JP2012023566A - Monitoring system, and behavior monitoring program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a monitoring system and a behavior monitoring program which can be expected to perform integral and efficient monitoring on a culpable act, human sufferings, and property damage, with criminality, in a facility which is individually used by miscellaneous users and where protection of privacy is required.SOLUTION: A monitoring system includes a monocular camera 11 to take an image of a use area in a facility used by monitoring objects, which are users of the facility; a capture unit 12 to capture an image for each frame unit taken by the camera 11; an image processing unit 13 to monitor the behavior (movement) of the objects in the use area based on the image captured by the capture unit 12 and to discriminate abnormal behavior which meets a specific behavior pattern preliminary registered; and an alarm output unit 15 to output an alarm when abnormal behavior is discriminated by the image processing unit 13.

Description

  The present invention relates to a monitoring device that monitors the behavior of an object to be monitored based on an image captured by a monocular imaging camera, and that uses the imaging camera as a sensor for alarm output in a facility where a user uses the individual unit, and behavior monitoring Related to the program.

  In facility facilities (for example, toilet rooms, bathrooms, various service rooms, etc.) that are used by individual users on an individual basis by a large number of users who have specified the purpose of use, such as toilets, bathrooms, and various service rooms, Various restrictions are imposed from the viewpoint of protection. For example, strict regulations and restrictions are required for video output of surveillance cameras from the viewpoint of privacy protection. For this reason, in monitoring the behavior of various facilities described above, a monitoring system using various security sensors such as a human sensor, an illuminance sensor, a motion sensor, and a vibration sensor is widely applied in place of the monitoring camera for imaging. Yes.

  However, in this type of surveillance system using security sensors, for example, criminal, responsible acts (for example, violence), human damage (for example, killing damage), physical damage (for damage to property), etc. However, there was a problem that integrated and effective monitoring could not be expected.

  Conventionally, as a behavior monitoring system using an imaging camera, there has been a camera system characterized by an alarm generation function.

JP 2008-199390 A

  As mentioned above, in facilities that require privacy protection, which is used by many users on an individual basis, it has traditionally been against criminal, responsible acts, human damage and property damage. There was a problem that integrated and effective monitoring could not be expected.

  Embodiments of the present invention are integrated for criminal, responsible actions, human damage, physical damage, etc. in facilities that require privacy protection for use by many users on an individual basis. An object of the present invention is to provide a monitoring device and behavior monitoring program that can be expected to be effective and effective monitoring.

  The present invention is a monitoring device that uses an imaging camera as an alarm output sensor for a facility used by a user, and obtains an inter-frame difference processed image based on a plurality of time-series frame images input from the camera. , Processing means for determining whether or not the difference processed image corresponds to a behavior pattern registered in advance, and output means for generating an alarm when the processing means determines that the difference processed image corresponds to the behavior pattern And a monitoring device using an imaging camera as a sensor for alarm output of a facility used by a user.

  It is integrated and effective for criminal, responsible acts, human damage, physical damage, etc. in facilities that require the protection of privacy where many users use imaging cameras. Surveillance can be expected.

The block diagram which shows the structure of the monitoring apparatus which concerns on embodiment of this invention. The figure which shows the operation | movement concept of the monitoring apparatus which concerns on the said embodiment. The block diagram which shows the more detailed structure of the monitoring apparatus which concerns on the said embodiment. The figure for demonstrating the background image creation process which concerns on the said embodiment. The figure for demonstrating the background image creation process which concerns on the said embodiment. The figure for demonstrating the background image creation process which concerns on the said embodiment. The flowchart which shows the process sequence of the monitoring apparatus which concerns on the said embodiment. The flowchart which shows the process sequence of the monitoring apparatus which concerns on the said embodiment. The figure for demonstrating the example of a behavior determination process which concerns on the said embodiment.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows an overall configuration of a monitoring apparatus according to an embodiment of the present invention, and FIG. 2 shows an operation concept of the system.

  The monitoring apparatus according to the above embodiment realizes a sensor-type alarm output camera. As shown in FIG. 1, a monocular camera 11 that takes a person as a monitoring target object and shoots a usage area in a use facility of the monitoring target object. A capture unit 12 that captures an image of the frame unit captured by the camera 11, and the behavior (behavior) of the object existing in the use area is monitored based on the image captured by the capture unit 12, and registered in advance. The image processing unit 13 that determines the abnormal behavior corresponding to the specific behavior pattern and the alarm output unit 15 that outputs an alarm when the image processing unit 13 determines the abnormal behavior.

  The camera 11 shoots the use area of the use facility in facility facilities (for example, a toilet room, a bathroom, various service rooms, etc.) used by an unspecified number or a large number of users who have specified the purpose of use on an individual basis. Installed as possible.

  As shown in FIG. 2, the image processing unit 13 detects a person (human body) entering the use area using a background image processing technique based on an image input from the camera 11 via the capture unit 12. The abnormal behavior (behavior) of the detected person is monitored. The monitoring of the abnormal behavior (abnormal behavior) is executed according to the definition of the behavior pattern registered in advance. In the example shown in FIG. 2, the stationary state defined by the behavior pattern 1 and the human body hardly moves, and the behavior pattern The human body lying down (falling down) defined by 2 and the human body lying down (falling down), and the behavioral patterns defined by the behavior pattern 3 are fast-moving and intense radical behavior states. The flame detection state defined by 4 for detecting a flame in the vicinity of the human body can be recognized for each registered pattern.

  The determination of the abnormal behavior in the image processing unit 13 is realized using a background image processing technique. In this embodiment, in addition to the moving object detection function (function for detecting short-term movement on the screen) for extracting a region including a change pixel from a plurality of time-series images taken by the camera 11, this short-term operation is performed. In contrast to a simple motion detection function on a screen, a detection function for a stationary object (stationary object) that has been stationary for a long time (a function for detecting a change in a pixel (background) without fluctuation) is realized.

  In this embodiment, pixels of an image with a lot of movement have a large variation (dispersion) in luminance value, and pixels of an image with no movement have a small variation in luminance value and stably fall within a narrow luminance range of low luminance. Focusing on the above, calculate the luminance variance value and standard deviation per pixel for multiple images within a certain period of input, and use this luminance variance value or standard deviation as a threshold value, and the luminance value variance is large However, the pixel set is classified and identified stepwise according to the speed of movement (moving speed), and is used to determine the behavior (movement) of a person who is a moving object to be detected.

  In addition, using the above threshold, pixels where the variance of luminance values is large are discarded, and only pixels that are stably within a low luminance range with a low luminance are regarded as background pixels and extracted. Creating an image. Here, a pixel having no fluctuation is determined as a background pixel, and a luminance value of the background pixel is registered to create a background image. The difference between the created background image and the latest luminance that is considered to be the current background is executed, and background change pixels are extracted. The extracted background change pixels are subjected to a regionization process and can be tracked by using a tracking function, in the same manner as the detection of a human body. If necessary, the setting values required for this function can be changed using external parameters. This detection function makes it possible to distinguish between a stationary object (a human body that is regarded as a stationary state) and an animal body (a human body that is considered to be in motion) and to monitor the behavior of the person. In addition, it is possible to recognize the presence of a human body that is regarded as a stationary state and a plurality of human bodies that are fast and intense, and notify the outside of the abnormal behavior.

  As shown in FIG. 3, the image processing unit 13 that determines the abnormal behavior includes an image buffer 131, a background image creation processing unit 132, a background image creation buffer 133, and a behavior determination pixel extraction processing unit 134. The regionalization processing unit 135, the object model registration unit 136, the object detection processing unit 137, the tracking processing unit 138, the behavior pattern registration unit 139, and the behavior determination processing unit 140 are configured.

  The image buffer 131 sequentially inputs frame-by-frame images captured by the camera 11 at regular intervals, and each time a current image is input, a plurality of images within a certain period of the input current image and a plurality of past images are input. Update and memorize. Here, as an example, 30 frames of images (frame images) are stored while being updated each time an image is input in units of frames.

  The background image creation processing unit 132 has a processing function for creating a background image of the use area by performing difference processing on a plurality of images stored in the image buffer 131 within a certain period. The background image creation processing unit 132 sequentially inputs images captured by the camera 11 via the capture unit 12, and calculates a luminance average value and a standard deviation per pixel for a plurality of input images within a certain period. Based on the standard deviation or luminance dispersion value obtained by this calculation, for a corresponding pixel of the plurality of images within the certain period, a plurality of pixels that fall within a certain luminance fluctuation range of low luminance are specified, A background change pixel is extracted by background image creation means for creating a background image using the average luminance of the specified pixel, and difference processing for each pixel between the background image created by the background image creation means and the input current image. And background change pixel extraction means for updating the background image.

  The processing function of the background image creation means and the processing function of the background change pixel extraction means detect a person (human body) that has entered the use area, and the behavior of the detected person (behavior) is a normal dynamic with gentle movement. Distinguishing between fast and intense abnormal dynamics, stationary state where the movement within a certain time is within the fixed amount, etc., and comparing with various registered behavior patterns shown in FIG. It is possible to recognize, for each registered behavior pattern, a recumbent state lying down (falling down) with almost no movement in the human body, an extreme action state in which movement is fast and intense for a plurality of human bodies. The processing of the background image creation processing unit 132 will be described later with reference to FIGS.

  The background image creation buffer 133 is a work buffer that provides various storage areas used for the background image creation processing of the background image creation processing unit 132, and a typical background image creation buffer structure is shown in FIG. Show.

  As shown in FIG. 4, the background image creation buffer 133 includes a buffer 321 for storing the background luminance sum (ΣBi), a buffer 322 for storing the background luminance (Ave_Bi) that is the average luminance of a plurality of background pixels, and for 30 screens. Buffer 323 for storing the background image of each image = background luminance (B0 to B29) in units of pixels, a buffer 324 for storing the background luminance buffer number (BufNum) for each pixel, and a buffer 325 for storing the background luminance registration number (EntryNum) for each pixel. Each region such as a buffer 326 for storing background luminance registration frame counter information (EntryFrmCtr) for each pixel is provided. Further, the background image creation buffer 132 is an image buffer memory for storing image data (color) for each screen input from the capture unit 12, a background image buffer memory for storing the created background image, and data for creating a background image. Including input / output data, various data in the middle of calculation, an internal parameter storage area, an external parameter storage area, and the like. In this embodiment, the average luminance of a plurality of background pixels (average luminance) is calculated and used as the background luminance (the background luminance update timing varies depending on the pixel). In addition to the background image (average average brightness) (Ave_Bi) and the background brightness total (ΣBi), the background update timing differs depending on the pixel. Therefore, a buffer (B0-29) for creating the background brightness (B0-29) for each pixel ( 30) 323, a buffer 324 for storing the background luminance buffer number, a buffer 325 for storing the background luminance registration number (29 at the maximum), and a buffer 326 for storing the background luminance registration frame counter information (frame count value). In the buffer 153 for creating the background luminance (B0-29), the low luminance stable pixels extracted by the background image creation processing unit 14 are registered in units of one pixel according to the designation of the buffer number, frame counter information, etc. A background image based on the pixel set is generated.

  The behavior determination pixel extraction processing unit 134 uses the background image created by the background image creation processing unit 132 and includes a set of low-luminance stable pixels to stop the object (a pixel set of pixels that are considered to be in a stationary state). And a detection process of a moving object (a pixel set of pixels that are considered to be in motion).

  The region processing unit 135 performs the stop object region processing based on the pixel set of pixels considered to be in a stationary state, which is processed by the behavior determination pixel extraction processing unit 134, and determines that there is motion. The moving object region is processed based on the pixel set of the trapped pixels. Since the region processing unit 135 can be realized by a known technique (Japanese Patent No. 4427052), the description thereof is omitted here.

  The object model registration unit 136 is used for the processing of the object detection processing unit 137 that uses a person (human body) as a monitoring target object. The object model registration unit 136 has a screen in which the camera 11 captures the use area. In the above, the human body image in the use area in the use facility is modeled as an object area (for example, a rectangular area), and is registered in advance including a determination allowable range (allowable size). For example, in a usage area used by a person who is a monitoring target object, by registering a rectangular area that is considered to be a human body in landscape and portrait orientation, the landscape object model is referred to in the usage area. It is possible to recognize the lying state of the human body (monitoring target object) (falling down), and the normal usage state of the human body (monitoring target object) in the usage area with reference to the vertically long object model ( Standing position).

  The object detection processing unit 137 uses the object model registered in the object model registration unit 136 based on the background image created by the background image creation processing unit 132 and the current image stored in the image buffer 131. It has a processing function for individually detecting objects to be monitored in the area. In this embodiment, an object in a lying state and a plurality of objects in a standing state are detected as monitoring target objects in the use area.

  The tracking processing unit 138 performs tracking processing for tracking the area of each object detected by the object detection processing unit 137. Since the tracking processing unit 138 can be realized by the above-described known technique (Japanese Patent No. 4427052) in the same manner as the above-described regionalization processing unit 135, the description thereof is omitted here.

  The behavior pattern registration unit 139 is used for the processing of the behavior determination processing unit 140, and the behavior pattern registration unit 139 includes one or a plurality of behavior patterns that define specific behaviors for the monitoring target object in the use area. Is registered. In this embodiment, a plurality of behavior patterns (behavior patterns 1 to 4) for monitoring determination shown in FIG. 2 are registered in advance. As the behavior pattern registration data, for example, position information on the screen indicating the state transition of the object and time information at the position are defined as parameters in time series. The time information can be given by the number of frames in units of frames of the input video.

  The behavior determination processing unit 140 determines whether or not the behavior of the object corresponds to the behavior pattern registered in the behavior pattern registration unit 139 for the object detected by the object detection processing unit 137 and tracked by the tracking processing unit 138. It has a processing function to determine whether or not. This processing procedure will be described later with reference to the flowcharts shown in FIGS. 7 and 8 and the behavior determination processing example shown in FIG.

  The processing of the background image creation processing unit 132 will be described with reference to FIGS. The functional blocks of the background image creation processing unit 132 are shown in FIG.

  As illustrated in FIG. 5, the background image creation processing unit 132 includes an image buffer memory 131 that stores an image input from the capture unit 12, a variable pixel extraction unit 332 having an automatic threshold value calculation function using a standard deviation, and a background pixel. A background image update unit 333 that performs the update process, a background image update management unit 334 that controls the update process of the background image update unit 333, a background image refresh unit 335 that refreshes the background pixels at intervals, and a background It comprises a buffer 336 for storing image creation data and a background image buffer memory 331 for storing the created background image.

  The image buffer memory 131, the background image creation data buffer 336, the background image buffer memory 331, and the like are placed in the background image creation buffer 133 shown in FIG. The variable pixel extraction unit 332 receives, as external parameters, the minimum threshold value (upper and lower limit values), the number of frames used, and the like for extracting stable pixels with low luminance.

  The variable pixel extraction unit 332 reads (R) images for a plurality of screens from the image buffer memory 131, and calculates each value of luminance average, luminance variance, and standard deviation for each pixel. Based on the current frame counter information, the background image update management unit 334 checks the elapsed time since the previous update in units of pixels, and refreshes the background image refresh unit 335 for pixels that have not been updated for a long time. Instruct. In accordance with this instruction, the background image refresh unit 335 refreshes the corresponding background pixel. On the other hand, the background image update unit 333 reads the background pixel and its luminance value and the luminance variance value or standard deviation from the variable pixel extraction unit 332, and the luminance variance value or standard deviation of the background pixel (target pixel) is low. If it is determined whether the pixel is a stable pixel of low brightness, the pixel is determined to be a background pixel, and the background image creation data is updated using the average image data. Then, the updated background image is registered in the background image buffer memory 331. The background image buffering in the background image buffer memory 331 is performed at intervals specified by parameters.

  The operation concept in the background image creation processing unit 132 is shown in FIG. The automatic threshold setting technique realized by the present applicant (see Japanese Patent Application Laid-Open No. 2008-250892) calculates a luminance average value and a standard deviation per pixel for a plurality of input pixels within a certain period, and calculates the calculated pixels. Each standard deviation is set to the threshold value of the corresponding pixel for extracting the change pixel. The automatic threshold value is automatically set to a high threshold value for a pixel with a large variation in luminance, and a low and stable threshold value is set for a pixel with a small variation = a pixel considered to be a background.

  Using this characteristic, it is determined that the automatic threshold is low = no background fluctuation = background pixel, the average luminance at that time is used as the background luminance, an average image is created from the set, and used as the background image . Then, a difference process between the latest registered background luminance and the current background image is performed to extract background change pixels. This concept works even in moving images, and makes it possible to recognize a stationary object at the same location where the moving object is detected with a high recognition rate.

  On the other hand, a pixel with a long luminance variation time has a long non-updated state from the state determined as the background pixel, and there is a possibility that a deviation from the current state has occurred. It has a refreshing function. In FIG. 6, reference symbol Ta is the lowest threshold value, and SA is the low luminance stable range equal to or lower than the lowest threshold value (Ta).

  The background image creation processing unit 132 determines whether or not the target pixel is in the low luminance stable range (SA) equal to or lower than the minimum threshold (Ta). When the target pixel is equal to or lower than the minimum threshold (Ta), the background pixel update processing is performed. To implement.

  An example of the processing procedure of the behavior determination processing unit 140 provided in the image processing unit 13 in the monitoring apparatus shown in FIGS. 1 to 5 is shown in FIGS.

  The camera 11 takes a person as a monitoring target object and captures a usage area in the facility where the monitoring target object is used. The capture unit 12 captures an image (video) captured by the camera 11 and sends it to the image processing unit 13 in units of frames. Deliver.

  In the image processing unit 13, the image buffer 131 sequentially inputs frame-by-frame images captured by the camera 11 at regular time intervals, and a plurality of images (30 within a certain period of the input current image and a plurality of past images). Each time an image is input, it is updated and stored.

  The background image creation processing unit 132 performs binarization conversion (grayscale conversion), noise removal, etc. on an image (video in frame units) captured by the camera 11 and input to the image buffer 131 via the capture unit 12. After performing a predetermined pre-process, the background image creation buffer 133 is used to calculate the luminance average value and standard deviation per pixel for a plurality of inputted images within a certain period, and this calculation Based on the standard deviation or luminance dispersion value obtained in step 1, a plurality of pixels that fall within a constant luminance fluctuation range of low luminance are specified for the corresponding pixels of the plurality of images within the predetermined period, and this specification is performed. A background image is created using the average luminance of the pixels, background change pixels are extracted by a difference process for each pixel between the created background image and the input current image, and the background image is updated (steps A11 to A). 5).

  Based on the background image created by the background image creation processing unit 132, the behavior determination pixel extraction processing unit 134 uses a background image based on a set of low-luminance stable pixels to detect a stationary object (a pixel regarded as a stationary state). And a detection process of a moving object (a pixel set of pixels that are considered to be in motion), and the region processing unit 135 has a pixel set of pixels that are considered to be stationary. The object detection processing unit 137 registers the object in the object model registration unit 136. The object detection processing unit 137 registers the object in the object model registration unit 136. The monitored object is individually detected in the usage area using the object model thus determined, and it is detected that one or more monitoring target objects have entered the usage area (steps A16 to A18). .

  When it is detected that the monitoring target object has entered the use area, the behavior determination processing unit 140 recognizes whether the monitoring target object is single (one person) or plural (two or more), and When the object is singular (one person), it recognizes the position on the screen where the object exists and the movement in the set time (Ta) unit, and refers to the behavior pattern registered in the behavior pattern registration unit 139, It is determined whether the object is in a stationary state or in a normal dynamic state with a slow movement (steps A19 to A20).

  Here, when it is determined that the object is in a normal dynamic state, the monitoring process for the current input video is terminated. If it is determined that the object is in a stationary state, this state is regarded as occurrence of abnormal behavior, and an output instruction to that effect is given to the alarm output unit 15 (step A21).

  Thereby, the alarm output unit 15 operates to notify the outside that an abnormal behavior has occurred in the use area of the use facility monitored by the camera 11.

  Further, the behavior determination processing unit 140 determines whether or not the object in the stationary state is in a lying state (step A22). When the object is in a lying state, this state is regarded as occurrence of a severe abnormal behavior. Then, an output instruction to that effect is given to the alarm output unit 15 (step A23).

  As a result, the alarm output unit 15 operates to notify the outside that a severe abnormal behavior has occurred in the usage area of the usage facility monitored by the camera 11.

  In addition, when there are a plurality of monitoring target objects (two or more), the behavior patterns registered in the behavior pattern registration unit 139 are referred to, and the objects are each in a normal dynamic state with a slow motion or a fast and intense motion. It is determined whether or not there is abnormal kinetics (steps A31 to A34).

  Here, when it is determined that the object is in a normal dynamic state, the monitoring process for the current input video is terminated. If it is determined that the object cancer abnormal kinetics is present, this state is regarded as occurrence of abnormal behavior such as assault, and an output instruction to that effect is given to the alarm output unit 15 (step A35).

  As a result, the alarm output unit 15 operates to notify the outside that an extreme abnormal behavior has occurred in the use area of the use facility monitored by the camera 11.

  FIG. 9 shows an example of the interframe feature of the video frame accompanying the occurrence of abnormal behavior such as assault. Here, the normal behavior, the abnormal behavior, and the pixel variation state of the abnormal sample in the monitoring target object, the vertical axis indicates the ratio of the variation pixels, and the horizontal axis indicates time (seconds).

  The behavior determination processing unit 140 executes the monitoring process shown in FIG. 8 in addition to the monitoring process shown in FIG.

  In this process, similar to the process shown in FIG. 7 described above, the background image creation process by the background image creation process unit 132, and the pixel set and motion of the pixels that are regarded as a still state by the behavior determination pixel extraction process unit 134 After the detection processing of the pixel set of pixels considered to be present, the regionization processing by the region processing unit 135, and the detection processing of the monitoring target object by the object detection processing unit 137 (steps B11 to B17), the surroundings of the monitoring target object An image is acquired, the behavior pattern of the ignition operation registered in the behavior pattern registration unit 139 is referred to determine whether or not it is in the behavior state of the ignition operation, and when it is in the behavior state of the ignition operation, this state is This is regarded as the occurrence of abnormal behavior accompanied by fire, and an output instruction to that effect is given to the alarm output unit 15 (steps B18 to B20).

  As a result, the alarm output unit 15 operates to notify the monitoring target object (user) using the usage area that the usage area of the usage facility monitored by the camera 11 is strictly prohibited.

  Furthermore, the behavior determination processing unit 140 acquires the surrounding image of the monitoring target object as an RGB color image, refers to the behavior pattern of the ignition operation registered in the behavior pattern registration unit 139, and the flame condition continues for a set time. When the flame condition continues, this condition is regarded as the occurrence of abnormal behavior accompanied by flame, and an output instruction to that effect is given to the alarm output unit 15 (steps B21 to B21). B26).

  In accordance with this instruction, the alarm output unit 15 operates to notify the outside that the use area of the use facility monitored by the camera 11 is in the fire state.

  In this manner, the behavior determination processing unit 140 provided in the image processing unit 13 has the behavior of the object detected by the object detection processing unit 137 and the tracking processing unit 138 tracking the behavior pattern registration. Processing for determining whether or not the behavior pattern registered in the unit 139 is applicable is detected, an abnormal behavior state is detected, and the fact is notified to the outside via the alarm output unit 15.

  According to the above-described embodiment, in a behavior monitoring apparatus that monitors facilities used by a large number of users on an individual basis, human beings are responsible for criminal, responsible actions, human damage, and physical damage. Integrated and effective monitoring can be expected with an economically advantageous configuration that uses a monocular camera as an alarm output sensor without using various security sensors such as a sensor, illumination sensor, motion sensor, vibration sensor, etc. .

  In the above-described embodiment, the background image creation processing unit 132, the behavior determination pixel extraction processing unit 134, the regionalization processing unit 135, the object detection processing unit 137, the tracking processing unit 138, and the behavior determination processing unit 140 are each performed by software modules. It is feasible.

  Further, in the above-described embodiment, the camera 11 is used in facilities such as a toilet room, a bathroom, and various service rooms of a facility that is used by an unspecified number or a large number of users who specify the purpose of use on an individual basis. The use area of the facility is photographed, and the image processing unit 13 moves in a stationary state in which the human body has almost no movement, a lying state in which the human body has almost no movement and is lying (falls down), and a plurality of human bodies. It is possible to recognize the state of fast and intense extreme behavior and the flame detection state by the detection of flames near the human body for each registered pattern, but not limited to this, for example, in the daytime and nighttime, the equipment in a specific facility It can be applied to various facilities and equipment that require monitoring, such as monitoring of traffic, monitoring of invasion of moving objects, and traffic monitoring.

  DESCRIPTION OF SYMBOLS 11 ... Camera (monocular camera), 12 ... Capture part, 13 ... Image processing part, 15 ... Alarm output part, 16 ..., 17 ..., 18 ..., 19 ..., 131 ... Image buffer, 132 ... Background image creation process part, 133 ... Background image creation buffer, 134 ... Behavior determination pixel extraction processing unit, 135 ... Regionalization processing unit, 136 ... Object model registration unit, 137 ... Object detection processing unit, 138 ... Tracking processing unit, 139 ... Behavior pattern registration 140, a behavior determination processing unit.

Claims (8)

A monitoring device that uses an imaging camera as a sensor for alarm output of a facility used by a user,
Processing means for acquiring an inter-frame difference processed image based on a plurality of time-series frame images input from the camera, and determining whether the difference processed image corresponds to a pre-registered behavior pattern;
A monitoring apparatus comprising: output means for generating an alarm when the processing means makes a determination that the difference processed image corresponds to the behavior pattern. A monocular camera that takes a person as a monitoring object and shoots a use area in a use facility of the monitoring object;
Frame-by-frame images captured by the camera are sequentially input at regular time intervals, and a plurality of images within a certain period of the input current image and a plurality of past images are updated and stored each time the current image is input. An image buffer;
A behavior pattern registration means for registering a behavior pattern defining a specific behavior for the monitored object in the use area;
Background image creation processing means for creating a background image of the use area by performing a difference process on a plurality of images within a predetermined period stored in the image buffer;
Object detection processing means for individually detecting monitoring target objects in the use area based on the background image created by the background image creation processing means and the current image stored in the image buffer;
Determination processing means for determining whether or not the behavior of the object corresponds to the behavior pattern registered in the behavior pattern registration means for the object detected by the object detection processing means;
An output processing means for notifying the outside that an abnormal behavior has occurred in the use area when the determination processing means makes a determination corresponding to the behavior pattern;
A monitoring device comprising: The background image creation processing means includes
The luminance average value and standard deviation per pixel are calculated for a plurality of images within a certain period stored in the image buffer, and the certain period is calculated based on the standard deviation or luminance dispersion value obtained by the calculation. A background image creating means for identifying a plurality of pixels that fall within a constant brightness fluctuation range of low brightness, and creating a background image using the average brightness of the identified pixels And a background change pixel extracting means for extracting a background change pixel and updating the background image by pixel-by-pixel difference processing between the background image created by the background image creation means and the input current image. The monitoring device according to claim 2.   The background image creation processing unit calculates a luminance variance value and a standard deviation per pixel for the plurality of pixels within the input fixed period, and uses the luminance variance value or the standard deviation as a threshold value to obtain a luminance. There is a processing means for creating a background image by a pixel set by discarding pixels where the variance of values is large and extracting only pixels that are stably within a narrow luminance range of low luminance as background pixels. The monitoring apparatus according to claim 3. Using the background image created by the background image creation processing means, which is a set of low-luminance stable pixels, there is a detection process for detecting a pixel set of pixels that are considered to be stationary as a stationary object, and movement A behavior determination pixel extraction processing means for performing a detection process of detecting a pixel set of pixels regarded as a moving object;
Based on the pixel set of pixels that are considered to be stationary, processed by the pixel determination processing means for behavior determination, the stop object regionization process is performed, and the pixel set of pixels that are considered to have motion is obtained. 5. The monitoring apparatus according to claim 4, further comprising: a region processing unit that performs a region processing of the moving object.   The determination processing means is a processing means for discriminating the behavior of a person entering the monitoring area into a stationary state in which movement within a certain time is within a fixed amount of time, normal dynamics with slow movement, and abnormal dynamics with quick movement. The monitoring apparatus according to claim 2, wherein the monitoring apparatus is provided. A monitoring program for causing a computer to function as a monitoring device used as an alarm output sensor of a facility where a user uses an imaging camera,
A processing function for acquiring an inter-frame difference processed image based on a plurality of time-series frame images input from the camera, and determining whether the difference processed image corresponds to a pre-registered behavior pattern;
An alarm output of a facility where a user uses an imaging camera, which causes a computer to realize an output function that generates an alarm when the difference processing image is determined to correspond to the behavior pattern by the processing function Program for monitoring behavior of monitoring device used as sensor. Using a person as a monitoring target object, a monocular camera that captures the usage area in the facility where the monitoring target object is used, and frame-by-frame images captured by the camera are sequentially input at regular intervals, and the input current image and a plurality of past images An image buffer that updates and stores a plurality of images within a certain period of time each time a current image is input, and a behavior in which a behavior pattern that defines a specific behavior for a monitored object in the usage area is registered A computer for realizing an apparatus for monitoring the behavior of the monitored object from the input current image and past image, comprising a pattern registration means;
A background image creation processing function for creating a background image of the use area by performing a difference process on a plurality of images within a predetermined period stored in the image buffer;
An object detection processing function for individually detecting objects to be monitored in the use area based on the background image created by the background image creation processing function and the current image stored in the image buffer;
Behavior monitoring characterized in that, for an object detected by the object detection processing function, a determination processing function for determining whether the behavior of the object corresponds to a behavior pattern registered in the behavior pattern registration unit is realized. Program.

Images