Location-specific traffic analysis with traffic path detection

Abstract

Device for location-based traffic analysis, comprising an arrangement of at least one radar sensor (10) on a traffic space (20), and an evaluation unit (12) configured to determine object trajectories (34) from temporal sequences of object positions (32) of respective objects (30) detected by the radar sensor (10) and moving in the traffic space (20), to recognize the position of at least one traffic path (24; 40; 42; 44; 46) based on a cluster (36) of object trajectories (34), and to assign further detected objects (30) to a respective traffic path (24; 40; 42; 44; 46) whose position has been recognized. wherein the evaluation unit (12) is configured to assess the traffic situation associated with a lane with regard to the risk of congestion, based on the recorded object speeds of several objects (30) assigned to the same traffic path (40) in the form of a lane, and / or the evaluation unit (12) is set up to perform a summary evaluation of data for several objects (30) assigned to the same traffic path (40) in order to determine at least one traffic parameter from: an average speed (v i,avg ), a minimum speed (v i,min ), a minimum distance between the objects (30), an average distance between the objects (30), a maximum distance between the objects (30), a traffic flow (p i ) and a traffic density (d i ).

Description

[0001] The invention relates to a device for location-based traffic analysis, comprising an arrangement of at least one radar sensor in a traffic area and an evaluation unit. STATE OF THE ART

[0002] To detect vehicles in a lane before a traffic light, induction loops (also known as contact loops) embedded in the road surface are known. These loops detect the presence of a vehicle by changing its inductance when it passes over them. However, laying the wire loop in the road surface is complex and time-consuming. Furthermore, an induction loop can only detect sufficiently large metallic objects.

[0003] Furthermore, light barrier measuring devices are known in which a light barrier is aligned perpendicular to the roadway. A vehicle crossing the light barrier can thus be detected. A transmitter must be positioned on one side of the roadway, while a reflector or receiver is positioned on the other side. If a light barrier extends across a roadway with multiple lanes, it is not readily possible to distinguish which lane a detected object is in.

[0004] DE 10 2007 032 091 B3 describes a method for monitoring a level crossing, in which cameras directed in one direction and cameras directed in the opposite direction first capture an image of an object located on a road section on one side of a danger zone, and by comparing this image with the image from another camera, the object is identified in a road section on the other side of the danger zone. This is intended to enable a clearance signal for the level crossing.

[0005] DE 10 2006 040 542 A1 describes a device for monitoring a level crossing with a video camera, the signal of which is transmitted to a train of vehicles by means of radio signals. This is intended to enable a train driver to recognize the area of ​​the level crossing even though there is no visual contact yet.

[0006] DE 196 12 579 A1 describes an arrangement for monitoring a danger zone at level crossings with full barriers, using a rotating radar rangefinder that scans the danger zone horizontally. Reference markers are arranged on the boundary of the danger zone to limit the scanning to the area within these markers. Alternatively, sector elements of the danger zone can be stored according to length and angle, and the scanning of the danger zone can be limited electronically to the stored sector elements.

[0007] In the well-known method of monitoring hazardous areas using a rotating radar rangefinder, the configuration of the area to be monitored must be elaborately adapted to the existing hazardous area.

[0008] DE 10 2011 113 019 A1 describes a method for identifying and assessing the hazards of a traffic situation involving at least two road users at a road intersection. To determine the magnitude of the hazards, a probability-based method is used to cognitively assess potential and actual hazards of the traffic situation. This involves, in a first step, a probabilistic interpretation of the traffic situation, in a second step a probabilistic hypothesis estimation of basic hypotheses for estimating future trajectories of the road users before crossing the intersection, and in a third step a probabilistic hazard assessment of the hazards for pairs of road users in relative motion to each other.

[0009] EP 2 011 103 B1 describes a traffic control system with a sensor. Traffic is detected using a signal from a wave reflection detection device in order to classify road users into different types. REVELATION OF THE INVENTION

[0010] The object of the invention is to create a device for location-based traffic analysis that allows for a more reliable and accurate traffic analysis than known systems.

[0011] This problem is solved according to the invention by a device for location-based traffic analysis according to claim 1, comprising an arrangement of at least one radar sensor on a traffic area, and an evaluation unit which is configured to determine object trajectories from temporal sequences of object positions of respective objects moving in the traffic area detected by the radar sensor, to recognize the position of at least one traffic path based on a cluster of object trajectories, and to assign further detected objects to a respective traffic path whose position has been recognized.

[0012] The traffic path in question can be, for example, a roadway, a footpath, or a track. In the case of a traffic path for a means of transport, such as vehicles or trains, it is single-track and thus corresponds to a roadway or a track.

[0013] By being designed to detect the location of a traffic path, the device is particularly easy to install. This allows the location of a traffic path to be detected automatically, without requiring precise alignment or positioning of the radar sensor relative to the traffic area being monitored. By assigning the detected objects to a recognized traffic path, objects belonging to the same path can be analyzed more accurately. For example, objects moving outside a traffic path can be excluded from the traffic analysis. Assigning each detected object to its respective traffic path thus enables a traffic analysis specific to that path. For instance, if the locations of multiple lanes are determined, a traffic analysis can be performed separately for each lane.Therefore, it is not necessary to specify the number of lanes. Instead, the individual lanes can be detected based on the recorded object movements.

[0014] An object trajectory is determined from the temporal progression of the object positions of a detected object. For example, determining object trajectories can consist of combining temporal sequences of object positions of respective objects moving in the traffic area and detected by the radar sensor into object trajectories.

[0015] The problem is further solved by a corresponding method for traffic analysis according to the dependent patent claim using location data from at least one radar sensor fixed to a traffic area, comprising the steps: - Determining object trajectories from temporal sequences of object positions of respective objects moving in the traffic area and detected by the radar sensor; - Identifying the location of at least one traffic route based on a cluster of object trajectories, and - Assigning each object detected by at least one radar sensor to a respective traffic path whose location has been determined. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] A preferred embodiment of the invention is explained in more detail below with reference to the drawing.

[0017] They show: Fig. 1 a schematic top view of a traffic area with a device for location-based traffic analysis; Fig. 2. An exemplary representation to explain how to determine an object trajectory; Fig. 3. A schematic representation for determining a characteristic trajectory of a cluster of trajectories; and Fig. 4 A schematic representation of the location and type of lanes and footpaths detected by the device. DETAILED DESCRIPTION OF AN EXAMPLE OF EXECUTION

[0018] Fig. Figure 1 schematically shows a radar sensor 10 and an evaluation unit 12 connected to or integrated into the radar sensor 10. The radar sensor 10 is an FMCW radar sensor with a transmit / receive device comprising a patch antenna array and has a schematically depicted field of view 14, which covers an azimuth angle range of at least 90°. Several antenna elements are arranged horizontally offset from one another. Preferably, the field of view 14 covers an azimuth angle range of at least 160°.

[0019] The radar sensor 10 is located at the edge of a traffic area 20, which is covered by the field of view 14.

[0020] The traffic area 20 includes a road 22 and can, for example, contain a railway line with tracks 24 and / or a pedestrian path 26 ( Fig. 2) include.

[0021] In Fig. Figure 1 also schematically shows barriers 28 of a level crossing.

[0022] Fig. Figure 2 shows an example situation with an object 30 in the form of a vehicle detected by the radar sensor 10. The evaluation unit 12 receives data on object positions 32 for the detected object 30, several of which are located in the traffic area 20. Fig. 2 are shown. Furthermore, the evaluation unit 12 receives data on the object velocity of a detected object 30 in a manner known per se, in particular the relative velocity with respect to the stationary radar sensor 10.

[0023] Optionally, the evaluation unit 12 can receive further object data for a detected object 30, such as object reflection data, for example, backscattering power, horizontal extent, and / or the height of the object 30. Height information can be determined, for example, from elevation-angle-dependent reflection data of the object 30. Horizontal extent can be determined, for example, by assigning several reflection centers to an object 30.

[0024] An object trajectory 34 is determined from the temporal progression of the determined object position data 32.

[0025] The specified object trajectories 34 are stored.

[0026] Fig. Figure 3 schematically shows clusters 36 of spatially close object trajectories 34. These clusters 36 are determined. For this purpose, for example, groups of similar or approximately spatially corresponding trajectories 34 are formed from object trajectories 34 determined for several objects 30 based on their respective similarities. For this purpose, for example, a characteristic trajectory 38 can be determined for each object. The characteristic trajectory 38 is optionally smoothed or straightened and is then determined as the location of a traffic path.

[0027] The type and / or width of the traffic path can be differentiated, for example, based on spatial dispersion within a cluster of trajectories 34, determined object dimensions, and / or other object backscatter properties. Trajectories 34 can optionally be distinguished based on their direction of movement. Unidirectional traffic paths, such as lanes and tracks, can be differentiated from bidirectional traffic paths, such as footpaths.

[0028] For example, during a learning phase of the device, traffic paths can be determined according to the procedure described below.

[0029] The detected objects 30 are classified and assigned, for example, to the groups vehicle, person, and, if a railway line 24 is present, train. The classification can be based on the radar backscatter characteristics, the size, and / or the height of the detected objects.

[0030] The object positions of the recorded objects 30 are determined over a period of time.

[0031] From the temporal changes of the object positions assigned to an object 30, a trajectory 34 of object 30 is determined.

[0032] By determining an object trajectory 34 for each of several objects 30, random errors of a single trajectory determination can be compensated.

[0033] For each object 30, for example, object properties such as an extent and / or a height are determined.

[0034] For each group of objects, characteristic trajectories 38 and optionally characteristic properties are determined. Thus, in the Fig. For example, on the road 22 shown in Figure 2, two characteristic trajectories 38 can be determined for the group "vehicle" corresponding to the two carriageways of road 22. Similarly, for example, two characteristic trajectories 38 can be determined for the group "person", which correspond to the courses of the respective pedestrian paths 26 on both sides of road 22.

[0035] From the determined trajectories 34 to the respective groups of objects, conclusions are drawn about driving lanes 40, 42 and footpaths 44, 46.

[0036] Since the determination of the characteristic trajectories 38 incorporates a large number of trajectories 34 of individual objects 30, outliers, such as an overtaking vehicle, are negligible. Thus, the positions and, if applicable, the spatial extent of the lanes 40, 42 can be reliably determined.

[0037] The classification of objects into individual groups can also take into account, for example, the object speed, provided that a particular object speed suggests a specific group or excludes a group.

[0038] After the device's learning phase, the following are thus available in evaluation unit 12: Fig.The positions of the traffic lanes 40 and 42, as well as the footpaths 44 and 46, are stored in four schematically represented locations. The learning phase can be terminated once sufficient data is available, or continuous learning can be permitted even during operation of the device following the initial learning phase.

[0039] After determining the location of the traffic lanes 40, 42 and optionally footpaths 44, 46 and railway tracks 24, the recorded objects 30 are assigned to the respective traffic paths L1, ..., Ln.

[0040] When assigning the recorded objects to the traffic paths, objects located outside the traffic paths are disregarded.

[0041] These objects 30 assigned to individual traffic paths are then classified as person, passenger car, truck, etc., based on the aforementioned properties such as backscatter power, extent and / or height, as described above. For example, the classes passenger car and truck can also be combined into the class motor vehicles.

[0042] For each traffic route, a traffic analysis specific to that route is then carried out, i.e., an evaluation of the object data of the other recorded objects. Several examples are explained below using lanes 40 and 42 as examples. A corresponding evaluation can also be carried out for the other traffic routes.

[0043] For the detected objects assigned to each lane 40, 42, the average speed v is calculated. i,avg and the minimum speed v i,mindetermined, where i indicates the number of the respective lane.

[0044] For each lane, for example, a minimum distance, an average distance and / or a maximum distance of the objects following the lane can be determined.

[0045] For each lane, for example, the average speed v can be determined. i,avg be determined.

[0046] For each lane, the traffic volume or traffic flow can be determined, for example, as p. i = M / Δt, where M denotes the number of vehicles detected on the lane in a period of time Δt.

[0047] The traffic density for each lane can also be determined as d i = N / Δs = p i / V i . Here, N denotes the number of vehicles that are simultaneously on the track segment of length Δs.

[0048] Depending on the assessment model used, the traffic situation can be evaluated based on the above parameters, in particular classified into one of the categories "free traffic", "slow-moving traffic" and "congestion". This evaluation can be carried out individually for each lane.

[0049] In the manner described, an automated analysis of traffic in the traffic area can be carried out. In particular, the risk of congestion can be assessed.

[0050] Lane-specific traffic analysis and automatic detection and position recognition of lanes provide better and more accurate information for subsequent traffic analysis. As described, characteristics of the respective lanes, such as their width, can also be determined.

[0051] The described method and device are not limited to the detection of spatially separated lanes. The described algorithm can also detect, for example, intersecting lanes, such as the intersection of railway tracks 24 with lanes 40 and 42 and footpaths 44 and 46. Furthermore, intersecting traffic paths can be detected at road junctions or other intersecting traffic routes, and their location can be determined.

[0052] While the described example only mentions one radar sensor 10, it is also conceivable that the device comprises several radar sensors whose fields of view can optionally overlap. In particular, several sensors 10 can be fixedly arranged at different positions. The determined parameters of the monitored traffic can be used, for example, for the automatic determination of speed limits, such as on highways. The presented device can also be used to control traffic signals at intersections, where it can replace or supplement conventional induction loops.

Claims

[1] Device for location-based traffic analysis, comprising an arrangement of at least one radar sensor (10) on a traffic space (20), and an evaluation unit (12) configured to determine object trajectories (34) from temporal sequences of object positions (32) of respective objects (30) detected by the radar sensor (10) and moving in the traffic space (20), to recognize the position of at least one traffic path (24; 40; 42; 44; 46) based on a cluster (36) of object trajectories (34), and to assign further detected objects (30) to a respective traffic path (24; 40; 42; 44; 46) whose position has been recognized, wherein the evaluation unit (12) is configured to assess, based on recorded object speeds of several objects (30) assigned to the same traffic path (40) in the form of a lane, a traffic situation assigned to that lane with regard to the risk of congestion, and / or the evaluation unit (12) is set up to perform a summary evaluation of data for several objects (30) assigned to the same traffic path (40) in order to determine at least one traffic parameter from: an average speed (v i,avg ), a minimum speed (v i,min ), a minimum distance between the objects (30), an average distance between the objects (30), a maximum distance between the objects (30), a traffic flow (p i ) and a traffic density (d i ). [2] Device according to claim 1, wherein the evaluation unit (12) is configured to determine object extensions on the basis of object reflections associated with the respective detected objects (30) and to recognize the location and type of at least one traffic path (24; 40; 42; 44; 46) on the basis of a cluster (36) of object trajectories (34) and on the basis of object extensions of the objects (30) associated with the respective object trajectories (34), in particular comprising types: a driving lane (40; 42) and a footpath (44; 46) and / or a track (24). [3] Device according to one of the preceding claims, wherein the evaluation unit (12) is configured to classify objects (30) moving in the traffic area (20) detected by the radar sensor (10) into object classes of objects of different size and / or type based on object reflections belonging to the respective objects (30), wherein preferably passenger cars and persons are assigned to different object classes. [4] Device according to claim 3, wherein the evaluation device (12) is configured to recognize the position of at least one traffic path (24; 40; 42; 44; 46) on the basis of a cluster (36) of object trajectories (34) relating to objects (30) of a certain object class. [5] Device according to one of the preceding claims, wherein the evaluation unit (12) is configured to classify a traffic situation associated with a traffic lane in at least three classes with regard to the risk of congestion, based on the detected object speeds of several objects (30) assigned to the same traffic path (40) in the form of a lane. [6] Device according to one of the preceding claims, wherein the at least one radar sensor (10) comprises a wide-angle radar sensor with a stationary antenna arrangement whose field of view includes an azimuth angle range of at least 90°. [7] Device according to one of the preceding claims, wherein the at least one radar sensor (10) comprises an FMCW radar sensor configured to determine an object position and object velocity of a detected object (30). [8] Device according to one of the preceding claims, wherein the evaluation unit (12) is configured to detect the location of several traffic paths (24; 40; 42; 44; 46) and to assign further detected objects (30) to a respective traffic path (24; 40; 42; 44; 46) whose location has been detected. [9] Method for traffic analysis using location data of at least one radar sensor (10) fixed to a traffic space (20), comprising the steps: - Determining object trajectories (34) from temporal sequences of object positions (32) of respective objects (30) moving in the traffic space (20) detected by the radar sensor (10); - Identifying the location of at least one traffic path (24; 40; 42; 44; 46) based on a cluster (36) of object trajectories (34), and - Assigning each object (30) detected by the at least one radar sensor (10) to a respective traffic path (24; 40; 42; 44; 46) whose position has been determined, the procedure includes: based on recorded object speeds of several objects (30) assigned to the same traffic path (40) in the form of a lane, assessing a traffic situation assigned to this lane with regard to the risk of congestion; and / or for several objects (30) assigned to the same traffic path (40), performing a summary evaluation of data to determine at least one traffic parameter from: an average speed (v i,avg ), a minimum speed (v i,min ), a minimum distance between the objects (30), an average distance between the objects (30), a maximum distance between the objects (30), a traffic flow (p i ) and a traffic density (di ).

Images