnitude is calculated from the direction and magnitude of the movement, and the amount of movement of the person is compared with the person's rampage threshold to compare the person's movement. Since the rampage is judged, in various places such as an enclosed space such as in an elevator car, the false detection of the rampage detection is performed with respect to the installation condition of the camera 1, the change of the environment, and the person entering and leaving the room due to the movement of the door. It is hard to occur and a simple rampage detection device can be obtained.

  Even if the rampage determination threshold is corrected using the statistical model, the effect is the same as when the amount of variation in human motion is corrected using the statistical model. Further, the determination result of the open / closed state of the door may be used by either the motion calculation unit 2 or the correction unit 3. Furthermore, it is possible to determine whether the person is rampant by combining a plurality of types of movement variations of the person. Further, the rampage determination unit 5 may determine the rampage of the person by using the amount of variation in the movement of the person after a predetermined time has elapsed after the door is closed and before the door is opened. In this case, the motion calculation unit 2 may calculate the amount of variation in the movement of the person after a predetermined time has elapsed since the door was closed and before the door was opened.

Embodiment 2. FIG.
FIG. 5 is a block diagram showing a rampage detection apparatus showing Embodiment 2 for carrying out the present invention. 5 is different from the first embodiment in that the number of persons estimation unit 7 is provided. In the drawings, the same reference numerals denote the same or corresponding parts, and this is common throughout the entire specification. Furthermore, the description of the constituent elements appearing in the whole specification is merely an example, and is not limited to these descriptions. Specifically, when there is a possibility that many people exist in the monitoring area, the number estimating unit 7 which is a number estimating means for estimating the number of people in the monitoring area is added. The number of persons estimation unit 7 estimates the number of persons from the image taken by the camera 1 and outputs the result to the correction unit 3. The statistical model DB 4 includes, as a database constituting the statistical model, an average value and a variance value of the amount of variation in the movement of the person classified for each estimated number of people.

  FIG. 6 is a flowchart showing the operation of the rampage detection apparatus according to the second embodiment. First, after an image is taken by the camera 1 in ST10, the number estimating unit 7 estimates the number of persons appearing in the image in ST20. For example, it is only necessary to compare a previously acquired image of only a background without a person and the current image to extract a region having a large difference, and estimate the number of people based on the number or area of the regions. ST11 to ST13 are the same as in the first embodiment. Next, in step ST21, the correction unit 3 corrects the variation amount of the person's movement or the rampage determination threshold for each estimated number of persons using the statistical model for each number of persons prepared in advance in the statistical model DB 4. Finally, in ST15, the ramp determination unit 5 determines rampage.

  As described above, the number-of-people estimation unit 7 is added and correction is performed for each estimated number of people. Therefore, it is possible to correctly perform the determination of the person's rampage with respect to the movement of many people.

Embodiment 3 FIG.
FIG. 7 is a block diagram showing a rampage detection apparatus showing Embodiment 3 for carrying out the present invention. 7 is different from the first embodiment in that a model update unit 8 is provided. When the environment of the monitoring area changes from moment to moment with respect to changes of day and night or changes of the four seasons, the model update unit 8 is added to cope with it. The model update unit 8 receives the variation amount of the person's motion calculated by the motion calculation unit 2, generates a new statistical model, and outputs it to the statistical model DB 4. As a result, the statistical model DB 4 is updated with the variation amount of the person's motion calculated by the motion calculation unit 2.

  FIG. 8 is a flowchart showing the operation of the rampage detection apparatus according to the third embodiment. First, in step ST30, a counter for counting image inputs is reset. Subsequent ST10 to ST15 are the same as in the first embodiment. Next, in step ST31, it is determined whether or not the counter has reached the predetermined number N. If not, the model updating unit 8 determines the variation amount of the person's motion calculated by the motion calculation unit 2 in step ST32. Are stored, the counter is incremented by 1 in ST33, and the process returns to ST10. On the other hand, when the counter reaches the predetermined number N, in step ST34, the model update unit 8 calculates the average value M ′ and the variance value σ ′ from the variation amounts of the movements of the N persons stored so far. And is stored in the statistical model DB 4 as a new statistical model in step ST35. Finally, in step ST36, the counter is reset, and the process returns to ST10. In this way, the statistical model can be updated, and the amount of variation in human movement corresponding to the latest environment can be corrected.

  As described above, since the statistical model is updated, it is possible to perform correction following a long-term change in the environment due to day or night or the four seasons, and it is possible to more accurately determine a person's rampage.

Embodiment 4 FIG.
FIG. 9 is a block diagram showing a rampage detection apparatus showing Embodiment 4 for carrying out the present invention. In FIG. 9, the configuration other than the motion amount calculation unit 9 and the rampage amount determination unit 10 is the same as that of the first embodiment. The motion amount calculation unit 9 is a motion amount calculation means for calculating the magnitude of a person's motion from two images taken by the camera 1 and having different shooting times for a predetermined time. The rampage amount determination unit 10 is a rampage amount determination unit that compares the magnitude of the person's movement with the rampage amount determination threshold value of the person to determine the rampage of the person. Operations other than the method of calculating the magnitude of the person's movement are the same as those in the first embodiment. Since only the magnitude of the person's movement is calculated, the calculation processing speed can be increased.

  As described above, since the magnitude of the person's movement is calculated from two images whose shooting times are different by a predetermined time, the processing speed of the rampage detection can be increased.

It is a block diagram which shows the rampage detection apparatus which shows Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the rampage detection apparatus in Embodiment 1 of this invention. It is explanatory drawing of the window for determination in Embodiment 1 of this invention. It is explanatory drawing of the optical flow in Embodiment 1 of this invention. It is a block diagram which shows the rampage detection apparatus which shows Embodiment 2 of this invention. It is a flowchart which shows operation | movement of the rampage detection apparatus in Embodiment 2 of this invention. It is a block diagram which shows the rampage detection apparatus which shows Embodiment 3 of this invention. It is a flowchart which shows operation | movement of the rampage detection apparatus in Embodiment 3 of this invention. It is a block diagram which shows the rampage detection apparatus which shows Embodiment 4 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Camera, 2 motion calculation part, 3 correction | amendment part, 4 statistical model DB, 5 rampage determination part, 6 door opening / closing determination part, 7 person number estimation part, 8 model update part, 9 motion amount calculation part, 10 rampage amount determination part.

Claims (10)

A camera that captures the surveillance area;
The direction and magnitude of the movement of each point in the image is calculated from two images taken by the camera at different shooting times for a predetermined time, and the direction of the movement of the person or the person from the direction and magnitude of the movement of each point Movement calculating means for calculating the amount of variation in movement of at least one of the persons,
A rampage detection apparatus comprising rampage determination means for comparing the amount of variation in the movement of a person with a rampage determination threshold of the person to determine the rampage of the person. The rampage detection apparatus according to claim 1, wherein the monitoring area is a closed space, and the motion calculation unit calculates a variation amount of the movement of the person excluding the person entry / exit data into the monitoring area. The ramp detection device according to claim 2, further comprising door open / close determining means for determining an open / closed state of the door leading to the inside and outside of the monitoring area which is a closed space. The monitoring area is a closed space, and includes a door open / close determining means for determining the open / closed state of the door leading to the inside / outside of the monitoring area,
The rampage determination means determines the rampage of the person using a variation amount of the movement of the person between a predetermined time after the door is closed and a predetermined time before the door is opened. The rampage detection apparatus according to 1. Based on a database of average and variance values of person movement variations classified for each door open / close state, a correction means is provided for correcting the amount of person movement variation or the person rampage determination threshold used in the rampage determination means. The rampage detection device according to claim 3, wherein the device is a rampage detection device. Based on a database of average values and variance values of person movement variations classified for each type of closed space, a correction means for correcting a person movement variation amount or person rampage determination threshold used in the rampage determination means is provided. The rampage detection device according to any one of claims 2 to 5, wherein: Based on a database of average and variance values of person movement variability classified for each camera angle, a correction means is provided for correcting the amount of person movement variability or the person riot determination threshold used in the riot determination means. The rampage detection device according to any one of claims 1 to 6, wherein It has a person estimation means to estimate the number of people in the monitoring area,
8. The rampage detection apparatus according to claim 5, wherein the database includes an average value and a variance value of a variation amount of a person's movement classified for each estimated number of persons. 9. The rampage detection apparatus according to claim 5, wherein the database is updated with a variation amount of the movement of the person calculated by the movement calculating means. A camera that captures the surveillance area;
A motion amount calculating means for calculating the magnitude of a person's motion from two images taken by the camera at different shooting times by a predetermined time;
A rampage detection apparatus comprising rampage amount determination means for comparing a magnitude of a person's movement with a rampage amount determination threshold of a person to determine a rampage of the person.

Images