JP5956248B2 - Image monitoring device - Google Patents

Description

  The present invention relates to an image monitoring apparatus in which a monitor visually monitors captured images of a monitoring location captured by a plurality of monitoring cameras.

  In recent years, image monitoring apparatuses that centrally monitor abnormal states in a plurality of areas at a monitoring location by sequentially receiving captured images captured by a plurality of monitoring cameras and displaying them on a monitoring terminal used by a supervisor are used. ing. With regard to such an image monitoring device, in particular, Patent Document 1 discloses that tracking is performed between a plurality of cameras by associating objects captured in images taken from a plurality of monitoring cameras in the preceding and following times based on their feature values. ), And displaying the movement trajectory of the object on the plan view of the monitoring place, an image monitoring apparatus is disclosed that makes it easy for the observer to visually recognize the action state of the object in the entire monitoring place in a wide range. Yes.

Japanese Patent Application Laid-Open No. 07-049952

  However, as surveillance locations become more widespread, such as shopping streets and shopping malls, it is difficult to install surveillance cameras to cover all areas of the surveillance location without gaps for reasons such as installation costs and operational costs. In general, a monitoring camera is installed only in an area to be monitored, particularly in a monitoring place. Therefore, in the conventional technique, when an object moves outside the range captured by the surveillance camera (hereinafter referred to as “imaging area”), the monitor loses sight of the object and the image monitoring apparatus can continue tracking. There was a problem of disappearing. Therefore, even when an object moves outside the imaging area, there is a need for monitoring to predict where the object may exist.

  Therefore, even if an object moves outside the imaging area, the observer can calculate the probability of existence of the target moving object for each area based on the building structure and camera installation information, etc. It is an object of the present invention to predict and grasp the approximate positional relationship of a target moving object that has gone outside, and at the same time, to continue tracking between a plurality of surveillance cameras with high accuracy.

In order to achieve the above-described object, an image monitoring apparatus according to claim 1 of the present application includes:
An imaging unit that sequentially acquires captured images of the monitoring location, a structure in the virtual space about the object that configures the monitoring location, and structure information that indicates an attribute of whether or not the opening is at least, and the imaging unit in the virtual space A storage unit that stores an imaging area and area information indicating a non-imaging area other than the imaging area, an image processing unit that temporally tracks a moving object extracted from the captured image, and moves outside the imaging area An image monitoring apparatus comprising: a display control unit that calculates a probability of existence of the mobile body in the non-imaging area and displays a current location of the mobile body in the non-imaging area based on the presence probability. ,
The display control unit obtains a vacant position in the virtual space from a position where the moving body has moved out of the imaging area, and performs non-imaging of a moving destination of the moving body from the vacant position, the area information, and the structure information. The openings constituting the outer periphery of the area are specified, and the existence probability is calculated so as to decrease as the elapsed time from the lost time of the moving body and the number of openings increase.

  With this configuration, the image processing unit of the present invention extracts a change area from a captured image acquired from a plurality of imaging units installed at a monitoring location using a conventional technique such as background difference. Then, based on image features such as the position and size of each extracted change area on the image, image tracking is performed on the captured image using the change area as a moving body. Then, the display control unit of the present invention detects the moving body that has gone out of the imaging area of the imaging unit and is no longer displayed in the captured image (hereinafter referred to as a “lost moving body”) in the captured image. A position (a missing position) in the virtual space corresponding to the image position is obtained. Then, the non-imaging area at the movement destination (rejection destination) from the lost position is specified using the area information, and the opening that forms the outer periphery of the non-imaged area is specified using the structure information. And considering that the probability (existence probability) that the vacant moving body exists in the non-imaging area becomes lower as the number of the specified openings is larger (and the elapsed time from the lapse time becomes larger), the elapsed time and The calculation is performed so that the existence probability decreases as the number of openings increases. Using the existence probability calculated in this way, for example, the symbol representing the lost moving object is changed in color, transparency, or the like according to the existence probability, and is superimposed and displayed on the plan view (map) of the monitoring place. Thus, when tracking a moving body between a plurality of imaging units, the monitor can estimate the location of the moving body that has moved outside the imaging area. Alternatively, by using the calculated existence probability and the image feature (feature amount) of the moving object, it is determined whether or not the unnoticed moving object in one imaging unit is the same as the moving object in the other imaging unit. The moving body can be accurately tracked between the imaging units.

The image monitoring apparatus according to claim 2 of the present application is the image monitoring apparatus of claim 1,
In the non-imaging area, the display control unit can move the movable body from the missing position through the opening without passing through the imaging area by using the lost position, the structure information, and the area information. A predicted route calculating means for calculating a predicted route;
A probability distribution is obtained based on the moving speed before the relapse of the moving object and the elapsed time, and a path-specific area existence probability is calculated for each predicted path using the probability distribution and the structure information, and the path-specific area Presence probability calculating means for calculating the existence probability from the existence probability.

  With such a configuration, the predicted route calculation unit of the present invention uses the lost position, the structure information, and the area information to predict that the lost moving body is one or a plurality of routes that can move from the lost position via the opening. Calculate the route. At this time, in consideration of the fact that when the lost moving body passes through the imaging area, it is determined that the moving body does not pass through the imaging area, the route is limited to the non-imaging area that can be moved. The route is calculated as a predicted route. Then, the existence probability calculating means of the present invention obtains a probability distribution on the predicted route based on the movement speed before the remnant of the remnant moving body and the elapsed time from the relapse time, and determines each probability based on the probability distribution and the structure information. The existence probability in the non-imaging area is calculated. Thus, by calculating the predicted path that changes depending on the number and position of openings based on the structure information and area information, and calculating the probability of existence for each predicted path, The existence probability can be calculated more accurately in consideration of the position of the part. In addition, by obtaining the probability distribution in consideration of the moving speed of the lost moving body before the remnant, it is possible to calculate a more accurate existence probability reflecting the movement mode of the remnant moving body before the remnant.

The image monitoring apparatus according to claim 3 of the present application is the image monitoring apparatus according to claim 1 or 2,
The image processing unit obtains a similarity between the image feature of the moving body and feature amount information stored in the storage unit in advance, and the moving body uses the similarity and the existence probability to detect the feature amount information. It has the determination means which determines whether it is the same as the mobile body corresponding to.

  With this configuration, for example, an image feature of a moving object (an unrecognized moving object) that has moved out of an imaging area of a certain image capturing unit is stored in advance in the storage unit as feature amount information, and the image of the unrecognized moving object and another image is captured. The existence probability calculated by the existence probability calculation means as well as the similarity between the image feature of the moving object and the feature amount information as to whether or not the moving object detected / tracked by the unit is the same Also determine by using. Alternatively, an image feature of a mobile object (for example, a specific person) to be authenticated is stored in advance as feature amount information in the storage unit, and whether or not the mobile object being tracked is the specific person is determined. The determination is made by using not only the degree of similarity between the image feature and the feature amount information stored in the storage unit in advance, but also the existence probability calculated by the existence probability calculating means. Thereby, since it is possible to accurately determine whether the mobile body currently being tracked is a remnant mobile body, the tracking accuracy of the mobile body in a monitoring place where a plurality of imaging devices are installed, The authentication accuracy of the mobile object can be improved.

The image monitoring apparatus according to claim 4 of the present application is the image monitoring apparatus according to any one of claims 1 to 3,
The display control unit includes display control means for displaying and outputting an image display representing the location of the moving object on the map of the monitoring location stored in advance in the storage unit based on the existence probability. Yes.

  With this configuration, the display control means of the present invention arranges a symbol representing the position of the moving object on the monitoring place map (plan view or three-dimensional map) using the virtual space position of the moving moving object. To do. At this time, if the moving body is a remnant moving body that exists outside the imaging area (non-imaging area), for example, the symbol has the highest existence probability based on the existence probability calculated by the existence probability calculating means. The symbol is arranged at a predetermined position in the non-imaging area, and the color, transparency, shape and the like of the symbol are changed and displayed according to the magnitude of the existence probability. Thereby, the supervisor can easily visually recognize the current location and the existence probability of the moving body that has moved out of the imaging area.

The image monitoring device according to claim 5 of the present application is the image monitoring device according to any one of claims 1 to 4,
The structure of the object in the structure information is three-dimensional shape data,
The storage unit further stores imaging condition information in which an installation position, an optical axis, and an angle of view of the imaging unit are associated with the structure information,
The image processing apparatus further includes an area setting unit that obtains the imaging area and the non-imaging area using the structure information and the imaging condition information and stores the area in the storage unit as the area information.

  With such a configuration, the area setting unit of the present invention, for example, even when the imaging condition information is changed by performing camera control such as pan, tilt, and zoom on the imaging unit, the changed imaging condition The imaging area and the non-imaging area can be dynamically calculated based on the structure information (three-dimensional model) expressing the three-dimensional shape data in the virtual space. Therefore, even when camera control or the like is performed on the imaging unit, it is possible to calculate the predicted path and the existence probability of the remnant moving body based on the changed latest imaging condition.

The image monitoring apparatus according to claim 6 of the present application is the image monitoring apparatus according to any one of claims 2 to 5,
The probability distribution is characterized by being expressed by a normal distribution using as parameters the average value obtained from the moving speed and the elapsed time and the standard deviation obtained from the elapsed time.

  With this configuration, the average value (peak value) of the probability distribution (probability density function) of the moving object can be suitably obtained from the moving speed of the moving object and the passage of time. Also, considering that the moving body does not necessarily move to the non-imaging area at the end, the standard deviation of the normal distribution may be increased with the passage of time, so that it may be located in any non-imaging area. Can be shown to the monitor as a more probable value.

  According to the image monitoring apparatus according to the present invention, by calculating the existence probability of the remnant moving body from the captured image and the building structure (structure information) and using the existence probability, the supervisor can image the remnant moving body. An action mode outside the area (non-imaging area) can be assumed, and tracking between a plurality of surveillance cameras can be performed with high accuracy.

Block configuration diagram showing the overall configuration of the image monitoring device Table showing structure information Table showing area information Table showing moving body information Diagram explaining generation of imaging area Diagram explaining generation of imaging area The figure explaining generation of a non-imaging area Flow chart showing processing procedure of overall operation by image monitoring apparatus A diagram representing the structural model of the monitoring location The figure explaining calculation of virtual space position The flowchart which shows the specific process sequence of a prediction path | route calculation process The figure explaining prediction route calculation processing Flow chart showing specific processing procedure of graph generation processing A diagram representing a graph Diagram showing probability distribution The figure explaining display control processing

  Hereinafter, an embodiment in which the image monitoring apparatus according to the present invention is applied to the building shown in FIG. 9 will be described in detail with reference to the accompanying drawings. FIG. 9 shows the structure of a building (monitoring place) as a monitoring target of the image monitoring apparatus as a three-dimensional model, and corresponds to structure information 211a described later. As shown in FIG. 9, the monitoring place in this embodiment is composed of a plurality of rooms, and each room is partitioned by a wall surface 51 on which a person cannot move and an opening 52 such as a door on which the person can move. In addition, four imaging devices 3 a to 3 d are installed in the monitoring place in the present embodiment, and each imaging device 3 is installed so as to capture a person moving on the floor surface 53.

(About the overall configuration of the image monitoring device)
As shown in FIG. 1, the image monitoring apparatus 1 according to the present embodiment includes a monitoring terminal 2 and an imaging apparatus 3. Note that the monitoring terminal 2 in the image monitoring apparatus 1 is installed in a security room or the like away from the monitoring place, and is not shown in FIG.

  The imaging device 3 is a so-called surveillance camera including an imaging element such as a CCD element or a C-MOS element, an optical system component, and the like, and functions as an imaging unit in the present invention. The imaging device 3 is installed on the upper part or ceiling of the indoor wall, and is installed so as to take an image while looking down at the monitoring place. The imaging device 3 captures the monitoring location at predetermined time intervals and sequentially transmits the captured images to the monitoring terminal 2. The time interval at which the captured image is captured is, for example, 1/5 second. In the present embodiment, as shown in FIG. 9, it is assumed that four imaging devices 3 are installed in the monitoring place.

  The monitoring terminal 2 is schematically configured to include a storage unit 21, a control unit 22, a communication unit 23, a display unit 24, and an input unit 25. The communication unit 23 is a communication interface such as a LAN or USB, and communicates with the imaging device 3. In the present embodiment, the captured images transmitted from the four imaging devices 3 are received via the communication unit 23. The input unit 25 is an information input device such as a keyboard, a mouse, a touch panel, or a portable storage medium reading device. The monitor or the like can use the input unit 25 to set information regarding various imaging conditions such as the installation position of each imaging device 3 or to select the imaging device 3 that is desired to be displayed on the screen.

  The storage unit 21 is an information storage device such as a ROM, RAM, or HDD. The storage unit 21 stores various programs and various data, and inputs / outputs such information to / from the control unit 22. The various data includes structure information 211, imaging condition information 212, area information 213, and moving body information 214. In addition to the above, as various data, various information used for the processing of the control unit 22 (for example, a captured image for each frame acquired by the imaging device 3, in order to extract a change area from the captured image by the extraction unit 221. A reference image to be used and an extraction threshold).

  The structure information 211 is information representing the building structure of the monitoring place. The structure information of the structural objects such as walls, floors / doors, pillars, and fixtures in the real world existing in the monitoring place as shown in FIG. This is information composed of three-dimensional shape data represented by dimensional coordinates (X, Y, Z) and attribute information given to each constituent object. As an example of the structure information 211, FIG. 2 shows the structure information 211 of this embodiment. As shown in FIG. 2, the structure information 211 includes an object identifier that is an identifier of a constituent object at a monitoring location, three-dimensional coordinates that represent the structure (size, shape, position) of the constituent object, and attributes of the constituent object. It is expressed as a table that associates information. Various types of attribute information, such as the type of component (wall / door / camera, etc.), whether it is an opening, whether it is movable, and the like are assigned to the attribute information. As will be described later, in the present embodiment, at least an attribute indicating whether or not the constituent object is an opening needs to be given. The three-dimensional shape data in the structure information 211 may be created by three-dimensional CAD based on the shape information of the monitoring place, or the data obtained by capturing the three-dimensional shape of the monitoring place by a three-dimensional laser scanner or the like is used. Also good. The structure information 211 created in this way is stored in the storage unit 21 by being set and registered from the input unit 25 by a supervisor or the like. Note that reference numeral 50 in FIG. 9 represents the structure information 211 as a three-dimensional shape model, and hereinafter, the three-dimensional model of the structure information 211 is referred to as a “structure model”.

  The imaging condition information 212 includes installation condition information related to the installation position and optical axis (attitude) of the imaging device 3 at the current time, and field angle condition information related to the focal length, the number of pixels, the pixel size, and the lens distortion. Individually set for each. The imaging condition information 212 is set as a value converted into a value in the coordinate system of the virtual space, and the structure information 211 and the imaging condition information 212 are associated with each other as values in the same coordinate system. Here, the installation condition information regarding the installation position represents the inside of the monitoring place (real space) as a three-dimensional orthogonal coordinate system, and the coordinate value of the reference point whose coordinates are known in the orthogonal coordinate system of the real space is changed from the reference point to the coordinate value. This is information expressed as coordinate values in the virtual space calculated using a known technique such as measuring and correcting the relative distance and direction. The installation condition information regarding the optical axis is information regarding the rotation angle of the optical axis of the imaging device 3 with respect to the coordinate axis, and can be obtained from the so-called pan angle and tilt angle of the imaging device 3. The imaging condition information 212 is stored in the storage unit 21 by being set and registered from the input unit 25 by a monitor or the like at the time of initial setting.

  The area information 213 is coordinate information that represents a range in a virtual space between an imaging area that represents a range captured by the imaging device 3 and a non-imaging area that is an area other than the imaging area at the monitoring location. It is converted into a coordinate system value and stored. In the present embodiment, area information 213 is calculated from the structure information 211 and the imaging condition information 212 by the area setting unit 22B described later, and stored in the storage unit 21. In the present embodiment, the area information 213 includes coordinates indicating the range in the virtual space calculated by the area setting unit 22B described later as shown in FIG. 3 (two-dimensional on the floor surface when the floor surface is Z = 0). (Coordinates) and area attributes indicating whether the imaging area / non-imaging area is different are stored in the storage unit 21 as a table associated with an area identifier that is an identifier of each area.

  The moving body information 214 is information on a moving body (change area) tracked by the tracking unit 222 described later, and is registered in the storage unit 21 by the tracking unit 222 and the virtual space position calculation unit 224 described later. It is. As shown in FIG. 4, the moving body information 214 includes a tracking label that is a unique label given to each tracking moving body in the tracking process by the tracking unit 222, and feature amount information that is an image feature of the moving body. The virtual space position, which is the position (coordinates) in the virtual space of the moving body, and the detection time, which is the time when the moving body is detected, are stored in the storage unit 21 as a table in which they are associated with each other. Here, the feature amount information of the moving object is a vector quantity that quantitatively represents the image feature unique to the moving object, and includes color and texture features, shape features, size and aspect ratio features, and movement (moving speed). ) And a feature of a position in the captured image. For example, the mode luminance values of the lower body and the upper body can be used as the color feature of the moving object, and in this case, the feature amount is a two-dimensional vector. The moving body information 214 is used in tracking processing in the tracking unit 222 and determination processing in the determination unit 223 as described later.

  The control unit 22 is configured by a computer having a CPU and the like, receives an input of a digitized image from the imaging device 3, and performs an area generation process, a captured image reading process, and an extraction as a series of processes shown in FIG. An image processing unit 22A including an extraction unit 221, a tracking unit 222, and a determination unit 223, and an area setting unit to execute processing, tracking processing, predicted route calculation processing, existence probability calculation processing, determination processing, and display control processing 22B and a display control unit 22C including a virtual space position calculation unit 224, a predicted route calculation unit 225, an existence probability calculation unit 226, and a display control unit 227.

  The extraction unit 221 of the image processing unit 22A performs an extraction process of extracting an area having a luminance change from the captured image acquired by the imaging apparatus 3 as a change area. In the present embodiment, a past captured image in which no moving object exists for each imaging device 3 is stored in the storage unit 21 in advance as a reference image, and a difference value between luminance values of the latest acquired captured image and the reference image is determined. An area that is equal to or greater than a predetermined extraction threshold stored in the storage unit 21 is extracted as a change area. At this time, a captured image of the background of the monitoring place, a captured image acquired in the past, or the like can be appropriately selected and used as the reference image. Further, the extraction unit 221 performs a process of labeling the extracted change area as the extraction process. At this time, when attention is paid to an extraction pixel having a change area, the extraction pixels adjacent to the extraction pixel of interest are regarded as a group of extraction pixel areas. Assign a unique label. Hereinafter, a group of extracted pixel areas to which a label is assigned is referred to as a label area.

  The tracking unit 222 temporally traces the change area extracted by the extraction unit 221, and the tracking label region (tracking) of the captured image of the same imaging device 3 acquired last time with respect to the label region obtained by the extraction unit 221. A tracking process for associating with a label area (labeled area) is performed. The tracking process will be described in more detail. First, the feature quantity of the tracking label area of the captured image acquired last time is compared with the feature quantity of the label area of the captured image currently being processed. Determine whether or not. If it is determined that the tracking object is the same tracking object, the same tracking label as the tracking label area of the previously acquired image is applied again. For tracking label areas that existed in the previous tracking label area that is not associated with any of the label areas that appear in the currently processed image, the corresponding moving body is considered to have gone out of the imaging area, and the moving body Recognized as a lost mobile. Note that the feature amount information of the moving object used for the tracking process is associated with the tracking label and the detection time so that it can be used in the tracking process in a later captured image frame or the determination process in the determination unit 223 described later. The moving part information 214 is stored in the storage unit 21.

  The area setting unit 22B performs area generation processing for calculating the area information 213 using the structure information 211 and the imaging condition information 212 and setting the area information 213 in the storage unit 21. Details of the area generation process will be described below with reference to FIGS.

  FIG. 5 shows an example of calculating the imaging area in the area information 213 based on the imaging condition information 212 and the structure information 211 of the imaging device 3. In FIG. 5, reference numerals 51 a and 53 a represent a part of the structural model 50 a used for explaining the present processing. Of these, reference numeral 51 a represents a wall surface of the structural model 50 a, and 53 a represents a floor of the structural model 50 a. It represents the surface. In calculating the imaging area, the area setting unit 22B first reads the installation position (X, Y, Z) of the imaging device 3 from the imaging condition information 212 of the storage unit 21, and the structure information 211 ( The optical center O on the (virtual space) is obtained. The area setting unit 22B obtains the optical axis of the imaging device 3 in the virtual space from the optical center O and the installation condition information regarding the optical axis (posture) read from the imaging condition information 212 of the storage unit 21. Further, the area setting unit 22B reads the imaging condition information 212 such as the focal length f, the actual size of the CCD pixel, the number of pixels in the vertical and horizontal directions of the image, and the lens distortion, from the storage unit 11, and the structure of the imaging device 3 On the information 211, a projection plane abcd perpendicular to the optical axis is obtained. Then, a quadrangular pyramid Oa'b'c'd 'passing through the four vertices of the projection surface abcd from the optical center O is generated. The height of the quadrangular pyramid is an arbitrary height that allows at least four vertices (a ', b', c ', d') on the bottom surface of the quadrangular pyramid to penetrate the structural model 50a. Then, the area setting unit 22B obtains an interference surface ABCD between the quadrangular pyramid and the floor surface 53a of the structural model 50a by a known geometric calculation, and calculates an area composed of the interference surface ABCD as the imaging area 30a.

  As described above, the shape of the monitoring range model 213 changes depending on the value of the imaging condition information 212. For example, when the imaging device 3 performs a zoom operation, the length of the side of the bottom surface a′b′c′d ′ of the quadrangular pyramid decreases as the focal length f increases, and the size of the interference surface also increases. It changes to be smaller. Further, when the pan / tilt operation is performed on the image pickup device 3 and the optical axis of the image pickup device 3 is moved so that the floor surface 53a and the wall surface 51a are included in the monitoring range as shown in FIG. , Calculated as an area 30b made of ABEF.

  Next, the area setting unit 22B calculates a non-imaging area using the calculated imaging area 30 and the structure information 211. In the present embodiment, an area in a monitoring place other than the imaging area 30 and where the movement of a person is restricted by an object such as a wall or a door is calculated as a non-imaging area. FIG. 7 is a diagram illustrating processing for generating a non-imaging area. FIG. 7 a is a diagram illustrating the structural model 50 b and the non-imaging area 31 when the imaging device 3 is not installed. That is, the non-imaging area 31 that is an area where movement of a person is restricted by walls and doors in the figure can be obtained as two areas 31a and 31b. FIG. 7 b is a diagram illustrating the structural model 50 b, the imaging area 30 c, and the non-imaging area 31 when the imaging device 3 is installed. That is, as shown in the figure, by calculating the imaging area 30c, the non-imaging area 31 can be obtained as three areas 31c to 31e. Thus, when the imaging area 30 is generally obtained, the non-imaging area 31 other than the imaging imaging area 30 can be uniquely calculated using the structure information 211.

  The virtual space position calculation means 224 obtains the virtual space position that is the position in the virtual space from the image position on the captured image for all the moving objects tracked by the tracking means 222, and the tracking label of the moving object A virtual space position calculation process to be stored in association with the moving body information 214 of the storage unit 21 is performed. Details of the virtual space position calculation process will be described below with reference to FIG.

  FIG. 10 illustrates an example of calculating the virtual space position of the label area based on the captured image of the imaging device 3 that captures the ground direction from above the monitoring location. In the virtual space position calculation process, similarly to the area calculation process described above, the optical axis on the structural model 50c of the imaging apparatus 3 and the projection plane abcd perpendicular to the optical axis are obtained based on the imaging condition information 212. In FIG. 10, it is assumed that an image area denoted by reference numeral 40 on the captured image projected on the projection plane abcd is a label area of a moving object that is extracted by the extracting unit 221 and tracked by the tracking unit 222. Next, a straight line connecting the lower end position 40a of the label region 40 and the optical center O is obtained, and an intersection 41a with the structural model 50c on the extension line of the straight line is obtained. In the present embodiment, when there are a plurality of pixels located at the lower end of the label area 40, the pixel at the center position of the label area 40 is set as the lower end position 40a. However, the present invention is not limited to this. ) May be the lower end position 40a. Since the intersection 41a corresponds to a position on the structural model 50c (virtual space) corresponding to the position 40a on the captured image of the label area 40, the coordinates of the position in the virtual space are calculated as the virtual space position.

  The predicted path calculation means 225 uses the virtual space position (lost position) immediately before the lapse, the area information 213, and the structure information 211 when the tracked moving body moves out of the imaging area and is lost. A predicted route calculation process is performed to calculate a predicted route in the non-imaging area 31 in which the moving body can move from the lost position without passing through the imaging area 30. Details of the predicted route calculation process will be described later.

  The existence probability calculating unit 226 uses the movement speed and the elapsed time from the lapse time before the remnant time, the predicted route obtained by the predicted route calculating unit 225, and the like for each non-imaging area 31. Existence probability calculation processing for calculating the existence probability is performed. The details of the existence probability calculation process will be described later.

  The determination unit 223 performs a determination process for determining whether or not a moving body that has appeared in a captured image of a certain imaging device 3 is the same as any of the remnant moving bodies tracked by the tracking unit 222 in the past. The determination process will be described more specifically. The feature amount information of the unrecognized moving body tracked by the tracking unit 222 is read from the moving body information 214 of the storage unit 21, and the feature amount information and the movement currently detected are read. The image feature (feature amount) of the body is compared and the similarity between the two is calculated. As described above, the feature amount of the moving object is expressed as an n-dimensional vector amount corresponding to the number of image features used for determination. Accordingly, the degree of similarity can be calculated by obtaining the Euclidean distance between the feature amount information of the lost moving body and the feature amount of the currently detected moving body. When the similarity obtained in this way is equal to or smaller than a predetermined similarity determination threshold value stored in advance in the storage unit 21, both are determined to be the same moving object, assuming that they are sufficiently close feature amounts. In the present embodiment, the similarity is calculated using the Euclidean distance, but other than the Euclidean distance, for example, the Mahalanobis distance may be used.

  In the determination process, by using the existence probability of the sight moving body calculated by the existence probability calculating means 226, the identity between the sight moving body and the currently detected moving body is determined. Specifically, the existence probability of the non-imaging area 31 (hereinafter referred to as “adjacent non-imaging area”) adjacent to the imaging area 30 of the imaging apparatus 3 that is imaging the moving body currently detected is calculated. The process of referring to and changing the magnitude of the similarity determination threshold according to the existence probability is performed. That is, if the existence probability of the adjacent non-imaging area is small, it means that there is a low possibility that there is a remnant moving body in the adjacent non-imaging area, and that the newly detected moving body is unlikely to be a remnant moving body. Therefore, the similarity determination is performed more strictly than usual so that the erroneous determination is less likely to occur by changing the similarity determination threshold value to be smaller. On the other hand, if the existence probability of the adjacent non-imaging area is large, the adjacent non-imaging area is likely to have a retinal moving body, and the newly detected moving body is likely to be a retinal moving body. For this reason, the similarity determination is performed so that the loss determination is less likely to occur by changing the similarity determination threshold to a larger value. As described above, in this embodiment, whether or not the sight moving body is the same as the currently tracked moving body, whether or not the image feature of the moving body and the feature amount information of the sight moving body stored in the storage unit 21 is determined. By using the existence probability calculated by the existence probability calculation unit 226 as well as the degree of similarity, it is possible to make a determination with high accuracy. Therefore, in a monitoring place where a plurality of imaging devices are installed The tracking accuracy of the moving object can be improved. In the determination process, when it is determined that the moving object that appears in the imaging area is the same as the moving object, the process of initializing the existence probability of the moving object in the non-imaging area to “0” is performed. Therefore, in this case, since the probability of existence of the remnant moving body in the non-imaging area adjacent to another imaging area (different from the imaging area determined to be the same) is necessarily 0, the other imaging area This means that even if a new mobile object appears, the similarity determination threshold value between the feature quantity of the mobile object and the feature quantity of the lost mobile object is minimized, and it is most difficult to determine that they are the same. To do.

  The display control means 227 superimposes and displays a symbol representing the moving body at a position on the map of the monitoring location corresponding to the virtual space position of the moving body, and displays a display output for displaying the moving locus of the moving body. Display control processing to be performed. At this time, the symbol of the moving object is displayed in a color different from that of the moving object existing in the imaging area, and the symbol whose transparency is changed based on the existence probability calculated by the existence probability calculating unit 226 is displayed. It is arranged at the center of gravity of the non-imaging area with the highest existence probability. Thereby, the supervisor can presume suitably the location of the moving body which moved out of the imaging area. In the present embodiment, the map is set and registered in advance in the storage unit 21 by an administrator or the like, but the map may be dynamically generated using the structure information 211. Details of the display control process will be described later.

  The display unit 24 is an information display device such as a display, and displays a captured image of a monitoring location imaged by the imaging device 3 and a map image output by the display control unit 227 (a symbol of a moving object is superimposed and displayed). indicate.

  As described above, the image monitoring apparatus according to the present embodiment calculates the existence probability of the remnant moving body based on the structure information 211, the imaging condition information 212, the image feature (moving speed) of the tracking moving body, and the like. Even if the imaging device 3 cannot cover all areas of the monitoring location by clearly indicating the current position of the sighting moving body to the monitoring person by the display output according to the existence probability, the monitoring person can move outside the imaging area. It is possible to predict and grasp the approximate positional relationship of the moving body (a missing moving body) that appears on the screen. Further, by using the existence probability to determine the identity of the moving body between the plurality of imaging devices 3, tracking between the plurality of imaging devices 3 can be performed with high accuracy.

(About processing executed by the control unit 22 of the monitoring terminal 2)
Hereinafter, an example of the flow of processing executed by the control unit 22 of the monitoring terminal 2 according to the image monitoring apparatus 1 of the present embodiment will be described in detail with reference to the flowchart of FIG.

  Prior to the operation, various initial settings such as setting of the imaging condition information 212 and registration of the structure information 211 are performed by the monitor or the like using the input unit 25 of the monitoring terminal 2 (ST1). In this embodiment, since it is assumed that four imaging devices 3 are installed in the building that is the monitoring place as described above, the imaging condition information 212 is registered by default for each imaging device 3. Is done. The imaging condition information 212 registered at the time of initial setting may be changed by performing camera control such as pan / tilt / zoom on the imaging device 3. For this reason, in the present embodiment, when camera control is performed, the latest imaging condition information 212 is received from the imaging device 3 via the communication unit 23 and the storage unit 21 is updated. In addition, the following processing will be described on the assumption that the structure information 211 related to the structure model 50 in FIG.

  Next, the area setting unit 22B of the control unit 22 executes area generation processing for calculating the area information 213 using the structure information 211 and the imaging condition information 212 (ST2). Note that the details of the area generation processing have been described above, and a description thereof will be omitted here. As described above, the area information 213 is updated by using the latest imaging condition information 212 stored in the storage unit 21, so that even if the camera control is performed on the imaging device 3, the information is changed. Based on the latest imaging condition information 212, it is possible to calculate the predicted path and the existence probability of the lost moving body.

  Next, the control unit 22 performs a captured image reading process for reading the captured image received from the imaging device 3 and stored in the storage unit 21 for each imaging device 3 (ST3). In the captured image reading process, the latest captured image (closest to the current time) is read for each imaging device 3.

  Next, extraction processing is executed by the extraction means 221 of the control unit 22 (ST4). In the extraction process, the change areas are extracted from the captured images of the respective image pickup devices 3 acquired in the captured image reading process, and the extracted change areas are labeled. Since the details of the extraction process have been described above, the description thereof is omitted here.

  Next, a tracking process is executed by the tracking unit 222 of the control unit 22 (ST5). Since the details of the tracking process have been described above, the description thereof is omitted here.

  Next, the virtual space position calculation means 224 executes virtual space position calculation processing for obtaining the virtual space position of each label area tracked by the tracking means 222 (ST6). Note that details of the virtual space position calculation processing have been described above, and a description thereof will be omitted here.

  Next, the predicted route calculation unit 225 executes a predicted route calculation process for calculating a predicted route in the non-imaging area 31 in which the lost moving body can move from the lost position without passing through the imaging area 30 (ST6). Details of the predicted route calculation process will be described below with reference to the flowchart of FIG. In addition, the loop 1 of FIG. 11 means that each process of ST30 to ST33 is executed by the number of the remnant moving bodies recognized in the tracking process. In the description of this processing, the selected moving body refers to a retinal moving body that is a processing target in the loop 1.

  In the predicted route calculation process, first, it is determined whether or not the initial movement destination area of the selected moving body is an imaging area (ST30). That is, since the lost position of the selected moving body that is the lost moving body corresponds to the virtual space position detected at the end of the selected moving body in the moving body information 214, the area adjacent to the lost position is set as the initial movement destination area. Therefore, it is determined with reference to the area attribute in the area information 213 whether the initial movement destination area is an imaging area or a non-imaging area. When it is determined in ST30 that the initial moving destination area of the selected moving body is the imaging area (ST30-Yes), the other lost moving body that has not yet been processed in loop 1 is changed to the selected moving body. Then, the process from ST30 is performed.

  When it is determined in ST30 that the initial moving destination area of the selected moving body is not the imaging area (ST30-No), the initial moving destination area of the selected moving body is set as the moving source area (ST31). FIG. 12 is a diagram for explaining the predicted route calculation process, in which the imaging area 30 and the non-imaging area 31 are superimposed and displayed on a drawing that simply represents the structural model 50 of FIG. 9 in two dimensions. . In the same figure, areas B, D, J, and O painted with diagonal lines are imaging areas 30, and other areas A, C, E, F, G, H, I, K, L, M, and N Is a non-imaging area 31. A line / arrow indicated by reference numeral 33 represents a movement trajectory in area D, which is an imaging area before the remnant moving body, and reference numeral 34 represents a retirement position of the remnant moving body. . In the example of the figure, since the initial destination area of the lost moving body is the area G that is a non-imaging area, it is determined in ST30 that it is not an imaging area, and the area G is set as a source area in ST31. The

  When the movement source area is set in ST31, the predicted route calculation means 225 performs graph generation processing (ST32). The graph generation process is a process of generating a graph having a data structure in which a node representing a non-imaging area to which a lost moving body can move from the movement source area is connected to a node representing the movement source area by an edge. The details of the graph generation processing will be described below with reference to the flowchart of FIG.

  In the graph generation process, first, the structure information 211 of the storage unit 21 is referred to, and a configuration having an “opening” attribute that configures the outer periphery of the movement source area (that is, the non-imaging area of the vacant destination) set in ST31. An object is specified (ST40). In the example of FIGS. 12 and 9, the object having the opening attribute that forms the outer periphery of the area G that is the movement source area of the remnant moving body is only the opening 52 f (door) that leads to the area H. Only 52f is identified. Loop 2 in FIG. 13 means that the processes of ST41 to ST43 are executed by the number of openings specified in ST40. In the description of this process, the selected opening refers to an opening that is a processing target in the loop 2.

  Next, it is determined whether or not the previous area passing through the selected opening is an imaging area (ST41). When it is determined in ST41 that the previous area that has passed through the selected opening is an imaging area (ST41-Yes), another opening that is not yet processed in Loop 2 is changed to the selected opening. Then, the process from ST42 is performed. On the other hand, when it is determined in ST41 that the previous area via the selected opening is not the imaging area (ST41-No), the node of the previous area via the selected opening is connected to the source node. The graph is updated, and the updated graph is stored in the storage unit 21 (ST42). In the example of FIG. 12, since the area H that is the previous area via the selected opening is a non-imaging area, the graph indicating the area H is connected to the source node indicating the area G and the graph is updated.

  Next, the previous area that has passed through the selected opening is reset as the movement source area, and the graph generation processing from ST40 is recursively processed (ST43). That is, in the example of FIG. 12, area H is set as the movement source area in ST43. And in ST40 processed recursively, the opening parts 52h, 52i, and 52j adjacent to the movement origin area (area H) are specified. Since the previous areas L, M, and N that have passed through these openings 52h, 52i, and 52j are non-imaging areas, the nodes indicating these areas are moved to the movement source area in ST42 that has been recursively processed. The graph is updated by connecting to the nodes in (area H). A graph in the example of FIG. 12 generated in this way is shown in FIG. FIG. 14 shows a moving route of the non-imaging area 31 in which the sight moving body in the example of FIG. 12 can move from the sight position 34 without passing through the imaging area 30. That is, it can be seen that the remnant moving body in the example of FIG. 12 can move on routes that move in the areas of G → H → L, G → H → M, and G → H → N. In this way, when the sight moving body passes through the imaging area, it is limited to only the route in the non-imaging area that can move without passing through the imaging area, considering that it is always found in the imaging area. Thus, by obtaining the travel route, it is possible to limit an unlimited number of travel routes depending on the combination, and to reduce the amount of calculation.

  When the graph generation process is completed in ST32, the predicted path calculation unit 225 advances the process to ST33 of the predicted path calculation process of FIG. In ST33, a process of calculating a predicted route is performed using the graph generated by the graph generation process and the structure information 211. Here, the predicted route means a route in which it is predicted through which position the lost moving body will move in the movement between the non-imaging areas shown in the graph. In this embodiment, it is assumed that the lost moving body moves along the shortest path between the lost position and the opening of the destination area indicated by the graph, and the area corresponding to the terminal node of the graph is set to the center of gravity position of the area. A predicted route is calculated on the assumption that the vehicle moves toward the vehicle. That is, in the example of FIG. 12, in the movement route of G → H → L in the graph of FIG. 14, the lost moving body moves along the predicted route of the lost position 34 → the opening 52f → the opening 52h → the area centroid position 36a. I decided to. Similarly, in the G → H → M movement route in the graph, the movement moves along the predicted route of the remnant position 34 → the opening 52f → the opening 52i → the area centroid position 36b, and in the G → H → N movement route in the graph. Is moved along the predicted path of the lost position 34 → the opening 52f → the opening 52j → the area gravity center position 36c. The position of the opening 52 is calculated from the gravity center position of the constituent object having the “opening” attribute with reference to the structure information 211.

  When the predicted route is calculated in ST33, the predicted route calculation means 225 finishes the predicted route calculation process in ST7 of FIG. 8, and advances the process to ST8 of FIG. In ST8, the existence probability calculation means 226 executes an existence probability calculation process. Details of the existence probability calculation process will be described below.

In the existence probability calculation process, first, a probability distribution on each predicted route is obtained using the movement speed before the remnant moving body and the elapsed time from the relapse time. In the present embodiment, when the probability distribution is defined as the movement speed v before the remnant moving body and the elapsed time T from the relapse time, the average value μ = vT, and P = ∫p (x) dx = 1 / It is assumed that the distribution is a normal distribution obtained by √ (2πσ ² ) × ∫exp (− (x−μ) ² / (2σ ² )) dx (1). Here, the moving speed v before the remnant moving body is a value read from the feature amount information in the moving body information 214 of the storage unit 21 and is calculated from the moving distance between frames of the captured image. It is a calculated value. The elapsed time T is a value calculated from the difference between the detection time in the mobile object information 214 of the storage unit 21 and the current time. In the present embodiment, the maximum value of μ is set as the path length of each predicted path, and μ does not increase any more when μ reaches the path length. The standard deviation σ is a parameter calculated by a function σ (T) that increases as the elapsed time T increases. FIG. 15 shows the probability distribution obtained by Equation (1) in the predicted route of the G → H → L movement route in FIG. In the figure, reference numeral 37a, 37b represents the probability distribution at the time of each elapsed time _{T 1, T 2 (T 1} <T 2). In the present embodiment, the standard deviation σ is obtained as a function of the elapsed time T as described above, but a fixed value experimentally predetermined by an administrator or the like may be used.

Next, the existence probability (area-specific probability of each route) in each non-imaging area (area A, C, E, F, G, H, I, K, L, M, N) is calculated for each predicted route. At this time, the path-specific area existence probability is calculated using the integration interval obtained from the entry position and the exit position in each non-imaging area on the predicted route with respect to the probability distribution of Expression (1). Here, the intrusion position in the area G corresponds to the lost position 34, and the exit position corresponds to the position of the opening 52f. Therefore, when the distance between the position of the opening 52f from the lost position 34 is X1, the integration interval in the equation (1) is set to [0, X ₁ ], and the area-specific probability P _g1 for each route is calculated. The entry position in the area H corresponds to the position of the opening 52f, and the exit position corresponds to the position of the opening 52h. Therefore, when the distance between the position of the opening 52h from the position of the opening 52f and the _{X 2,} the integral interval in Equation (1) as _{[X 1, X 1 + X} 2], the path-specific area existence probability _{P h1} calculate. Further, the entry position in the area L corresponds to the position of the opening 52h, and the exit position corresponds to the area gravity center position 36a. Therefore, when the distance from the position of the opening 52h between areas centroid position 36a was _{X 3,} as the integral interval in Equation (1) is _{[X 1 + X 2, X} 1 + X 2 + X 3], path-specific area existence probability _P11 is calculated. In this way, the area-specific area existence probabilities P _g1 , P _h1 , and P ₁₁ of the predicted route of the G → H → L movement route can be calculated using the equation (1) and the integration interval. In the same way, G → H → presence path-specific area in the prediction path of the moving route of the M probability _{P g2,} P _h2, P l2 _and,, G → H → N paths by area existence probability in the prediction path of the moving route of P _g3 , P _h3 and P _l3 are calculated.

Next, the path-specific area existence probability obtained for each predicted path is normalized to calculate the existence probability of the retinal moving body for each non-imaging area. At this time, normalization is performed so that the sum of the existence probabilities in all non-imaging areas is 1. Specifically, the sum P of the calculated route by area existence probability _'= when the _{P g1 + P g2 + P g3} + P h1 + P h2 + P h3 + P l1 + P l2 + P l3, the existence probability _{P g} of the non-imaging area G is P _g = (P _g1 + P _g2 + P _g3 ) / P ′ Similarly, the existence probability _{P h = (P h1 + P} h2 + P h3) of the non-imaging area H / P is calculated by 'the existence probability _P l of the non-imaging area _{L = (P l1 + P l2} + P l3) / P' .

  Next, determination processing is executed by the determination means 223 of the image processing unit 22A (ST9). In the determination process, it is determined whether or not a mobile object that has newly appeared in a captured image of a certain imaging device 3 is the same as any of the retinal mobile objects that have been tracked by the tracking unit 222 in the past, and is determined to be the same. When this is done, the process of associating the tracking label of the newly appearing mobile object with the tracking label of the remnant mobile object is executed by setting / changing it. Since the details of the determination process have been described above, the description thereof is omitted here.

  Next, display control processing is executed by the display control means 227 of the display control unit 22C (ST10). In the display control process, the display control means 227 reads a map of the monitoring location stored in advance in the storage unit 21, and places a symbol representing the moving object at a position on the map corresponding to the current virtual space position of the moving object. To do. At this time, a symbol whose display has been changed based on the existence probability is arranged for the lost mobile body. Hereinafter, a specific example of the display control process will be described with reference to FIG.

  FIG. 16 shows an example of a map screen showing the current position of the moving object displayed and output on the display unit 24 by the display control means 227. In the figure, symbols S1 to S6 represent symbols at times t1 to t6 of the moving body tracked by the tracking means 222. In S1 and S2, since the virtual space position is calculated in the virtual space position calculation process of ST6, the symbol of the moving object is arranged at a position on the map corresponding to the virtual space position. Since the mobile object has been lost at time t2 (position S2), a symbol is arranged at the barycentric position of the non-imaging area with the highest existence probability on each predicted path after such a lost. In the example of FIG. 16, at t3 after time t2, the symbol S3 corresponding to the lost moving body is arranged at the center of gravity of the area G, which is the non-imaging area with the highest existence probability, and is connected and displayed by the movement locus 33. doing. At this time, in the present embodiment, the symbol display of the symbol S3 is changed depending on the opacity according to the existence probability. That is, the smaller the existence probability, the more transparent the symbol is displayed. In this way, by displaying the current position of the sighted moving body according to the existence probability, even if the entire area of the monitoring place cannot be monitored by the imaging device 3, the monitor goes out of the imaging area. It is possible to predict and grasp the approximate positional relationship of the target moving body. In the present embodiment, symbols are not displayed when the existence probability is 50% or less. The symbol S5 in FIG. 16 represents that even if the existence probability of the non-imaging area G is the maximum at time t5, the existence probability is 50% or less, so that the symbol is not displayed (for explanation) For the sake of convenience, but not actually displayed). In this way, by not displaying the symbol of the sight moving body having a predetermined existence probability or less, the position of the sight moving body is not misleaded to the monitor.

  Note that, as shown in FIG. 16, when the moving body detected in the imaging area O at time t6 is determined to be the same moving body as that corresponding to S1 and S2 in the determination process of ST9. The symbol S6 representing the moving object and the symbol S5 representing the lost moving object are connected by a movement locus. Thereby, the supervisor can recognize that the moving bodies related to the symbols S1 to S6 are the same, and can estimate the behavior mode of the moving body outside the imaging area of the imaging device 3.

  By the way, the present invention is not limited to the above-described embodiment, and may be implemented in various different embodiments within the scope of the technical idea described in the claims. Further, the effects described in the embodiments are not limited to this.

  In the above embodiment, a predicted route is calculated based on the position and number of openings specified by the predicted route calculating unit 225 of the display control unit 22C, and the existence probability calculating unit 226 calculates the moving speed and elapsed time of the moving object. Based on the obtained probability distribution, a path-specific area existence probability is obtained for each predicted route, and the existence probability in each non-imaging area is calculated. That is, in the above-described embodiment, the number of predicted route candidates increases as the number of openings increases, the presence of a plurality of predicted routes according to the connection relationship between areas via the openings, and the physical properties of the openings. This is because the moving distance of the lost moving body on the predicted route differs depending on the positional relationship. However, the present invention is not limited to this, and another embodiment is used in which the existence probability is simply calculated from the number of openings and the elapsed time, considering only that the number of predicted path candidates increases as the number of openings increases. May be. That is, in the other embodiment, the greater the number of openings in the non-imaging area that is the destination of the lost moving body (remaining destination), the more the existence probability in the non-imaging area with the passage of time from the lapse time. A method of calculating so as to be small is used. Specifically, the predicted route calculation unit 225 in the other embodiment is a destination (a relapse destination) of the remnant moving body without calculating a predicted route in the predicted route calculation process as in the above embodiment. Only the opening part which comprises the outer periphery of a non-imaging area is specified. Then, the existence probability calculation means 226 in the other embodiment calculates the existence probability P in the destination non-imaging area as P = exp (−αNT), where N is the number of identified openings and T is the elapsed time. ... Calculated according to (2). Here, α is a fixed parameter experimentally determined in advance, and is a value set by an administrator or the like. In the expression (2), for example, when the moving body moves to the non-imaging area of the narrow path where the number of openings is “0”, the existence probability is always “1 (always exists) regardless of the elapsed time. "Is calculated. On the other hand, the existence probability is calculated to decrease as the number of openings increases and as the elapsed time elapses. Then, the display control means 227 in the other embodiment arranges the symbol indicating the lost moving body at the center of gravity position of the non-imaging area on the map, and changes the opacity of the symbol according to the calculated existence probability. The obtained map image is output to the display unit 24.

  In the above embodiment, the coordinate information of the three-dimensional shape data is stored as the structure information 211 stored in the storage unit 21, but the present invention is not limited to this, and the coordinate information of the two-dimensional shape data may be simply used. . In this case, the area information 213 (imaging area and non-imaging area) is not calculated by the area calculation process in the area setting unit 22B, and the calibration of the imaging apparatus 3 is performed in advance by an administrator or the like. The setting is registered in the storage unit 21. That is, in the other embodiment, the imaging condition information 212 of the storage unit 21 is omitted without being stored in the storage unit 21, and the area setting unit 22B is also omitted. In addition, the virtual space position calculation process in the other embodiment is configured to calculate the virtual space position (including the lost position) from the image position on the captured image based on the information regarding the imaging area that has been calibrated in advance. .

  In the above-described embodiment, the determination unit 223 determines the identity of a newly appearing moving body and a previously-acquired moving body based on the similarity and presence probability of the image features of both. It is carried out. However, the present invention is not limited to this, and the degree of similarity between the feature quantity of a newly appearing mobile object and the registered feature quantity of a specific person registered in advance by an administrator or the like is obtained, and the mobile object is calculated from the similarity and the existence probability. A process for determining whether or not is the specific person (so-called authentication process) may be performed in the determination process.

In the above-described embodiment, the predicted path calculation unit 225 uses the predicted path in the non-imaging area where the retinal moving body can move from the lapse position without passing through the imaging area as the “opening” in the non-imaging area of the movement destination (lost destination) It is calculated according to the number of constituent objects having the “part” attribute. However, the predicted route may be calculated in consideration of not only the number of constituent objects having the “opening” attribute but also the “opening time” attribute of the constituent objects having the “opening” attribute. That is, the opening time (the time when “open”) of a door or window, which is a constituent object having the “opening” attribute, is stored as attribute information, and the opening / closing time is taken into account when calculating the predicted route. Specifically, for example, a sensor such as a magnet sensor is installed in advance in the constituent object having the opening attribute so that the opening state of the constituent object having the opening attribute can be detected, and the monitoring terminal 2 passes through the communication unit 23. So that the state can be received. And when the constituent object which has an opening part attribute opens, the process which changes the attribute information (structure information 211 of the memory | storage part 21) of the open constituent object based on the time of the state received via the communication part 23 Do. Then, when calculating the predicted path of a certain moving object, the constituent object having the “opening” attribute in the non-imaging area of the moving destination (the lost object) of the moving object is specified, and Processing to read the opening time is performed. Then, with reference to the read opening time, when it is determined that the identified constituent object is not opened between the lapse time and the time when the predicted route is calculated, the non-opening constituent object Processing is performed so that the imaging area is not included in the predicted path. In other words, in the loop 2 of the graph generation process (ST32) in the predicted route calculation process (ST7), processing is performed so that the constituent object that is not open is not selected as the selected opening. By processing in this way, it is possible to calculate the existence probability with higher accuracy without calculating the existence probability of the non-imaging area in which the remnant moving body cannot move before the opening that is not open. .

  In the above-described embodiment, the display control unit 227 performs display control processing so as to display the sighted moving body as a symbol on the map of the monitoring place based on the existence probability. However, the present invention is not limited to this, and the non-imaging area 31 may be colored and displayed based on the existence probability without arranging symbols. For example, the display control process may be performed so that the non-imaging area 31 having a higher existence probability is colored in dark red, and the area having a lower existence probability is colored in light red.

DESCRIPTION OF SYMBOLS 1 ... Image monitoring apparatus 2 ... Monitoring terminal 3 ... Imaging device 21 ... Memory | storage part 22 ... Control part 23 ... Communication part 24 ... Display part 25 ... Input part 211 ... Structural information 212 ... Imaging condition information 213 ... Area information 214 ... Moving body information 215 ... Position information 22A ... Image processing unit 22B ... Area setting unit 22C ... Display Control unit 221 ... Extraction means 222 ... Tracking means 223 ... Determination means 224 ... Virtual space position calculation means 225 ... Predicted path calculation means 226 ... Presence probability calculation means 227 ... Display Control means 30 ... Imaging area 31 ... Non-imaging area 33 ... Movement locus 34 ... Lost position 36 ... Area centroid position 37 ... Probability distribution 40 ... Change area 41 ... Virtual space position 50 ・..Structural model 51 ... Wall 52 ... Opening 53 ... Floor surface

Claims (6)

An imaging unit that sequentially acquires captured images of the monitoring location;
The structure information indicating the structure in the virtual space and the attribute indicating whether or not the object is at least an opening, and the imaging area of the imaging unit and the non-imaging area other than the imaging area in the virtual space. A storage unit storing the area information;
An image processing unit that temporally tracks the moving object extracted from the captured image;
A display control unit that calculates an existence probability in the non-imaging area of the moving object that has moved out of the imaging area, and displays a current location of the moving object in the non-imaging area based on the existence probability; An image monitoring device,
The display control unit
A lost position in the virtual space is obtained from a position where the moving body has moved outside the imaging area, and an outer periphery of a non-imaging area to which the moving body is moved is configured from the lost position, the area information, and the structure information. An image monitoring apparatus characterized by specifying an opening, and calculating such that the existence probability decreases as the elapsed time from the lapse time of the moving object and the number of openings increase. The display control unit
Prediction for calculating a predicted path in a non-imaging area in which the moving body can move from the lapse position through the opening without passing through the imaging area using the lost position, the structure information, and the area information. Route calculation means;
A probability distribution is obtained based on the moving speed before the relapse of the moving object and the elapsed time, and a path-specific area existence probability is calculated for each predicted path using the probability distribution and the structure information, and the path-specific area The image monitoring apparatus according to claim 1, further comprising presence probability calculating means for calculating the presence probability from the presence probability. Wherein the image processing unit obtains the degree of similarity between the images features of mobile and stored in advance in the storage unit feature quantity information, the moving body is the feature amount information using said existence probability and the degree of similarity The image monitoring apparatus according to claim 1, further comprising a determination unit configured to determine whether or not the moving object corresponds to.   The display control unit has display control means for displaying and outputting an image display indicating the position of the moving object on the map of the monitoring location stored in advance in the storage unit based on the existence probability. The image monitoring apparatus according to claim 3. The structure of the object in the structure information is three-dimensional shape data,
The storage unit further stores imaging condition information in which an installation position, an optical axis, and an angle of view of the imaging unit are associated with the structure information,
5. The apparatus according to claim 1, further comprising an area setting unit that obtains the imaging area and the non-imaging area using the structure information and the imaging condition information and stores the imaging area and the non-imaging area in the storage unit as the area information. The image monitoring apparatus described. The image monitoring apparatus according to claim 2, wherein the probability distribution is represented by a normal distribution using an average value obtained from the moving speed and the elapsed time and a standard deviation obtained from the elapsed time as parameters.

Images