JP2024102674A - Setting device, counting system, setting method, and program - Google Patents

Abstract

To provide a setting device capable of appropriately setting a path for counting a moving object, a counting system, a setting method, and a program.SOLUTION: A setting device sets a path for counting a moving object and includes: a clustering unit configured to classify a plurality of trajectories along which the object moves into a predetermined number of clusters by clustering based on trajectory information indicating the trajectories; and a path setting unit configured to set the path for each of the clusters.SELECTED DRAWING: Figure 13

Description

The present disclosure relates to a setting device, a counting system, a setting method, and a program for setting a path for counting objects.

Technology has been developed for counting objects (such as people) moving within various facilities such as stores. For example, Patent Document 1 discloses a people counting device that counts the number of people who pass through a measurement area set within an image of a store or facility. Patent Document 2 discloses an image processing system that can calculate a number of people appropriate for the purpose when calculating the number of people who have passed through a specified position.

International Publication No. 2012/132437 JP 2019-061665 A

For example, in a store, it may be possible to obtain information that is useful for store management by counting the number of customers moving along each aisle and analyzing the counting results. For this reason, when a facility has multiple routes of movement, there is a demand to be able to count objects moving along each route.

When multiple routes exist within a facility, it is desirable to set the route along which the greatest number of objects move as the route to be counted in order to obtain more effective information. However, when there are a large number of routes within a facility, it is difficult to determine which route is appropriate for counting.

The present disclosure aims to provide a setting device, counting system, setting method, and program that can appropriately set a path for counting moving objects.

A setting device according to one aspect of the present disclosure is a setting device that sets a route for counting moving objects, and includes a clustering unit that classifies a plurality of trajectories into a predetermined number of clusters by clustering based on trajectory information indicating the trajectory of the object's movement, and a route setting unit that sets the route for each of the clusters.

A counting system according to one aspect of the present disclosure includes an image generating device that captures an image of a moving object and generates an image, a trajectory generating device that generates the trajectory information based on the image, a setting device according to claim 1 that sets the path based on the trajectory information for a predetermined period of time, and a counting device that counts the objects that have moved along the path based on trajectory information newly acquired after the predetermined period of time.

A setting method according to one aspect of the present disclosure is a setting method executed by a computer included in a setting device that sets a predetermined route when an object that has moved along the route is to be counted, in which the computer acquires trajectory information indicating the trajectory of the object's movement, classifies the trajectories into a predetermined number of clusters by clustering based on a plurality of pieces of trajectory information, selects a predetermined number of selected trajectories for each classified cluster, and sets the route based on the selected trajectories.

A program according to one aspect of the present disclosure is a program executed by a computer included in a setting device that sets a predetermined path when counting an object that has moved along the path, and causes the computer to execute the following steps: acquire trajectory information indicating the trajectory along which the object moves, classify the trajectories into a predetermined number of clusters by clustering based on a plurality of pieces of trajectory information, select a predetermined number of selected trajectories for each of the classified clusters, and set the path based on the selected trajectories.

The route for counting moving objects can be set appropriately.

FIG. 1 is a diagram for explaining the overall configuration of a counting system. FIG. 1 is a diagram for explaining a functional configuration of a trajectory generation device; FIG. 1 is a diagram showing how an image acquisition unit sets multiple regions within an image. FIG. 1 is a diagram showing how a trajectory generating unit generates trajectory information; FIG. 1 is a diagram for explaining trajectory information when the same object passes through the same area multiple times; FIG. 2 is a diagram for explaining a functional configuration of a setting device; FIG. 1 is a diagram for explaining a specific example of a trajectory; A conceptual diagram showing that trajectories 1 and 4 are medoids. FIG. 13 is a diagram showing how the distance between a trajectory 1 and a trajectory 2 is calculated. FIG. 13 is a diagram showing how the distance between a trajectory 1 and a trajectory 5 is calculated; FIG. 1 is a diagram for explaining a functional configuration of a counting device. A flowchart for explaining an example of the overall operation of the counting system. A flowchart for explaining an example of the operation of a setting device in a setting phase of a counting system. A flowchart for explaining an example of the operation of a counting device in a counting phase of a counting system.

Embodiments of the present disclosure will be described below with reference to the drawings. However, the scope of the invention is not limited to the examples shown in the drawings. In the following description, parts having the same functions and configurations are given the same reference numerals and their description will be omitted.

Below, a counting system 100 according to an embodiment of the present disclosure will be described. The counting system 100 is a system that detects various moving objects within a predetermined range and counts the objects that have moved along a set route. Examples of predetermined ranges include various facilities such as retail stores, banks, stations, and factories, as well as roads such as public roads, private roads, and crosswalks, entrances and exits of private homes, and corridors of apartment buildings. Examples of objects detected by the counting system 100 include people, animals, mobile robots, automobiles, and bicycles.

<Overall configuration of counting system 100>
1 is a diagram for explaining the overall configuration of a counting system 100. The counting system 100 includes an image generating device 1, a trajectory generating device 2, a setting device 3, and a counting device 4.

The counting system 100 operates differently in a setting phase, in which a path along which moving objects are to be counted is set during a predetermined period before the actual start of counting, and in a counting phase, in which counting is actually performed using the path set in the setting phase. In this disclosure, a path is a path along which the counting device 4 counts an object when the object moves along that path.

The setting phase is a phase in which the setting device 3 determines the route after the system is installed and before the counting device 4 starts counting. On the other hand, the counting phase is a phase in which the counting device 4 actually performs counting using the route after the route is set.

The image generating device 1 is, for example, a camera installed in various locations inside and outside a facility. The counting system 100 may have multiple cameras. One or multiple image generating devices 1 are installed in positions that allow them to capture images of the range within which an object that a user of the counting system 100 wants to count can move. This allows the image generating device 1 to generate images of one or multiple ranges.

In the following, a case where the imaging range of the image generating device 1 is fixed will be described, but in this disclosure, the imaging range of the image generating device 1 may change over time. For example, the image generating device 1 may have a swivel function.

As a specific example, the image generating device 1 is installed in a position where it can capture an image of an area including at least a portion of an aisle in a store where display shelves are lined. As another example, the image generating device 1 is installed in a position where it can capture an image of an area including the entrance and exit of a store. As another example, the image generating device 1 is installed in a position where it can capture an image of an area including a ticket gate of a station. As another example, the image generating device 1 is installed in a position where it can capture an image of at least a portion of a pedestrian crossing.

The image generated by the image generating device 1 is, for example, a moving image. In this disclosure, a moving image means an image that includes multiple images captured at different times of the same area in a chronological order. This disclosure does not limit the number of images (frames) included per unit time in a moving image.

The trajectory generating device 2 detects an object moving within the image generated by the image generating device 1, and generates trajectory information indicating the trajectory of the object. The trajectory generating device 2 generates trajectory information in both the setting phase and the counting phase, and outputs the trajectory information to at least one of the setting device 3 and the counting device 4.

In the setting phase, the setting device 3 sets a route for counting moving objects based on the trajectory information.

In the counting phase, the counting device 4 counts the objects that have moved along the path based on the new trajectory information.

In the counting system 100 having such a configuration, the setting device 3 sets an appropriate path, so that the counting results of the counting device 4 can be made into more useful information. The configurations and operations of the trajectory generating device 2, the setting device 3, and the counting device 4 are described in detail below.

<Trajectory generating device 2>
The trajectory generating device 2 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), and a RAM (Random Access Memory), and realizes various functions by reading out various processing programs such as a system program stored in the ROM, expanding them in the RAM, and executing the expanded programs. The ROM is made up of non-volatile memory such as a semiconductor, and stores the system program and various processing programs that can be executed on the system program. These programs are stored in the form of computer-readable program codes, and the CPU sequentially executes operations in accordance with the program codes. The RAM forms a work area that temporarily stores various programs executed by the CPU and data related to these programs.

FIG. 2 is a diagram for explaining the functional configuration of the trajectory generating device 2. The trajectory generating device 2 includes an image acquisition unit 21, an area setting unit 22, an object detection unit 23, an object tracking unit 24, a trajectory generating unit 25, an operation unit 26, a display unit 27, and a memory unit 28. These functional configurations of the trajectory generating device 2 may be realized in software by a processor executing a program, or may be realized as a hardware circuit such as an integrated circuit.

The image acquisition unit 21 acquires images from the image generation device 1. The image acquisition unit 21 may acquire images from the image generation device 1, for example, by wired or wireless communication. Furthermore, when the image acquisition unit 21 acquires images of multiple ranges from multiple image generation devices 1, it may integrate the images of the multiple ranges to generate an integrated image.

The region setting unit 22 sets multiple regions in the image acquired by the image acquisition unit 21. FIG. 3 is a diagram showing how the image acquisition unit 21 sets multiple regions in an image. FIG. 3A shows an example in which a rectangular image I is divided by straight grid lines GL, thereby setting multiple regions A in the image I. FIG. 3B shows an example in which a circular image I is divided by straight or curved grid lines GL, thereby setting multiple regions A. The image I shown in FIG. 3 includes three display shelves and a cash register.

The region setting unit 22 assigns identification information to each region in order to distinguish between multiple regions set in the image. The example shown in FIG. 3A shows how the region setting unit 22 assigns identification information [x, y] to each region. x is set to 0 at the top left of image I, and the value increases toward the right for each region A. y is set to 0 at the top left of image I, and the value increases toward the bottom for each region A. Note that for the sake of explanation, FIG. 3A shows identification information [0,0], [0,1], etc. in each region, but in reality, identification information does not need to be displayed in each region.

When the imaging range of the image generating device 1 is fixed, the region setting unit 22 may set a region using one frame of the moving image. When the imaging range of the image generating device 1 varies over time, the region setting unit 22 may set a region for each of all frame images included in the moving image, or may set a region for images at intervals of a predetermined number of frames.

When setting an area in an integrated image obtained by integrating images of multiple ranges acquired from multiple image generating devices 1, the area setting unit 22 sets a global coordinate system in the integrated image and assigns identification information for each area based on the global coordinate system. The area setting unit 22 may set the identification information for each area in the integrated image by, for example, converting the identification information for each area in a local coordinate system set for each image generating device 1 into the global coordinate system.

The area setting unit 22 may determine the size of an area to be set within an image based on, for example, the operation of the user of the counting system 100.

The object detection unit 23 detects objects moving within the image generated by the image generation device 1. The object tracking unit 24 tracks the objects detected by the object detection unit 23 and generates time-based position information of the objects. The object detection unit 23 and the object tracking unit 24 may detect and track objects moving within the image using a publicly known object detection method and object tracking method (e.g., SORT (Simple Online and Realtime Tracking)).

When the object detection unit 23 and the object tracking unit 24 are no longer able to track the same object over time, they terminate detection and tracking of the object.

The trajectory generating unit 25 identifies the trajectory of an object based on the position information of the same object over time, and generates trajectory information indicating the trajectory. FIG. 4 is a diagram showing how the trajectory generating unit 25 generates trajectory information for the image I shown in FIG. 3A. FIG. 4A shows the temporal transition of the position information generated by the object tracking unit 24. FIG. 4B shows the trajectory of an object identified by the trajectory generating unit 25 based on the position information. As shown in FIG. 4B, the trajectory generating unit 25 connects areas including the object's position to generate the trajectory of the object. In FIG. 4B, the area corresponding to the trajectory is shown with diagonal lines. Note that, for the purpose of explanation, in FIG. 4A and FIG. 4B, numbers corresponding to the x value and y value of each area are shown on the outer edge of the image I.

4A and 4B, the trajectory generating unit 25 determines that the multiple regions through which the object has passed are the trajectory of the object's movement. That is, the trajectory generating unit 25 identifies the trajectory of the object by connecting the multiple regions through which the object has passed. In the example shown in FIG. 4B, the trajectory of the object is composed of region [0,5], region [1,5], region [2,5], region [3,5], region [4,5], region [5,5], region [5,4], region [6,4], region [6,3], region [6,2], region [6,1], region [7,1], region [8,1], region [9,1], region [10,1], and region [11,1]. The trajectory generating unit 25 generates trajectory information indicating the identified trajectory based on the identification information of these regions. The trajectory generation unit 25 may include in the trajectory information the order of the areas through which the object has moved.

Figure 5 is a diagram for explaining trajectory information when the same object passes through the same area multiple times. Figure 5A shows an example of the temporal transition of position information generated by the object tracking unit 24 when an object moves through the same area multiple times. Such object movement can occur, for example, when a customer, as an object, selects a product while walking back and forth in front of a display shelf in a store.

In such a case, as shown in FIG. 5B, even if the same object passes through the same area multiple times, the trajectory generation unit 25 identifies the trajectory of the object by treating each area as having passed through it once. In FIG. 5B, an example of an identified trajectory is shown as a series of shaded areas. This makes it possible to avoid a situation in which the trajectory information becomes unnecessarily complicated, and also reduces the amount of data in the trajectory information.

The trajectory generating unit 25 may not generate trajectory information for a trajectory whose length from the start point to the end point is less than a predetermined value, i.e., whose number of regions is less than a predetermined number. This is because an extremely short trajectory is highly likely to have been generated by, for example, erroneous detection by the object detecting unit 23.

The trajectory generating unit 25 may also be configured not to generate trajectory information for a trajectory whose start point or end point is not in a specific area. A specific area is an area where an object enters the imaging range of the image generating device 1 from outside the imaging range into the imaging range, or an object exits from the imaging range to the outside of the imaging range. More specifically, a specific area is an area where, for example, a passageway captured in the imaging range reaches the edge of the image. Or, if there is an entrance to a facility in the image, a specific area is an area where the entrance is located.

In this way, the trajectory generation unit 25 does not generate trajectory information for a trajectory that starts or ends in an area other than the specific area. The reason for this is that it is usually impossible for an object to appear and start moving outside the specific area of the imaging range, or to suddenly disappear from outside the specific area, and such a trajectory is highly likely to have been generated by, for example, erroneous detection by the object detection unit 23.

The trajectory generating unit 25 may also be configured not to generate trajectory information for objects moving within a preset exclusion area within the imaging range. The exclusion area is a number of areas included in the imaging range in which it is not necessary to detect objects. As a specific example, when the imaging range is the interior of a small glass-walled store, the road outside the store that is visible through the glass may be set as the exclusion area. This is because people and bicycles moving on the road outside the store do not correspond to objects that should be detected inside the store.

The specific area and the excluded area may be set appropriately based on the operation of the user of the counting system 100.

The operation unit 26 is, for example, any of various operation devices such as a mouse, a trackball, a touchpad, a button, or a keyboard. The display unit 27 is, for example, any of various display devices such as a liquid crystal display, an organic EL (Electro-Luminescence) display, or a CRT (Cathode Ray Tube) display. The trajectory generating device 2 may have a touch panel in which a touchpad as the operation unit 26 and a liquid crystal display or an organic EL display as the display unit 27 are superimposed.

The memory unit 28 stores various information necessary for the operation of the trajectory generating device 2, specifically trajectory information, and programs for implementing each functional configuration.

The trajectory generating device 2 outputs trajectory information indicating the generated trajectory to the setting device 3 and the counting device 4.

<Setting device 3>
In the setting phase, the setting device 3 sets a route along which the counting device 4 will count moving objects.

FIG. 6 is a diagram for explaining the functional configuration of the setting device 3. The setting device 3 includes a clustering unit 31, a route setting unit 32, an operation unit 33, a display unit 34, and a storage unit 35. The setting device 3 may realize various functions in the form of hardware or software, similar to the trajectory generating device 2.

The clustering unit 31 performs clustering of multiple trajectories based on the trajectory information generated by the trajectory generating device 2. As a clustering method used by the clustering unit 31, a known method such as k-medoids or k-means can be adopted.

Figure 7 is a diagram for explaining specific examples of trajectories. In Figure 7, examples of trajectories are shown by lines connecting the center points of the areas that make up each trajectory. Six trajectories, trajectories 1 to 6, are shown as examples in Figure 7. Note that while Figure 7 shows an example in which different trajectories do not overlap with each other, different trajectories may overlap at least in part. In other words, different trajectories may be at least in part made up of the same area.

As an example, we will explain the operation of clustering trajectories 1 to 6 in Figure 7 when the clustering unit 31 adopts k-medoids.

The k-medoids method is a method used for clustering a plurality of clustering target points into k clusters. The k-medoids method includes, for example, the following process.
(1) Among a plurality of points, k points are randomly selected and are assumed to be medoids. A medoid is a point in a cluster that has a minimum sum of distances to other points in the cluster.
(2) Each point other than the k medoids is tentatively assigned to the cluster of the nearest medoid.
(3) In each cluster, the point with the smallest total distance to all other points in the cluster is determined as a new medoid.
(4) If there is a new medoid in (3), return to (2); if not, end the process.

Below, a specific example of the process when the above steps (1) to (4) of k-medoids are applied to the clustering of the trajectories shown in FIG. 7 will be described. The number of clusters, k, may be set to an appropriate number based on, for example, the operation of the user of the counting system 100. In the following example, k is set to 2.

In step (1), in the initial setting of medoids, it is assumed that trajectories 1 and 4 are selected from trajectories 1 to 6 shown in FIG. 7. FIG. 8 is a conceptual diagram showing that trajectories 1 and trajectory 4 are set as medoids. In FIG. 8, trajectories 1 and 4 as medoids are shown with thicker lines than the other trajectories. In the following explanation, the cluster to which trajectory 1 provisionally belongs is referred to as cluster 1, and the cluster to which trajectory 6 provisionally belongs is referred to as cluster 2.

In step (2), the distance between trajectory 1 and trajectory 4, which are set as medoids, and the other trajectories 2, 3, 5, and 6 is calculated, and trajectories 2, 3, 5, and 6 are provisionally assigned to either cluster 1 or 2. In the example shown in FIG. 8, trajectories 2 and 3 are provisionally assigned to cluster 1, and trajectories 5 and 6 are provisionally assigned to cluster 2.

Figures 9 and 10 are conceptual diagrams for explaining a method for calculating the distance between trajectories. Figure 9 is a diagram showing how the distance between trajectories 1 and 2 shown in Figure 7 is calculated. Figure 10 is a diagram showing how the distance between trajectories 1 and 5 shown in Figure 7 is calculated. As shown in Figures 9 and 10, the distance between trajectories is calculated by matching the areas included in each trajectory with each other, calculating the two-dimensional distance on the image between the center points of the matched areas, and calculating the sum of the distances in all areas. In Figures 9 and 10, arrows connect the center points of the matched areas.

9, when the number of regions included in the two trajectories is the same, all the regions constituting the two trajectories are associated with each other, and the distance between the center points is calculated. When the number of regions included in the two trajectories is different, as in the example shown in FIG. 10, one region of one trajectory may be associated with multiple regions of the other trajectory as in FIG. 10A, or some regions may not be associated as shown in FIG. 10B. The distance between the trajectories may also be calculated using a known technique, two-dimensional DTW (Dynamic Time Warping). FIG. 10A shows an example in which the regions of the start points and the regions of the end points of the two trajectories are associated with each other, and the regions between the start points and the end points are evenly associated with each other. FIG. 10B shows an example in which each region constituting trajectory 1 is associated with the region of trajectory 2 closest to each region.

In step (3), trajectory 2 is assumed to be a new medoid in provisional cluster 1 consisting of trajectories 1, 2, and 3. Also, in provisional cluster 2 consisting of trajectories 4, 5, and 6, trajectory 5 is assumed to be a new medoid.

In step (4), since there are trajectories that were newly designated as medoids in step (3), the process returns to step (2), and from now on, the distances between trajectories 2 and 5 and the other trajectories are calculated, and new provisional cluster classification is performed.

By repeating this process, the trajectories are clustered. In the example shown in Figure 7, trajectories 2 and 5 are finally classified as medoids, and trajectories 1 to 6 are classified into cluster 1 consisting of trajectories 1, 2, and 3, and cluster 2 consisting of trajectories 4, 5, and 6.

In the above example, the clustering unit 31 employs k-medoids as the clustering method. In the present disclosure, the clustering unit may employ other known clustering methods, such as k-means, a method in which the centroids of k temporary clusters are continually updated so that the distance between the centroids and other points is minimized.

The route setting unit 32 selects a representative trajectory for each cluster classified by the clustering unit 31, and sets the selected trajectory as a route for counting moving objects.

The route setting unit 32 may use the following method to select a representative trajectory for each cluster:

As a first method, when the clustering unit 31 uses k-medoids or k-means as a clustering method, there is a method of selecting trajectories that ultimately become medoids or centroids. Medoids are trajectories that are located closest to the center of all trajectories belonging to a cluster, and centroids are trajectories that are located closest to the average of all trajectories in the cluster. Therefore, by using the first method, a trajectory that is close to the median or average value within each cluster can be selected as a representative trajectory.

A second method is to display at least one of the trajectories classified into each cluster on the display unit 34, and select a trajectory based on the user's operation via the operation unit 33. In the second method, when the number of trajectories included in each cluster is large, it is more preferable from the viewpoint of reducing the load on the user to display only a predetermined number (2 to 3) of trajectories that are close to the final medoids or centroids in the clustering, rather than displaying all of the trajectories on the display unit 34.

The route setting unit 32 sets the trajectory selected using the first method or the second method as a route for counting moving objects, and displays it on the display unit 34. When a user who has viewed the route displayed on the display unit 34 performs an operation to modify the route via the operation unit 33, the route setting unit 32 may modify the route based on the operation.

The operation unit 33 is, for example, various operation devices such as a mouse, a trackball, a touch pad, a button, a keyboard, etc. The display unit 34 is, for example, various display devices such as a liquid crystal display, an organic EL display, a CRT display, etc. The setting device 3 may have a touch panel in which a touch pad as the operation unit 33 and a liquid crystal display or an organic EL display as the display unit 34 are superimposed.

The memory unit 35 stores various information necessary for the operation of the setting device 3, specifically, trajectory information, route information, and programs for implementing each functional configuration.

When there are multiple types of objects moving within the imaging area of the image (e.g., people and robots, or bicycles and automobiles, etc.), the setting device 3 may set a route for each type of object. As a specific example, in an image that includes a road as an imaging range, the setting device 3 may set separate routes for people, bicycles, and automobiles.

The setting device 3 outputs route information indicating the set route to the counting device 4.

<Counting device 4>
After the setting device 3 sets the route, the counting device 4 counts the number of objects moving along the route based on the route information.

FIG. 11 is a diagram for explaining the functional configuration of the counting device 4. The counting device 4 includes a determination unit 41, a counting unit 42, an operation unit 43, a display unit 44, and a storage unit 45. The counting device 4 may realize various functions in a hardware or software manner, similar to the trajectory generating device 2 and the setting device 3.

The determination unit 41 determines whether the object's trajectory matches the set path based on multiple pieces of trajectory information acquired from the trajectory generating device 2 and the path information acquired from the setting device 3.

As a method for the determination unit 41 to determine whether or not a trajectory and a route match, for example, the following method can be adopted. The determination unit 41 calculates the similarity between each route and all trajectories. Then, the determination unit 41 determines whether or not the similarity is higher than a predetermined threshold for each trajectory. The determination unit 41 may determine that all trajectories whose similarity is higher than the predetermined threshold match the route.

The determination unit 41 may determine the distance between the route indicated by the route information and the trajectory indicated by the trajectory information as the similarity. The distance between the route and the trajectory may be calculated in the same manner as in the case of calculating the distance between the trajectories described in FIG. 9 and FIG. 10.

The counting unit 42 counts the objects that have moved along the path set by the setting device 3. More specifically, the counting unit 42 counts the objects that have moved along a trajectory that has been determined by the determination unit 41 to have a similarity to the path higher than a predetermined threshold (i.e., objects that have moved along a path). This makes it possible to accurately count which path and how much of an object moving within the range captured by the image generating device 1 has moved. The counting unit 42 stores counting result information indicating the counting result in the memory unit 45 and displays it on the display unit 44.

The operation unit 43 is, for example, various operation devices such as a mouse, a trackball, a touch pad, a button, a keyboard, etc. The display unit 44 is, for example, various display devices such as a liquid crystal display, an organic EL display, a CRT display, etc. The counting device 4 may have a touch panel in which a touch pad as the operation unit 43 and a liquid crystal display or an organic EL display as the display unit 44 are superimposed.

The memory unit 45 stores various information necessary for the operation of the counting device 4, specifically, trajectory information, route information, counting result information, and programs for implementing each functional configuration.

[Operation example of counting system 100]
<Overall operation>
Fig. 12 is a flowchart for explaining an example of the overall operation of the counting system 100. As shown in Fig. 12, step S1 is a setting phase, in which the setting device 3 sets a route. Then, step S2 is a counting phase, in which the counting device 4 counts objects moving along the route. The counting phase of step S2 is executed repeatedly, for example, at predetermined time intervals.

For example, if the installation location of the counting system 100 is changed, the setting phase of step S1 will be performed again.

Note that the counting device 4 does not have to operate in the setting phase of step S1, and the setting device 3 does not have to operate in the counting phase of step S2. However, the present disclosure is not limited to this, and for example, after a route is set once in the setting phase and the counting phase is started, the setting phase may be executed again, for example, after a predetermined time has elapsed, to reset the route. Also, the setting phase and the counting phase may be executed in parallel at all times, and the route may be continuously updated in real time.

<Operation during the configuration phase>
FIG. 13 is a flowchart for explaining an example of the operation of the setting device 3 in the setting phase of the counting system 100.

In step S11, the setting device 3 acquires multiple pieces of trajectory information for a predetermined period from the trajectory generating device 2.

In step S12, the setting device 3 clusters the multiple trajectories over a specified period into a specified number of clusters.

In step S13, the setting device 3 selects a representative trajectory for each cluster.

In step S14, the setting device 3 sets the selected trajectory as a route and generates route information. The generated route information is output to the counting device 4.

<Counting Phase Operation>
FIG. 14 is a flowchart for explaining an example of the operation of the counting device 4 in the counting phase of the counting system 100.

In step S21, the counting device 4 acquires the route information from the setting device 3. After acquiring the route information once, the counting device 4 does not need to execute this step again until new route information is generated by the setting device 3.

In step S22, the counting device 4 acquires new trajectory information from the trajectory generating device 2. Note that in this specification, new trajectory information refers to trajectory information in the counting phase, and is distinguished from trajectory information in the setting phase.

In step S23, the counting device 4 calculates the similarity between the route indicated by the route information and the trajectory indicated by the trajectory information for each trajectory.

In step S24, the counting device 4 counts the trajectories whose similarity is higher than a threshold value. The counting result indicates the number of objects that have moved along the path. This allows the counting device 4 to accurately count the objects that move along the path set in the setting phase.

As described above, according to the counting system of the present disclosure, in the setting phase, the setting device classifies multiple trajectories into at least one cluster by clustering, and then sets a representative trajectory for each cluster as a path for counting moving objects. In the counting phase, the counting device counts the objects moving along the path.

This allows the setting device to automatically set a route even when multiple trajectories are generated from multiple images, significantly reducing the burden on the user compared to when the user selects a route from multiple trajectories.

The setting device also clusters multiple trajectories and sets a representative trajectory of each cluster as a route, making it possible to appropriately set a route that is more effective in counting objects. In particular, by adopting known methods such as k-medoids and k-means as the clustering method, it is possible to appropriately select a trajectory that is close to the median or average value for each cluster. A trajectory that is close to the median or average value for each cluster is a trajectory that is likely to have been traveled by many objects among all the trajectories, so by setting such a trajectory as a route, it is possible to increase the effect of counting objects moving along the route compared to when a trajectory located in the periphery of each cluster is set as a route.

[Modification]
Although the embodiments of the present disclosure have been described above, the present disclosure is not limited to the above-described embodiments and may be modified in various ways.

In the above-described embodiment, when the setting device 3 performs clustering, the number of clusters used is a value set by the user of the counting system 100. The counting system according to the present disclosure may have a function of automatically determining an appropriate number of clusters based on the results of counting using the route, for example by setting the number of clusters to various numbers and actually setting a route. Specifically, the counting system according to the present disclosure may set various numbers of clusters and have the setting device set a route, evaluate the results of counting by the counting device using the route, and use the number of clusters with the highest evaluation as the number of clusters in the transition setting phase.

The present disclosure is useful for setting devices that set paths for counting moving objects.

REFERENCE SIGNS LIST 100 Counting system 1 Image generating device 2 Trajectory generating device 21 Image acquiring section 22 Area setting section 23 Object detecting section 24 Object tracking section 25 Trajectory generating section 26 Operation section 27 Display section 28 Memory section 3 Setting device 31 Clustering section 32 Path setting section 33 Operation section 34 Display section 35 Memory section 4 Counting device 41 Determination section 42 Counting section 43 Operation section 44 Display section 45 Memory section

Claims (12)

A setting device for setting a path for counting a moving object, comprising:
a clustering unit that classifies a plurality of trajectories into a predetermined number of clusters by clustering based on trajectory information indicating trajectories along which an object moves;
a route setting unit that sets the route for each of the clusters;
A setting device comprising: the route setting unit sets, as the route, a trajectory selected from the plurality of trajectories based on a distance between the trajectories included in the cluster;
The setting device according to claim 1 . the route setting unit sets, as the route, a trajectory among the trajectories included in the cluster, the trajectory having the smallest sum of the distances from all other trajectories;
The setting device according to claim 2 . An operation unit that accepts an operation to correct the route is further provided,
The route setting unit modifies the set route based on the operation.
The setting device according to claim 1 . An image generating device that captures an image of a moving object and generates an image of the object;
a trajectory generating device that generates the trajectory information based on the image;
The setting device according to claim 1 , wherein the route is set based on the trajectory information for a predetermined period of time;
A counting device that counts the number of objects that have moved along the path based on trajectory information newly acquired after the predetermined period of time;
A counting system comprising: The counting device is
a determination unit that determines whether a similarity between the trajectory indicated by the newly acquired trajectory information and the route is higher than a threshold;
a counting unit that counts the number of the objects that have moved along a trajectory for which the similarity is determined to be higher than the threshold;
6. The counting system of claim 5, further comprising: The trajectory generating device is
A region setting unit that sets a plurality of regions within an image of a moving object;
a trajectory generating unit that generates, as the trajectory information, information indicating an area including the trajectory through which the object has passed, among the plurality of areas;
The counting system of claim 6 further comprising: The plurality of regions are regions obtained by dividing the image into a grid pattern.
8. The counting system according to claim 7. the trajectory generation unit does not generate the trajectory information when the number of the regions included between the start point and the end point of the trajectory is less than a threshold value;
8. The counting system according to claim 7. the trajectory generation unit does not generate the trajectory information when a start point or an end point of the trajectory does not correspond to a specific area among the plurality of areas;
8. The counting system according to claim 7. A setting method executed by a computer included in a setting device for setting a predetermined path when counting an object that has moved along the path, comprising:
The computer,
Acquire trajectory information indicating a trajectory along which an object moves;
classifying the trajectories into a predetermined number of clusters by clustering based on a plurality of pieces of trajectory information;
selecting a predetermined number of selection trajectories for each of the classified clusters, and setting the route based on the selection trajectories;
Setting method. A program executed by a computer included in a setting device for setting a predetermined path when counting an object that has moved along the predetermined path,
Acquire trajectory information indicating a trajectory along which an object moves;
classifying the trajectories into a predetermined number of clusters by clustering based on a plurality of pieces of trajectory information;
selecting a predetermined number of selection trajectories for each of the classified clusters, and setting the route based on the selection trajectories;
A program that causes the computer to execute the procedure.

Images