Door passenger counting device, counting system, vehicle and counting method therefor

Abstract

The invention relates to a device (10) for counting passengers at the door, configured so as to: for each image of a video stream representative of a monitored door, implement a person detection algorithm; determine the detection results which correspond to the same identified passenger; for each identified passenger, calculate a respective trajectory from each image on which said identified passenger appears; implement a partitioning algorithm to assign each trajectory to a respective class among at least two classes; depending on dimensions of the identified passengers, define one class as an incoming class, and another class as an outgoing class; and estimate the number of incoming passengers according to a size of the incoming class and / or the number of outgoing passengers according to a size of the outgoing class. Figure 1

Description

Title of the invention: Door passenger counting device, counting system, vehicle and counting method associated therewith Technical field

[0001] The present invention relates to a device for counting passengers at the door.

[0002] The invention also relates to a counting system, a vehicle embarking such a counting system, and an associated counting method.

[0003] The invention applies to the field of transport, in particular rail transport. State of the art

[0004] It is known to equip a vehicle, in particular a rail transport vehicle, with a passenger counting system at the door.

[0005] By "door counting" (also called "cumulative counting") for a given passenger access, such as a vehicle door, is meant, in the sense, the count of the number of passengers entering and exiting through said passenger access.

[0006] Such a counting system generally comprises at least one camera, positioned so as to acquire a video stream representative of a given passenger access, and a computer connected together.

[0007] Conventionally, the calculator is configured to receive the video stream and to: • for each image of the video stream, detect passengers and calculate, for each passenger detected, coordinates of a respective bounding box; • identify and track, in the video stream, each passenger detected; • for each detected passenger, compare a position, over time, of the respective bounding box to a predetermined crossing line associated with passenger access; • for each passenger detected, determine whether the passenger exited or re-entered the vehicle via the passenger access, based on a comparison result.

[0008] However, such a counting system is not satisfactory.

[0009] Indeed, to achieve efficient counting, such a counting system requires the use of omnidirectional cameras, arranged above the passenger access so that their optical axis is perpendicular to the direction of movement of the passengers.

[0010] This arrangement of the cameras allows the crossing line to extend in the horizontal plane representing the floor of the railway vehicle. In the same way, the bounding boxes around the detected passengers are also defined in the horizontal plane representing the floor of the railway vehicle. Thus, the comparison of the position of passengers in relation to the crossing line is carried out in the same horizontal plane and takes into account the physical realities of the railway vehicle.

[0011] Otherwise, i.e. when using directional cameras, a person inside the vehicle may be counted as exiting (or vice versa), the relative position of the respective bounding box being underestimated due to perspective effects. This therefore distorts the counting.

[0012] Therefore, the need for the use of omnidirectional cameras, for the known counting system, excludes the implementation of such a system in railway vehicles pre-equipped with directional cameras, which represents a large part of the cameras present in railway video surveillance systems.

[0013] An aim of the present invention is to remedy at least one of the drawbacks of the state of the art.

[0014] Another object of the invention is to propose a counting system which is more versatile, and which exhibits satisfactory operation even in the case of the use of directional cameras. Statement of the invention

[0015] To this end, the invention relates to a counting device of the aforementioned type, comprising a processing unit configured to receive at least one video stream representative of a monitored door of a vehicle between an instant of opening of the monitored door and an instant of closing of the monitored door, each video stream having been, preferably, acquired by means of a directional camera, the processing unit being, in addition, configured so as to, for each video stream received: • implement a person detection algorithm for each frame of the video stream; • determine the detection results which correspond to the same passenger, forming an identified passenger; • for each identified passenger, calculate a respective trajectory from a position of at least one pixel of interest associated with said identified passenger, on each image of the video stream on which said identified passenger appears; • implement a partitioning algorithm to assign each calculated trajectory to a respective class among at least two distinct classes; • based on dimensions associated with passengers identified in at least one image of the video stream, defining one of the at least two classes as an inbound class corresponding to inbound passengers, and another of the at least two classes as an outbound class corresponding to passengers outgoing; and • Estimate the number of incoming passengers based on the size of the incoming class and / or the number of outgoing passengers based on the size of the outgoing class.

[0016] Indeed, in such a device, for each video stream, the implementation of the partitioning algorithm based on the calculated trajectories allows for a grouping of said trajectories into trajectory classes having similar characteristics. In particular, among the classes resulting from such a partitioning, one class is representative of the passengers boarding the vehicle during the time interval during which the door shown on the video stream is open, while the other is representative of the outgoing passengers.

[0017] Knowledge of the variation in the apparent dimensions of the passengers, depending on whether they are boarding the vehicle or leaving it, then allows the definition, with great reliability, of each class as being one or the other of the entering class and the exiting class.

[0018] Advantageously, the counting device according to the invention has one or more of the following characteristics, taken in isolation or in any technically possible combination:

[0019] to define the incoming class and the outgoing class, the processing unit is configured to: • determine, for each class, and for each identified passenger corresponding to said class, an area of ​​said passenger on the first image of the video stream on which said identified passenger appears; • calculate, for each class, an average of the determined surfaces; and • define : • the class for which the calculated average is the smallest as being the entering class; and • the class for which the calculated average is the highest as being the outgoing class;

[0020] to define the incoming class and the outgoing class, the processing unit is configured so that, for at least one class: • determine a temporal evolution of the size of the identified passengers of said class; and • define : • said class as the incoming class if the size of the corresponding identified passengers increases over time; and • define said class as the outgoing class if the size of the corresponding identified passengers decreases over time;

[0021] each detection result is associated with a respective bounding box, the processing unit being configured to calculate the trajectory associated with each identified passenger as being a trajectory of at least one pixel depending on a position of the bounding box;

[0022] the processing unit is configured to, prior to the implementation of the partitioning algorithm, carry out a normalization of the lengths of the calculated trajectories, the normalized trajectories having the same length;

[0023] standardization includes: • a truncation of each trajectory to the length of the shortest trajectory, by deleting the positions corresponding to the most recent images of the video stream; or • an addition, to each trajectory, of a number of positions equal to a difference between a length of said trajectory and a length of the longest trajectory, each added position having coordinates: • null; or • equal to the coordinates of the last position of the calculated trajectory, before its normalization;

[0024] the processing unit is configured to implement the partitioning algorithm to assign each trajectory to a respective class among three distinct classes, the processing unit being, in addition, configured to define the class which is neither the incoming class nor the outgoing class as being the representative class of the passengers remaining on the platform and / or on board the vehicle.

[0025] The device according to the invention can be any type of device such as a server, a computer, a tablet, a calculator, a processor, a computer chip.

[0026] According to another aspect of the invention, there is provided a counting system comprising a counting device as defined above and at least one camera, each camera being connected to the counting device to transmit a video stream to said counting device, each camera preferably being a directional camera.

[0027] Advantageously, the counting system according to the invention is configured to, in use, trigger and interrupt an acquisition of the video stream by at least one corresponding camera as a function of a control signal from at least one associated monitored door.

[0028] According to another aspect of the invention, there is provided a vehicle, in particular a rail transport vehicle, comprising a counting system as defined above, each camera of the counting system being on board the vehicle and positioned so that a corresponding door of the vehicle, forming a monitored door, is located in a field of view of said camera.

[0029] According to another aspect of the invention, there is provided a method of counting passengers at the gate, the method being computer implemented and comprising the steps: • implementation of a person detection algorithm for each image of at least one received video stream, each received video stream being representative of a monitored door of a vehicle between an instant of opening of the monitored door and an instant of closing of the monitored door, each video stream having been, preferably, acquired by means of a directional camera; • determination of detection results which correspond to the same passenger, forming an identified passenger; • for each identified passenger, calculation of a respective trajectory from a position of at least one pixel of interest associated with said identified passenger, on each image of the video stream on which said identified passenger appears; • implementation of a partitioning algorithm to assign each trajectory to a respective class among at least two distinct classes; • based on dimensions associated with the passengers identified in at least one image of the video stream, characterizing one of the at least two classes as an incoming class corresponding to the incoming passengers, forming an incoming class, and another of the at least two classes as a class outgoing corresponding to outgoing passengers; and • estimation of the number of incoming passengers based on an incoming class size and / or the number of outgoing passengers based on an outgoing class size.

[0030] According to another aspect of the invention, there is provided a computer program comprising executable instructions which, when executed by computer, implement the steps of the method as defined above.

[0031] The computer program can be in any computer language, such as for example machine language, C, C++, JAVA, Python, etc. Brief description of the figures

[0032] The invention will be better understood on reading the description which follows, given solely by way of non-limiting example and made with reference to the appended drawings in which:

[0033] [Fig.l] is a schematic representation of a railway vehicle according to the invention; and

[0034] [Fig.2] is a flowchart of a counting method implemented by a counting system of the railway vehicle of [Fig.l].

[0035] It is understood that the embodiments which will be described below do not are in no way limiting. In particular, it is possible to imagine variants of the invention comprising only a selection of characteristics described below isolated from the other characteristics described, if this selection of characteristics is sufficient to confer a technical advantage or to differentiate the invention compared to the state of the prior art. This selection includes at least one preferably functional characteristic without structural details, or with only part of the structural details if it is this part which is only sufficient to confer a technical advantage or to differentiate the invention compared to the state of the prior art.

[0036] In particular, all the variants and all the embodiments described can be combined with each other if nothing prevents this combination from a technical point of view.

[0037] In the figures and in the remainder of the description, the elements common to several figures retain the same reference. Detailed description

[0038] A vehicle 2 according to the invention is illustrated by [Fig.l].

[0039] The vehicle 2 is, in particular, a passenger transport vehicle, preferably a rail transport vehicle such as a train.

[0040] The vehicle 2 comprises at least one door 4, intended to allow passengers to board the vehicle 2, and to disembark from the vehicle 2.

[0041] Furthermore, the vehicle 2 comprises a counting system 6 intended to carry out a passenger count at the door.

[0042] The counting system 6 comprises at least one camera 8 and one counting device 10 connected to each other.

[0043] Camera 8

[0044] Each camera 8 is mounted on board the vehicle 2. Furthermore, each camera 8 is positioned so that, in its field of view, there is a corresponding door 4 of the vehicle 2, called the “monitored door”.

[0045] Preferably, each camera 8 is a directional camera (as opposed to an omnidirectional camera). In this case, each camera 8 is, for example, oriented so that a respective optical axis forms an angle of between 30° and 60° with a floor of the vehicle 2. In this way, each camera 8 is able to collect front images of the passengers present in the corresponding field of view.

[0046] Each camera 8 is configured to acquire a video stream from the corresponding monitored door 4, and to transmit the acquired video stream to the counting device 10.

[0047] Such a video stream comprises N images of dimensions (X,Y) each. For example, in each of the images of the video stream, each corresponding pixel is, in addition, associated with its amplitude for each channel among the red, green and blue channels.

[0048] Preferably, each camera 8 is an IP technology camera (also called an “IP camera”) connected to a port of a network switch (also called a “switch”) of the vehicle 2 of a communication network of the vehicle 2, such as an Ethernet IP wired communication network having a ring topology. In this case, each camera 8 is configured to transmit the acquired video stream to the counting device 10 via said communication network.

[0049] Preferably, the counting system 6 is installed in the vehicle 2 so as to receive, for at least one monitored door 4, a control signal from said monitored door 4. More precisely, such a control signal is a control signal for the opening and / or closing of the monitored door 4. In this case, the counting system 6 is configured so that, for at least one camera 8, the triggering and / or interruption of the acquisition of the corresponding video stream depends on the control signal. More precisely still, the counting system 6 is configured so that, for at least one camera 8, the acquisition of the corresponding video stream is triggered when the associated monitored door is opened and / or interrupted when said monitored door 4 is closed.

[0050] Counting device 10

[0051] The counting device 10 is intended to determine, for each monitored door 4, the number of incoming and outgoing passengers based on the video stream received from each corresponding camera 8.

[0052] The counting device 10 is on board the vehicle 2. Alternatively, the counting device 10 is a ground-based device. In this case, the transmission of each video stream acquired from each camera 8 to the counting device 10 is likely to be carried out by means of any appropriate communication means, such as a wireless communication network.

[0053] The counting device 10 comprises a memory 12 and a processing unit 14 connected together.

[0054] Memory 12

[0055] The memory 12 is configured to store a trajectory table 16, described later.

[0056] The memory 12 is also configured to store a person detection algorithm 18 (called a “detection algorithm”), a tracking algorithm 20 and a partitioning algorithm 22.

[0057] Preferably, the detection algorithm 18 implements a detection model based on the so-called “YOLO” architecture, described by Joseph Redmon et al. in the digital preprint “You Only Look Once: Unified, Real-Time Object Detection”, referenced arXiv: 1506.02640. Such an architecture is advantageous, insofar as it is designed to provide, as output, a bounding box for each person detected on an image.

[0058] Preferably, the tracking algorithm 20 is based on the so-called “DeepSORT” architecture, described by Nicolai Wojke et al. in the digital preprint “Simple online and real time tracking with a deep association metric”, referenced arXiv: 1703.07402.

[0059] Preferably, the partitioning algorithm 22 is a k-means partitioning algorithm, a DBSCAN algorithm (for “Density-Based Spatial Clustering of Applications with Noise”, or noisy data partitioning algorithm based on spatial density), or even a partitioning algorithm called “mean shift”.

[0060] Processing unit 14

[0061] The processing unit 14 is configured to receive the video stream acquired from each camera 8, and to determine, for each door 4, the number of incoming passengers and the corresponding number of outgoing passengers.

[0062] Preferably, the processing unit 14 has a compact shape guaranteeing a small footprint in the vehicle 2, and is, in particular, compliant with railway standards for electromagnetic compatibility, resistance to fire and smoke, and resistance to shock and vibrations.

[0063] In order to determine, for each door 4, the number of incoming passengers and the corresponding number of outgoing passengers, the processing unit 14 is configured to implement a counting method 30 ([Fig.2]).

[0064] As illustrated by [Fig.2], the counting method comprises a step 32 of person detection (called “detection step”), a step 34 of passenger identification (called “identification step”), a step 36 of trajectory calculation (called “calculation step”), a partitioning step 38, a step 40 of class characterization (called “characterization step”) and a step 42 of estimation of the number of incoming and outgoing passengers (called “estimation step”).

[0065] For each video stream received, the processing unit 14 is configured to implement, during the detection step 32, the detection algorithm 18 stored in the memory 12 for each image of said video stream.

[0066] Preferably, for each image, each detection result is associated with a respective bounding box.

[0067] By "bounding box" of a detected person, it is understood, within the meaning of the present invention, the smallest rectangle in which said detected person is inscribed.

[0068] Such a bounding box is associated with corresponding coordinates in the image. Preferably, the coordinates of a bounding box are defined as the coordinates of two corners of the corresponding rectangle that are not connected. by the same side of said rectangle, for example, the upper left corner and the lower right corner.

[0069] According to another alternative, the processing unit 14 is configured to, during the detection step 32, implement a segmentation algorithm providing, as output, an outline (also called a “segmentation mask”) of each person detected.

[0070] Furthermore, the processing unit 14 is configured to, during the identification step 34, determine the detection results which correspond to the same passenger, called “identified passenger”. More precisely, for each image, a given identified passenger appears, at most, only once.

[0071] To make such a determination, the processing unit 14 is configured to implement the tracking algorithm 20 stored in the memory 12.

[0072] Preferably, each identified passenger is associated with a unique identifier, and mapped to a respective row of the trajectory table 16. For example, the unique identifier assigned to an identified passenger is the number of the respective row of the trajectory table 16.

[0073] The processing unit 14 is also configured to, during the calculation step 36, calculate, for each identified passenger, a respective trajectory. More precisely, for each identified passenger, the processing unit 14 is configured to calculate the respective trajectory from a position of at least one pixel (called a “pixel of interest”) associated with said identified passenger. More precisely, the trajectory associated with an identified passenger is the position taken, on the successive images on which said identified passenger appears, by each corresponding pixel of interest.

[0074] Preferably, for a given identified passenger, each pixel of interest is a pixel of the corresponding bounding box, for example a pixel corresponding to a corner of said bounding box. In this case, the trajectory associated with said passenger is given by the position of at least one pixel of the associated bounding box.

[0075] Alternatively, the processing unit 14 is configured to calculate the position of a barycenter of the bounding box. In this case, the trajectory associated with an identified passenger is defined by the positions successively occupied by the barycenter of the corresponding bounding box.

[0076] According to the alternative in which the processing unit 14 is configured to implement a segmentation algorithm during the detection step 32, the processing unit 14 is advantageously configured to calculate, during the calculation step 36, for each identified passenger, a position of a barycenter of the corresponding contour. In this case, the associated trajectory is defined by the positions occupied by said barycenter from image to image.

[0077] Further, for each identified passenger, the processing unit 14 is configured to write the corresponding trajectory in the trajectory table 16.

[0078] In this case, for a given line associated with an identified passenger, the successive components correspond to the coordinates of the or each pixel of interest on the successive images in which said identified passenger appears.

[0079] Furthermore, the processing unit 14 is configured to, during the partitioning step 38, implement the partitioning algorithm 22 stored in the memory 12, on the basis of each trajectory calculated during the calculation step 36, with a view to assigning each trajectory to a respective class among at least two distinct classes.

[0080] Advantageously, prior to the implementation of the partitioning algorithm 22, the processing unit 14 is configured to carry out a normalization of the lengths of the calculated trajectories. This results in normalized trajectories having the same length.

[0081] Such normalization is advantageous, insofar as it leads to providing, to the partitioning algorithm 22, objects which all have the same size.

[0082] Preferably, to perform such normalization, the processing unit 14 is configured to truncate each trajectory to the length of the shortest trajectory. More precisely, in this case, to perform such truncation, the processing unit 14 is configured to delete, in each trajectory, the positions corresponding to the most recent images of the video stream.

[0083] Alternatively, to carry out such normalization, the processing unit 14 is configured to add, to each trajectory, a number of positions equal to a difference between a length of said trajectory and a length of the longest trajectory. In this case, each added position has zero coordinates (a technique called “zero-padding”). Alternatively, each added position has coordinates equal to the coordinates of the last position of the trajectory calculated before its normalization (i.e. the position of the pixels of interest of the last image on which the identified passenger associated with the trajectory considered appears).

[0084] Furthermore, the processing unit 14 is configured to define (i.e. characterize), during the characterization step 40, one of the at least two classes as a class corresponding to the incoming passengers (called "incoming class"), and another of the at least two classes as a class corresponding to the outgoing passengers (called "outgoing class"). More precisely, the processing unit 14 is configured to define each class according to dimensions associated with the passengers identified in at least one image of the video stream.

[0085] Advantageously, to carry out such a definition of the incoming class and the outgoing class, the processing unit 14 is configured so as to, for each class obtained at the end of the partitioning step 38, and for each identified passenger of said class, first determine a surface area of ​​said passenger on the first image of the video stream on which said identified passenger appears. For example, the surface area of ​​an identified passenger is the surface area of ​​the corresponding apparent box.

[0086] In this case, the processing unit 14 is also configured to calculate, for each class, an average of the determined surfaces.

[0087] Furthermore, in this case, the processing unit 14 is configured to define: • the class for which the calculated average is the smallest as being the entering class; and • the class for which the calculated average is the highest as being the outgoing class.

[0088] This is advantageous, insofar as such an approach leads to an easy assignment of each class to the corresponding passenger category. Indeed, when they first appear in the field of vision of the cameras 8, the passengers who are about to board the vehicle 2 are further away from the cameras 8 than the passengers who are about to disembark. The average initial apparent size of the incoming passengers is therefore less than the average initial apparent size of the outgoing passengers.

[0089] Alternatively, to define the incoming class and the outgoing class, the processing unit 14 is configured so as to, for at least one class obtained at the end of the partitioning step 38, determine an evolution of the size of the identified passengers of said class, and to define the incoming class and / or the outgoing class from such a determination.

[0090] More specifically, the processing unit 14 is configured to: • define said class as the incoming class if the size of the corresponding identified passengers increases over time; and • define said class as the outgoing class if the size of the corresponding identified passengers decreases over time.

[0091] This is advantageous, insofar as such an approach leads to an easy assignment of each class to the corresponding passenger category. Indeed, the passengers who board the vehicle 2 move closer to the cameras 8, so that their apparent size, in the field of vision of the cameras 8, increases over time. Conversely, the passengers who get off the vehicle move away from the cameras 8, and their apparent size therefore decreases over time.

[0092] By way of example, to implement such a definition, the processing unit 14 is configured so as to, for at least one class, and for each identified passenger of said class, determine the corresponding surface in M ​​images on which said identified passenger appears. In this case, the processing unit 14 is also configured to calculate, for the identified passengers of said class, an average respective surfaces determined from the images of rank k (k being between 1 and M). Furthermore, in this case, the processing unit 14 is configured to determine an evolution, over time, of the calculated average surface.

[0093] Preferably, in the case where the processing unit 14 is configured to assign, during the partitioning step 38, each trajectory to a respective class among three distinct classes, the processing unit 14 is, in addition, configured to identify the remaining class as being representative of the passengers remaining on the platform and / or on board the vehicle.

[0094] The processing unit 14 is also configured to determine, or estimate, during the estimation step 42, the number of incoming passengers as a function of a size of the incoming class and / or the number of outgoing passengers as a function of a size of the outgoing class.

[0095] Preferably, the number of incoming passengers is equal to the size of the incoming class and / or the number of outgoing passengers is equal to the size of the outgoing class.

[0096] Operation

[0097] The operation of the counting system 6 will now be described with reference to [Fig.2].

[0098] Each camera 8 acquires a video stream of the corresponding monitored door 4. Preferably, each camera 8 acquires the video stream only between an opening time and a closing time of the corresponding monitored door 4.

[0099] Then, each camera 8 transmits the acquired video stream to the counting device 10.

[0100] Then, during the detection step 32, for each video stream received, and for each image of said video stream, the processing unit 14 applies the detection algorithm 18 to said image.

[0101] Then, during the identification step 34, the processing unit 14 determines the detection results which correspond to the same identified passenger. Preferably, each identified passenger is associated with a unique identifier.

[0102] Then, during the calculation step 36, the processing unit 14 calculates, for each identified passenger, the respective trajectory, from a position of at least one pixel of interest associated with said identified passenger. Furthermore, for each identified passenger, the processing unit 14 writes the corresponding trajectory in the trajectory table 16.

[0103] Then, during the partitioning step 38, the processing unit 14 applies the partitioning algorithm 22 to the calculated trajectories to assign each trajectory to a respective class among at least two distinct classes.

[0104] Then, during the characterization step 40, the processing unit 14 defines the incoming class and the outgoing class as a function of dimensions, in at least one image of the video stream, associated with the identified passengers whose respective trajectory is present in each class.

[0105] Then, during the estimation step 42, the processing unit 14 determines the number of incoming passengers as a function of the size of the incoming class and / or the number of outgoing passengers as a function of the size of the outgoing class.

[0106] Of course, the invention is not limited to the examples which have just been described.

Claims

Claims

1. Device (10) for counting passengers at the door comprising a processing unit (14) configured to receive at least one video stream representative of a monitored door of a vehicle between an opening time of the monitored door and a closing time of the monitored door, each video stream having been, preferably, acquired by means of a directional camera, the processing unit (14) being, in addition, configured so as to, for each video stream received: • implement a person detection algorithm for each image of the video stream; • determine the detection results which correspond to the same passenger, forming an identified passenger; • for each identified passenger, calculate a respective trajectory from a position of at least one pixel of interest associated with said identified passenger, on each image of the video stream on which said identified passenger appears;• implementing a partitioning algorithm to assign each calculated trajectory to a respective class among at least two distinct classes; • based on dimensions associated with the passengers identified in at least one image of the video stream, defining one of the at least two classes as an incoming class corresponding to the incoming passengers, and another of the at least two classes as an outgoing class corresponding to the outgoing passengers; and • estimating the number of incoming passengers based on a size of the incoming class and / or the number of outgoing passengers based on a size of the outgoing class.;

2. Counting device (10) according to claim 1, wherein, to define the incoming class and the outgoing class, the processing unit (14) is configured to: • determine, for each class, and for each identified passenger corresponding to said class, an area of said passenger on the first image of the video stream on which said identified passenger appears; • calculate, for each class, an average of the determined areas; and • define: • the class for which the calculated average is the smallest as the incoming class; and • the class for which the calculated average is the largest as the outgoing class.

3. Counting device (10) according to claim 1, wherein, to define the incoming class and the outgoing class, the processing unit (14) is configured so as to, for at least one class: • determine a temporal evolution of the size of the identified passengers of said class; and • define: • said class as being the incoming class if the size of the corresponding identified passengers increases over time; and • define said class as being the outgoing class if the size of the corresponding identified passengers decreases over time.

4. A counting device (10) according to any one of claims 1 to 3, wherein each detection result is associated with a respective bounding box, the processing unit (14) being configured to calculate the trajectory associated with each identified passenger as being a trajectory of at least one pixel dependent on a position of the bounding box.

5. Counting device (10) according to any one of claims 1 to 4, in which the processing unit (14) is configured to, prior to the implementation of the partitioning algorithm, carry out a normalization of the lengths of the calculated trajectories, the normalized trajectories having the same length.

6. A counting device (10) according to claim 5, wherein the normalization comprises: • truncating each trajectory to the length of the trajectory the shortest, by deleting the positions corresponding to the most recent images of the video stream; or • an addition, to each trajectory, of a number of positions equal to a difference between a length of said trajectory and a length of the longest trajectory, each added position having coordinates: • zero; or • equal to the coordinates of the last position of the calculated trajectory, before its normalization.

7. A counting device (10) according to any one of claims 1 to 6, wherein the processing unit (14) is configured to implement the partitioning algorithm to assign each trajectory to a respective class among three distinct classes, the processing unit being further configured to define the class which is neither the incoming class nor the outgoing class as being the representative class of the passengers remaining on the platform and / or on board the vehicle.

8. A counting system (6) comprising a counting device (10) according to any one of claims 1 to 7 and at least one camera (8), each camera being connected to the counting device (10) for transmitting a video stream to said counting device (10), each camera preferably being a directional camera.

9. A counting system (6) according to claim 8, configured to, in use, trigger and interrupt an acquisition of the video stream by at least one corresponding camera (8) as a function of a control signal from at least one associated monitored door (4).

10. Vehicle (2), in particular a rail transport vehicle, comprising a counting system (6) according to claim 8 or 9, each camera (8) of the counting system (6) being on board the vehicle (2) and positioned so that a corresponding door (4) of the vehicle, forming a monitored door (4), is located in a field of view of said camera (8).

11. Method for counting passengers at the door, the method being implemented by computer and comprising the steps: • implementing (32) a person detection algorithm for each image of at least one video stream received, each video stream received being representative of a monitored door of a vehicle between a time of opening of the monitored door and a time of closing of the monitored door, each video stream having been, preferably, acquired by means of a directional camera; determination (34) of the detection results which correspond to the same passenger, forming an identified passenger; for each identified passenger, calculation (36) of a respective trajectory from a position of at least one pixel of interest associated with said identified passenger, on each image of the video stream on which said identified passenger appears; implementing (38) a partitioning algorithm to assign each calculated trajectory to a respective class among at least two distinct classes; based on dimensions associated with the passengers identified in at least one image of the video stream, characterizing (40) one of the at least two classes as an incoming class corresponding to the incoming passengers, forming an incoming class, and another of the at least two classes as an outgoing class corresponding to the outgoing passengers; and estimating the number of incoming passengers based on a size of the incoming class and / or the number of outgoing passengers based on a size of the outgoing class.

12. A computer program comprising executable instructions which, when executed by a computer, implement the steps of the method according to claim 11.

Images