System and method for calibrating a video surveillance camera in a railway vehicle
The method automates the calibration of non-motorized railway vehicle cameras using a calculator and robotic arm to adjust image acquisition units, addressing the inefficiencies of manual calibration and high costs of motorized systems, achieving precise and cost-effective camera alignment.
Patent Information
- Application Number
- FR2022005906
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-06-16
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-06-16
AI Technical Summary
Existing video surveillance cameras in railway vehicles require manual, time-consuming, and imprecise calibration due to vibrations, necessitating two operators and being costly for motorized solutions, which are not suitable for environments with many cameras.
A method for calibrating non-motorized cameras using a calculator to determine a mathematical transformation function, enabling automated calibration by a single operator, involving a computer, automaton, and robotic arm to adjust the image acquisition unit based on reference and current images.
Enables precise and rapid calibration of multiple cameras without replacing existing equipment, reducing operator effort and costs, using a single operator and a compact, inexpensive setup.
Smart Images

Figure 00000017_0000 
Figure 00000017_0001 
Figure 00000017_0002
Abstract
Description
Title of the invention: System and method for calibrating a video surveillance camera in a railway vehicle. Technical field
[0001] The present invention relates to the field of video surveillance devices and more particularly to a system and method for calibrating a video surveillance camera in a railway vehicle.
[0002] As is known, with reference to [Fig. 1], a railway vehicle V comprises several railway cars 1 for accommodating passengers. In each railway car 1, it is known to mount one or more video surveillance cameras 2 to monitor the interior of the railway car 1 and ensure the protection of the passengers.
[0003] In practice, in the railcar 1, the camera 2 must be positioned and oriented precisely to obtain an optimal field of vision and view a predetermined target area. For example, the camera 2 can be positioned and oriented to view an access door from the outside in order to see passengers entering and exiting and all the safety devices positioned near the access door, such as the alarm button or the emergency manual release handle for the door.
[0004] As is known, with reference to [Fig. 2], the camera 2 includes an image acquisition device 20 which allows for real-time acquisition of an image of the target area. The camera 2 is connected to a computer 3 to analyze the image and, for example, compare it with an expected image or to store it.
[0005] The current image can also be used by a mathematical algorithm, for example for facial recognition of passengers. Such an algorithm can only work if the current image conforms to an expected image.
[0006] However, during the installation of the video surveillance camera in the railway car or during maintenance operations, the field of view of the image acquisition device may shift, resulting in an image offset and potentially leading to errors in the interpretation of the image received by the computer. When the railway vehicle is in motion, vibrations may also cause the camera to shift or become misaligned, resulting in an actual image that is offset from a predetermined, expected image.
[0007] Therefore, it is necessary to calibrate the video surveillance camera to reposition it precisely so that the current image corresponds to the predetermined expected image. Such a calibration operation is carried out, in a known manner, manually by two operators. One operator manually orients the one camera while a second monitors a display screen to which the computer is connected until the current image matches the expected image.
[0008] However, such a manual operation, performed iteratively, is imprecise and tedious. Furthermore, such calibration is time-consuming and requires the intervention of two operators, which is undesirable.
[0009] A motorized video surveillance camera is known in the prior art, comprising automatic calibration means capable of modifying the position and angle of the image acquisition element without manual intervention by an operator. However, such a motorized camera is expensive and cannot be used in a railway vehicle, which typically has a large number of video surveillance cameras per railway car. Furthermore, such a motorized camera is not suitable for mounting in an environment subject to significant vibrations.
[0010] The invention thus aims to eliminate at least some of these drawbacks by proposing a calibration method for a non-motorized video surveillance camera in a railway vehicle that is simple, fast and efficient, while being feasible by a single operator. PRESENTATION OF THE INVENTION
[0011] The invention relates to a method for calibrating a fixed video surveillance camera in a railway vehicle, • the rail vehicle comprising at least one target area to be monitored, the target area including at least one reference device, • the camera comprising an image acquisition element configured to be moved mechanically, • the camera being initially configured to view the target area according to a reference setting and to acquire a reference image of the target area, the reference image comprising a reference pattern having a reference position in the reference image, the reference pattern representing said reference equipment of the target area, • the camera being configured to routinely view the target area according to a current setting and acquire a current image of the target area, the current image including the reference pattern having a current position in the current image.
[0012] The process comprises the steps of: • determine, using a calculator, a mathematical transformation function that transforms the current position of the reference pattern into the reference position of the reference pattern, • determine at least one mechanical displacement control from the mathematical transformation function, and • move the image acquisition unit of the video surveillance camera using an automated system based on the mechanical movement control.
[0013] The calibration method according to the invention allows for automated calibration while eliminating the need for expensive motorized cameras. The method enables precise calibration performed by an automated system, thus limiting the number of iterations between two operators. The computer determines an optimal mechanical movement command for the camera, which is advantageously moved by the automated system, ensuring rapid calibration.
[0014] Furthermore, the method according to the invention, using an automated system, can advantageously be applied to existing railway vehicles in which known non-motorized cameras are installed. In other words, thanks to the invention, it is not necessary to replace the existing cameras for them to be quickly and accurately calibrated.
[0015] The use of an automated system also helps to limit the operator's physical strain.
[0016] In a first embodiment, the reference motif represents an access door to the outside of the railway vehicle.
[0017] In a second embodiment, the reference motif represents a row of seats in the railway vehicle.
[0018] Preferably, the camera is non-motorized, which allows the use of a compact and inexpensive camera.
[0019] In one embodiment, the camera includes a mechanical element connected to the image acquisition element, making it easier to grasp.
[0020] Preferably, the automaton comprises a movable arm and a gripping member for the image acquisition member. The automaton can easily move the gripping member, which can grasp the image acquisition member in a simple manner.
[0021] In one embodiment, the camera being positioned at a fixed camera position in the railway vehicle, the automaton being positioned at a fixed automaton position in the railway vehicle, the mechanical movement control is determined from the mathematical transformation function, the camera position and the automaton position.
[0022] Preferably, the automaton is mounted removably in the railway vehicle, which limits the space taken up in the railway vehicle while allowing the use of a single automaton to calibrate several cameras in the railway vehicle.
[0023] In one embodiment, the calibration process includes, prior to the step of determining, by means of a computer, a mathematical function The transformation process involves a step where a signal is sent to activate the computer by a piece of equipment on the railway vehicle. The entire process can thus be launched automatically directly by a piece of equipment on the railway vehicle.
[0024] The invention also relates to a railway vehicle configured to implement the method as described above, the railway vehicle comprising: • at least one target area to be monitored, the target area comprising at least one reference device, • at least one camera comprising an image acquisition unit configured to be mechanically moved, • the camera being initially configured to view the target area according to a reference setting and to acquire a reference image of the target area, the reference image comprising a reference pattern having a reference position in the reference image, the reference pattern representing said reference equipment of the target area, • the camera being configured to routinely view the target area according to a current setting and acquire a current image of the target area, the current image including the reference pattern having a current position in the current image, • at least one calculator configured for: • determine at least one mathematical transformation function that transforms the current position of the reference pattern into the reference position of the reference pattern, • determine at least one mechanical displacement control from the mathematical transformation function, and • send the mechanical movement command, • at least one automated system configured to receive the mechanical movement command from the computer and to move the camera's image acquisition unit from the mechanical movement command.
[0025] In one embodiment, the railway vehicle comprises: • a plurality of target areas to be monitored, each target area including at least one specific reference piece of equipment, • a plurality of CCTV cameras, each camera being in a fixed and distinct camera position in the railway vehicle, each camera being configured to view one of the target areas out of the plurality of target areas, • Each camera is initially configured to view the associated target area according to a reference setting and acquire a reference image of said target area, the reference image comprising a reference pattern having a reference position in the reference image, the reference pattern representing said reference equipment of the associated target area, each camera being configured to routinely view the associated target area according to a current setting and acquire a current image of said target area, the current image comprising the reference pattern having a current position in the image current, • a computer configured for each camera to: • determine at least one mathematical transformation function that transforms the current position of the reference pattern into the reference position of the reference pattern, • determine at least one mechanical movement control associated with the camera from the mathematical transformation function, • a single automaton, mobile in the railway vehicle, the automaton being configured to be moved successively to a plurality of automaton positions in the railway vehicle and to move the image acquisition unit of each camera from the associated mechanical movement control. PRESENTATION OF THE FIGURES
[0026] The invention will be better understood upon reading the following description, given by way of example, and referring to the following figures, given by way of non-limiting examples, in which identical references are given to similar objects.
[0027] Fig. 1 is a schematic representation of a railway vehicle comprising a plurality of railway cars.
[0028] Fig. 2 is a schematic representation of a railway car of the railway vehicle of Fig. 1.
[0029] Fig. 3 is a schematic representation of a video surveillance camera mounted in the railway car of Fig. 2.
[0030] The [Fig.4] is a representation of a reference image acquired by an image acquisition device of the camera of the [Fig.3] according to a reference parameterization.
[0031] Fig. 5 is a representation of a current image acquired by the image acquisition organ of the camera in Fig. 3 according to a current parameter setting.
[0032] Figure 6 is a representation of a current image following calibration according to the invention of the camera.
[0033] Fig. 7 is a representation of an automaton mounted in the railway car used for calibration according to the invention of the camera.
[0034] Figure 8 is a diagram of the steps of a calibration process for the camera in Figure 3.
[0035] It should be noted that the figures set out the invention in detail to implement the invention, said figures being of course able to serve to better define the invention where appropriate. DETAILED DESCRIPTION OF THE INVENTION
[0036] As explained previously, with reference to [Fig. 1], a railway vehicle V comprises several railway cars 1 to accommodate passengers. For the sake of brevity and clarity, the invention will be described for a single railway car 1.
[0037] According to the invention, with reference to [Fig. 2], the railcar 1 has a target area Z to be monitored, the target area Z having at least one reference device ER. In this example, the reference device ER is an access door P to the outside of the railcar 1. Alternatively, the reference device ER could be an interior door PI providing access to an adjacent railcar 1, a row of seats S, or any equipment normally mounted in a railcar 1.
[0038] The railway car 1 includes one or more video surveillance cameras 2 to monitor the interior of the railway car 1. In this example, for the sake of brevity, a railway car 1 comprising a single camera 2 will be described; however, it is understood that the railway car 1 could include a number of cameras 2 greater than one.
[0039] Still with reference to [Fig.2], the camera 2 is fixed in the railway car 1 and is mounted at a predetermined and known POScam camera position, for example, on the ceiling.
[0040] With reference to [Fig. 2] and [Fig. 3], the camera 2 comprises an image acquisition element 20 mechanically connected to a base 21 fixed to the railway car 1. Preferably, the image acquisition element 20 is configured to be mechanically moved to orient the field of view of the camera 2. By way of example, the image acquisition element 20 is configured to be adjustable in translation and rotation, as will be described in more detail later. In this example, the acquisition element 20 is in the form of an acquisition sensor.
[0041] In one embodiment, the camera 2 includes a mechanical element 22 connected to the image acquisition element 20, so as to facilitate its grasping. In this example, the image acquisition element 20 having a cylindrical shape, the element The mechanical component 22 is in the form of a hollow cylinder mounted around the image acquisition element 20. The mechanical component 22 extends longitudinally along an axis of revolution X of the cylinder and is longer than the image acquisition element 20 along this same X axis, in order to facilitate its grasping. Adjusting the position and orientation of the image acquisition element 20 is thus simpler and faster. It goes without saying that the mechanical component 22 could have another form suitable for grasping, particularly via an automated system as will be described later.
[0042] According to the invention, the camera 2 is initially configured to look at the target area Z in the railway car 1 according to a reference PARref setting, that is to say according to a predetermined and known reference orientation and positioning of the image acquisition organ 20.
[0043] With reference to [Fig. 4], the camera 2 is configured to acquire a reference image IMref of the target area Z. The reference image IMref includes a reference pattern MR that represents the reference equipment ER of the target area Z. In the reference image IMref, the reference pattern MR has a reference position POSref. In this example, where the reference equipment ER is an access door P to the outside of the railcar 1, the reference pattern MR representing the reference equipment ER comprises four points Po and has a "T" shape, as shown in [Fig. 4].
[0044] The reference pattern MR is preferably determined by image processing, for example, by determining the intersection of lines specific to the reference equipment ER. The reference pattern MR can also be determined by a learning algorithm, in particular, by a neural network. In this example, the reference pattern MR has a T-shape, but it is understood that its shape could be different. Furthermore, the reference position POSref corresponds to the position of a plurality of pixels of the reference pattern MR in the reference image IMref, for example, the 4 black dots illustrated in [Fig. 4].
[0045] Similarly, with reference to [Fig. 5], camera 2 is configured to view the target area Z in a current PARcour setting, i.e., according to an orientation and position at a given time. Camera 2 is then configured to acquire a current IMcour image of the target area Z. The current IMcour image also includes the reference pattern MR, which represents the reference equipment ER of the target area Z. In the current IMcour image, the reference pattern MR has a current POScour position. Similar to the previous example, in the current IMcour image, the reference pattern MR is preferably determined by image processing, for example, by determining the intersection of lines specific to the reference equipment ER. The reference pattern MR can also be determined by a learning algorithm, in particular, by a neural network. Thus, the current position POScour of the reference pattern MR can be determined automatically and conveniently.
[0046] In a preferred embodiment, the camera 2 is non-motorized, which allows for a robust, inexpensive camera 2 with limited bulk.
[0047] According to the invention, as shown in [Fig. 2], the rail vehicle V also includes a computer 3, mounted, for example, in one of the rail cars 1. Preferably, the computer 3 is in the form of an on-board computer of the rail vehicle V, for example, mounted in a power car. Alternatively, the computer 3 can be in the form of a nano-computer (for example, a Raspberry Pi®) which can be mobile within the rail vehicle V.
[0048] The computer 3 is configured to be electronically connected to the camera 2 and to exchange data with the camera 2 via a data network, for example wired, via an Ethernet network, or wirelessly, via a wifi network for example.
[0049] Calculator 3 is configured to: • receive, via the data network, the reference image IMref and the current image IMcour, • compare the reference image IMref and the current image IMcour, and more specifically compare the reference position POSref of the reference pattern MR in the reference image IMref with the current position POScour of the reference pattern MR in the current image IMcour, • determine a mathematical transformation function F-IM to transform the current position POScour of the reference pattern MR into the reference position POSref of the reference pattern MR, • determine a mechanical displacement control COM from the mathematical transformation function F-IM, and • send the COM mechanical displacement command.
[0050] Preferably, the mathematical transformation function F-IM is a homography matrix which includes the set of three-dimensional geometric transforms to be applied to realign the current image IMcour with the reference image IMref.
[0051] Preferably, the mechanical displacement control COM is determined from the mathematical transformation function F-IM, the camera position POScam and a PLC position POSaut, which will be described below.
[0052] According to the invention, with reference to [Fig. 7], the rail vehicle V also includes an automaton 4 configured to mechanically move the image acquisition unit 20, as will be described in more detail later. The automaton 4 has a POSaut automaton position in the rail car 1. predetermined. Preferably, the automaton 4 is configured to be mounted removably in the rail vehicle 1, so as to limit the space taken up in the rail car 1. In other words, the automaton 4 is configured to be placed in the automaton position POSaut only during the calibration of the camera 2.
[0053] In this example, the railway vehicle V comprises a single automaton 4 suitable for being mounted consecutively in each railway car 1. It goes without saying that the railway vehicle V could just as well comprise a plurality of automata 4.
[0054] The automaton 4 is configured to receive the mechanical displacement command COM sent by the computer 3 and to move the image acquisition element 20 of the camera 2 from the mechanical displacement command COM, so as to reposition the reference pattern MR to its reference position POSref, as shown in [Fig.6].
[0055] As shown in [Fig.7], the automaton 4 is preferably in the form of a robotic arm and as such comprises a movable arm 41 and a gripping member 42 of the image acquisition member 20. The movable arm 41 comprises a fixed end in the railway car 1 and a free end, the gripping member 42 is mounted on the free end of the movable arm 41. The movable arm 41 allows the gripping member 42 to be moved to the camera 2, in particular, to grasp the mechanical member 22 as will be shown later.
[0056] A railway car 1 has been described comprising a single target zone Z and a single camera 2; however, it is understood that the railway car 1 could include a number of target zones Z and cameras 2 greater than or equal to two, for example to monitor the door P providing access to the outside, an interior door PI providing access to an adjacent car, each row of seats S, ... Preferably, the railway car 1 includes as many cameras 2 as there are target zones Z.
[0057] In the case of a plurality of target zones Z and cameras 2, each camera 2 being configured to view one of the target zones Z respectively, each camera 2 is initially configured to view the associated target zone Z according to its own PARref reference parameterization, and to acquire a reference image IMref (including the reference pattern MR) of the associated target zone Z. In the reference image IMref acquired by each camera 2, the reference pattern MR has its own POSref reference position.
[0058] Each camera is also configured to continuously view the associated target area Z according to its own current PARcour setting, and to acquire a current IMcour image (including the reference pattern MR) of the associated target area Z. In the current IMcour image acquired by each camera 2, the reference pattern MR has its own current POScour position.
[0059] Preferably, the calculator 3 is then configured to repeat the previous steps previously mentioned for each of the cameras 2, independently of each other, and to determine a mechanical movement command COM specific to each camera 2.
[0060] The automaton 4 is in this case configured to receive a plurality of mechanical movement commands COM, in particular a specific mechanical movement command COM for each camera 2. The automaton 4 is also configured, preferably, to be moved successively to different automaton POSaut positions in the railway vehicle 1 (a POSaut automaton position can be determined for one or more camera(s) 2 in the same railway car 1) and to move the image acquisition element 20 of each camera 2 from the associated mechanical movement command COM.
[0061] A calibration method for a camera 2 mounted in a railway vehicle V as described above will now be described, according to an embodiment of the invention, with reference to [Fig.8].
[0062] In this example, the calculator 3 has previously, in a step E0: • Received and recorded the POScam camera position of camera 2, • received, via the data network, the reference image IMref acquired by the image acquisition unit 20 of camera 2 according to the reference parameter PARref, • determined the reference position POSref of the reference pattern MR in the reference image IMref, • recorded the reference image IMref and stored in memory the reference position POSref of the reference pattern MR in the reference image IMref.
[0063] At a given moment, the camera 2 is located according to a current PARcour parameter and is looking at the target area Z. In this example, due to the vibrations of the railway vehicle V, the acquisition element 20 has moved.
[0064] The calibration process in this example includes a preliminary step of emitting an activation signal to the computer 3. Such a signal is, for example, a signal to close the access door P to the outside.
[0065] Following the emission of the signal, the camera 2 sends to the computer 3, via the data network, in a first step El, a current image IMcour of the target area Z.
[0066] The computer 3, having received the current image IMcour, compares it in a step E2 to the reference image IMref. More precisely, in this step E2, the computer 3 compares the reference position POSref of the reference pattern MR in the reference image IMref with the current position POScour of the reference pattern MR in the current image IMcour. The reference position POSref is determined here by image processing.
[0067] In a third step E3, the calculator 3 determines a mathematical function The F-IM transformation function transforms the current position POScour of the reference pattern MR into the reference position POSref of the reference pattern MR. In one embodiment, the mathematical function of the F-IM transformation of the camera 2 is stored by the computer 3 for subsequent calibration, for example, for the successive calibration of several cameras 2. In this example, the mathematical function of the F-IM transformation is a homographic function that transforms the coordinates of a plurality of pixels in the current position POScour into the coordinates of a plurality of pixels in the reference position POSref. Such a mathematical function is known and will not be described in detail.
[0068] The method then includes a fourth step E4, positioning the automaton 4 near the camera 2. The automaton 4 is attached to the railcar 1 and positioned at a predetermined POSaut automaton position. Preferably, the POSaut automaton position is previously known to the computer 3 based on the camera 2. Alternatively, after the positioning of the automaton 4, the latter can send its POSaut automaton position to the computer 3 via the data network.
[0069] The computer 3 then determines, in a step E5, a mechanical movement command COM from the stored mathematical transformation function F-IM, the previously stored camera position POScam, and the PLC position POSaut, and sends this mechanical movement command COM to the PLC 4 via the data network. Preferably, the mechanical movement command COM is determined from a database that associates a mechanical movement command COM with each mathematical transformation function F-IM. Preferably, the mechanical movement command COM depends on the relative position of the camera 2 and the PLC 4.
[0070] In a sixth step E6, the automaton 4 moves the image acquisition element 20 of the camera 2 using the mechanical movement control COM, in order to calibrate the camera 2, that is, to move the image acquisition element 20 from the current PARcour setting to the reference setting by changing the current POScour position of the reference pattern MR to its reference position POSref. For this purpose, in this example, the gripping element 42 of the automaton 4 grasps the mechanical element 22 connected to the acquisition element 20 in order to prevent any damage to the camera 2. Preferably, the mechanical movement control COM performs a translation and / or a rotation of the acquisition element 20 in order to orient it optimally ([Fig. 6]).
[0071] In one embodiment, for the successive calibration of several cameras 2, the set of steps E1 to E3 are repeated for each camera 2 and a plurality of mathematical transformation functions F-IM are preferably successively recorded by the computer 3, a mathematical transformation function F-IM being determined and recorded for each pair of a reference image IMref and a current image IMcour received by camera 2.
[0072] In this embodiment, the computer 3 thus stores a plurality of mechanical movement commands COM, one mechanical movement command COM respectively for each camera 2. The automaton 4 is then preferably successively placed on several automaton positions POSaut (one automaton position POSaut corresponding to each camera 2) and successively moves the image acquisition element 20 of each camera 2 during a single calibration operation of all the cameras 2 of the railway vehicle V. Preferably, each automaton position POSaut is previously determined and recorded by the computer 3.
[0073] The calibration method according to the invention can easily be carried out by a single operator to calibrate all the cameras of the railway vehicle accurately and without using a motorized camera, allowing for smaller and less expensive cameras.
[0074] The calibration process can be carried out during maintenance steps of the railway vehicle, during a particular event or periodically.
Claims
Claims
1. Method for calibrating a non-motorized video surveillance camera (2) fixedly mounted in a railway vehicle (1), • the railway vehicle (1) comprising at least one target zone (Z) to be monitored, the target zone (Z) comprising at least one reference equipment (ER), • the camera (2) comprising an image acquisition member (20) configured to be moved mechanically, • the camera (2) being initially configured to look at the target area (Z) according to a reference setting (PARref) and acquire a reference image (IMref) of the target area (Z), the reference image (IMref) comprising a reference pattern (MR) having a reference position (POSref) in the reference image (IMref), the reference pattern (MR) representing said reference equipment (ER) of the target area (Z), • the camera (2) being configured to currently view the target area (Z) according to a current setting (PARcour) and acquire a current image (IMcour) of the target area (Z), the current image (IMcour) comprising the reference pattern (MR) having a current position (POScour) in the current image (IMcour), • the method comprising the steps of: • Determine, using a calculator (3), a mathematical transformation function (F-IM) transforming the current position (POScour) of the reference pattern (MR) into the reference position (POSref) of the reference pattern (MR), • Determine at least one mechanical displacement command (COM) from the mathematical transformation function (F-IM), and • Move the image acquisition device (20) of the video surveillance camera (2) using an automaton (4) from the mechanical movement control (COM).
2. Calibration method according to claim 1, in which the reference pattern (MR) represents an access door (P) to the exterior of the railway vehicle (1).
3. A calibration method according to claim 1, wherein the reference pattern (MR) represents a row of seats (S) in the railway vehicle (1).
4. Calibration method according to one of claims 1 to 3, in which the automaton (4) comprises a movable arm (41) and a gripping member (42) of the image acquisition member (20).
5. Calibration method according to one of claims 1 to 4, wherein the camera (2) being positioned at a fixed camera position (POScam) in the railway vehicle (1), the automaton (4) being positioned at a fixed automaton position (POSauto) in the railway vehicle (1), the mechanical movement command (COM) is determined from the mathematical transformation function (F-IM), the camera position (POScam) and the automaton position (POSauto).
6. Calibration method according to one of claims 1 to 5, in which the automaton (4) is removably mounted in the railway vehicle (1).
7. Calibration method according to one of claims 1 to 6 comprising previously to the step of determining, by means of a calculator (3), a mathematical transformation function (F-IM), a step of transmitting an activation signal of the calculator (3) by equipment of the railway vehicle (1).
8. Railway vehicle (1) comprising: • at least one target area (Z) to be monitored, the target area (Z) comprising at least one piece of reference equipment (ER), • at least one non-motorized camera (2) comprising an image acquisition member (20) configured to be moved mechanically, • the camera (2) being initially configured to look at the target area (Z) according to a reference parameter setting (PARref) and acquire a reference image (IMref) of the target area (Z), the reference image (IMref) comprising a reference pattern (MR) having a reference position (POSref) in the reference image (IMref), the reference pattern (MR) representing said reference equipment (ER) of the target area (Z), • the camera (2) being configured to currently view the target area (Z) according to a current setting (PARcour) and acquire a current image (IMcour) of the target area (Z), the current image (IMcour) comprising the reference pattern (MR) having a current position (POScour) in the current image (IMcour), at least one calculator (3) configured to: • determine at least one mathematical transformation function (F-IM) transforming the current position (POScour) of the reference pattern (MR) into the reference position (POSref) of the reference pattern (MR), • determine at least one mechanical movement command (COM) from the mathematical transformation function (F-IM), • send the mechanical movement command (COM), at least one automaton (4) configured to receive the mechanical movement command (COM) from the computer (3) and to move the image acquisition member (20) of the camera (2) from the mechanical movement command (COM).
9. Rail vehicle (1) according to claim 8 comprising: • a plurality of target zones (Z) to be monitored, each target zone (Z) comprising at least one specific reference equipment (RE), • a plurality of video surveillance cameras (2), each camera (2) being at a fixed and distinct camera position (POScam) in the railway vehicle (1), each camera (2) being configured to look at one of the target zones (Z) of the plurality of target zones (Z), • each camera (2) being initially configured to look at the associated target zone (Z) according to a pa reference image (PARref) and acquiring a reference image (IMref) of said target area (Z), the reference image (IMref) comprising a reference pattern (MR) having a reference position (POSref) in the reference image (IMref), the reference pattern (MR) representing said reference equipment (ER) of the associated target area (Z), • each camera (2) being configured to currently view the associated target area (Z) according to a current setting (PARcour) and acquire a current image (IMcour) of said target area (Z), the current image (IMcour) comprising the reference pattern (MR) having a current position (POScour) in the current image (IMcour), a calculator (3) configured for each camera (2) to: • determine at least one mathematical transformation function (F-IM) transforming the current position (POScour) of the reference pattern (MR) into the reference position (POSref) of the reference pattern (MR), • determine at least one mechanical movement command (COM) associated with the camera (2) from the mathematical transformation function (F-IM), a single automaton (4), movable in the railway vehicle (1), the automaton (4) being configured to be moved successively to a plurality of automaton positions (POSauto) in the railway vehicle (1) and to move the image acquisition member (20) of each camera (2) from the associated mechanical movement command (COM).
