Moving-object counter apparatus, moving-object counting method, and non-transitory computer readable medium

This application is based on and claims priority under 35 USC 119 from Japanese Patent Application No. 2016-061313 filed Mar. 25, 2016.

The present invention relates to a moving-object counter apparatus, a moving-object counting method, and a non-transitory computer readable medium.

According to an aspect of the invention, there is provided a moving-object counter apparatus including a first captured-image acquisition unit, a moving-object moving-path specifying unit, a virtual-line setting unit, and a first moving-object counter unit. The first captured-image acquisition unit acquires multiple images captured at respective times different from each other by a first imaging apparatus capturing images of a predetermined region. The moving-object moving-path specifying unit specifies, on the basis of the multiple images acquired by the first captured-image acquisition unit, one or more moving paths along which one or more respective moving objects have moved in the predetermined region. The virtual-line setting unit sets a virtual line on the basis of the one or more moving paths specified by the moving-object moving-path specifying unit, the virtual line being imagined on the images acquired by the first captured-image acquisition unit, and serving as a reference for counting of one or more moving objects that have passed through a predetermined position in the predetermined region. The first moving-object counter unit counts, by counting one or more moving paths that cross the set virtual line among the one or more moving paths specified by the moving-object moving-path specifying unit, the one or more moving objects that have passed through the predetermined position.

An exemplary embodiment of the present invention will be described in detail based on the following figures, wherein:

FIG. 1 is a diagram illustrating an example of the configuration of a moving-object counter apparatus;

FIG. 2 is a diagram illustrating an example of installation of imaging apparatuses;

FIG. 3 is a flowchart illustrating an example of a moving-object counting operation;

FIG. 4 is a flowchart illustrating an example of a moving-path specification process;

FIG. 5 is a flowchart illustrating an example of a virtual-line angle setting process;

FIG. 6 is a diagram illustrating an example of moving paths;

FIG. 7 is a diagram illustrating an example of moving-direction classification;

FIG. 8 is a diagram illustrating an example of a moving-path count table;

FIG. 9 is a flowchart illustrating an example of a virtual-line coordinate setting process;

FIG. 10 is a diagram illustrating an example of changes in person regions corresponding to sets of coordinates of a moving path;

FIG. 11 is a diagram illustrating an example of the areas of person regions corresponding to sets of coordinates constituting a moving path;

FIG. 12 is a diagram illustrating an example of a virtual line;

FIG. 13 is a diagram illustrating an example of an image captured by an imaging apparatus among the imaging apparatuses;

FIG. 14 is a diagram illustrating an example of an image captured by the other imaging apparatus among the imaging apparatuses;

FIG. 15 is a flowchart illustrating an example of a virtual-line coordinate setting process;

FIG. 16 is a diagram illustrating an example of changes in person regions and head-portion regions corresponding to sets of coordinates of a moving path;

FIG. 17 is a diagram illustrating, for each set of coordinates constituting a moving path, an example of the distance between a person region corresponding to the set of coordinates and a head-portion region of the person region; and

FIG. 18 is a diagram illustrating an example of virtual regions.

In the following, an exemplary embodiment of the present invention will be described with reference to the drawings.

1. Exemplary Embodiment

1-1. Configuration

FIG. 1 is a diagram illustrating an example of the configuration of a moving-object counter apparatus 1 according to an exemplary embodiment of the present invention. The moving-object counter apparatus 1 is an apparatus that counts, on the basis of images captured by an imaging apparatus 2 A, objects that are moving (hereinafter referred to as “moving objects”). Here, specifically, the moving objects are persons. The moving-object counter apparatus 1 includes a controller 10 , a memory 20 , an output unit 30 , and external interfaces 40 A and 40 B.

The controller 10 includes an arithmetic processing unit such as a central processing unit (CPU), and a memory, and executes a program stored in the memory 20 . The memory 20 is a storage device such as a hard disk, and stores various types of data in addition to the program executed by the controller 10 . The output unit 30 outputs the various types of data stored in the memory 20 . The output unit 30 is, for example, a display device such as a liquid crystal display, a communication device such as a data communication card, or a communication module that conforms to a short-range wireless communication standard such as Bluetooth®. The external interface 40 A is an interface for connecting the imaging apparatus 2 A, which is a first imaging apparatus, to the moving-object counter apparatus 1 . The external interface 40 B is an interface for connecting the imaging apparatus 2 B, which is a second imaging apparatus, to the moving-object counter apparatus 1 .

The imaging apparatuses 2 A and 2 B are cameras that capture images of a predetermined region. Here, specifically, the predetermined region is a region in front of an entrance of a facility. Here, specifically, the facility is a store.

FIG. 2 is a diagram illustrating an example of installation of the imaging apparatuses 2 A and 2 B. The imaging apparatus 2 A is installed at a ceiling C inside the store, and captures images of a region in front of an entrance E of the store from above. When being installed at the ceiling C, the imaging apparatus 2 A is installed in an inclined manner such that an obstruction O such as a shutter is not included in images. In other words, the imaging apparatus 2 A is installed such that the optical axis of the imaging apparatus 2 A does not become parallel to the normal to a floor surface F. The imaging apparatus 2 A outputs captured moving-image data to the external interface 40 A. In contrast, the imaging apparatus 2 B is installed inside the store, and captures, in a horizontal direction, images of the region in front of the entrance E of the store. The imaging apparatus 2 B outputs captured moving-image data to the external interface 40 B.

The controller 10 realizes, by executing the program stored in the memory 20 , functions that are a first captured-image acquisition unit 11 , a moving-object moving-path specifying unit 12 , a virtual-line setting unit 13 , a first moving-object counter unit 14 , a second captured-image acquisition unit 15 , a face-portion direction specifying unit 16 , and a second moving-object counter unit 17 .

The first captured-image acquisition unit 11 acquires multiple images captured at respective times different from each other by the imaging apparatus 2 A. Here, specifically, the multiple images are multiple still images that have been captured continuously. Note that the multiple images do not have to be continuous still images, and may also be discontinuous still images.

The moving-object moving-path specifying unit 12 specifies, on the basis of the multiple images acquired by the first captured-image acquisition unit 11 , one or more moving paths along which one or more respective persons have moved in the region in front of the entrance E of the store. Specifically, first, the moving-object moving-path specifying unit 12 detects, in each of the multiple images acquired by the first captured-image acquisition unit 11 , a moving-object region using a background subtraction method, the moving-object region being a region in which a moving object is positioned. Next, the moving-object moving-path specifying unit 12 extracts a Histograms-of-Oriented-Gradients (HOG) feature value from each of the detected moving-object regions, and determines, in the detected moving-object regions, person regions in which persons are positioned using a person recognition device constructed through prior learning to recognize persons. Finally, the moving-object moving-path specifying unit 12 specifies, on the basis of each of the moving-object regions determined to be the person regions, the moving paths of the persons using a particle filter.

The virtual-line setting unit 13 sets a virtual line on the basis of the moving paths specified by the moving-object moving-path specifying unit 12 , the virtual line being imagined on the images acquired by the first captured-image acquisition unit 11 , and serving as a reference for counting of persons who have passed through a predetermined position in the region in front of the entrance of the facility. Here, specifically, the predetermined position is a position directly below the imaging apparatus 2 A.

Specifically, the virtual-line setting unit 13 sets, on the basis of moving directions indicated by the moving paths specified by the moving-object moving-path specifying unit 12 , an angle of the virtual line that is the angle formed with respect to a reference line preset on the images acquired by the first captured-image acquisition unit 11 . Here, specifically, the reference line preset on the images is one side of each image. When the angle of the virtual line is set, the virtual-line setting unit 13 sets the angle of the virtual line such that the virtual line forms a predetermined angle with respect to the moving direction. Here, specifically, the predetermined angle is 90 degrees.

In addition, for each of the multiple images acquired by the first captured-image acquisition unit 11 , the virtual-line setting unit 13 sets, on the basis of attribute values of the multiple person regions detected by the moving-object moving-path specifying unit 12 , coordinates through which the virtual line passes on the image. Here, specifically, the attribute value of each person region is the area of the person region. In this case, the virtual-line setting unit 13 sets, as the above-described coordinates, coordinates of the person region having the smallest area among the multiple detected person regions. Here, specifically, the coordinates of the person region are the coordinates of the barycenter of the person region.

The first moving-object counter unit 14 counts, by counting moving paths that cross the virtual line set by the virtual-line setting unit 13 among the moving paths specified by the moving-object moving-path specifying unit 12 , persons who have passed through the position directly below the imaging apparatus 2 A.

The second captured-image acquisition unit 15 acquires an image captured by the imaging apparatus 2 B. The second captured-image acquisition unit 15 particularly acquires a captured image of persons who have passed through the position directly below the imaging apparatus 2 A.

On the basis of the image acquired by the second captured-image acquisition unit 15 , the face-portion direction specifying unit 16 specifies, for each of the persons who have passed through the position directly below the imaging apparatus 2 A, the direction in which the face portion of the person is facing. Specifically, the face-portion direction specifying unit 16 determines whether a frontal face of the person or the back of the person's head is detected from the acquired image using a frontal face recognition device constructed through prior learning to recognize frontal faces and a back-of-head recognition device constructed through prior learning to recognize the back of persons' heads. In other words, the face-portion direction specifying unit 16 determines whether the person is facing toward the imaging apparatus 2 B or in the direction opposite to the imaging apparatus 2 B.

In the case where the direction specified by the face-portion direction specifying unit 16 is a predetermined direction, the second moving-object counter unit 17 counts the person who has passed through the position directly below the imaging apparatus 2 A as a person who has entered the store. Here, specifically, the predetermined direction is the direction in which the imaging apparatus 2 B is present. In other words, the predetermined direction is a direction toward the inside of the store.

1-2. Operation

The operation of the moving-object counter apparatus 1 will be described. FIG. 3 is a flowchart illustrating an example of a moving-object counting operation executed by the moving-object counter apparatus 1 . In this moving-object counting operation, persons entering the store and persons leaving the store are counted on the basis of images captured by the imaging apparatuses 2 A and 2 B.

The first captured-image acquisition unit 11 of the moving-object counter apparatus 1 acquires multiple images continuously captured by the imaging apparatus 2 A (S 1 ). When the multiple images are acquired, the moving-object moving-path specifying unit 12 executes a moving-path specification process (S 2 ). FIG. 4 is a flowchart illustrating an example of the moving-path specification process.

In the moving-path specification process, the moving-object moving-path specifying unit 12 detects, in each of the multiple images acquired in step S 1 , a moving-object region that is a region in which a moving object is positioned (S 21 ). When the moving-object regions are detected, the moving-object moving-path specifying unit 12 extracts a HOG feature value from each of the detected moving-object regions, and determines, in the detected moving-object regions, person regions in which persons are positioned using the person recognition device constructed through prior learning to recognize persons (S 22 ). When the person regions are determined, the moving-object moving-path specifying unit 12 specifies the moving paths of persons by applying the particle filter to each of the moving-object regions (more specifically, the barycenter of each of the moving-object regions) determined to be the person regions (S 23 ).

After completion of the moving-path specification process, in the case where a virtual line has not been set (NO in S 3 ), the process proceeds to step S 4 . In contrast, in the case where the virtual line has been set (YES in S 3 ), the process proceeds to step S 7 .

In the case where it is determined in step S 3 that the virtual line has not been set, the virtual-line setting unit 13 executes a virtual-line angle setting process (S 4 ). FIG. 5 is a flowchart illustrating an example of the virtual-line angle setting process.

In the virtual-line angle setting process, the virtual-line setting unit 13 classifies the moving paths specified in step S 2 into multiple groups in accordance with the moving directions, and counts the moving paths on a group-by-group basis (S 41 ). FIG. 6 is a diagram illustrating an example of each of the moving paths specified in step S 2 . An arrow Ar illustrated in FIG. 6 indicates the moving path of a person. FIG. 6 illustrates the moving paths of 10 persons. Each moving path is expressed using the xy coordinate system in which the lower-left corner of the image is treated as the origin. FIG. 7 is a diagram illustrating an example of moving-direction classification to which reference is made when moving paths are classified. In the classification illustrated in FIG. 7 , moving directions are classified into eight directions from 0 degrees to 315 degrees in units of 45 degrees. The 0-degree moving direction starts from a barycenter G of the image, and is a direction parallel to the x axis. Note that in the case where the moving direction that matches a certain moving path is not any of the eight directions illustrated in FIG. 7 , the virtual-line setting unit 13 classifies the moving path into the moving direction closest to the moving path. In addition, in the case where a certain moving path is not straight (that is, in the case where the moving path is constituted by multiple vectors), a resultant vector is calculated, and the calculated resultant vector is classified into any one of the moving directions. FIG. 8 is a diagram illustrating an example of a count table generated in the case where the moving paths illustrated in FIG. 6 are classified into the eight directions illustrated in FIG. 7 . According to the count table illustrated in FIG. 8 , it is clear that the most moving paths correspond to the 90-degree moving direction. After classifying and counting the moving paths, the virtual-line setting unit 13 specifies the moving direction into which the most moving paths have been classified (S 42 ). According to the example illustrated in FIG. 8 , the moving direction “90 degrees” is specified. After specifying the moving direction, the virtual-line setting unit 13 calculates, as the angle of the virtual line (0 degrees≤θ<180 degrees), the angle that is a right angle with respect to the specified moving direction (S 43 ). If 90 degrees are determined to be the moving direction, the virtual-line setting unit 13 calculates 0 degrees as the angle of the virtual line.

After setting the angle of the virtual line, the virtual-line setting unit 13 executes a virtual-line coordinate setting process (S 5 ). FIG. 9 is a flowchart illustrating an example of the virtual-line coordinate setting process.

In the virtual-line coordinate setting process, about one moving path classified into the moving direction specified in step S 42 , the virtual-line setting unit 13 calculates the areas of person regions corresponding to respective sets of coordinates constituting the moving path (S 51 ). FIG. 10 is a diagram illustrating an example of changes in the person regions corresponding to the sets of coordinates of the moving path. An image IM 1 illustrated in FIG. 10 is an image captured by the imaging apparatus 2 A at the time of t seconds. In contrast, an image IM 2 is an image captured by the imaging apparatus 2 A at the time of t+1 seconds. Person regions A 1 and A 2 illustrated in the respective images are expressed using circumscribed rectangles with which image portions representing a person P are enclosed. The person regions A 1 and A 2 are expressed using the xy coordinate system in which the lower-left corner of the image is treated as the origin. A barycenter G 1 of the person region A 1 and a barycenter G 2 of the person region A 2 are positioned at respective sets of coordinates constituting the moving path. FIG. 11 is a diagram illustrating an example of the areas of the person regions corresponding to the sets of coordinates constituting the moving path. According to FIG. 11 , until 1.50 seconds have passed after the person P passes through the entrance E, as a y-coordinate value increases, that is, as the person P goes away from the entrance E, the area decreases. After 1.50 seconds have passed after the person P passes through the entrance E, as the y-coordinate value increases, that is, as the person P goes away from the entrance E, the area increases. After calculating the area of each person region, the virtual-line setting unit 13 specifies coordinates corresponding to the smallest area among the calculated areas (S 52 ). According to the example illustrated in FIG. 11 , coordinates “(200, 150)” corresponding to an area “100” are specified. Here, the coordinates corresponding to the smallest area are coordinates corresponding to the position directly below the imaging apparatus 2 A in the region in front of the entrance E of the store.

After setting coordinates of a virtual line, the virtual-line setting unit 13 sets the virtual line on the basis of the angle set in step S 4 and the coordinates set in step S 5 (S 6 ). Specifically, a line segment that forms the angle set in step S 4 with respect to the x axis of the image, and that passes through the coordinates set in step S 5 is set as the virtual line. FIG. 12 is a diagram illustrating an example of the set virtual line. In the example illustrated in FIG. 12 , a virtual line L set in the case where an angle “0 degrees” is set in step S 4 , and coordinates “(200, 150)” are set in step S 5 is illustrated. The virtual line L is expressed using the xy coordinate system in which the lower-left corner of the image is treated as the origin.

After the virtual line is set, the first moving-object counter unit 14 determines whether each of the moving paths specified in step S 2 crosses the virtual line set in step S 6 (S 7 ). As a result of this determination, in the case where the moving path does not cross the virtual line, that is, where no person has passed directly below the imaging apparatus 2 A (NO in S 7 ), the process returns to step S 1 . In contrast, as a result of this determination, in the case where the moving path crosses the virtual line, that is, where a certain person has passed directly below the imaging apparatus 2 A (YES in S 7 ), the first moving-object counter unit 14 counts the person as a person entering or leaving the store (S 8 ). Specifically, the value of the number of persons entering or leaving the store stored in the memory 20 is incremented. FIG. 13 is a diagram illustrating an example of a captured image of a person passing directly below the imaging apparatus 2 A. In the image illustrated in FIG. 13 , a person region A 3 of a person P 1 and a person region A 4 of a person P 2 are detected. A barycenter G 3 of the person region A 3 among these person regions A 3 and A 4 crosses the virtual line L at coordinates (150, 200).

When the person is counted as a person entering or leaving the store, the second captured-image acquisition unit 15 acquires an image of the person captured by the imaging apparatus 2 B (S 9 ). Specifically, an image captured by the imaging apparatus 2 B at the time when the person passes directly below the imaging apparatus 2 A is acquired. Here, the time when the person passes directly below the imaging apparatus 2 A is specified by referring to a time stamp of the image in which the barycenter of the person region of the person is positioned on the virtual line. FIG. 14 is a diagram illustrating an example of an image captured by the imaging apparatus 2 B at the time when the image illustrated in FIG. 13 is captured. In the image illustrated in FIG. 14 , the position of each person is expressed using a rectangular coordinate system composed of the x axis and the z axis. Here the x axis corresponds to the x axis illustrated in FIG. 13 . For example, the position of the person P 1 is recognized using the x coordinate “150”. According to the image illustrated in FIG. 14 , the direction in which the face portion of the person P 1 is facing and the direction in which the face portion of the person P 2 is facing are specified.

When the image of the person entering or leaving the store is acquired, the face-portion direction specifying unit 16 specifies, on the basis of the acquired image, the direction in which the face portion of the person is facing (S 10 ). Specifically, the face-portion direction specifying unit 16 determines whether a frontal face of the person or the back of the person's head is detected from the acquired images using the frontal face recognition device constructed through prior learning to recognize frontal faces and the back-of-head recognition device constructed through prior learning to recognize the back of persons' heads. In other words, the face-portion direction specifying unit 16 determines whether the person is facing toward the imaging apparatus 2 B or in the direction opposite to the imaging apparatus 2 B. As a result of this determination, in the case where the frontal face of the person is detected (YES in S 11 ), the second moving-object counter unit 17 counts the person as a person entering the store (S 12 ). Specifically, the value of the number of persons entering the store stored in the memory 20 is incremented. In contrast, as a result of this determination, in the case where the back of the person's head is detected (NO in S 11 ), the second moving-object counter unit 17 counts the person as a person leaving the store (S 13 ). Specifically, the value of the number of persons leaving the store stored in the memory 20 is incremented.

The description above is about the moving-object counting operation.

2. Modification

The above-described exemplary embodiment may be modified as in the following. In addition, the following modifications may each be combined with one or more other modifications.

2-1. First Modification

The persons according to the above-described exemplary embodiment are an example of moving objects. The moving objects may also be animals other than persons. Alternatively, the moving objects may also be objects that are moved through persons' operation (for example, cars).

2-2. Second Modification

The region in front of the entrance E of the store according to the above-described exemplary embodiment is an example of a predetermined region whose images are captured by the imaging apparatus 2 A. This predetermined region may be another region inside the store (for example, a corridor), or may also be a region outside the store. In addition, the store according to the above-described exemplary embodiment is an example of a facility for which the number of persons entering or leaving the facility is counted. The facility for which the number of persons entering or leaving the facility is to be counted may also be, for example, a house or an office other than the store.

2-3. Third Modification

The moving-object moving-path specifying unit 12 according to the above-described exemplary embodiment may specify, using a known technique different from the above-described technique, a moving path from the multiple images acquired by the first captured-image acquisition unit 11 . For example, the moving-object moving-path specifying unit 12 may extract, from each of the moving-object regions detected from the images, another feature value (for example, an Edge of Orientation Histograms (EOH) feature value or an Edgelet feature value) other than the HOG feature value, and determine a person region using the person recognition device. In addition, the moving-object moving-path specifying unit 12 may also determine the moving path of a person by applying another moving-object tracking technique (for example, mean shift) other than the particle filter to the person region.

2-4. Fourth Modification

In the virtual-line angle setting process according to the above-described exemplary embodiment, the angle of the virtual line is set on the basis of the moving paths of the 10 persons (see FIG. 6 ); however, the number of moving paths to which reference is made to set the angle of the virtual line may be less than 10 or greater than or equal to 11. In addition, in the virtual-line angle setting process, the moving paths of the 10 persons are classified into the eight directions (see FIG. 7 ); however, the number of directions into which the moving paths are classified may be less than eight or greater than or equal to nine. In addition, in the virtual-line angle setting process, the angle that is a right angle with respect to the moving direction into which the most moving paths are classified is calculated as the angle of the virtual line; however, the angle of the virtual line does not have to be strictly a right angle with respect to the moving direction. That is, preferably, the angle of the virtual line is a substantially right angle with respect to the moving direction. In addition, in the virtual-line angle setting process, the angle formed by the virtual line and one side of the image (specifically, the x axis) is set as the angle of the virtual line; however, the one side of this image is an example of the reference line preset in the image. The reference line preset in this image may also be a line segment (for example, a diagonal line of the image).

2-5. Fifth Modification

In the virtual-line coordinate setting process according to the above-described exemplary embodiment, the person region detected in the image is expressed using a circumscribed rectangle; however, the circumscribed rectangle is an example of the shape of the person region. The shape of the person region may also be another shape (for example, a circumcircle) other than the circumscribed rectangle. In addition, in the virtual-line coordinate setting process, the coordinates of the barycenter of the person region having the smallest area are set as coordinates of the virtual line; however, the coordinates of the barycenter of the person region are an example of coordinates indicating the person region. The coordinates indicating the person region may also be coordinates of a point (for example, a vertex of the person region) other than the barycenter. In addition, in the virtual-line coordinate setting process, the coordinates of the person region having the smallest area are set as coordinates of the virtual line; however, the coordinates of the person region having the smallest amount of change in area may also be set as coordinates of the virtual line. Here, the person region having the smallest amount of change in area is a person region having an area obtained before or after the amount of change becomes smallest. In addition, in the virtual-line coordinate setting process, the coordinates of the person region having the smallest area are set as coordinates of the virtual line; however, the area of the person region is an example of the attribute value to which reference is made to set coordinates of the virtual line. The attribute value to which reference is made to set coordinates of the virtual line may also be an attribute value (for example, the width, the height, or the length of a diagonal line of the person region) other than the area of the person region.

2-6. Sixth Modification

In the virtual-line coordinate setting process according to the above-described exemplary embodiment, the areas of the person regions corresponding to the sets of coordinates constituting the moving path are calculated, and the coordinates of the person region having the smallest area are set as coordinates of the virtual line. However, this method is an example of a method for setting coordinates of the virtual line. As another method, the virtual-line setting unit 13 may set coordinates of the virtual line on the basis of a relative positional relationship between a person region detected in each of multiple images by the moving-object moving-path specifying unit 12 and a head-portion region in which a head portion of the person is positioned. More specifically, among pairs of the person regions and the head-portion regions detected in the multiple respective images, the coordinates of the person region or head-portion region of the pair for which the distance between a center portion of the person region and a center portion of the head-portion region is shortest may also be set as coordinates of the virtual line. Here, the case where the distance between the center portion of the person region and the center portion of the head-portion region is shortest is, for example, the case where the center portion of the person region overlaps with the center portion of the head-portion region. Here, specifically, the center portion is the barycenter of each of the regions. Note that in the case where this virtual-line coordinate setting method is employed, the moving-object moving-path specifying unit 12 detects a head-portion region in addition to a person region in each of the multiple images acquired by the first captured-image acquisition unit 11 .

FIG. 15 is a flowchart illustrating an example of a virtual-line coordinate setting process according to the present modification. In the virtual-line coordinate setting process illustrated in FIG. 15 , about one moving path classified into the moving direction specified in step S 42 , the virtual-line setting unit 13 calculates, for each set of coordinates constituting a moving path, the distance between the person region corresponding to the set of coordinates and the head-portion region of the person region (S 53 ). FIG. 16 is a diagram illustrating an example of changes in person regions and head-portion regions corresponding to the sets of coordinates of the moving path. An image IM 3 illustrated in FIG. 16 is an image captured by the imaging apparatus 2 A at the time of t seconds. In contrast, an image IM 4 is an image captured by the imaging apparatus 2 A at the time of t+1 seconds. Person regions A 5 and A 7 illustrated in the respective images are expressed using circumscribed rectangles with which image portions representing the person P are enclosed. Head-portion regions A 6 and A 8 illustrated in the respective images are expressed using circumscribed rectangles with which image portions representing the head portion of the person P are enclosed. The person regions A 5 and A 7 and the head-portion regions A 6 and A 8 are expressed using the xy coordinate system in which the lower-left corner of the image is treated as the origin. A barycenter G 5 of the person region A 5 and a barycenter G 7 of the person region A 7 are positioned at respective sets of coordinates constituting the moving path. FIG. 17 is a diagram illustrating, for each set of coordinates constituting a moving path, an example of the distance between a person region corresponding to the set of coordinates and a head-portion region of the person region. According to FIG. 17 , until 1.50 seconds have passed after the person P passes through the entrance E, as a y-coordinate value of the person region increases, that is, as the person P goes away from the entrance E, the distance decreases. After 1.50 seconds have passed after the person P passes through the entrance E, as the y-coordinate value of the person region increases, that is, as the person P goes away from the entrance E, the distance increases. When, for each set of coordinates, calculating the distance between the person region and the head-portion region, the virtual-line setting unit 13 specifies coordinates corresponding to the shortest distance among the calculated distances (S 54 ). According to the example illustrated in FIG. 17 , coordinates “(200, 150)” corresponding to a distance “0” are specified. Here, the coordinates corresponding to the shortest distance are coordinates corresponding to the position directly below the imaging apparatus 2 A in the region in front of the entrance E of the store.

The description above is about the virtual-line coordinate setting process according to the present modification.

Note that in the virtual-line coordinate setting process described above, the shape of a head-portion region may also be another shape (for example, a circumcircle) other than a circumscribed rectangle. In addition, in the virtual-line coordinate setting process, the coordinates of the person region having the smallest amount of change in distance may also be set as coordinates of the virtual line. Here, the person region having the smallest amount of change in distance is a person region having a distance to a corresponding head-portion region obtained before or after the amount of change becomes smallest. In addition, in the virtual-line coordinate setting process, the coordinates of the head-portion region corresponding to the shortest distance instead of the coordinates of the person region corresponding to the shortest distance may also be set as coordinates of virtual line.

2-7. Seventh Modification

The imaging apparatus 2 A according to the above-described exemplary embodiment is installed at the ceiling C of the store such that the imaging apparatus 2 A is inclined (see FIG. 2 ). Thus, in an image captured by the imaging apparatus 2 A, the area of a person region detected at the center of the image is not smallest. Thus, in the moving-object counting operation according to the above-described exemplary embodiment, the coordinates of the person region having the smallest area are specified by performing the virtual-line coordinate setting process. However, in the case where the imaging apparatus 2 A is installed such that the imaging apparatus 2 A is not inclined (in other words, in the case where the imaging apparatus 2 A is installed such that the optical axis of the imaging apparatus 2 A is parallel to the normal to the floor surface), the area of a person region detected at the center of a captured image is smallest. Thus, in the case where the imaging apparatus 2 A is installed such that the imaging apparatus 2 A is not inclined, the virtual-line coordinate setting process may be omitted. Instead, coordinates (n, 200) (0≤n≤600) are set in a fixed manner as coordinates of the virtual line.

2-8. Eighth Modification

In the moving-object counting operation according to the above-described exemplary embodiment, after the virtual line is set, two virtual regions are set where the virtual line serves as a boundary, and the first moving-object counter unit 14 may count persons entering or leaving the store by counting moving paths passing through these two virtual regions. FIG. 18 is a diagram illustrating an example of the virtual regions. In the example illustrated in FIG. 18 , a virtual region A 9 is set on the side closer to the entrance E, and a virtual region A 10 is set on the side further from the entrance E. Each virtual region is expressed using the xy coordinate system in which the lower-left corner of the image is treated as the origin.

2-9. Ninth Modification

The program executed by the controller 10 of the moving-object counter apparatus 1 according to the above-described exemplary embodiment may be provided in the state of being stored in a storage medium such as a magnetic tape, a magnetic disk, a flexible disk, an optical disc, a magneto-optical disk, or a memory. In addition, the program may also be downloaded through communication lines such as the Internet.

The foregoing description of the exemplary embodiment of the present invention has been provided for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Obviously, many modifications and variations will be apparent to practitioners skilled in the art. The embodiment was chosen and described in order to best explain the principles of the invention and its practical applications, thereby enabling others skilled in the art to understand the invention for various embodiments and with the various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalents.