METHOD, DEVICE AND RAILWAY VEHICLE, IN PARTICULAR RAIL VEHICLE, FOR OBSTACLE DETECTION IN RAILWAY TRANSPORT, IN PARTICULAR RAIL TRANSPORT

Description

[0001] Method, device and railway vehicle, in particular rail vehicle, for obstacle detection in railway traffic, especially in rail traffic

[0002] The invention relates to a method for obstacle detection in railway traffic, in particular in rail traffic, according to the preamble of claim 1, a device for obstacle detection in railway traffic, in particular in rail traffic, according to the preamble of claim 6 and a railway vehicle for obstacle detection in railway traffic, in particular a railway vehicle for obstacle detection in rail traffic, according to the preamble of claim 12.

[0003] Rail vehicles, as part of a modern transport infrastructure, are track-bound means of transport that move, for example, by rolling on or under one or two guide rails (tracks), suspended above or below a magnetic field, or suspended from steel cables. Of these track-bound means of transport, rail vehicles based on a wheel-rail system are the most widespread. These vehicles are either self-propelled (multiple units) or pulled or pushed by a locomotive, and predominantly use flanged steel wheels guided on two steel rails or tracks.

[0004] From EP 2 993 105 A2, a method and optical route testing system for examining routes travelled by a vehicle, e.g., a train or a motor vehicle, is known, in which, by obtaining field-of-view image data from a camera mounted on board the vehicle and by pixel- / intensity-based examination of the image data in the vehicle, a route feature of interest, e.g., a measuring distance between route segments, or a specific object on the route, e.g., persons or obstacles, can be identified, and depending on this, a warning signal is generated, and as a result, vehicle control is carried out in such a way that the vehicle is automatically slowed down by manually or automatically applying the brakes.

[0005] German patent DE 10 2014 204 473 A1 discloses a method and assistance system for the automatic driver support of a track-bound vehicle, e.g., a rail vehicle, in which an impending collision is detected when the object is located within a partial volume occupied by the vehicle, and information about the collision is output. This involves detecting an area of ​​interest along the route of the track-bound vehicle and determining a free space within that area where the vehicle occupies the partial volume of the free space. The system then determines whether objects are present in this area with which a collision of the vehicle should be avoided. In this way, it is possible to detect impending object collisions, thereby improving the reliability of object collision detection in vehicles.

[0006] US Patent 2004 / 056182 A1 discloses a scanning laser beam obstacle detection system for rail vehicles, such as a train, comprising a processor, a light detector, and a laser source. The system processes electrical signals from the light detector to determine if an obstacle is present in front of the rail vehicle. The light detector receives light echoes from a scanned laser beam emitted by the laser source and converts these echoes into electrical signals. This improves the safety of rail vehicles and ensures safe railway operations by preventing derailments, reducing the risk of obstacles being run over, and detecting general threats to the train.

[0007] The object of the invention is to provide a method, a device and a railway vehicle, in particular a rail vehicle, for obstacle detection in railway traffic, in particular in rail traffic, with which obstacles in railway traffic are automatically detected when railway vehicles are traveling on railway lines in the railway network, respectively obstacles in railway traffic when railway vehicles are traveling on railway lines in the railway network.

[0008] The automatic detection of obstacles in rail traffic, in particular in rail transport, which is the subject of the present International Patent Application (Application No. PCT / EP2017 / 081834; Publication No. WO 2018 / 104454 A2) and the priority-establishing German Patent Application (Application No. 102016224344.6), is an indispensable necessity with regard to future automated (autonomous) or assisted driving of railway vehicles in rail transport.

[0009] For automated or assisted driving of rail vehicles, it is necessary to detect moving or stationary objects and people within the track area. At the same time, it is necessary to distinguish between permissible objects and people (e.g., buffer stops on the track, maintenance workers next to the track) and impermissible objects and people (e.g., uprooted trees or children playing).

[0010] The problem of automated or assisted driving has so far been solved through costly additional investments in track infrastructure, such as induction loops, trackside computers, and communication systems between the train and track components. Furthermore, special safety fences are used to prevent access to the track (e.g., those found at airports).

[0011] However, it is not only the aspect of automatic obstacle detection that is important for future automated (autonomous) or assisted driving, but also the following technical aspects, all of which are more or less in a technical context with the present patent application and are therefore listed and whose contents must be considered and possibly even included in this context.

[0012] These are the aspects in question: 1) The automatic detection of signals in railway / rail traffic according to the International Patent Application (Application No. PCT / EP2016 / 057804; Publication No. WO 2017 / 174155 A1) and the technical teaching disclosed therein. 2 ) The automatic detection of dangerous situations in rail transport according to the German patent application (application no. 102016224358.6) and the international patent application (application no. PCT / EP2017 / 081841; publication no. WO 2018 / 14460 A1) and the technical teaching disclosed therein. 3 ) The automatic recognition of lanes / tracks in rail traffic according to the German patent application (application no. 102016224335.7) and the international patent application (application no. PCT / EP2017 / 081890; publication no. WO 208 / 104477A1) and the technical teaching disclosed therein. 4) The alternative determination of positions in rail transport when conventional satellite-based positioning fails or is insufficient, according to the German patent application (application no. 102016224355.1) and the international patent application (application no. PCT / EP2017 / 081784; publication no. WO 2018 / 104427 A1) and the technical teaching disclosed therein. 5 ) Performing a track-based image analysis in rail transport according to the German patent application (application no. 102016224331.4) and the international patent application (application no. PCT / EP2017 / 081845; publication no. WO 2018 / 104462 A1) and the technical teaching disclosed therein.

[0013] The aforementioned context-related problem is solved, starting from the obstacle detection method defined in the preamble of claim 1, by the features specified in the characterizing portion of claim 1.

[0014] Furthermore, the aforementioned context-related problem is solved, starting from the obstacle detection device defined in the preamble of claim 6, by the features specified in the characterizing portion of claim 6.

[0015] Furthermore, the aforementioned context-related problem is solved, starting from the railway vehicle defined in the preamble of claim 12, in particular rail vehicle, by the features specified in the characterizing portion of claim 12.

[0016] The idea underlying the invention according to independent claims 1, 6 and 12 consists of using image analysis to identify, based on several images of a track area upstream of a railway vehicle, the lane positioned by the marking in a marked image area that essentially shows a lane used by the railway vehicle, and to compare this with stored known image meta-information or with stored known image meta-information and additional information, and to use an object recognition method to detect in a section of the marked image area whether an object, such as a person, an animal, a fallen tree, etc., is located on the lane, wherein an obstacle in the image area, preferably in the image area section, is marked when the object is detected by the object recognition method.The image metadata, in the literal sense, includes feature and property data of the images captured from the driving route area.

[0017] The basic principle of the invention is to use metadata about the route, e.g. the route profile, in combination with sensors in the railway vehicle as well as calculation and evaluation algorithms to improve the detection of objects and people and to enable the differentiation between permissible and impermissible objects and people.

[0018] The aim is to enable a contribution to fully automated driving without additional investments in track infrastructure.

[0019] Pattern recognition algorithms or pattern matching algorithms are used to detect objects or people on the track or in a critical area next to the track, the area "in front of the vehicle" (in the direction of travel).

[0020] 1. These algorithms are inefficient for detecting objects or people on the track at a great distance because only a small part of the image is relevant.

[0021] 2. These algorithms are unable to distinguish between permissible objects and persons and impermissible objects and persons.

[0022] In the automotive environment, with its focus on roads, it is known according to US 6,405,128 B1 to evaluate a so-called "electronic horizon".

[0023] The automated detection of objects and lanes / tracks, as well as the differentiation between permissible and impermissible objects / persons, can advantageously be achieved, at least partially, through the following steps: A. In a first step, several image recording devices (e.g. sensors) of different types (e.g. video camera, laser sensors, infrared camera, thermal imaging camera, radar equipment, other image acquisition devices, etc.) are used in the railway vehicle to generate images or other information of the railway / railway track in front of the railway / railway vehicle.

[0024] For example, radar for detecting metallic objects, even in bad weather, can be combined with video cameras and image acquisition devices such as thermal imaging cameras for detecting people.

[0025] B.In a second step, the currently used track is marked in the respective image by means of image analysis using external meta-formations. Option 1:

[0026] In a video image or video-like image, this can be achieved using edge detection algorithms starting from the rails directly in front of the vehicle. By using additional information, such as track diagrams, maps, etc., this detection can be made more robust. In this context, reference is made to the implementation of a track-based image analysis in rail transport according to the German patent application (application no. 102016224331.4) and the international patent application (application no. PCT / EP2017 / 081-845; publication no. WO 2018 / 104462 A1) and the technical teaching disclosed therein. Option 2:

[0027] In a radar-based image, this can be done approximately based on knowledge of the route travelled (the track / rail alignment relative to a geographical position is known).

[0028] C. In a third step, the system focuses on the occupied lane / track and uses object recognition methods to detect whether an object or person is present on the lane / track for each image recording device (e.g., an image acquisition device). This means that only the image section containing the occupied lane / track and the critical area to its left and right are considered. Depending on the image acquisition device, one or both of the following pattern-matching methods are used.

[0029] Here too, the integration of external information or additional information increases the recognition accuracy. C1. Positive matching

[0030] The system checks whether the relevant image area contains patterns that correspond to people or objects such as fallen trees or approaching railway vehicles, e.g., trams or trains. If so, an obstacle or potential obstacle is marked. C2. Negative matching

[0031] The system checks whether an expected pattern is detected, such as a solid track or regular rail supports in the image(s). If not, an image database is used to check whether the irregularity was expected (this information can be gathered, for example, during initial test runs with a train driver). If the irregularity was not expected, a potential obstacle is marked.

[0032] D.In a fourth step, the obstacle marking results from the different image acquisition devices are combined. Here, too, the different information sources are combined, for example through the use of probabilistic image processing methods such as hidden Markov models, in order to minimize false detection and eliminate "false negatives," i.e., the erroneous assumption that no object is in the lane / track area, even though it is actually present.

[0033] The analysis of images of the track in front of the train, as outlined above, can achieve the following: Objects and people in the relevant track area can be detected more efficiently than before. Permissible objects and people in the area in front of the train (but outside the traveling track and a critical area to the left and right) can be distinguished from impermissible objects and people on the traveling track or in the critical area to the left and right. Objects and people can be detected more reliably under poor visibility conditions than by the train driver. Train drivers are no longer needed to detect obstacles, so that the traveling track can be detected regardless of their availability.

[0034] In the course of an advantageous further development of the invention, the following additional components - a) to c) for the image recording device (e.g. the image acquisition device) - can be used with regard to the obstacle detection device according to claim 6: a. A correction component that incorporates weather and brightness data into the evaluation of the image material. This allows, for example, in heavy fog, the evaluation of video images to be limited to the first 50 meters in front of the train or rail vehicle, and the vehicle's speed to be reduced accordingly. b.A focal length adjustment component that selects the correct shooting angle depending on the environment (e.g., train station, urban area, countryside, etc.) and speed, thus optimally supporting image analysis. For example, this allows for suitable shooting situations on open track (requiring images from a great distance to allow for timely reactions due to speed) as well as shooting situations in station areas (requiring images with a high width). Additionally, by fusing image data and track data, particularly interesting areas, such as a level crossing, can be focused on. c. A lighting component, for example a spotlight, that operates inside or outside the human visible range, which improves the quality of the image material recorded by the image recording device or image acquisition device at night or in bad weather. d.A land-based evaluation station, connected via mobile network, receives images from an image storage device for which evaluation is only possible with a high degree of uncertainty. These images can then be evaluated by a human expert, and this information can then be fed back into the image storage device, which can be located either within the train / rail vehicle (option "A") or outside the obstacle detection device, e.g., as a storage database within the train / rail vehicle or as a data cloud. 1. With sufficient communication bandwidth and the availability of human experts, this can even be done in real time, such that the result of the evaluation can be used to control the train / rail vehicle. 2.Furthermore, the image data from rail vehicles of one or more fleets can be compared and distributed via the land-based evaluation station. e. A mobile device belonging to a train conductor or similar railway employee who is already traveling on the rail vehicle for passenger handling purposes and evaluates images with a high uncertainty factor, similar to d).

[0035] Furthermore, it is possible that an obstacle detection device is designed and functions as a virtual machine in the sense of a "Software Defined Signal Recognition of Rail Traffic Systems".

[0036] Further advantages of the invention will become apparent from the following description of an exemplary embodiment of the invention based on the FIGURES 1 and 2 These show: FIGURE 1 a railway vehicle-based detection of an obstacle in the form of a tree that has fallen across a railway line, FIGURE 2a basic structure of an obstacle detection device for the purpose of which FIGURE 1 Railway vehicle-based obstacle detection in the form of a tree that has fallen on the railway line.

[0037] FIGURE 1 BVK shows a train vehicle-based detection of an obstacle in rail traffic when, on a section-by-section railway line BST of a railway network BNE, a train vehicle BFZ approaches an object OBJ located as an obstacle on a track FS of the railway line BST, in the case shown a tree that has fallen onto the track FS.

[0038] According to the present embodiment, the track-related railway line BST of the railway network BNE is a railway line SST of a railway network SNE, on which, in rail traffic SVK, a rail vehicle SFZ travels on track GL for rail vehicle-based obstacle detection and approaches the object OBJ located on track GL as an obstacle, in the illustrated case the tree that has fallen onto track GL. As discussed at the beginning, any other arbitrary short- or long-distance rail transport system can be conceived as a further embodiment of the invention, instead of the illustrated rail traffic SVK with the rail vehicle SFZ traveling on the railway line SST of the railway network SNE. For example, a magnetic levitation train transport system (e.g., Transrapid, Maglev) with a correspondingly comparable infrastructure consisting of a railway network, railway line, and rail vehicle would also be suitable.

[0039] In the FIGURE 1 The depicted rail transport system includes a railcar TRW of the rail vehicle SFZ with a driver's cab TFS and an integrated display unit AZE, in which the driver's workplace FZF is located, and an obstacle detection device HEV for rail-based obstacle detection. The obstacle detection device HEV includes an image recording device BAZG, preferably designed as a sensor, which is configured, for example, as a conventional video camera, laser sensor, thermal imaging camera, radar device, infrared camera, etc., and is also referred to as an image acquisition device due to its image acquisition function.

[0040] With the image recording device BAZG, when the rail vehicle SFZ traveling on track GL approaches the object OBJ located on track GL as an obstacle, in the illustrated case the fallen tree, a multitude of images BI FSB representing the track area FSB can be captured from the rail vehicle SFZ, e.g. from the perspective of the train driver FZF in the driver's cab TFS of the train TRW and / or from a stationary, track-observing position in or on the vehicle SFZ, from a track area FSB located in front of the rail vehicle SFZ, preferably oriented to the speed of the rail vehicle SFZ.

[0041] The images BI FSB of the track area FSB contain an image area BIB with an image area section BIB AS, which represents the track GL in use as well as a critical area for rail traffic SVK. This section defines a critical radius for rail traffic SVK, essentially to the left and right of track GL, within the part of the track area FSB shown by image area BIB of the images BI FSB of the track area FSB. In other words, the track area FSB also includes the critical area for rail traffic SVK.

[0042] How obstacle detection is now carried out based on the images BI FSB of the track area FSB with the contained image area BIB and the image area section BIB AS is described below. FIGURE 2 explained.

[0043] FIGURE 2 shows the basic structure of the HEV obstacle detection device for the vehicle according to the FIGURE 1Rail vehicle-based obstacle detection of the rail vehicle SFZ, which is traveling on track GL and is approaching the object OBJ located as an obstacle on track GL, in the case shown the fallen tree.

[0044] The starting point for obstacle detection is, according to the explanations regarding the FIGURE 1 the image recording device BAZG, which captures the images BI FSB of the driving area FSB for obstacle detection.

[0045] The image recording device BAZG is preferably designed to be swivelled for alignment with the image object.

[0046] Furthermore, it is possible, and potentially also advantageous for data acquisition reasons, to integrate several image recording devices (BAZG) of the same type, e.g., multiple video cameras, or devices of different types, e.g., multiple video cameras, laser sensors, radar-based sensors, sensors based on radio-based positioning and distance measurement, infrared cameras, and / or thermal imaging cameras, into the obstacle detection device (HEV) to capture the images. Such multiple image recording or acquisition methods can be relevant, among other things, for redundancy purposes.

[0047] To further improve the quality of the images recorded or acquired with the BAZG image recording device, the BAZG image recording device preferably includes the following components: 1. A correction component, KOK, which incorporates weather and brightness data into the evaluation of the image material. With this component, it is possible, for example, to limit the evaluation of video images to the first 50 meters in front of the train in heavy fog and to reduce the train's speed accordingly. 2.A focal length adjustment component (BVK) selects the correct shooting angle depending on the environment (e.g., train station, urban area, countryside, etc.) and speed, thus optimally supporting image analysis. This allows for suitable shooting situations both on open track (requiring images from a great distance to allow for timely reactions due to speed) and in station areas (requiring images with a high width). Additionally, by fusing image data and track data, particularly interesting areas along the SST railway line within the SNE rail network can be focused on, such as a level crossing. 3.A lighting component BLK, which is designed, for example, as a spotlight that operates inside or outside the human visible range, and which improves the quality of the image material recorded by the image recording device or the image acquisition device BAZG at night or in bad weather.

[0048] The images thus captured are stored by the image recording device BAZG in an image storage device BSPE. This image storage device BSPE is either connected to the image recording device BAZG as a component of the obstacle detection device HEV, according to option "A", or, according to option "B", located outside the obstacle detection device HEV, e.g., as a storage database, in the railcar or in a data cloud, and assigned to or connectable with the image recording device BAZG.

[0049] For the evaluation of the recorded or acquired images to identify objects that pose obstacles to rail traffic along the railway line, e.g., a tree that has fallen onto the track according to the FIGURE 1The image recording device BAZG is connected to a calculation / evaluation unit BAWE, which is also a component of the obstacle detection device HEV. For this purpose, the calculation / evaluation unit BAWE, like the image recording device BAZG, is either connected to the image storage unit BSPE according to option "A" or assigned to or connectable to the image storage unit BSPE according to option "B". In this way, a functional subunit is created consisting of the calculation / evaluation unit BAWE, the image recording device BAZG, and the image storage unit BSPE, in which the aforementioned components of the obstacle detection device HEV partially cooperate for calculation / evaluation-based obstacle detection.

[0050] To form a complete functional unit for calculation- / evaluation-based obstacle detection, in which the participating subunits interact functionally, the aforementioned functional subunit is extended by a further subunit, an information database (IDB). The information database (IDB) can, for example, be integrated with the image storage device (BSPE) as a single structural unit in a common storage device. This in the FIGURE 2The storage device not explicitly shown can, in turn, like the image storage device BSPE, either be connected as a component of the obstacle detection device HEV to the image recording device BAZG and the calculation / evaluation device BAWE, according to option "A", or be assigned to or connectable to the image recording device BAZG and the calculation / evaluation device BAWE outside the obstacle detection device HEV in the railcar or in a data cloud, according to option "B". In this context, reference is made to the information storage device in the German patent application (application no. 102016224355.1) and the corresponding international patent application (application no. PCT / EP2017 / 081784; publication no. WO 2018 / 104427 A1) for the alternative determination of positions in rail transport when conventional satellite-based positioning fails or is insufficient.

[0051] In addition to image metadata (BMI), which literally contains characteristic and property data of the route area FSB captured in the images (BI FSB), the information database IDB also stores supplementary information (ZI), such as route plans or maps, etc. According to the representation in the FIGURE 2 The information database IDB of the obstacle detection device HEV is assigned to or connected to it in such a way that the calculation / evaluation unit BAWE accesses the image metadata BMI and additional information ZI stored in the information database IDB for calculation / evaluation-based obstacle detection. The information database IDB is preferably located outside the obstacle detection device HEV, e.g., as a database within the railcar, or is designed as a data cloud.

[0052] For calculation-based / evaluation-based obstacle detection, the calculation-based / evaluation unit BAWE preferably has a non-volatile, readable memory SP in which processor-readable control program instructions of a program module PGM controlling obstacle detection are stored, and a processor PZ that executes the control program instructions of the program module PGM for calculation-based / evaluation-based obstacle detection. In addition to accessing the image metadata BMI and the supplementary information ZI in the information database IDB, the processor PZ also accesses the image recording device BAZG and the image storage device BSPE for control purposes and to read data.

[0053] The BAWE calculation / evaluation unit or the PGM program module with the processor PZ, which executes the control program commands of the PGM program module for calculation / evaluation-based obstacle detection, is now designed with regard to calculation / evaluation-based obstacle detection in such a way that in the BI FSB images, the image area BIB is marked which shows the track GL used by the rail vehicle SFZ, whereby the track GL of the rail vehicle SFZ, which is positioned image-wise by the marking, is recognized by an image analysis and is compared either with the stored known image meta-information BMI or with the stored known image meta-information MMI and the additional information ZI.

[0054] The image analysis and thus the marking is preferably carried out using edge detection algorithms, in which, starting from the track GL recorded in the track area FSB, the course of the track GL used by the rail vehicle SFZ is recognized in the image area BIB by a changing image component of the track GL in the recorded image relative to the overall image.

[0055] Furthermore, preferably when the BI FSB images are taken with RADAR-based sensors based on radio-based positioning and distance measurement, the image analysis and thus the marking is carried out on the basis of knowledge of the track GL used, because the course of the track GL used is known relative to a geographical position.

[0056] If the track GL of the rail vehicle SFZ, positioned by the marker, is recognized by the image analysis and compared either with the stored known image meta-information BMI or with the stored known image meta-information BMI and the additional information ZI, then for the image area section BIB AS of the marked image area BIB, which represents the used track GL and the area critical for rail traffic SVK, an object recognition method is used to determine whether an object OBJ, such as a person, an animal, a fallen tree, etc., is located on the track GL, whereby an obstacle in the image area BIB, e.g., if it is located in the image area section BIB AS and / or if it is a potential obstacle, is marked if the object OBJ is recognized by the object recognition method.

[0057] The object recognition method performs a pattern comparison based on a positive comparison and / or negative comparison, in the case of a positive comparison it is checked whether object-specific patterns are contained in the image area section BIB AS and in the case of a negative comparison it is checked whether an expected pattern is contained in the image area section BIB AS, e.g. the solid track GL used by the rail vehicle SFZ or a regularity formed by track beams of track FS or track beams between the parallel tracks GL.

[0058] If this check results in a "NO" in the negative comparison, the detected irregularity is compared with reference information from previously recorded route images used in route initialization runs, and if the irregularity was not expected, an obstacle in the image area BIB, e.g. in the image area section BIB AS and / or as a potential obstacle, is marked.

[0059] The obstacle markings applied to all images BI FSB, specifically within the image area BIB or the image area section BIB AS, are preferably combined using image processing methods such as hidden Markov models to combine the different image sources. This minimizes the probability of false detection and prevents "false negatives," i.e., incorrect assumptions that no object is present in the lane or track area when it actually is.

[0060] Furthermore, for the obstacle detection device HEV with the integrated or associated image storage unit BSPE, a land-based evaluation station AWS is provided for images in the image storage unit BSPE for which evaluation is only possible with a high degree of uncertainty. This AWS is connected to the image storage unit via mobile network and receives the stored images from the unit for modified evaluation. These images can then be evaluated by a human expert, and this information can then be fed back into the image storage unit BSPE. 1. With sufficient communication bandwidth and the availability of human experts, this can even be done in real time, such that the evaluation results can be used to control the train / rail vehicle. 2. Furthermore, the image data from rail vehicles of one or more fleets can be compared and distributed via the land-based evaluation station AWS.

[0061] As an alternative to the AWS evaluation station for the modified evaluation of images, for which evaluation is only possible with a high degree of uncertainty, it is also possible for a train driver or a comparable railway employee, who is already on the train for passenger handling purposes, to evaluate images with a high degree of uncertainty using a mobile device, just as the human expert does with regard to the images in the AWS evaluation station.

[0062] With the obstacle detection device HEV described above, automated (autonomous) or assisted driving of the rail vehicle BFZ or the rail vehicle SFZ along a route can be assisted or even realized without additional infrastructure. This is particularly the case when the obstacle detection device HEV is implemented as a virtual machine that is designed and functions as a "Software Defined Signal Recognition of Rail Traffic Systems".

Claims

1. Method for obstacle recognition in railway traffic (BVK), in particular in rail traffic (SVK), characterized in that a) from a railway vehicle (BFZ), in particular a rail vehicle (SFZ), in particular from the perspective of a tractive unit driver (FZF, TFS, TRW) and / or from a stationary, track-observing position in or on the vehicle (BFZ, SFZ), a multiplicity of images (BIFSB) of a route region (FSB) ahead of the railway vehicle (BFZ, SFZ), and in particular oriented to the speed of the railway vehicle (BFZ, SFZ), said images representing the route region (FSB), are captured, b) an image region (BIB) is marked in each of the images (BIFSB), which image region shows a track (FS) used by the railway vehicle (BFZ, SFZ), in particular a rail track (GL), wherein the track (FS, GL) of the railway vehicle (BFZ, SFZ), which track is visually positioned by the marking, is recognized by means of an image analysis and is compared either with stored known image meta information (BMI) or with stored known image meta information (BMI) and additional information (ZI), such as e.g. route schedules or map material, c) for an image region segment (BIBAS) of the marked image region (BIB) representing the used track (FS, GL) and a region which is critical for railway traffic (BVK, SVK), an object recognition method recognizes whether an object (OBJ), such as e.g. a person, an animal, a fallen tree, is situated on the track (FS, GL), wherein an obstacle is marked in the image region (BIB), preferably in the image region segment (BIBAS) and / or as a potential obstacle, when the object (OBJ) is recognized by the object recognition method.

2. Method according to Claim 1, characterized in that the images (BIFSB) are recorded by a plurality of image recording appliances (BAZG) of varying design, e.g. by video cameras, laser sensors, RADAR-based sensors based on radio-based locating and distance measurement, infrared cameras, and / or thermal imaging cameras.

3. Method according to Claim 2, characterized in that if the images (BIFSB) are recorded by RADAR-based sensors based on radio-based locating and distance measurement, the image analysis is carried out on the basis of the knowledge of the used track (FS, GL), since the course of the used track (FS, GL) relative to a geographical position is known.

4. Method according to any of Claims 1 to 3, characterized in that the object recognition method carries out a pattern comparison based on a positive comparison and / or a negative comparison, wherein a) the positive comparison involves checking whether the image region segment (BIBAS) contains object-specific patterns, and b) the negative comparison b1) involves checking whether the image region segment (BIBAS) contains an expected pattern, preferably the continuous track (FS, GL) used by the railway vehicle (BFZ, SFZ) or a regularity formed by track supports of the track (FS) or rail track supports between the rail tracks (GL) running parallel, b2) for the case where the checking of the result ends with a "NO", the ascertained irregularity is compared, with regard to its expectation, with route images used as reference information and previously recorded in route initialization passes, b3) for the case where the irregularity was not expected, an obstacle is marked in the image region (BIB), preferably in the image region segment (BIBAS) and / or as a potential obstacle.

5. Apparatus (HEV) for obstacle recognition in railway traffic (BVK), in particular in rail traffic (SVK), characterized in that a) there is at least one image recording appliance (BAZG) with which, from a railway vehicle (BFZ), in particular a rail vehicle (SFZ), in particular from the perspective of a tractive unit driver (FZF, TFS, TRW) and / or from a stationary, track-observing position in or on the vehicle (BFZ, SFZ), a multiplicity of images (BIFSB) of a route region (FSB) ahead of the railway vehicle (BFZ, SFZ), and in particular oriented to the speed of the railway vehicle (BFZ, SFZ), said images representing the route region (FSB), are capturable and storable in an image storage device (BSPE), b) there is a calculation / evaluation device (BAWE) designed to be connected to and functionally cooperating with the image recording appliance (BAZG), the image storage device (BSPE) and an information database (IDB), wherein preferably both, the image storage device (BSPE) and the information database (IDB), are integrated as a structural unit in a common storage apparatus, in such a way, in particular with a non-volatile, readable memory (SP), in which processor-readable control program instructions of a program module (PGM) controlling the obstacle recognition are stored, and a processor (PZ), which executes the control program instructions of the program module (PGM) for calculation- / evaluation-aided obstacle recognition, that an image region (BIB) is marked in each of the images (BIFSB), which image region shows a track (FS) used by the railway vehicle (BFZ, SFZ), in particular a rail track (GL), wherein the track (FS, GL) of the railway vehicle (BFZ, SFZ), which track is visually positioned by the marking, is recognized by means of an image analysis and is compared either with stored known image meta information (BMI) or with stored known image meta information (BMI) and additional information (ZI), such as e.g. route schedules or map material, c) the calculation / evaluation device (BAWE) is designed in such a way that, for an image region segment (BIBAS) of the marked image region (BIB) representing the used track (FS, GL) and a region which is critical for railway traffic (BVK, SVK), an object recognition method recognizes whether an object (OBJ), such as e.g. a person, an animal, a fallen tree, is situated on the track (FS, GL), wherein an obstacle is marked in the image region (BIB), preferably in the image region segment (BIBAS) and / or as a potential obstacle, when the object (OBJ) is recognized by the object recognition method.

6. Apparatus (HEV) according to Claim 5, characterized in that it contains a plurality of image recording appliances (BAZG) of varying design, e.g. a plurality of video cameras, laser sensors, RADAR-based sensors based on radio-based locating and distance measurement, infrared cameras, and / or thermal imaging cameras, which record the images (BIFSB).

7. Apparatus (HEV) according to Claim 6, characterized in that the calculation / evaluation device (BAWE) is designed in such a way that, if the images (BIFSB) are recorded by RADAR-based sensors based on radio-based locating and distance measurement, the image analysis is carried out on the basis of the knowledge of the used track (FS, GL), since the course of the used track (FS, GL) relative to a geographical position is known.

8. Apparatus (HEV) according to any of Claims 5 to 7, characterized in that the calculation / evaluation device (BAWE) is designed in such a way that the object recognition method carries out a pattern comparison based on a positive comparison and / or a negative comparison, wherein a) the positive comparison involves checking whether the image region segment (BIBAS) contains object-specific patterns, and b) the negative comparison b1) involves checking whether the image region segment (BIBAS) contains an expected pattern, preferably the continuous track (FS, GL) used by the railway vehicle (BFZ, SFZ) or a regularity formed by track supports of the track (FS) or rail track supports between the rail tracks (GL) running parallel, b2) for the case where the checking of the result ends with a "NO", the ascertained irregularity is compared, with regard to its expectation, with route images used as reference information and previously recorded in route initialization passes, b3) for the case where the irregularity was not expected, an obstacle is marked in the image region (BIB), preferably in the image region segment (BIBAS) and / or as a potential obstacle.

9. Apparatus (HEV) according to any of Claims 5 to 8, characterized in that the image recording appliance (BAZG) is designed in pivotable fashion.

10. Railway vehicle (BFZ) for obstacle recognition in railway traffic (BVK), in particular a rail vehicle (SFZ) for obstacle recognition in rail traffic (SVK), characterized in that an apparatus (HEV) for obstacle recognition according to any of Claims 5 to 9 is integrated into the railway vehicle (BFZ, SFZ).