Method and apparatus for estimating maritime traffic conditions

Abstract

Procedure for estimating a maritime traffic condition (20), comprising the following steps: Detecting at least one object (30) using a sensor (7), Detecting the at least one object (30) by means of at least one further sensor (8), wherein the sensors (7, 8) are located at different locations, and wherein the sensor (7) and the at least one further sensor (8) each provide original position information (10, 11) of the at least one object (30) as sensor data, Transmitting the sensor data to at least one fusion device (3), Assigning the original data provided by the sensors Position information (10, 11) of the at least one object (30) relative to each other, Performing a match test on the corresponding position information, Merging the associated position information into a merged position information (12), provided the agreement test is positive, Estimating the state of at least one object (30) using a tracking device (4) based on the fused position information (12), Outputting the estimated object condition of at least one object (30) as the estimated traffic condition (20), where at least one of the sensors (7, 8) is a radar sensor, characterized by the fact that the original position information (10, 11) provided by the radar sensor is derived from a video image (40) provided by the radar sensor.

Description

[0001] The invention relates to a method and a device for estimating a maritime traffic condition.

[0002] The classic method for detecting and estimating the distance of objects within a specific search radius is radar. The basic idea is to transmit electromagnetic pulses via a rotating antenna and measure the reflected electromagnetic response for each angular segment. Radar operates autonomously, i.e., without the assistance of an external service, and is available in a wide range of designs on the maritime market. Its performance is limited by its object resolution and its range. Objects obscured by other ships or other obstacles (such as maritime infrastructure, headlands, etc.) cannot usually be detected by radar.

[0003] The Automatic Identification System (AIS), introduced by the IMO in 2004, serves the purpose of cooperative traffic situation assessment, but it is based on a different approach than radar. Using a broadcast system, individual vessels transmit their position, dynamic, static, and voyage-related data to other traffic users equipped with a corresponding receiving system. The system is based on an on-board Global Positioning System (GPS) receiver, from which the relevant dynamic data is derived. The AIS therefore does not operate autonomously. Furthermore, there is no plausibility assessment of the transmitted data before it is distributed to others. In addition, much of the static and / or voyage-related data can be manually changed, i.e., is fundamentally corruptible.Another disadvantage is that AIS is only mandatory for vessels larger than 300 GRT, meaning the system does not cover all shipping traffic. Another potential weakness of the system is its susceptibility to interference (e.g., deliberate insertion of false information, blocking of data transmission time slots).

[0004] US Pat. No. 7,817,079 B1 discloses a method and device for automatically collecting ship position information. WO 2012 / 017470 A1 discloses a system and device for detecting suspicious vessels.

[0005] The invention is based on the technical problem of creating a method and a device for estimating a maritime traffic condition, in which a maritime traffic condition can be detected and estimated in an improved manner.

[0006] The technical problem is solved according to the invention by a method having the features of patent claim 1 and a device having the features of patent claim 8. Advantageous embodiments of the invention emerge from the subclaims.

[0007] In particular, a method for estimating a maritime traffic condition is provided, comprising the following steps: detecting at least one object by means of a sensor, detecting the at least one object by means of at least one further sensor, wherein the sensors are located at different locations, and wherein the sensor and the at least one further sensor each provide original position information of the at least one object as sensor data, transmitting the sensor data to at least one fusion device, assigning the original position information of the at least one object provided by the sensors to one another, performing a match test on the mutually assigned position information, fusing the mutually assigned position information to form fused position information if the match test is positive,Estimating an object state of the at least one object by means of a tracking device on the basis of the fused position information, outputting the estimated object state of the at least one object as an estimated traffic state.,

[0008] Furthermore, a device for estimating a maritime traffic condition is provided, comprising a receiving device, a fusion device, and a tracking device, wherein the receiving device is configured to receive original position information of at least one object as sensor data from one sensor and original position information of the at least one object as sensor data from at least one further sensor, wherein the sensor and the at least one further sensor are located at different locations, and wherein the fusion device is configured to associate the received original position information of the at least one object with one another, further to carry out a consistency test on the mutually associated position information, and if the consistency test is positive, to merge the mutually associated position information into fused position information,and wherein the tracking device is configured to estimate an object state of the at least one object based on the fused position information and to output the estimated object state of the at least one object as an estimated traffic state.,

[0009] The basic idea of ​​the invention is to detect an object using at least two sensors, with the sensors located at different locations. The sensors each provide original position information relating to the detected object. The original position information of the object is then associated with each other using a fusion device. A consistency test is then performed on the associated position information. The consistency test checks whether the associated position information originates from the same object. If the consistency test is positive, i.e., if it shows that the associated position information originates from the same object, the associated position information is merged into a fused position information item.By fusing the associated original position information, increased accuracy is achieved, which is reflected in the fused position information. The fused position information is then fed to a tracking device. The tracking device estimates the state of the at least one object based on the fused position information. This estimated state of the at least one object is subsequently output as an estimated traffic state.

[0010] According to the invention, at least one of the sensors is a radar sensor. Radar is the classic method for detecting and estimating the distance of objects within a certain search radius and is particularly widespread on ships of a certain size, so that existing infrastructure can be utilized.

[0011] According to the invention, it is further provided that the original position information provided by the radar sensor is derived from a video image provided by the radar sensor. Commercially available radar sensors typically come with a video console on which the sensor data is graphically displayed. This video console provides a video signal that can be used to extract the detected radar echoes from objects within the detection range of the radar sensor and to derive original position information therefrom. The derived original position information comprises, in particular, a position of the at least one object in a coordinate system. The coordinate system of a radar sensor is typically configured in polar coordinates.Together with a reference position from the radar sensor, the position of the object can then be converted into an absolute position in a Cartesian coordinate system, for example, global coordinates, and provided as original position information. This can involve processing the video image and, for example, applying pattern recognition methods to detect objects.

[0012] Objects within the meaning of the present invention are in particular ships, parts of a Vessel Traffic Service (VTS) infrastructure and obstacles.

[0013] Original position information refers to a single piece of position information detected by a sensor or a sensor network, which is not assigned or fused to another piece of original position information. Such original position information comprises position coordinates of the at least one object in a coordinate system. Furthermore, original position information can also comprise a position of the associated sensor. The coordinate system can be based, for example, on polar coordinates or Cartesian coordinates. Original position information can also comprise dimensions of the at least one object and / or an intensity detected by the sensor with respect to the at least one object. Furthermore, the original position information can also comprise properties of the sensor with which the position coordinates were detected. Such properties are, in particular, properties relating to the measurement accuracy of the sensor.In particular, it can also be provided that original position information includes pre-processed data or information. For example, a sensor can already perform a preliminary evaluation to identify and select potential candidates for objects in its detection range.

[0014] Associated position information shall refer to two or more associated original position information.

[0015] A merged position information is a single piece of position information that has been merged from two or more original pieces of position information. In this context, "merging" means that the data underlying the original position information are combined in such a way that a single piece of position information is derived from it.

[0016] The mapping of the original position information to each other can be performed, for example, using the Global Nearest Neighbor (GNN) rule. This rule provides bijective links between the individual original position information based on a distance measure between the individual measurements, i.e., the positions of at least one object in the original position information of the individual sensors.

[0017] An object state is an estimate of at least one position of the associated object. Furthermore, the object state can also include the object's speed, heading, rate of turn, size, or shape, etc.

[0018] An object track refers to the linking of several individual positions of an object into a track. The object track can be described, for example, by a set of position coordinates. Additionally, it can be provided to assign a direction to the object track, which defines a temporal or spatial sequence of the position coordinates in the sense of a trajectory.

[0019] In one embodiment, the method is repeated cyclically. In this way, the at least one object is repeatedly detected, and a current object status is continually estimated. This allows a current maritime traffic status to be continuously estimated and made available.

[0020] Furthermore, it can be provided that not just one object is detected, but a plurality of objects simultaneously. The method is then carried out accordingly on each of the objects. In particular, the tracking device then estimates an object state for each of the objects, which is subsequently also output as part of the estimated traffic state. In this way, it is possible to estimate a traffic state that includes a plurality of objects. A current maritime traffic state can thus be detected and estimated more effectively.

[0021] It can be provided that the tracking device is supplied not only with fused position information of the at least one object, but also with unfused and / or unassigned original position information of other objects. The tracking device then estimates an object state for these objects as well. The output traffic state then also reflects the object states of those objects that were only detected by one sensor or for which no fused position information is available.

[0022] In one embodiment, the object state is estimated using a predictor-corrector filter of the tracking device. Such a filter estimates a future object state based on a model of the object dynamics and the previously fused position information. In the next time step, this estimate is then corrected based on more recent fused position information.

[0023] In a further embodiment, the predictor-corrector filter is an Interacting Multiple Model Joint Probabilistic Data Association (IMM-JPDA) filter. Such a filter comprises a dynamic model for the objects, which in turn comprises two submodels. Within the framework of one submodel, it is assumed that the objects have a constant speed and move on a straight-line path (constant velocity, straight-path motion). The other submodel takes into account a constant turn rate velocity (constant turn rate velocity, turn maneuver motion). The interaction of both submodels allows for accurate estimation of the movements of ships, in particular. Furthermore, individual position information is not necessarily uniquely assigned to an object, but can be used in a weighted form to correct several estimated object states.Each individual estimated object state can, in turn, be corrected using multiple weighted position information. This probabilistic approach is particularly suitable for tracking ships based on radar sensors, which are generally prone to misdetection due to wave swell or certain weather phenomena. A clear assignment of one or more measurements to one or more objects is often not possible.

[0024] In one embodiment, it is further provided that an object track is generated for an object in the tracking device if a generation criterion is met, and an object track for an object in the tracking device is deleted if a deletion criterion is met. For this purpose, the device comprises a track management device. The generation criterion is, for example, a predetermined minimum number of cycles in which the at least one object must be detected. If the method is carried out eight times in succession, for example, the minimum criterion can be that the at least one object must be detected at least seven times in these eight cycles for an object track to be generated.Accordingly, the deletion criterion can be provided that an object track of an object for which an object track exists in the tracking device is deleted if the object is no longer detected in eight or more consecutive cycles. Of course, other generation and deletion criteria can also be provided. In particular, the generation and deletion criteria can also depend on the position information or the estimated object states, for example, the size of the uncertainty range of an estimated object position, a height, an intensity detected by the sensor, etc.

[0025] In one embodiment, it is provided that, before the original position information is transmitted to the fusion device, a transformation of the original position information into a uniform reference system takes place. If, for example, the original position information of one sensor is based on polar coordinates, while the original position information of at least one other sensor is based on Cartesian coordinates, the original position information of one sensor can be converted from polar coordinates to Cartesian coordinates. Both original position information items are then available in Cartesian coordinates. For this purpose, the device comprises, for example, a transformation device that performs the transformation.

[0026] The agreement test can, in particular, be a statistical test in which the original position information from two or more sensors is compared. If the original position information whose agreement is to be checked includes, for example, two statistical distributions that represent an uncertainty in a position determination of the sensors, the statistical test can be used to determine a similarity between the two distributions. If the similarity exceeds a threshold, it is assumed that the statistical distributions originate from the same object.

[0027] In a further embodiment, the Bhattacharyya coefficient is determined as part of the agreement test, and then, as a test criterion, it is checked whether the Bhattacharyya coefficient lies above a predetermined threshold. The Bhattacharyya coefficient is a measure of the overlap between two statistical draws or populations. Accordingly, the method determines an overlap between the statistical distributions of the original position information. If the threshold is then exceeded, it is assumed that the original position information originates from the same object, and the original position information is merged to form a fused position information.

[0028] In one embodiment, it is provided that when estimating the object state of the at least one object, position information from an Automatic Identification System (AIS) is additionally taken into account. In particular, it can be provided that the position information from the AIS is provided as original position information for the at least one object and is then treated like original position information acquired by one of the sensors. The advantage is that, in addition to the sensors, other already established infrastructure can be used. By additionally taking into account the AIS data, an object state or the traffic state can be estimated more accurately.

[0029] In particular, one embodiment can provide for both stationary and non-stationary sensors to be used as sensors. This makes it possible to access all available sensors and, for example, also to consider sensors, e.g., radar sensors, on ships when estimating the traffic situation. The original position information of such non-stationary sensors then additionally includes a current position of the non-stationary sensor, e.g., the current position of the ship, so that an absolute position of an object in a coordinate system, in particular a global coordinate system, can be calculated from the original position information.

[0030] In one embodiment, the associated position information is approximated as normal distributions for merging, and the approximated normal distributions are summarized. As a rule, measurement inaccuracies of sensors can be approximated as normal distributions. Since the positions of maritime traffic are usually two-dimensional positions on the water surface, the position of an object can be uniquely described using two coordinates in the global system. These coordinates are subject to a measurement inaccuracy of the sensor that takes the measurement. The measurement inaccuracy is then described for the two coordinates by a two-dimensional normal distribution. During the merging process, two (or more) two-dimensional normal distributions are combined by summarizing the means and covariances.

[0031] It can be provided that the receiving device for receiving the sensor data comprises at least one interface tailored to the sensors. This interface is configured, in particular, to convert sensor data provided by a sensor into original position information. This can, for example, involve data conversion.

[0032] Otherwise, the embodiments of the device are identical in content to the respective embodiments of the method.

[0033] The invention will be explained in more detail below using preferred embodiments with reference to the figures. Fig. 1 is a schematic representation of an embodiment of the apparatus for estimating a maritime traffic condition; Fig. 2 is a schematic flow diagram of an embodiment of the method for estimating a maritime traffic condition; Fig. 3 a schematic representation of two original position information provided by two radar sensors to illustrate the method; Fig. 4 a schematic representation of the original position information from Fig. 3 and a position information fused therefrom; Fig. 5a a representation of the aspect angle between two radar sensors over time of a series of measurements; Fig. 5b a representation of the norm of the covariance of the merged normal distributions in the merged position information for the measurement series from Fig. 5a; Fig. 5c a representation of the Fig. 5a and Fig. 5b merged data; Fig. 6 shows a comparison between the method according to the invention and a method in which the original position information is not fused; Fig. 7a is a schematic diagram explaining the processing of a video image provided by a radar console to derive original position information; Fig. 7b is a schematic diagram explaining the processing of a video image provided by a radar console to derive original position information; Fig. 7c is a schematic diagram explaining the processing of a video image provided by a radar console to derive original position information.

[0034] In Fig. Figure 1 shows a schematic representation of an embodiment of the device 1 for estimating a maritime traffic condition 20. The device 1 comprises a receiving device 2, a fusion device 3, and a tracking device 4. The receiving device 2 can further comprise two interfaces 5, 6.

[0035] The receiving device 2 receives original position information 10, 11 of an object from a sensor 7 and another sensor 8 as sensor data. The sensors 7, 8 are designed, for example, as radar sensors, with the sensors 7, 8 being located at different locations so that an object is detected by these sensors 7, 8 from different directions. The interfaces 5, 6 ensure that the receiving device 2 can receive and process the sensor data from both sensors 7, 8.

[0036] The receiving device 2 forwards the original position information 10, 11 to the fusion device 3. The fusion device 3 is configured to associate the received original position information 10, 11 of the at least one object with one another. The association is performed, for example, using the global nearest neighbor rule. The fusion device 3 then performs a match test on the associated original position information. As part of this match test, the Bhattacharyya coefficient, for example, can be determined. As a hypothesis, the fusion device 3 then checks whether the Bhattacharyya coefficient exceeds a predetermined threshold. If this is the case, the associated original position information is merged into one fused position information item 12.For example, the positions of the object, approximated as two-dimensional normal distributions, determined by the individual sensors 7, 8, are combined into a single two-dimensional normal distribution. The combined two-dimensional normal distribution is then much more accurate than that of the individual measurements.

[0037] The fused position information 12 is then fed to the tracking device 4. The tracking device 4 is configured to estimate an object state of the at least one object based on the fused position information 12 and to output the estimated object state of the at least one object as an estimated traffic state 20. Provision can be made for the original position information 10, 11 to be fed to the tracking device 4 in addition to the fused position information 12.

[0038] The device 1 operates cyclically, meaning that current original position information 10, 11 is continuously received by the receiving device 2, processed in the fusion device 3 as described, and a current object state or a current traffic state 20 is continuously estimated by the tracking device 4 based on the current fused position information 12. In this way, a current estimate of the traffic state 20 is continuously available.

[0039] In particular, it is provided that the device 1 estimates a traffic state 20 with a plurality of objects. For this purpose, the method is carried out as described for each of the objects detected by the sensors 7, 8. The tracking device 4 then estimates a multi-object state, which is output as the current traffic state 20.

[0040] Furthermore, additional sensors can be provided so that a larger detection area is covered overall.

[0041] In one embodiment, the device comprises a track management device 9. The track management device 9 is configured to generate an object track for an object in the tracking device 4 if a generation criterion is met. Accordingly, the track management device 9 deletes an object track for an object in the tracking device 4 if a deletion criterion is met. The generation criterion is, for example, a predetermined minimum number of cycles in which the at least one object must be detected. If the method is carried out eight times in succession, for example, the minimum criterion can be that the at least one object must be detected at least seven times in these eight cycles for an object track to be generated.Accordingly, the deletion criterion can be provided such that an object track of an object for which an object track is present in the tracking device 4 is deleted if the object is no longer detected in eight or more consecutive cycles. Of course, other generation and deletion criteria can also be provided. In particular, the generation and deletion criteria can also depend on the position information or the estimated object states, for example, on the size of the uncertainty range of an estimated object position, etc.

[0042] In Fig. Figure 2 shows a schematic flow diagram of an embodiment of the method for estimating a maritime traffic condition. In method step 100, a radar sensor acquires a radar image at a time k in method step 101. This radar image is then examined for potential objects in method step 102. If such an object is found, original position information 10 for this object is transferred to a fusion device. If multiple objects are detected, original position information 10 is transferred for each of the objects. In parallel, analogous method steps 103, 104, 105 are performed using another radar sensor.

[0043] In method step 106, the original position information 10, 11 is fused together. For this purpose, the original position information 10, 11 of the object is assigned to one another in a method step 107, for example, using the global nearest neighbor rule. The original position information 21 assigned in this way is then checked for agreement in method step 108. For this purpose, a agreement test is performed. For example, the Bhattacharyya coefficient of the two-dimensional normal distributions of the position coordinates in the two original position information items 10, 11 is determined. Subsequently, a check is performed to determine whether the Bhattacharyya coefficient exceeds a certain threshold. If the test is positive, the original position information 10, 11 is fused to form a fused position information item 12.This is done, for example, by combining the two-dimensional normal distributions by merging the means and the covariance matrices.

[0044] The fused position information 12 is then used in method step 109 by a tracking device to estimate the object's state and output this as an estimated traffic state. For this purpose, the tracking device comprises, for example, an IMM-JPDA filter. In this filter, an object is modeled using a two-part model. Within the framework of one sub-model, it is assumed that the objects have a constant speed and move on a straight-line path (constant velocity, straight-path motion). The other sub-model takes into account a constant turning speed (constant turn rate velocity, turn maneuver motion). This model can be used to accurately estimate the movements of ships, in particular. In method step 110, the filter assigns current fused position information 12 to an object already modeled in the filter.In process step 111, the two submodels for each of the objects are combined. In process step 112, the object state of the object is estimated, and a current traffic state is estimated based on this. This current traffic state is then output (not shown).

[0045] It can be provided that original position information 10, 11 which has not been assigned to other original position information 10, 11 is also supplied to the tracking device as position information 22 which is not assigned to one another, so that this is also available to the filter for estimating the traffic condition.

[0046] In some embodiments, it is provided that, in method step 113, object tracks are managed in a track management device based on the respectively fused position information 12. In method step 114, an object track is generated for an object in the tracking device if a generation criterion is met, and an object track is deleted for an object in the tracking device if a deletion criterion is met. The generated object is then also mapped by the internal state of the filter of the tracking device. The generation criterion is, for example, a predetermined minimum number of cycles in which the at least one object must be detected.Accordingly, the deletion criterion can be set to delete an object track of an object for which an object track exists in the tracking device if the object is no longer detected for a specified number of consecutive cycles. The object to be deleted is then deleted from the internal state of the tracking device's filter.

[0047] In Fig. Figure 3 shows a schematic representation of two original position information items 10, 11 in a Cartesian global coordinate system 13. The two original position information items were provided by two radar sensors (sensors 7, 8). The radar sensors detect an object 30 in their respective detection areas 14, 15. It is clearly visible that the radar echoes are subject to uncertainty, which results from a measurement error of the radar sensors. This is expressed in the fact that the object 30 appears as a banana-shaped lobe. The original position information items provided by these radar sensors each include, for example, the two-dimensional intensity distributions 23, 24 corresponding to the banana-shaped lobes.

[0048] In Fig. Figure 4 is a schematic representation of the original position information 10, 11 (intensity distributions 23, 24) from Fig. 3 and a fused position information item. Here, it is assumed that the original position information items have already been assigned to each other and the agreement test has been positive, so that the two assigned original position information items 10, 11 are to be fused. The banana-shaped lobes of the associated intensity distributions 23, 24 are then approximated, for example, as two-dimensional normal distributions 16, 17 in order to reduce the mathematical effort when merging the two original position information items 10, 11. The two two-dimensional normal distributions 16, 17 are then merged by merging the respective mean values ​​and covariance matrices. This results in a fused position information item 12, which is shown, for example, in the Fig. 4 is shown as another two-dimensional normal distribution. It can be clearly seen that by merging the two original position information items 10, 11, a much more accurate estimate of the object's position is possible, which translates into a more accurate estimate of the mean and a significantly lower covariance of the normal distribution of the fused position information 12. The fused position information 12 is then fed to the tracking device.

[0049] The degree of improvement in the estimate depends in principle on the resulting viewing angle between the two radar systems on the corresponding object (aspect angle). Fig. 3 and Fig. 4 belonging to the measurement series is in the Fig. 5a, Fig. 5b and Fig. 5c shown. Fig. Figure 5a shows the aspect angle between the two radar sensors over time during the measurement series. Related to this, the Fig. 5b the norm of the covariance of the merged normal distributions (fused position information) over time. In Fig. 5c the data from the Fig. 5a and the Fig. 5b. At an aspect angle of 90° between the radar sensors, the value for the norm of the covariance of the merged normal distributions is at a minimum, so that the measurement uncertainty of the radar measurements is minimized at this aspect angle. At an aspect angle of 90° between the two radar sensors, the detection of an object is thus optimal.

[0050] Fig. Figure 6 shows a comparison between the inventive method for estimating a traffic state comprising multiple objects (multi-object state) and a corresponding method in which only the individual original position information is supplied to the tracking device (as a sequential update) rather than the fused position information. The OSPA metric with c = 250 m and p = 2 is used as the metric for comparison. In this metric, the lower the resulting value, the better the result. Fig. 6 it can be clearly seen that the inventive method 18 (“measurement-to-measurement”) performs better than the comparison method 19 (“measurement-to-track”).

[0051] In an embodiment of the method according to the invention, it is provided that the original position information provided by the radar sensor is derived from a video image 40 provided by the radar sensor. Fig. 7a, Fig. 7b and Fig. 7c explain a processing of the video image 40 in this context. The Fig. Figure 7a schematically shows a video image 40, such as that provided by a radar console. This video image 40 includes, in addition to a representation of the actual measurement data from the radar sensor, a graphical user interface 41 in which further data from the radar sensor or a user interface is displayed. Using image processing, this graphical user interface 41 is removed, leaving only the measurement data isolated from it ( Fig.7b). The measurement data available as a graphic image can then be processed, for example, using image processing tools, and pattern recognition methods can be used to identify objects. The position associated with an object 30 detected by means of a pattern recognition method is then transmitted, together with a position of the radar sensor, as original position information to the device according to the invention or, as part of the method, fed to the fusion device. The position can be calculated, for example, using the corresponding position of the identified object 30 in the image and the known position 31 of the radar sensor in the original image center. The advantage of using video images 40 is that a variety of radar sensors can be used, since commercially available radar consoles generally all provide such a video image 40. List of reference symbols 1 device 2 Reception device 3 Fusion facility 4 Tracking device 5 Interface 6 Interface 7 Sensor 8 additional sensors 9 Lane management device 10 original position information 11 original position information 12 merged position information 13 Global coordinate system 14 Detection range 15 Detection range 16 two-dimensional normal distribution 17 two-dimensional normal distribution 18 inventive method 19 settlement proceedings 20 Traffic conditions 21 assigned original position information 22 unassigned original position information 23 Intensity distribution 24 Intensity distribution 30 objects 31 Position of the radar sensor 40 video images 41 graphical user interface 100-115 process steps

Claims

[1] A method for estimating a maritime traffic condition (20), comprising the following steps: Detecting at least one object (30) by means of a sensor (7), Detecting the at least one object (30) by means of at least one further sensor (8), wherein the sensors (7, 8) are located at different locations, and wherein the sensor (7) and the at least one further sensor (8) each provide original position information (10, 11) of the at least one object (30) as sensor data, Transmitting the sensor data to at least one fusion device (3), Assigning the original data provided by the sensors Position information (10, 11) of the at least one object (30) relative to each other, Performing a match test on the associated position information, Merging the associated position information into a fused position information (12) if the match test is positive, Estimating an object state of the at least one object (30) by means of a tracking device (4) on the basis of the fused position information (12), Outputting the estimated object state of the at least one object (30) as an estimated traffic state (20), wherein at least one of the sensors (7, 8) is a radar sensor, characterized by , that the original position information (10, 11) provided by the radar sensor is derived from a video image (40) provided by the radar sensor. [2] Method according to claim 1, characterized by that the estimation of the object state is carried out by means of a predictor-corrector filter of the tracking device (4). [3] Method according to claim 2, characterized bythat the predictor-corrector filter is an IMM-JPDA filter. [4] Method according to one of the preceding claims, characterized by that an object track is generated for an object (30) in the tracking device (4) when a generation criterion is met, and wherein an object track for an object (30) in the tracking device (4) is deleted when a deletion criterion is met. [5] Method according to one of the preceding claims, characterized by that before the original position information (10, 11) is transmitted to the fusion device (3), a transformation of the original position information (10, 11) into a uniform reference system takes place. [6] Method according to one of the preceding claims, characterized bythat the Bhattacharyya coefficient is determined as part of the agreement test and then checked as a test criterion whether the Bhattacharyya coefficient is above a predetermined threshold value. [7] Method according to one of the preceding claims, characterized by that when estimating the object state of the at least one object (30), position information from an automatic identification system is additionally taken into account. [8] Device (1) for estimating a maritime traffic condition (20), comprising: a receiving device (2), a fusion device (3), and a tracking device (4), wherein the receiving device (2) is configured to receive original position information (10, 11) of at least one object (30) as sensor data from a sensor (7) and original position information (10, 11) of the at least one object (30) as sensor data from at least one further sensor (8), wherein the sensor (7) and the at least one further sensor (8) are located at different locations, and wherein the fusion device (3) is configured to associate the received original position information (10, 11) of the at least one object (30) with each other, to carry out a match test on the mutually associated position information, and if the match test is positive, to merge the mutually associated position information into a fused position information (12), and wherein the tracking device (4) is configured to estimate an object state of the at least one object (30) on the basis of the fused position information (12) and to output the estimated object state of the at least one object (30) as an estimated traffic state (20), wherein at least one of the sensors (7, 8) is a radar sensor, characterized by , that the original position information (10, 11) provided by the radar sensor is derived from a video image (40) provided by the radar sensor.

Images