SYSTEM AND METHOD FOR AUTOMATICALLY DETECTING AND READING RAILWAY TRACK SIGNS

Abstract

SYSTEM AND METHOD FOR AUTOMATICALLY DETECTING AND READING RAILWAY TRACK SIGNS The invention relates to a system and method for automatically detecting and reading railway track signal signs comprising: a step (E1) of acquiring a 3D point cloud of the environment of the track; a concomitant step (E2) of acquiring an image of the environment of said track; a step (E3) of partitioning said acquired 3D point cloud into a plurality of slices; a step (E4) of orthogonal projection of the points of each slice; a step (E6) of comparing said projected 2D point cloud with a plurality of predetermined panel shapes; a step (E7) of selecting a panel, called recognized panel, chosen from said plurality of predetermined panels;a step (E8) of detecting, on said acquired image corresponding to said 3D point cloud from which the 2D point cloud corresponding to said recognized panel is derived, the light information provided by said recognized panel. Figure for the abstract: figure 4;

Description

Title of the invention: SYSTEM AND METHOD FOR AUTOMATICALLY DETECTING AND READING RAILWAY TRACK SIGNS Technical field of the invention

[0001] The invention relates to a system and a method for automatically detecting and reading traffic signs arranged in the vicinity of a track of a railway vehicle. The invention extends to a railway vehicle equipped with such a system for automatically detecting and reading traffic signs. The invention also extends to a drone equipped with such a system for automatically detecting and reading railway traffic signs. Technological background

[0002] The movement of a railway vehicle on a track is restricted and made safe by the presence of a certain number of signs arranged along the track and intended to provide traffic information to the train driver. These signals are generally light signals which authorize or not the passage of the train, depending on the railway traffic on the portion of track that the train is about to take. These signs are arranged in the vicinity of the track, and in particular on the lateral edges of the track or on gantries which overhang the track. These signs are read and interpreted by the train driver.

[0003] With the aim of deploying an autonomous train, it is necessary to equip the trains with a means of accessing traffic information and in particular traffic signs, in particular for autonomous trains intended to run on tracks not compatible with the European system known by the acronym ETCS for European Train Control System. This system is ultimately intended to replace the signaling of the traffic tracks, which should allow trains to directly access the signaling information. That being said, certain portions of track will not be quickly compatible with this ETCS system so that there is a need to propose a system and a method for automatically detecting and reading the traffic signs currently arranged along the tracks.

[0004] One of the solutions already proposed is to have a precise map of the location of the traffic signs along the tracks and to precisely geolocate the train. Thus, when the train approaches an area where a traffic sign is present, a camera on board the train is activated to acquire the side traffic sign. An image processing module can then analyze the acquired image to deduce the signaling information. emitted by the sign. One of the disadvantages of this solution is, on the one hand, that it is necessary to first create a map of all railway signal signs and, on the other hand, that this map requires a high level of precision. In addition, the location of the train must be as precise as the location of the signs, otherwise the camera on board the train may be triggered too early or too late and miss the sign and therefore not be able to access the signaling information.

[0005] The inventors therefore sought to propose an alternative solution for automatically detecting and reading signaling information from a railway track. Objectives of the invention

[0006] The invention aims to provide a system and a method for automatically detecting and reading railway track signaling signs.

[0007] The invention also aims to provide, in at least one embodiment, a method and a system for automatically detecting and reading traffic signs which do not require any prior knowledge of the mapping of the railway track.

[0008] The invention also aims to provide, in at least one embodiment, a method and a system for automatically detecting and reading traffic signs which do not require access to an absolute location of the train on the railway track.

[0009] The invention also aims to provide, in at least one embodiment, a method and a system for automatically detecting and reading traffic signs which can be fully implemented on board the train, without access to an external server.

[0010] The invention also aims to provide, in at least one embodiment, a method and a system for automatically detecting and reading traffic signs which can be certified for use on board autonomous trains.

[0011] The invention also aims to provide, in at least one embodiment, such a system and method for automatically detecting and reading traffic signs which do not require the implementation of artificial intelligence routines. Presentation of the invention

[0012] To do this, the invention relates to a method for automatically detecting and reading the signaling signs of a railway track comprising: - a step of acquiring a 3D point cloud of the track environment from a 3D sensor mounted on a railway vehicle traveling on said track and oriented towards the track environment likely to house traffic signs, - a concomitant step of acquiring an image of said environment of said railway track from an optical sensor on board said railway vehicle running on said track and oriented towards the environment of the track likely to house signal panels, - a step of partitioning said 3D point cloud acquired by the 3D sensor into a plurality of 3D slices each extending perpendicular to the track, - a step of orthogonal projection of the points of each 3D slice onto a plane perpendicular to the track so as to obtain a 2D point cloud of said track environment, - a step of comparing said 2D point cloud to a plurality of predetermined panel shapes, - a step of selecting a panel, called a recognized panel, chosen from said plurality of predetermined panels based on the results of the comparison step, - a step of detecting, on said image acquired by said optical sensor corresponding to said 3D point cloud from which the 2D point cloud corresponding to said recognized panel is derived, the light information provided by said recognized panel.

[0013] The method according to the invention therefore has the particularity of acquiring the environment of the track both by a 3D sensor, such as an active sensor of the lidar, radar, telemeter type or a passive sensor of the stereoscopic system type, and by an optical sensor such as an optical camera. The 3D sensor and the optical sensor simultaneously acquire the same area of ​​the environment of the track.

[0014] The 3D sensor thus provides a 3D point cloud of the track environment which constitutes a 3D imprint of the train environment. This 3D point cloud is projected in slices into 2D point clouds on planes perpendicular to the track. The 2D point cloud is then compared to a plurality of predetermined shapes so as to be able to determine, where appropriate, the type of panel imaged by the 3D sensor.

[0015] Advantageously and according to a variant of the invention, each 2D point cloud is transformed into a continuous 2D shape and this continuous 2D shape is compared to the plurality of predetermined shapes so as to be able to determine, where appropriate, the type of panel imaged by the 3D sensor.

[0016] In other words and according to this variant, the method comprises, after the step of orthogonal projection of the points of each slice making it possible to obtain a 2D point cloud of said environment of the track, a step of constituting a continuous 2D shape of said environment from each projected 2D point cloud, and said step of comparing said 2D point cloud to a plurality of predetermined panel shapes consists of comparing said continuous 2D shape to said plurality of predetermined panel shapes.

[0017] According to the invention, the signs arranged at the edges of the road have different shapes, but the plurality of shapes is known. These traffic signs each contain one or more items of information identifying the driving instruction to be adopted. All of this information can be detected by the method according to the invention. Also, the method according to the invention advantageously takes advantage of this prior knowledge of the shape of the signs which may be present at the edges of the roads by detecting 2D shapes in the data provided by the 3D sensor and by comparing these shapes with the known predetermined shapes of the traffic signs. This comparison makes it possible to classify and recognize the sign imaged by the 3D sensor.

[0018] Once the panel is recognized, the light information provided by the panel is determined by analyzing the optical image acquired by the optical sensor simultaneously with the 3D data from the 3D sensor.

[0019] This information can then be used by the vehicle to adapt its driving. In other words, the information detected by the method according to the invention makes it possible to provide driving instructions to be adopted by the vehicle.

[0020] The method according to the invention thus implements algorithmic routines which may be devoid of artificial intelligence routines, which opens the way to the certification of such a solution. That being said, in other embodiments, solutions based on artificial intelligence could be implemented.

[0021] In particular, the method according to the invention makes recognition of 2D shapes derived from the 3D data acquired by the 3D sensor and the recognized 2D shape can then be verified by analyzing the 3D data to compare them to the known dimensions of the reference panel.

[0022] The invention therefore makes it possible, in summary, to detect and identify in real time a sign arranged at the edges of a track (side edges or gantries overhanging the track) and to detect the area of ​​interest within an image so as to read the luminous information of the sign. It should be noted, however, that in other embodiments and for specific applications (such as for example the verification of the integrity of the road signs of a track) the analysis of the acquired data could be done in a delayed manner and not in real time.

[0023] The method according to the invention can advantageously measure, during the step of selecting the recognized panel, the dimensions of the recognized panel by an analysis of the 3D point cloud provided by the 3D sensor so as to confirm the recognized panel. In particular, the comparison of the 2D shapes makes it possible to form a detection hypothesis by a simple contour detection and the analysis of the 3D points makes it possible to confirm that the detected 2D shape does indeed correspond to one of the panels of the predetermined list of signs detectable around the track.

[0024] According to one embodiment, the method according to the invention can also, in parallel or following this reading of the luminous information from the road signs, process the other information (for example the signs) which can be interpreted by the active sensors. A particular embodiment consists of reading the inscriptions on the signs by Lidar.

[0025] The method according to the invention can advantageously minimize, during the selection step, a difference function between each 2D shape derived from the 3D point cloud and said predetermined panel shapes.

[0026] This advantageous variant makes it possible to obtain an ordered list of the panels potentially recognized in the 3D point cloud.

[0027] The method according to the invention can advantageously measure, during the selection step, the dimensions of the panel recognized by reprocessing the 3D data from the 3D sensor so as to confirm or not the recognition of said panel.

[0028] This variant makes it possible to confirm or deny the recognition of the panel by returning to the 3D data from the 3D sensor. Thus, the detection algorithm which is based on shape recognition is confirmed by a physical measurement of the standardized object.

[0029] The method according to the invention can advantageously comprise, during the selection step, the calculation of a distance, in the mathematical sense of the term, between the recognized shape and the next closest shape and a step of comparing this distance to a predetermined threshold.

[0030] The invention extends to a system for automatically detecting and reading railway track signaling signs comprising: - a 3D sensor (for example an active sensor of the lidar or radar type or a passive sensor of the stereoscopic system type) intended to be mounted on a railway vehicle running on said track and configured to acquire a 3D point cloud of the track environment likely to house signaling signs, - an optical sensor, such as a camera, intended to be mounted on a railway vehicle and configured to acquire an image of said environment of said railway track capable of housing traffic signs, - a module for partitioning said 3D point cloud acquired by said 3D sensor into a plurality of slices each extending perpendicular to the track, - an orthogonal projection module of the points of each slice onto a plane perpendicular to the track so as to obtain a 2D point cloud of said track environment, - a module for comparing said 2D point cloud to a plurality of predetermined panel shapes, - a module for selecting a panel, called a recognized panel, chosen from said plurality of predetermined panels based on the results of the comparison step, - a detection module, on said image acquired by the optical sensor and corresponding to said 3D point cloud from which the 2D point cloud corresponding to said selected panel is derived, of the light information provided by said detected panel.

[0031] The advantages and technical effects of the method according to the invention apply mutatis mutandis to the system according to the invention.

[0032] The 3D sensor and the optical sensor can be mounted outside the train or inside the train. The 3D sensor and the optical sensor are mounted in such a way that they can image the same area of ​​the train's environment. In general, the panels are arranged to the left of the track. Also, the sensors are preferably arranged to the left. However, there are some tracks where the panels are arranged to the right of the track or on gantries that overhang the track. If necessary, sensors are also arranged to the right of the train or towards the front of the train to be able to acquire this environment.It is also possible to provide sensors on the left and / or right of the train and / or towards the front, oriented so as to be able to simultaneously acquire the lateral environments on both sides of the train and towards the front of the train to detect all the panels likely to house signaling panels (right lateral environment, left lateral environment, gantries overhanging the track, etc.).

[0033] Generally speaking, the position of the different sensors is adapted to the known location of the signaling panels (right side, left side, gantry, etc.) depending on the traffic lane taken by the train. The system and method according to the invention can be adapted to all known configurations of the location of the panels.

[0034] It should be noted that the exact location of the panels along the track is not known and does not need to be known for the implementation of the invention. It is only the location of the panels in the environment of the train that needs to be known (left lateral environment, right lateral environment, gantry environment). If this environment of the track is not known, then the system must include sensors oriented in all directions likely to accommodate panels).

[0035] Throughout the text, module or sub-module refers to a software element, a subset of a software program, which can be compiled separately, either for independent use or for assembly with other modules of a program, or a hardware element, or a combination of a hardware element and a software subroutine. Such a hardware element may include an application-specific integrated circuit (better known by the acronym ASIC for the English term Application-Specific Integrated Circuit) or a programmable logic circuit (better known by the acronym FPGA for the English term Field-Programmable Gate Array) or a specialized microprocessor circuit (better known by the acronym DSP for the English term Digital Signal Processor) or any equivalent hardware or any combination of the aforementioned hardware. Generally speaking, a module or a sub-submodule is therefore an element (software and / or hardware) which makes it possible to ensure an algorithmic function.

[0036] According to an advantageous variant of the invention, the system further comprises a module for constituting a continuous 2D shape of the environment of the track from each projected 2D point cloud and said comparison module is configured to compare said continuous 2D shape with said plurality of predetermined panel shapes,

[0037] Advantageously and according to the invention, the selection module of said panel comprises one or more of the following sub-modules: - a sub-module for minimizing a difference function between each constituted 2D shape and said predetermined shapes, - a sub-module for measuring the dimensions of the panel recognized by reprocessing the 3D data from the 3D sensor so as to confirm or not the recognition of said panel, - a sub-module for calculating a distance (in the mathematical sense of the term) between the recognized shape and the next closest shape and a sub-module for comparing this distance to a predetermined threshold.

[0038] The system according to the invention can also be used to compare the technical plan of a track with the panels detected during the movement of the train, which makes it possible to determine whether panels have been added / removed or broken because they were not detected by the signal reading system. This comparison also makes it possible to determine whether the method has made an error by comparing the detections with the track plan.

[0039] According to an advantageous variant of the invention, the 3D sensor is an active sensor of the lidar or radar type or a passive sensor of the stereoscopic imaging system type.

[0040] The advantage of an active sensor of the lidar or radar type is that it is not very sensitive to atmospheric conditions so that the acquisition of data from the track environment can be done whatever the climatic conditions (rain, snow, fog, wind, etc.).

[0041] The invention also extends to a railway vehicle intended to travel on a railway track comprising a plurality of signaling panels, characterized in that the vehicle comprises a system for automatically detecting and reading the signaling panels according to the invention.

[0042] The technical advantages and effects of the method and system according to the invention apply mutatis mutandis to the vehicle according to the invention.

[0043] The invention also relates to a method, a system and a vehicle characterized in combination by all or part of the characteristics mentioned above or below. List of figures

[0044] Other aims, characteristics and advantages of the invention will appear on reading the following description given solely for non-limiting purposes and which refers to the appended figures in which: • [Fig.l] is a schematic view of a railway vehicle equipped with a system for automatically detecting and reading railway track signaling signs according to the invention, • [Fig.2] is a schematic view of the 3D point cloud acquired by a 3D sensor of the system according to an embodiment of the invention and its processing by the system according to the invention, • [Fig. 3] is a schematic view of a plurality of railway signaling panels recognizable by a system according to the invention, • [Fig.4] is a schematic view of a method for automatically detecting and reading railway track signs according to one embodiment of the invention, • [Fig.5] is a schematic view of a processing unit of a detection system according to one embodiment of the invention.

[0045] Detailed description of an embodiment of the invention

[0046] In the figures, the scales and proportions are not strictly respected, for the purposes of illustration and clarity.

[0047] Identical, similar or analogous elements are designated by the same references in all the figures.

[0048] [Fig. 1] schematically illustrates a train 10 running on a railway track 8, the train 10 carrying a system for automatically detecting and reading traffic signs according to one embodiment of the invention. In [Fig. 1], a traffic sign 20 is schematically represented and extends into the lateral environment of the railway track 8.

[0049] The detection system according to one embodiment of the invention comprises a lidar 12 (which forms the 3D sensor) and a camera 14 (which forms the optical sensor) mounted on the train and oriented towards the lateral environment of the track capable of housing traffic signs. This lateral environment of the track corresponds to the lateral zone of track 8 within which the traffic signs for this traffic lane are installed. As indicated previously, the 3D sensor and the optical sensor can be oriented towards another zone of the environment of the track depending on the knowledge of the environment of the location of signs on this track (on the right, on the left and / or on a gantry). According to the embodiment shown, the lidar 12 and the camera 14 can acquire data from the zone in which the traffic signs are arranged and in particular the sign 20.

[0050] The lidar 12 is configured to acquire a 3D point cloud of the lateral environment of the track and the camera 14 is configured to acquire an image of this same lateral environment of the track.

[0051] The data acquired by the lidar 12 and by the camera 14 are then transmitted to a processing unit 16, which is preferably on board the train 10. The data transmissions from the lidar 12 and the camera 14 to the processing unit 16 can be of any type. They can be wired transmissions or wireless transmissions regardless of the communication protocol.

[0052] Wired transmissions can be made up of electrical networks, optical networks, magnetic networks and generally any type of network enabling data to be transmitted.

[0053] Wireless transmissions can be formed from a network of any type, secure or unsecured. Such a wireless network is for example a Wi-Fi network (i.e. according to the IEEE 802.11 standard), but it will be understood that the invention applies to any wireless technology. In particular, other radio wave technologies such as WiMax, Bluetooth, 3G, etc. will be mentioned.

[0054] The wireless network may be secured by a security key. In the case of a Wi-Fi network, the security may be implemented by using one of the solutions known to those skilled in the art, for example WEP, WPA, WPA2, etc. security.

[0055] This processing unit 16, shown schematically in Figures 1 and 5, comprises for example a computer device which must be understood in the broad sense (computer, plurality of computers, virtual server, server networks, etc.). This computer device typically comprises one or more processors 15, one or more memories 17 and a human-machine interface. The processing unit also comprises a database and / or storage means 13 making it possible to save the results of the processing and to access information relating to previous processing.

[0056] The processing unit 16 comprises in particular, and as shown schematically in [Fig.5], a module 16a for partitioning the 3D point cloud acquired by the lidar 12 in a plurality of slices each extending in a plane perpendicular to the track.

[0057] The processing unit 16 also comprises a module 16b for orthogonal projection of the points of each slice onto a plane perpendicular to the track so as to obtain a 2D point cloud of the lateral environment of the track.

[0058] The processing unit 16 also comprises a module 16c for constituting a continuous 2D shape of the lateral environment from each projected 2D point cloud. The continuous 2D shape thus obtained is schematically represented in [Fig.2],

[0059] In other words, the modules 16a, 16b and 16c are configured to provide a set of continuous 2D shapes from the 3D point cloud provided by the lidar 12 mounted on the train.

[0060] [Fig.2] schematically illustrates the different stages of processing of the 3D point cloud by the modules 16a, 16b and 16c. The 3D point cloud acquired by the lidar 12 is represented on the left of [Fig.2] by the reference 22. This 3D point cloud is cut into slices, one slice 24 of which is represented in [Fig.2]. Each slice 24 is then projected onto a plane orthogonal to the track to obtain a continuous shape 26 of the slice, represented on the right of [Fig.2].

[0061] The processing unit 16 also comprises a comparison module 16d of each shape 26 determined by the module 16c with a plurality of predetermined panel shapes. In other words, this module 16d makes it possible to compare each continuous shape derived from the 3D point cloud with predetermined shapes of side signage panels.

[0062] [Fig. 3] schematically illustrates a plurality of shapes 26a, 26b, 26c, 26d, 26e, 26f of predetermined panels. Of course, the invention is not limited to the shapes illustrated in [Fig. 3]. Any shape of panel used for lateral signaling can be integrated into this predetermined list of panels. This predetermined list of panels and the characteristics of each panel are for example saved in the memory 17 and / or the storage means 13 of the computer device housing the processing unit 16.

[0063] The processing unit 16 also comprises a module 16e for selecting a panel, called a recognized panel, chosen from the plurality of predetermined panels based on the results of the comparison from the module 16e.

[0064] This selection module consists for example in selecting the panel whose distance measurement, in the mathematical sense of the term, with respect to each shape extracted from the point cloud is minimum and less than a predetermined threshold. This distance measurement can be of any type. In addition, other comparison methods can be implemented to select the panel closest to the panel imaged by lidar 12.

[0065] Finally, the processing unit 16 comprises a detection module 16f, on the image acquired by the camera 14 and corresponding to the 3D point cloud from which the recognized 2D shape is derived, the light information provided by the detected panel. This module implements image processing routines and consists, for example, of performing contour detection associated with thresholding of the image to determine which light is lit within the panel and what its color is, if applicable.

[0066] For example, assuming that the recognized panel is panel 26a shown in [Fig. 3], the detection module 16f is configured to determine which of the three lights on panel 26a is lit, and possibly its color, so that the train can then take the corresponding driving measures (stopping the train, continuing to drive at normal speed, slowing down, etc.). The position of the lights on each panel is known so that the image processing routine then makes it possible, based on knowledge of the shape of the panel, to detect the areas where the lights are located. A software routine then determines which area of ​​the panel is lit. These image processing routines are within the immediate reach of those skilled in the art and are not described here in detail.

[0067] [Fig.4] schematically illustrates a detection method according to one embodiment of the invention.

[0068] The method is implemented by the system according to the embodiment shown and comprises the following steps: - a step E1 of acquiring a 3D point cloud of the lateral environment of the track from an active sensor of the lidar or radar type mounted on a railway vehicle traveling on said track and oriented towards the lateral environment of the track, - a concomitant step E2 of acquiring an image of said lateral environment of said railway track from a camera on board said railway vehicle traveling on said track and oriented towards the lateral environment of the track, - a step E3 of partitioning said acquired 3D point cloud into a plurality of slices each extending in a plane perpendicular to the track, - a step E4 of orthogonal projection of the points of each slice onto a plane perpendicular to the track so as to obtain a 2D point cloud of said lateral environment of the track, - a step E5 of constituting a continuous 2D shape of said lateral environment from each projected 2D point cloud (as indicated previously, this step is optional), a step E6 of comparing each 2D shape thus formed with a plurality of predetermined panel shapes (in the case where the previous step is not implemented, the comparison step directly compares the projected 2D point cloud with the plurality of predetermined panels), a step E7 of selecting a panel, called a recognized panel, chosen from said plurality of predetermined panels based on the results of the comparison step, a step E8 of detecting, on said acquired image corresponding to said 3D point cloud from which the constituted 2D shape corresponding to said recognized panel is derived, the light information provided by said recognized panel.

[0069] The step of detecting the light information on the panel is configured to detect the area of ​​interest on the image likely to present signaling information.

[0070] Each step of the method according to the invention is implemented by the corresponding module of the detection system according to the invention.

[0071] A system and method according to the invention can thus usefully equip a train with the aim of making it autonomous and being able to detect the traffic instructions provided by the illuminated panels arranged on the lateral areas of the train's running track. These panels are initially intended to be read and interpreted by a driver. The system according to the invention makes it possible to make them readable and interpretable automatically by an autonomous train, without physical modification of the panels.

[0072] It should be noted that the principles of the invention could also be used in the context of the automatic detection of road signs by a motor vehicle, or by an aircraft, drone, robot, etc. in a structured or semi-structured environment and comprising predetermined road signs.

Claims

Claims

1. Method for automatically detecting and reading railway track signaling signs comprising: • a step (El) of acquiring a 3D point cloud of the track environment from a 3D sensor mounted on a railway vehicle (10) traveling on said track and oriented towards the track environment likely to house signaling panels, • a concomitant step (E2) of acquiring an image of said environment of said railway track from an optical sensor (14) on board said railway vehicle (10) traveling on said track and oriented towards the environment of the track likely to house signaling panels, • a step (E3) of partitioning said acquired 3D point cloud into a plurality of slices each extending in a plane perpendicular to the track, • a step (E4) of orthogonal projection of the points of each slice onto a plane perpendicular to the track so as to obtain a 2D point cloud of said track environment, • a step (E6) of comparing said 2D point cloud with a plurality of predetermined panel shapes, • a step (E7) of selecting a panel, called a recognized panel, chosen from said plurality of predetermined panels based on the results of the comparison step, • a step (E8) of detecting, on said image acquired by said optical sensor and corresponding to said 3D point cloud from which said projected 2D point cloud corresponding to said recognized panel is derived, the light information provided by said recognized panel.

2. Method according to claim 1, characterized in that it further comprises, after the step (E4) of orthogonal projection of the points of each slice making it possible to obtain a 2D point cloud of said environment of the track, a step (E5) of constituting a continuous 2D shape of said environment from each projected 2D point cloud, and in that said step (E6) of comparing said point cloud 2D points to a plurality of predetermined panel shapes comprises comparing said continuous 2D shape to said plurality of predetermined panel shapes.

3. Method according to claim 2, characterized in that said step (E7) of selecting said recognized panel comprises a step of minimizing a difference function between each 2D shape constituted and said predetermined shapes.

4. Method according to one of claims 1 to 3, characterized in that said selection step (E7) further comprises measuring the dimensions of the recognized panel by reprocessing the 3D data from the 3D sensor so as to confirm or not the recognition of said panel.

5. Method according to one of claims 1 to 4, characterized in that said step (E7) of selecting the recognized panel further comprises the calculation of a distance, in the mathematical sense of the term, between the recognized shape and the next closest shape and a step of comparing this distance to a predetermined threshold.

6. System for automatically detecting and reading the signaling signs of a railway track comprising: • a 3D sensor intended to be mounted on a railway vehicle traveling on said track and configured to acquire a 3D point cloud of the environment of the track capable of housing signaling signs, • an optical sensor (14) intended to be mounted on a railway vehicle and configured to acquire an image of said environment of said railway track capable of housing signaling signs, • a module for partitioning (16a) said 3D point cloud acquired by said 3D sensor into a plurality of slices extending perpendicular to the track, • a module for orthogonal projection (16b) of the points of each slice onto a plane perpendicular to the track so as to obtain a 2D point cloud of said environment of the track,• a module (16d) for comparing said 2D point cloud to a plurality of predetermined panel shapes, • a module (16e) for selecting a panel, called a recognized panel, chosen from said plurality of predetermined panels, completed from the results provided by the comparison module, • a detection module (16f), on said image acquired by said camera and corresponding to said 3D point cloud from which the 2D point cloud corresponding to said selected panel is derived, of the light information provided by said detected panel.

7. System according to claim 6, characterized in that it further comprises a module (16c) for constituting a continuous 2D shape of said track environment from each projected 2D point cloud and in that said comparison module (16d) is configured to compare said continuous 2D shape to said plurality of predetermined panel shapes.

8. System according to claim 7, characterized in that said selection module (16e) of said panel comprises a sub-module for minimizing a difference function between each 2D shape constituted and said predetermined shapes.

9. System according to one of claims 6 to 8, characterized in that said selection module (16e) comprises a sub-module for measuring the dimensions of the panel recognized by reprocessing the 3D data from the 3D sensor so as to confirm or not the recognition of said panel.

10. System according to one of claims 6 to 9, characterized in that said selection module (16e) comprises a sub-module for calculating a distance, in the mathematical sense of the term, between the recognized form and the next closest form and a sub-module for comparing this distance to a predetermined threshold.

11. System according to one of claims 6 to 10, characterized in that said 3D sensor is an active sensor of the lidar or radar type or a passive sensor of the stereoscopic imaging system type.

12. Railway vehicle (10) intended to travel on a railway track comprising a plurality of signaling panels, characterized in that it comprises a system for automatically detecting and reading the signaling panels of said railway track according to one of claims 6 to 11.

Images