WO2018099480A1 - 车辆行驶轨迹监测方法及系统 - Google Patents

车辆行驶轨迹监测方法及系统 Download PDF

Info

Publication number
WO2018099480A1
WO2018099480A1 PCT/CN2017/114305 CN2017114305W WO2018099480A1 WO 2018099480 A1 WO2018099480 A1 WO 2018099480A1 CN 2017114305 W CN2017114305 W CN 2017114305W WO 2018099480 A1 WO2018099480 A1 WO 2018099480A1
Authority
WO
WIPO (PCT)
Prior art keywords
grid
coordinates
trajectory
historical
point
Prior art date
Application number
PCT/CN2017/114305
Other languages
English (en)
French (fr)
Inventor
董振江
刘丽霞
白雪
张帆
范小朋
须成忠
Original Assignee
中兴通讯股份有限公司
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司, 中国科学院深圳先进技术研究院 filed Critical 中兴通讯股份有限公司
Publication of WO2018099480A1 publication Critical patent/WO2018099480A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/123Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01CMEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34Route searching; Route guidance
    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/01Detecting movement of traffic to be counted or controlled
    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/01Detecting movement of traffic to be counted or controlled
    • G08G1/0104Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125Traffic data processing
    • G08G1/0129Traffic data processing for creating historical data or processing based on historical data

Definitions

  • the present disclosure relates to the field of intelligent transportation, and in particular, to a method and system for monitoring a vehicle travel trajectory.
  • the taxi industry is the main transportation service in modern urban areas, providing a lot of benefits and convenience for our daily life.
  • some taxi drivers are out of greed, they deliberately bypass some non-essential sections to increase the charges for passengers. Therefore, some taxi passengers, especially those from other cities in the city, suffer double losses of time and money.
  • a technical solution capable of detecting and punishing such fraudulent activities is needed.
  • the means of detecting taxi driving fraud is very limited, mainly based on the passengers' active complaints, relying on experienced staff to manually detect the taxi's driving trajectory. This approach is costly and inefficient, and even many deceptive behaviors are not discovered by passengers at all. Therefore, designing the taxi abnormal behavior detection system, through the detection of abnormal behavior of taxis, accurately detecting the deliberate detour of the taxi driver is of great significance, which is beneficial to the overall reputation of the city taxi company, effectively supervising and restraining The driver regulates the behavior and creates a civilized city image. At the same time, it can protect the legitimate interests of passengers and save the time and money spent by passengers on the journey.
  • the related art detection methods for taxi abnormal trajectories are roughly classified into distance-based and statistical-based methods.
  • the distance-based anomaly detection algorithm the main principle is that there are not enough neighbor objects in the given threshold range.
  • the common distance metrics are Mahalanobis distance, Manhattan distance, Euclidean distance. Distance from Hausdorff.
  • the distance-based anomaly detection algorithm can be divided into types based on cell-based, index-based, and nested-loop methods. These methods are computationally intensive, difficult to implement, and detectable. The result is a big error.
  • the statistical method based detection has the following defects: First, the abnormal points can be detected by different distribution models, and the abnormal mechanism is not unique, resulting in uncertainty of the meaning of the abnormal points.
  • the second need to know in advance the distribution or probability model of the data set obey, the actual environment is usually difficult to obtain, the implementation is more difficult, and the error of the detection result is large.
  • the vehicle travel trajectory monitoring method and system provided by the embodiments of the present disclosure mainly solve the technical problem that the existing vehicle travel path detection method is difficult to implement and the detection result error is large.
  • an embodiment of the present disclosure provides a vehicle travel trajectory monitoring method, including: when detecting a new travel of a vehicle, acquiring a geographic coordinate of a starting point of the travel and a geographic coordinate of the end point; The coordinates and the terminal geographic coordinates are respectively converted into a starting grid coordinate and an ending grid coordinate; and a track set of all historical tracks including the starting mesh coordinate and the ending mesh coordinate is found in the mesh track library; The vehicle has passed the journey The geographic coordinates of the current location in the process are sampled and converted into sample grid coordinates; the historical track information in the track set corresponding to the grid coordinates of the sample point on the vehicle is obtained, and the current sample point is not included in the track set.
  • the historical trajectory of the grid coordinates is cleared, and the historical trajectory information in the trajectory set corresponding to the grid coordinates of the current sampling point is obtained; the trajectory set corresponding to the grid coordinates of the current sampling point and the trajectory set corresponding to the grid coordinates of the previous sampling point are calculated.
  • the embodiment of the present disclosure further provides a vehicle travel trajectory monitoring system, including a detection subsystem, a grid subsystem, a database, and a real-time data acquisition subsystem; wherein the detection subsystem is configured to acquire the travel when detecting that the vehicle starts a new trip
  • the geographic coordinates of the starting point and the geographic coordinates of the end point, and the starting point geographic coordinate and the terminal geographic coordinate are respectively converted into the starting grid coordinate and the ending grid coordinate by the grid subsystem, according to the starting grid coordinate and the destination network Grid coordinates find a trajectory set of all historical trajectories including the starting grid coordinates and the ending grid coordinates from the grid trajectory of the database
  • the real-time data acquisition subsystem is configured to The geographic coordinates of the current location during the journey are collected and converted to the sampling point grid coordinates by the grid subsystem, and then sent to the detection subsystem; the detection subsystem is further configured to acquire a sampling point on the vehicle.
  • Historical track information in the track set corresponding to the grid coordinates, and the current sample is not included in the track set The historical trajectory of the grid coordinates is cleared, and the historical trajectory information in the trajectory set corresponding to the grid coordinates of the current sampling point is obtained, and the trajectory set corresponding to the grid coordinates of the current sampling point and the trajectory set corresponding to the grid coordinates of the previous sampling point are calculated.
  • the support value of the current sampling point grid coordinate is compared with the preset support degree threshold, and the grid coordinate of the current sampling point is determined to be abnormal according to the comparison result.
  • the vehicle travel trajectory monitoring method and system when detecting that the vehicle starts a new trip, acquiring the geographical coordinates of the starting point of the trip and the geographic coordinates of the end point, and then starting point geographic coordinates and terminal geographic coordinates of the trip Converting to the starting grid coordinates and the ending grid coordinates respectively, and finding a set of traces of all historical traces including the start grid coordinates and the end grid coordinates in the grid track library; then, the vehicle is currently in the course of the trip
  • the geographical coordinates of the location are sampled and converted into sampling grid coordinates, and the historical trajectory information in the trajectory set corresponding to the grid coordinates of a sampling point on the vehicle is obtained, and the history of the grid coordinates of the current sampling point is not included in the trajectory set.
  • the track is cleared, and the historical track information in the track set corresponding to the current sample point grid coordinate is obtained, and then the current sample point grid coordinate is calculated according to the track set corresponding to the current sample point grid coordinate and the track set corresponding to the grid point coordinate of the previous sample point.
  • Support value, the obtained support value and the preset support threshold Compared to determine whether the current sampling point grid coordinates exception.
  • the disclosure can directly monitor each position point in the current travel trajectory based on the historical trajectory of the trip, determine which position points in the driving process are abnormal, simple and easy to implement, and can maintain a good trajectory abnormality recognition effect, response time Short, the overall detection accuracy is high.
  • FIG. 1 is a schematic diagram of track completion according to Embodiment 1 of the present disclosure.
  • FIG. 2 is a schematic diagram of another trajectory complementation according to Embodiment 1 of the present disclosure.
  • FIG. 3 is a schematic diagram of setting a grid trajectory library according to Embodiment 1 of the present disclosure.
  • FIG. 4 is a schematic flow chart of a method for monitoring a vehicle travel trajectory according to Embodiment 1 of the present disclosure
  • FIG. 5 is a schematic flowchart of a search history track according to Embodiment 1 of the present disclosure.
  • FIG. 6 is a schematic diagram of a grid neighborhood according to Embodiment 1 of the present disclosure.
  • FIG. 7 is a schematic structural diagram of a vehicle travel trajectory monitoring system according to Embodiment 2 of the present disclosure.
  • FIG. 8 is a schematic diagram of networking of a vehicle travel trajectory monitoring system according to Embodiment 2 of the present disclosure.
  • FIG. 9 is a schematic flow chart of a vehicle travel trajectory monitoring method according to Embodiment 2 of the present disclosure.
  • FIG. 10 is a schematic diagram of a vehicle travel trajectory according to Embodiment 2 of the present disclosure.
  • Embodiment 1 is a diagrammatic representation of Embodiment 1:
  • the geographic coordinates in this embodiment refer to the coordinates that characterize a position by longitude and latitude.
  • the grid coordinates are the coordinates in the grid system obtained by transforming the geographic coordinates based on the grid algorithm.
  • This embodiment illustrates an example of a map meshing algorithm in the Bing Maps Tile System of Mercator projection.
  • the main principle of the Bing Maps Tile System is to map the geographic coordinates to the screen coordinates of the two-dimensional plane according to the Mercator projection, and then divide the two-dimensional screen into meshes, and encode each grid to map the screen coordinates to Grid coordinates.
  • Calculating grid coordinates can be divided into two steps:
  • the screen coordinate system is similar to the grid coordinate system, in which the top left corner of the map is the coordinate origin (0,0), the right is the pixelX positive direction, and the downward is the pixelY positive direction.
  • floor(x) is a "round-rounding" function, that is, taking the largest integer not greater than x, ensuring that tileX and tileY are within the valid range 2 level -1 and are integers. See Table 1 below for the corresponding parameters in Equations 1 through 5.
  • the grid mapping function ⁇ (p) R 2 ⁇ G, where R is the geographic coordinate, G is the grid set after the map mapping, p is a point in the two-dimensional continuous space, there are countless kinds of values, and the function value A domain is a point in a two-dimensional discrete space with only a finite number of values.
  • the formula for calculating the grid mapping function ⁇ is as follows:
  • Equation 6 can be directly applied to the mesh mapping component, input the latitude and longitude of any GPS coordinate point within the specified effective range, and specify the map zoom level level value that affects the grid size, and the corresponding mesh can be obtained. Coordinates to achieve discretization of continuous domain trajectory points.
  • the present embodiment exemplifies the trajectory of the grid system and the concept of trajectory completion.
  • the function of the trajectory complement function aug is to insert a plurality of supplementary points between two non-adjacent grid points g 1 and g 2 until a path is obtained such that g 1 and g 2 are connected.
  • each grid g Ai follows the direction from S to D, and any two adjacent grids in the path satisfy the neighborhood relationship, that is, g ai+1 ⁇ N(g ai ), 1 ⁇ i ⁇ n-1.
  • AE-AUG Algmented method of angle and edge
  • the track t is from S ⁇ D, and all black grids ⁇ g 25 , g 68 , g 511 , g 714 , g 715 > are sequentially composed from left to right.
  • the black grid is mapped to the GPS coordinates of the real vehicle, and all gray grids (ie, the complement grid) are obtained according to the AE-AUG algorithm provided by this embodiment, that is, each segment complements the trajectory.
  • the track t is complemented by all ordered colored meshes of S ⁇ D, and the sequence is from left to right and top to bottom.
  • mapping algorithms and completion algorithms in this embodiment are not limited to the above-described examples.
  • Other mapping algorithms or completion algorithms can also be flexibly adopted according to actual needs.
  • the trajectory monitoring algorithm provided in this embodiment is an online abnormal trajectory detection algorithm theory based on isolated characteristics.
  • the basic idea is to utilize the isolation of abnormal points, that is, the probability that the abnormal points appear to be small and distinctive.
  • Abnormal trajectories are usually separated from the main course, while normal trajectories are supported by a large number of similar historical trajectories. The number of historical trajectories supported is directly reflected in the trajectory support rate, and the trajectory with less support will have higher outliers.
  • the algorithm does not depend on the distance and density distribution of the trajectory group. It can overcome the shortcomings of relying on the feature and cannot identify some abnormal conditions. At the same time, it has the ability to identify the abnormal sub-track. When a track is identified as abnormal, the algorithm can locate the abnormality. Track segment or subtrack. In addition, the algorithm can be executed online, and it is possible to detect and retrieve the results in real time without acquiring all the track points.
  • the grid trajectory library can be obtained based on the historical itinerary data of the vehicle based on the above principle.
  • the process is shown in Figure 3 and includes the following steps.
  • a historical itinerary of each vehicle and historical location geographic coordinates in each historical itinerary are acquired.
  • the historical data of the vehicles in each area can be obtained by region.
  • vehicles in the cities of Shenzhen, Huizhou, Dongguan, etc. can be obtained in units of “city”, thereby counting the historical travels in the urban area.
  • Historical trajectory For example, vehicles in the cities of Shenzhen, Huizhou, Dongguan, etc. can be obtained in units of “city”, thereby counting the historical travels in the urban area. Historical trajectory.
  • the geographic coordinates of each historical position of each vehicle in a historical itinerary are replaced with corresponding grid coordinates, and the starting point geographic coordinates and the terminal geographic coordinates of the historical itinerary are respectively converted into starting grid coordinates and ending grid coordinates.
  • ⁇ g 25 , g 68 , g 511 , g 714 , g 715 > are the geographical coordinates of the respective historical positions of the vehicle during the strokes S to D.
  • the trajectory completion coordinate of each vehicle in a historical journey, the grid coordinates of each historical position, and the end grid coordinates are obtained to obtain a historical trajectory of each vehicle completing the historical itinerary, and The number of trips for each historical trajectory.
  • the S to D and the ⁇ g 25 , g 68 , g 511 , g 714 , g 715 > grids in FIG. 2 are complemented, and a historical trajectory of the strokes S to D is obtained, and the historical trajectory can be obtained.
  • the number of vehicles that is, the number of vehicles in the historical trajectory.
  • historical data within one year or half a year can be selected, and the time range can be flexibly set according to actual needs.
  • the processing of the start grid coordinate of a vehicle in a historical stroke, the grid coordinates of each historical position, and the end grid coordinates may include: the starting grid coordinates
  • the grid coordinates of each historical position and the end grid coordinates are mapped to the grids corresponding to the grid subsystem; for example, grids S, D and ⁇ g 25 , g 68 , g 511 , g 714 in FIG.
  • Grid as a complement grid then take the most diagonal All the grids between the rows or columns of the next grid to another grid in the adjacent grid are used as the completion grid; it should be noted that the grid of the resulting inner rectangle may also be concentrated in a certain row or For a column, each grid between the two grids of this row is also taken directly as a complement grid.
  • the vehicle travel trajectory monitoring method provided by this embodiment is shown in FIG. 4, and the method may include the following steps.
  • the starting point geographic coordinate and the terminal geographic coordinate of the trip are respectively converted into a starting grid coordinate and an ending grid coordinate; the conversion process may adopt the mapping algorithm of the above example.
  • a track set of all historical tracks including the start grid coordinates and the end grid coordinates is found in the grid track library, and the track set at this time is also the initial track set of the run.
  • the geographic coordinates of the current location of the vehicle during the trip are sampled and converted to sample point grid coordinates.
  • the geographic coordinates may be geographic coordinates reported by the vehicle at any time or location during the trip.
  • the historical trajectory information in the trajectory set corresponding to the grid coordinate of the sampling point on the vehicle is acquired, and the historical trajectory of the trajectory set not including the grid coordinates of the current sampling point is cleared, and the trajectory corresponding to the grid coordinate of the current sampling point is obtained. Collection Historical track information.
  • the support value of the grid coordinates of the current sampling point is calculated according to the trajectory set corresponding to the grid coordinates of the current sampling point and the trajectory set corresponding to the grid coordinates of the previous sampling point.
  • the obtained support value is compared with a preset support threshold, and it is determined whether the current sampling point grid coordinate is abnormal according to the comparison result.
  • the real-time location reporting information currently sent by the vehicle is received, where the real-time location reporting information includes the geographic coordinates of the current location and the geographic coordinates of the starting point of the current itinerary and the geographic coordinates of the destination. Then, it is determined whether the geographic coordinates of the starting point geographic coordinate and the ending point in the current real-time position reporting information of the vehicle are the same as the geographic coordinates of the starting point geographic coordinate and the ending point in the last-time real-time position reporting information, and if so, it is determined that the vehicle is executing the original Itinerary; if not, it is judged that the vehicle starts a new trip.
  • the vehicle may also be specifically sent a new trip start notification to start the new trip when starting a new trip.
  • searching for the trajectory set of all the historical trajectories including the start grid coordinates and the end grid coordinates in the grid trajectory may be performed by using a reverse lookup method, thereby reducing the workload and improving the search efficiency. It can also improve the accuracy of the search.
  • the process is shown in Figure 5 and includes the following steps.
  • an intersection of the starting point trajectory set and the ending point trajectory set is obtained to obtain a trajectory set of all historical trajectories including the starting grid coordinates and the ending grid coordinates.
  • the historical track clearing that does not include the grid coordinates of the current sampling point in the track set corresponding to the grid coordinates of the previous sampling point includes: determining a historical track corresponding to the track set corresponding to the mesh coordinate of the previous sampling point. Whether the grid set corresponds to the grid corresponding to the grid coordinates of the current sampling point, and if so, it is determined that the historical track contains the grid coordinates of the current sampling point; otherwise, it is determined whether the current sampling point network is included in the grid set corresponding to the historical track.
  • An adjacent mesh in the mesh neighborhood corresponding to the lattice coordinate, and the adjacent mesh satisfies pos(N(gk-1)) ⁇ pos(N(gk)), and if so, it is determined that the historical track includes the current Sampling point grid coordinates, otherwise, it is determined that the historical trajectory does not contain the grid coordinates of the current sampling point.
  • the first parameter in the function is the candidate track set T, and the second parameter is the target track t.
  • the set T is filtered with t, returning all traces in T that are similar to the given target track t. Its mathematical expression is as follows:
  • N(g i ) the mesh neighborhood N(g i ) is required to exist at least one point on the trajectory t', and the subscript position of N(g i ) in the trajectory t' i monotonically increments.
  • the hasPath function filter track process is described as follows:
  • Input track set T ⁇ e i
  • 1 ⁇ i ⁇ n, i ⁇ N + ⁇ , e i ⁇ a j
  • 1 ⁇ j ⁇ m,j ⁇ N + ⁇ , test track t ⁇ g k
  • the set T is the filtered object, and the track t is used as the filter condition, and both are used as input parameters of the hasPath function. Perform step (2).
  • T set is actually stored for each historical track contains only the number of e i, the search grid for each historical track by track library composed of grid points e i (a 1, a 2, ising , a m). Perform step (3).
  • step (8) Traverse all grid points g k in the test trace t. If the traversal is completed, step (8) is performed, otherwise step (4) is performed.
  • N(g k ) g k calculated grid mesh neighborhood N (g k).
  • the purpose of finding N(g k ) is to allow certain fault tolerance when trajectory comparison is made when discriminating whether the mapping trajectory e i contains the grid g k , as long as any grid of N(g k ) is on e i
  • the track e i goes through the grid g k . Then, step (5) is performed.
  • the grid neighborhood N is defined as follows: for a given grid g, with g as the center, N is g itself and the most M adjacent thereto (the value of M can be flexibly changed, for example, 8)
  • N is g itself and the most M adjacent thereto (the value of M can be flexibly changed, for example, 8)
  • N(g) G ⁇ P(G), returning the grid neighborhood of g for a given input grid g.
  • N(g ij ) ⁇ g mn
  • g ij is the given input grid
  • i and j are the grid coordinates of their corresponding x and y directions, respectively.
  • step (3) Traverse all the tracks e i in the set T. If the traversal is completed, step (3) is performed, otherwise step (6) is performed.
  • POS represents the grid position. This formula indicates that the points in the neighborhood of this grid are smaller than the grid positions of the points in the neighborhood of the previous grid.
  • the support value of the current sampling point grid coordinate is calculated according to the trajectory set corresponding to the grid point coordinates of the current sampling point and the trajectory set corresponding to the grid coordinates of the previous sampling point, but is not limited to any one of the following two methods. .
  • Manner 1 The number of historical trajectories in the trajectory set corresponding to the grid coordinates of the current sampling point is divided by the number of historical trajectories in the trajectory set corresponding to the grid coordinates of the previous sampling point, and the support value is obtained; for example, the current sampling point grid coordinates are assumed.
  • the number of historical trajectories in the corresponding trajectory set is 2, and the number of historical trajectories in the trajectory set corresponding to the grid coordinates of the previous sampling point is 6, and the support value is 2/6.
  • At least one of the trajectory distance and the abnormal value corresponding to each sampling point may be calculated for display, and the subsequent statistical management is also facilitated.
  • the following calculation formula may be used to calculate the trajectory distance corresponding to the grid coordinates of the current sampling point:
  • p i-1 and p i are the last sample point and the current sample point respectively;
  • R E is the radius of the earth, and acos is the inverse cosine function;
  • t 1 cos(a i-1 ) ⁇ cos(a i ) ⁇ cos(b i-1 ) ⁇ cos(b i );
  • x i-1 and y i-1 are the longitude and latitude of the geographic coordinate p i-1
  • x i and y i are described as the longitude and latitude of the geographic coordinate p i .
  • the following calculation formula can also be used to calculate the abnormal value corresponding to the grid coordinates of the current sampling point:
  • Equation 9 can reflect the current abnormal degree of the trajectory from the front, depending on the distance between the two points and the current support. The larger the value, the more abnormal the trajectory.
  • the trajectory completion method AE-AUG provided by the embodiment of the present disclosure has a simple and simple step. In actual application, a path can be quickly obtained to connect two non-adjacent grids.
  • the embodiment is based on a large number of vehicle history GPS records, generates historical trajectory data, and combines AE-AEG completion algorithm, anomaly detection algorithm, Bing Maps Tile System map grid calculation algorithm to realize a vehicle with fast response speed and high overall detection accuracy. Trajectory monitoring method.
  • Embodiment 2 is a diagrammatic representation of Embodiment 1:
  • the present embodiment provides a vehicle travel trajectory monitoring system, as shown in FIG. 7, comprising: a detection subsystem 61; a grid subsystem 62, which includes a grid manager component, which may have a grid mapping mapping and a trajectory complement Full Augmenting function; database 63, set to store the maintenance grid track library; and real-time data acquisition subsystem 64.
  • the detecting subsystem 61 is configured to acquire the geographic coordinates of the starting point geographic coordinates and the ending point of the trip when the vehicle starts to start a new trip, and convert the starting point geographic coordinate and the terminal geographic coordinate into the starting grid coordinate and the grid respectively by the grid subsystem 62.
  • the end point grid coordinates (as an example, the conversion algorithm may use the above mapping algorithm), look up from the grid trajectory of the database 63 according to the starting grid coordinates and the ending grid coordinates (as an example, the reverse lookup method described above may be used) A set of trajectories for all historical trajectories containing the starting grid coordinates and the ending grid coordinates.
  • the real-time data acquisition subsystem 64 is configured to collect the geographic coordinates of the current location of the vehicle during the trip and convert it to the sampling point grid coordinates by the grid subsystem 62 and send it to the detection subsystem 61.
  • the sampling rules can be flexibly set according to the application scenario.
  • the detection subsystem 61 is further configured to acquire historical trajectory information in the trajectory set corresponding to the grid coordinates of a sampling point on the vehicle, and clear the historical trajectory of the trajectory set that does not include the grid coordinates of the current sampling point, to obtain a current sampling point grid.
  • the historical trajectory information in the trajectory set corresponding to the coordinate, the support value of the grid coordinates of the current sampling point is calculated according to the trajectory set corresponding to the grid coordinates of the current sampling point and the trajectory set corresponding to the grid coordinates of the previous sampling point, and the obtained Comparing the support value with the preset support threshold, determining whether the current sampling point grid coordinate is abnormal according to the comparison result, and determining the current sampling point grid when the comparison result is that the support value is less than the preset support threshold
  • the coordinates are abnormal, and the track set corresponding to the grid point coordinates of the current sampling point is updated to a track set of all historical tracks including the start grid coordinates and the end grid coordinates.
  • the detection subsystem 61, the grid subsystem 62, and the real-time data acquisition subsystem 64 implement respective functions to perform, for example, the method in the first embodiment.
  • the detection subsystem 61 can also perform the calculation of the track distance and/or the outlier value in the manner shown in the first embodiment.
  • This embodiment describes an example of the networking structure of each of the foregoing systems. Referring to FIG. 8, the Hadoop platform and the Web server are respectively configured to perform an offline processing phase and an online processing phase task respectively. Combined with the actual data, each node is described below.
  • the historical vehicle 71 refers to a group of taxis (including, of course, other operating or private vehicles) in a certain area (for example, Shenzhen City) at any time in the past. This embodiment can focus only on the historical position data generated by the present embodiment.
  • GPS Packet Collector 72 A data merge node that collects historical GPS data records uploaded by all rental vehicles in the area.
  • HDFS Distributed File System
  • Data Cleaner 74 Raw number due to network instability, equipment aging, etc. According to a large number of abnormal conditions, such as missing fields, record reporting delay, GPS coordinate drift, intermittent vehicle loss, etc., greatly reducing data quality. For such problems, the cleaner removes the exception record while ensuring data integrity as much as possible.
  • the GPS Packet Receiver 75 is a part of the real-time data acquisition subsystem 64. It is a real-time data relay node configured to receive and detect GPS data packets transmitted by the vehicle.
  • a data formatter 76 (Data Formatter), which is a part of the real-time data acquisition subsystem 64, configured to extract and detect a vehicle data package license plate number, latitude and longitude, reporting time, passenger status, and the like, and format it as a semantic object. For easy data transfer and analysis.
  • Grid Manager 77 (Grid Manager), which is part of the Grid Subsystem 62, is a common component of the Hadoop platform and the Web server, and provides grid operation related functions, such as Grid Neighborhood N (gi),
  • the index position pos(t, gi) of the grid in the trajectory can also include two sub-components: mesh mapping and trajectory.
  • the mesh mapping component 78 (Mapping) is a part of the mesh subsystem 62.
  • the latitude and longitude coordinates are mapped to the grid coordinates by the mesh mapping function ⁇ , and the track points are discretized.
  • Augmenting component 79 (Augmenting), a part of the mesh subsystem 62, through the trajectory complement function aug, inserts a mesh into two meshes that are not adjacent, so that the two are connected.
  • Database Manager 710 (Database Manager): In the offline phase, the grid trajectory data generated by the Grid Manager is stored in the trajectory database, and the trajectory query service is provided for the detection engine in the online phase.
  • Trajectories Database 711 Trajectories Database
  • the database contains two parts of data: positive sequence and reverse order.
  • Positive sequence refers to indexing all track points by track number
  • reverse order means indexing all tracks passing through the point by grid coordinates.
  • Detection Engine 712 (Detection Engine), part of the detection subsystem 61: As a core component in the detection subsystem 61, a detection algorithm is implemented. At the same time, interact with Grid Manager and Database Manager, input track points and historical track sets to detect whether the point is abnormal.
  • Web Controller 713 (Web Controller), a part of the detection subsystem 61: interacts with the terminal device to issue a detection result.
  • the present embodiment is described by taking the entire monitoring process of the vehicle as an example. Referring to FIG. 9, the following steps are included.
  • a structured real-time vehicle status record is entered.
  • the taxi vehicle terminal device uploads the current driving state information of the vehicle to the detection server, the server extracts the valid detection field, and structures the data, and finally transmits the structured data to the detection engine.
  • Step S802 is performed.
  • the detection engine determines whether the vehicle starts a new operation track according to whether the record includes the end point latitude and longitude coordinates, that is, starts a new trip. If yes, go to step S803, otherwise go to step S808.
  • the taxi starts a new operational trajectory, which is the newly loaded guest, and takes the start point and the end point latitude and longitude coordinates from the record. Step S804 is performed.
  • the detection engine sends a command to the grid manager to map the start point and end point coordinates to corresponding grid points. Step S805 is performed.
  • the detection engine sends a retrieval command to the database manager, transmits the start and end grids, and obtains a historical trajectory set.
  • the trajectory full traversal method is not used in this embodiment, that is, whether the start and end points are simultaneously determined one by one, but all the trajectory IDs respectively falling in the grid are reversely indexed through the grid, and the intersections of the two are intersected. Seek.
  • the set consists of all tracks that happen to pass through or through the start and end grids. Then, step S806 is performed.
  • step S806 an initial detection result is set.
  • the starting coordinate point defaults to normal, the support degree is 1, and the abnormal value and track distance are 0. Then, step S807 is performed.
  • Step S809 the last detection result status information of the vehicle is read.
  • the taxi is still in the process of passenger operation.
  • the current track point is the latest in the online track, and this test is affected by the previous test result, so the last state information needs to be read.
  • Step S809 is performed.
  • Step S809 Read the recorded latitude and longitude coordinate points, and the detection engine sends a command to the grid manager to map the coordinate points to the corresponding grid. Step S810 is performed.
  • step S810 It is determined whether the grid corresponding to the latest coordinate point is the same as the grid of the previous coordinate point object. If they are the same, step S811 is performed. Otherwise, step S812 is performed.
  • Step S811 The latest track point and the previous track point fall into the same grid, and all detection states remain unchanged. If the previous track point is normal, the current track point is normal, otherwise it is abnormal. The two points before and after contain the same support, outliers and abnormal distance. Step S807 is performed.
  • Step S812 The newly received trajectory grid is different from the previous trajectory grid, and the current grid support degree needs to be recalculated, and the previous state trajectory set needs to be read. Step S813 is performed.
  • Step S813 Record the total number of tracks included in the track set before filtering, that is, the number of tracks of the previous state track set count(T i-1 ). Step S813 is performed.
  • Step S814 Filter the current working track set according to the hasPath function.
  • the two input parameters of the hasPath function are the filtered candidate track set and the reference track as the filter condition.
  • Step S815 is performed.
  • Step S815 set the trajectory is derived filtration step S814, the recording track number count (T si) included in the filtered set. Step S816 is performed.
  • step S817 Determine whether the current track point support degree is lower than a set threshold. If not, it is determined that the current track point is normal, and step S818 is performed; otherwise, abnormally, step S822 is performed.
  • step S818 According to the criterion of step S817, the latest received track point is normal. Normally, the number of historical trajectories of the vehicle traveling from the previous trajectory point gi to the current trajectory point gj exceeds at least the set reference value, that is, the travel from gi to gj belongs to the conventional route. Step S819 is performed.
  • Score(i-1) is the previous state outlier, which is negatively correlated with the current support degree and positively correlated with the distance from the previous track point to the current track point. Step S820 is performed.
  • Step S820 Calculate a track distance corresponding to the current track point.
  • the trajectory distance is the sum of all abnormal points and the spherical distance between the abnormal point and the normal point.
  • the abnormal distance remains unchanged only when the current and previous state track points are normal; otherwise, dist(p i , p i-1 ) needs to be accumulated. Step S821 is performed.
  • Step S821 Set the current track point detection result.
  • the detection result of the current track point is reset according to the support degree, the abnormal value, and the abnormal distance calculated above. Step S807 is performed.
  • step S822 According to the criterion of step S817, the newly received track point is abnormal.
  • the abnormality means that the number of historical trajectories of the vehicle traveling from the previous trajectory point gi to the current trajectory point gj does not exceed the set reference value, that is, the travel from gi to gj belongs to an unconventional route. Go to step S823.
  • Step S823 Reset the track set to the initial track set. Since the previous state track set is filtered by the current track point, the number of tracks is too small to be lower than the preset value. If the track set is not reset to the initial state, all track points received thereafter are determined to be abnormal. Step S819 is performed.
  • Fig. 10 it is assumed that there are three sets of conventional routes, that is, the preferred route from most of the taxi driver's passengers from the starting point S form to the ending point D, the arrow direction represents the driving direction, and the gray grid represents the area occupied by the conventional line.
  • the H2 line there are 40 drivers driving along the H2 line, 30 drivers along the H3 line, and 30 drivers along the H1 line.
  • the H4 line is the target test track t
  • each black point is the GPS coordinate point actually received by the server
  • the black point number represents the order in which the server receives the data.
  • the server receives the ⁇ g 1 , g 2 , g 3 , g 4 > points in turn, and this segment has H1 historical track support, and the 4-point detection is normal.
  • ⁇ g 5 , g 6 > is received, only the H3 trajectory of the three conventional lines passes through the grid area, but the two directions of travel are opposite, that is, no historical trajectory conforms to the movement, so ⁇ g 5 , g 6 >
  • Two-point detection is abnormal.
  • ⁇ g 7 , g 8 > completely fall into the new grid, not in any historical trajectory, so ⁇ g 7 , g 8 > is also an abnormal point.
  • the previous state working set T i-1 is filtered and the support is obtained.
  • the support is below the abnormal threshold, the current working set needs to be reset to the T 1 state.
  • the final output normal point set is ⁇ g 1 , g 2 , g 3 , g 4 , g 9 , g 10 , g 11 >, and the abnormal point set is ⁇ g 5 , g 6 , g 7 , g 8 >.
  • the trajectory completion method AE-AUG provided by the embodiment of the present disclosure can easily and quickly find a path to connect two non-adjacent grids.
  • the system provided by the embodiment of the present disclosure can generate historical trajectory data based on a large number of vehicle history GPS records, and
  • the AE-AEG complement algorithm, anomaly detection algorithm, and Bing Maps Tile System map grid calculation algorithm realize fast and reliable detection of vehicle trajectory, avoiding malicious roaming of drivers and improving user experience satisfaction.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Automation & Control Theory (AREA)
  • Traffic Control Systems (AREA)
  • Navigation (AREA)

Abstract

一种车辆行驶轨迹监测方法及系统,在车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标并分别转换为起点网格坐标和终点网格坐标,在网格轨迹库中查找出包含该起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;然后对车辆在该行程过程中当前所在位置的地理坐标进行采样并转换为采样点网格坐标,将该车辆上一采样点网格坐标对应的轨迹集合中不包含当前采样点网格坐标的历史轨迹清除得到当前采样点网格坐标对应的轨迹集合,进而根据该轨迹集合以及上一采样点对应的轨迹集合计算当前采样点的支持度值并与预设支持度阈值进行比较确定当前采样点是否异常,该监测方法能保持良好的轨迹异常识别效果,响应时间短,准度高。

Description

车辆行驶轨迹监测方法及系统 技术领域
本公开涉及智能交通领域,尤其涉及一种车辆行驶轨迹监测方法及系统。
背景技术
出租车行业是现代城市地区的主要交通服务,为我们日常生活提供了大量的好处和方便。但是,由于一些出租车司机出于贪心,他们故意绕行一些非必要的路段以增加对乘客的收费。因此,一些出租车乘客,特别是外地城市游客,就会遭受时间和金钱的双重损失。为了提高出租车服务质量,需要一种能够检测和惩罚此类欺诈行为的技术方案。
目前,检测出租车驾车欺骗行为的手段非常有限,主要是根据乘客主动投诉,依靠有经验的工作人员,手动检测出租车的行车轨迹。这种方式代价较大且低效,甚至很多欺骗行为根本就没有被乘客发现。因此,设计出租车异常行为探测系统,通过对出租车异常行为探测,准确地检测出出租车司机刻意的绕路行为,具有重大意义,既有利于城市出租车公司整体声誉,有效地监督和约束司机规范行为,营造文明城市形象,同时也能保护乘客合法利益,节省乘客在旅程所花费的时间和金钱。
相关技术中针对出租车异常轨迹检测方法大致分为基于距离和基于统计方法。
基于距离的异常检测算法,其主要原理是异常点在给定阈值范围内没有足够多的邻域对象,常见的距离度量有马氏(Mahalanobis)距离、曼哈顿(Manhattan)距离、欧式(Euclidean)距离和豪斯多夫(Hausdorff)距离。基于距离的异常检测算法从类型上划分可包括基于单元(cell-based)、基于索引(index-based)和嵌套循环(nested-loop)等方法,这些方法存在计算量大、难实现且检测结果误差大的问题。
基于统计的方法检测存在以下缺陷:第一,异常点可通过不同的分布模型检测得到,异常机制不唯一,导致异常点含义存在不确定性。第二需要预先知道数据集服从的分布或概率模型,实际环境通常难以得到,实现比较难,且导致检测结果误差较大。
因此提出一种能快速、精准的检测出车辆行驶路径是否正常的方法是目前急需解决的问题。
发明内容
本公开实施例提供的车辆行驶轨迹监测方法及系统,主要解决的技术问题是现有车辆行驶路径检测方法存在的实现较难且检测结果误差较大的问题。
为解决上述技术问题,本公开实施例提供一种车辆行驶轨迹监测方法,包括:检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标;将所述行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标;并在网格轨迹库中查找出包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;对所述车辆在所述行程过 程中当前所在位置的地理坐标进行采样并转换为采样点网格坐标;获取所述车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,并将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合中历史轨迹信息;根据所述当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值;将得到的所述支持度值与预设支持度阈值进行比较,根据比较结果确定所述当前采样点网格坐标是否异常。
本公开实施例还提供一种车辆行驶轨迹监测系统,包括检测子系统、网格子系统、数据库以及实时数据采集子系统;其中,检测子系统设置为检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标,并通过所述网格子系统将所述起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标,根据所述起点网格坐标和终点网格坐标从所述数据库的网格轨迹库中查找出包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;所述实时数据采集子系统设置为对所述车辆在所述行程过程中当前所在位置的地理坐标进行采集并通过所述网格子系统转换为采样点网格坐标后发给所述检测子系统;所述检测子系统还设置为获取所述车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,并将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合中历史轨迹信息,根据所述当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值,并将得到的所述支持度值与预设支持度阈值进行比较,根据比较结果确定所述当前采样点网格坐标是否异常。
根据本公开实施例提供的车辆行驶轨迹监测方法及系统,在检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标,然后将该行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标,并在网格轨迹库中查找出包含该起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;然后对车辆在该行程过程中当前所在位置的地理坐标进行采样并转换为采样点网格坐标,并获取车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合中历史轨迹信息,进而根据当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值,将得到的支持度值与预设支持度阈值进行比较即可确定当前采样点网格坐标是否异常。本公开可以直接基于行程的历史轨迹对当前行程行驶轨迹中的各个位置点进行监测,判断出行驶过程中哪些位置点是异常的,简单易实现,且能保持良好的轨迹异常识别效果,响应时间短,整体检测准度高。
附图说明
图1为本公开实施例一的轨迹补全示意图;
图2为本公开实施例一的另一轨迹补全示意图;
图3为本公开实施例一的网格轨迹库设置示意图;
图4为本公开实施例一的车辆行驶轨迹监测方法流程示意图;
图5为本公开实施例一的查找历史轨迹流程示意图;
图6为本公开实施例一的网格邻域示意图;
图7为本公开实施例二的车辆行驶轨迹监测系统结构示意图;
图8为本公开实施例二的车辆行驶轨迹监测系统组网示意图;
图9为本公开实施例二的车辆行驶轨迹监测方法流程示意图;
图10为本公开实施例二的车辆行驶轨迹示意图。
具体实施方式
下面通过具体实施方式结合附图对本公开实施例中的一些技术特征进行详细说明。应理解,以下细节的提供是为了帮助理解本发明的原理,而本发明并不限于实施例中的细节。
实施例一:
为了便于理解,首先对地理坐标和网格坐标的概念进行示例说明。
本实施例中的地理坐标是指通过经度和纬度表征一个位置的坐标。网格坐标是基于网格算法将地理坐标进行转换后得到的网格系统中的坐标。本实施例以墨卡托投影的Bing Maps Tile System中地图网格化算法进行示例说明。
Bing Maps Tile System的主要原理是先将地理坐标按墨卡托投影映射为二维平面的屏幕坐标,再将二维屏幕进行网格切分,为每个网格编码,从而将屏幕坐标映射到网格坐标。
计算网格坐标可以分为两个步骤:
(1)将轨迹点的经纬度坐标转换为屏幕坐标。公式如下:
Figure PCTCN2017114305-appb-000001
Figure PCTCN2017114305-appb-000002
Figure PCTCN2017114305-appb-000003
屏幕坐标体系与网格坐标体系类似,其中以地图左上角顶点为坐标原点(0,0),向右为pixelX正方向,向下为pixelY正方向。
(2)将屏幕坐标转换为网格坐标。公式如下:
Figure PCTCN2017114305-appb-000004
Figure PCTCN2017114305-appb-000005
公式4和公式5中,floor(x)为“向下取整”函数,即取出不大于x的最大整数,保证tileX和tileY在有效取值范围2level-1内,且为整数。上述公式1到公式5中相应参数说明见下表1。
表1
参数 含义
sinLatitude 为表示方便设定的中间变量
longitude 地理经度
latitude 地理纬度
pixelX 屏幕x方向坐标
pixelY 屏幕y方向坐标
tileX 网格x方向坐标
tileY 网格y方向坐标
定义网格映射函数ρ(p):R2→G,其中R是地理坐标,G是地图映射后的网格集合,p是二维连续空间下的一点,有无数种取值,而函数值域是二维离散空间下的一点,只有有限种取值。为了将GPS坐标点离散为网格点,以便于相似轨迹查找,令网格映射函数g=ρ(p),其中p=(longitude,latitude)为GPS坐标点,g=(tileX,tileY)为网格坐标点。根据上述推导,最终得出网格映射函数ρ的计算公式如下:
Figure PCTCN2017114305-appb-000006
公式6可直接应用到网格映射(Mapping)组件中,输入规定有效范围内的任一GPS坐标点的经纬度,同时指定影响网格大小的地图缩放水平level值,即可求出对应的网格坐标,实现连续域轨迹点离散化。
另外,为了更好地理解本公开的技术方案,本实施例对网格系统的轨迹以及轨迹补全概念进行示例说明。
为了保证每条轨迹经过网格映射后组成的网格能完整无缝地接合在一起,需要在断裂的单元中插入一个或多个补全网格,为异常检测有效查找相似轨迹做准备。本实施例提出一种简单快捷的轨迹补全算法AE-AUG(Augmented method of angle and edge),算法的描述与原理如下:
1、定义网格补全函数
定义网格补全函数aug(g1,g2):G×G→P(G),其中,输入参数g1、g2是映射后的网格点,P(G)=G×G×…×G,为多个网格构成集合的值域空间。
轨迹补全函数aug的作用是在两个不相邻的网格点g1、g2之间插入数个补充点,直到求出一条路径使得g1、g2连通。
2、AE-AUG算法描述
给定网格S和D,两者不相邻,求出一条从S到D的通路taug
通路由网格点构成,即taug=<gai,ga2,......,gan>,gai∈G,1≤i≤n,i∈N+,且各网格gai沿着从S向D接近的方向,通路中任意两相邻网格满足邻域关系,即gai+1∈N(gai),1≤i≤n-1。
3、AE-AUG算法图解说明
如图1所示,给定不相邻网格g31和g78,目标是求出一条从g31到g78的补充路径,使得两网格连通,算法步骤如下:
(1)找出由g31和g78确定的内矩形C,如图1灰白色区域所示。
(2)从起点网格出发g31,拾取以矩形C短边为边长的正方形的对角线网格,上图为<g42,g53,g64>。
(3)从正方形顶点网格g64出发,拾取沿内矩形C长边直到终点g78的所有网格,上图 为<g65,g66,g67>。
(4)最终输出补全轨迹taug为步骤2和步骤3按顺序拼接的网格集合,taug=<g42,g53,g64,g65,g66,g67>,即图1黑色网格坐标从左到右连成的轨迹。
若两不相邻网格内矩形退化为一条线段(也即两网格可能在同一行或同一列),则直接沿线段拾取网格。如给定网格为g31和g38,则输出为taug=<g32,g33,g34,g35,g36,g37>。
由于该补全算法先拾取对角线再拾取剩余边,因此称其为AE-AUG(Augmented method of angle and edge)。
如图2所示,轨迹t从S→D,由所有黑色网格<g25,g68,g511,g714,g715>依次从左到右构成。黑色网格为真实车辆的GPS坐标映射而来,而所有灰色网格(也即补全网格)是根据本实施例提供的AE-AUG算法求得的,即,各段补充轨迹。则轨迹t经过补全后为S→D的所有有序着色网格,序列依次从左到右从上到下。
应当理解的是,本实施例中网格坐标映射算法以及轨迹补全算法并不限于上述示例的算法。根据实际需求也可以灵活地采用其他的映射算法或补全算法。
本实施例提供的轨迹监测算法是基于孤立特性的在线异常轨迹检测算法理论,基本思想是利用异常点的孤立性,即异常点出现概率小且与众不同的性质。异常轨迹通常会从主体路线中分离出来,而正常的轨迹则会有大量相似的历史轨迹支撑。支撑的历史轨迹数量直接反应为轨迹支持率,支持率越少的轨迹将会有更高的异常值。该算法不依赖于轨迹群的距离和密度分布,能克服依赖该特征而无法识别某些异常情况的缺点,同时拥有异常子轨迹识别能力,当一条轨迹被识别为异常时,算法可定位到异常的轨迹片段或子轨迹。此外算法能在线执行,不需要获取全部轨迹点便可进行检测并实时返回结果。
因此在本实施例中,可以基于上述原理,根据车辆的历史行程数据得到网格轨迹库。该过程参见图3所示,包括以下步骤。
在S301,获取各车辆的历史行程以及在各历史行程中的历史位置地理坐标。
该步骤可以按区域获取各区域内的车辆历史数据,例如可以以“市”为单位,获取深圳市、惠州市、东莞市……等市区域的车辆,从而统计个市区内各历史行程的历史轨迹。
在S302,将各车辆在一个历史行程中的各历史位置地理坐标换成对应的网格坐标,并将该历史行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标。例如,假设某一行程是从图2中的S到D,则<g25,g68,g511,g714,g715>是车辆在行程S到D过程中的各历史位置地理坐标。
在S303,将每一车辆在一个历史行程中的起点网格坐标、各个历史位置的网格坐标,以及终点网格坐标进行轨迹补全处理得到各车辆完成所述历史行程的历史轨迹,并得到各历史轨迹的行车数量。
例如对图2中的S到D以及<g25,g68,g511,g714,g715>网格进行补全,得到行程S到D的一条历史轨迹,并可以得到行走该历史轨迹的车辆数量,也即该历史轨迹的行车数量。例如,可以选用一年内或半年内的历史数据,时间范围可以根据实际需要灵活设定。当然,经最终分析,在S到D之间可能会存在多条历史轨迹。
如上分析,本实施例中为一个车辆在一个历史行程中的起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标进行轨迹补全的处理可以包括:将所述起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标映射到网格子系统对应的各网格中;例如图2中的网格S、D以及<g25,g68,g511,g714,g715>;以起点网格坐标对应的网格为起点,终点网格坐标对应的网格为终点,按照各历史位置的网格坐标获取的时序,依次找到相邻两个网格;如果相邻两个网格在同一行或列,则以相邻两个网格在同一行或列之间的网格作为补全网格将两个网格连接;否则,确定所述相邻两个网格之间的网格组成的内矩形,然后以网格内矩形的短边为边长确定正方形,并以所述相邻网格中靠近起点的网格为起点取正方形对角上的各网格作为补全网格,然后再取对角线上最后一个网格所在行或列到所述相邻网格中另一个网格之间的所有网格作为补全网格;应当注意的是得到的内矩形的网格也可能集中在某一行或某一列,此时也直接取这一行的两网格之间的各网格作为补全网格。
基于上述处理,本实施例提供的车辆行驶轨迹监测方法参见图4所示,该方法可以包括以下步骤。
在S401,检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标。
在S402,将行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标;转换过程可以采用上述示例的映射算法。
在S403,在网格轨迹库中查找出包含起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合,此时的该轨迹集合也即该行程的初始轨迹集合。
在S404,对所车辆在所述行程过程中当前所在位置的地理坐标进行采样并转换为采样点网格坐标。该地理坐标可以是车辆在该行程过程中任意时刻或位置上报的地理坐标。
在S405,获取车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,并将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合 中历史轨迹信息。
在S406,根据当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值。
在S407,将得到的支持度值与预设支持度阈值进行比较,根据比较结果确定所述当前采样点网格坐标是否异常。
上述S401中,判断车辆是否开始新的行程可以采用以下方式进行判断。
首先,接收车辆当前发送的实时位置上报信息,该实时位置上报信息中包含当前所在位置的地理坐标以及当前行程的起点地理坐标和终点的地理坐标。然后,判断车辆当前实时位置上报信息中的起点地理坐标和终点的地理坐标与上一次发送的实时位置上报信息中的起点地理坐标和终点的地理坐标是否相同,如是,则判断该车辆在执行原行程;如否,则判断该车辆开始新的行程。
当然,本实施例中还可以采取车辆在开始一个新的行程时专门发送新行程开始通知以通知开始了新行程。
上述S403中,在网格轨迹库中查找出包含起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合可以采用反向查找的方式进行,这样既能减小工作量,提升查找效率,又能提升查找的准确率,该过程参见图5所示,包括以下步骤。
在S501,在网格轨迹库中查找出所有包含所述起点网格坐标的历史轨迹作为起点轨迹集合,并查找出所有包含终点网格坐标的历史轨迹作为终点轨迹集合。
在S502,取起点轨迹集合和终点轨迹集合的交集得到包含起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
在S405中,将上一采样点网格坐标对应的轨迹集合中不包含当前采样点网格坐标的历史轨迹清除包括:判断上一采样点网格坐标对应的轨迹集合中的某一历史轨迹对应的网格集合中是否包含当前采样点网格坐标对应的网格,如是,判断该历史轨迹包含当前采样点网格坐标;否则,判断该历史轨迹对应的网格集合中是否包含当前采样点网格坐标对应的网格邻域中的某一相邻网格,且该相邻网格满足pos(N(gk-1))<pos(N(gk)),如是,判断该历史轨迹包含当前采样点网格坐标,否则,判断该历史轨迹不包含当前采样点网格坐标。
为了更好地理解上述历史轨迹清除过程(或称为过滤过程),下面结合附图对轨迹过滤函数的定义及实现流程进行示例说明。
定义过滤函数hasPath(T,t):P(T)×T→P(T),P(T)=T×T×…×T,为多个T构成的集 合的值域空间。函数中第一个参数是候选轨迹集合T,第二个参数是目标轨迹t。用t过滤集合T,返回T中所有与给定目标轨迹t相似的轨迹。其数学表达如下:
Figure PCTCN2017114305-appb-000007
公式7中,轨迹t={g1,g2,......,gn}。其含义是对于轨迹t上的任意一点gi,要求其网格邻域N(gi)至少有一点存在于轨迹t’上,且N(gi)在轨迹t’中的下标位置随i单调递增。hasPath函数过滤轨迹过程描述如下:
(1)输入轨迹集合T={ei|1≤i≤n,i∈N+},ei={aj|1≤j≤m,j∈N+},测试轨迹t={gk|1≤k≤l,k∈N+}。集合T作为被过滤对象,轨迹t作为过滤条件,两者作为hasPath函数的输入参数。执行步骤(2)。
(2)实际存储的集合T仅包含每条历史轨迹ei的编号,通过网格轨迹库检索各历史轨迹ei组成的网格点(a1,a2,……,am)。执行步骤(3)。
(3)遍历测试轨迹t中所有网格点gk。如果遍历完成,则执行步骤(8),否则执行步骤(4)。
(4)求出网格gk的网格邻域N(gk)。求N(gk)的目的是当判别映射轨迹ei是否包含网格gk时,允许轨迹对比时具有一定容错性,只要N(gk)中任意一个网格在ei上,都认为轨迹ei经过网格gk。然后,执行步骤(5)。
本实施例中定义网格邻域N如下:对于给定网格g,以g为中心,则N为g自身及与其相邻的最多M(M的取值可以灵活变化,例如取8)个相邻网格所构成的网格集合,参见图6所示的几种情况:
对于给定的网格g:
(1)当g=g11时,邻域集合N={g11,g12,g21,g22},共4个元素。
(2)当g=g55时,邻域集合N={g44,g45,g46,g54,g55,g56,g64,g65,g66},共9个元素。
(3)当g=g79时,邻域集合N={g68,g69,g78,g79,g88,g89},共6个元素。
定义网格邻域函数N(g):G→P(G),对于给定输入网格g,返回g的网格邻域。N(gij)={gmn||i-m|≤1,|j-n|≤1.m,n∈N+}。其中gij是给定的输入网格,i和j分别是其对应的x和y方向的网格坐标。
(5)遍历集合T中所有轨迹ei。若遍历完成,则执行步骤(3),否则执行步骤(6)。
(6)判断轨迹ei是否至少包含网格邻域N(gk)中的一点且网格gk位置满足pos(N(gk-1))<pos(N(gk))。若是,则执行步骤(7),否则执行步骤(5)。
上述步骤中POS表示网格位置,此公式表示此网格邻域中的点都比前一网格邻域中的点的网格位置都小。
在网格位置的定义函数pos(t,g):T×G→N+中,两个输入参数取值范围均为N+。对于给定轨迹t和元素g,当存在下标i且为第一个时,使得ti=g,则函数值为i,公式如下:
Figure PCTCN2017114305-appb-000008
通过上述公式可以求出网格在轨迹首次出现的位置。
(7)若判定轨迹ei经过目标网格gk,满足过滤条件,则需要保留轨迹ei。将轨迹ei添加到gk对应结果集合Rk中。执行步骤(5)。
(8)求出所有Rk交集R,R即为T中所有包含轨迹t的相似轨迹集。执行步骤(9)。
(9)输出过滤后的轨迹集合R。
S407中,当比较结果为支持度值小于预设支持度阈值时,则判断为当前采样点网格坐标异常,并将当前采样点网格坐标对应的轨迹集合更新为包含起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;否则,判定当前采样点网格坐标正常。并可将判决结果进行显示。
S406中,根据当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值包括但不限于以下两种方式中的任意一种。
方式一:将当前采样点网格坐标对应的轨迹集合中的历史轨迹数量除以上一采样点网格坐标对应的轨迹集合中的历史轨迹数量,得到支持度值;例如假设当前采样点网格坐标对应的轨迹集合中的历史轨迹数量为2,上一采样点网格坐标对应的轨迹集合中的历史轨迹数量位6,则支持度值=2/6。
方式二:将当前采样点网格坐标对应的轨迹集合中的各历史轨迹的行车数量和除以上一采样点网格坐标对应的轨迹集合中的各历史轨迹行车数量和,得到支持度值。例如假设当前采样点网格坐标对应的轨迹集合中的历史轨迹数量为2,两条条历史轨迹的行车数分别为20和30;上一采样点网格坐标对应的轨迹集合中的历史轨迹数量位6,另外4条历史轨迹的行程数都为40,则支持度值=(20+30)/(20+30+40+40+40+40)=5/21。
采用哪一种计算方式可以根据需求灵活选定。
本实施例中,对于各个采样点,还可以计算各个采样点对应的轨迹距离和异常值中的进行至少一个进行显示,也便于后续统计管理。
在获取到当前采样点网格坐标后,可以采用以下计算公式计算所述当前采样点网格坐标对应的轨迹距离:
Figure PCTCN2017114305-appb-000009
公式8中pi-1、pi分别为上一采样点和当前采样点;RE为地球半径,acos是反余弦函数;
t1=cos(ai-1)×cos(ai)×cos(bi-1)×cos(bi);
t2=cos(ai-1)×sin(ai)×cos(bi-1)×sin(bi);
t3=sin(ai-1)×sin(bi-1),
Figure PCTCN2017114305-appb-000010
其中xi-1和yi-1为地理坐标pi-1的经度和纬度,xi和yi述为地理坐标pi的经度和纬度。
在获取到当前采样点网格坐标后,还可采用以下计算公式计算当前采样点网格坐标对应的异常值:
Figure PCTCN2017114305-appb-000011
公式9中
Figure PCTCN2017114305-appb-000012
其中x=support(i)*dist(pi-1,pi);λ为温度常量参数,其区域可以灵活变化,例如可以取150,θ为所述支持度阈值,dist(pi,pi-1)为采样点pi,pi-1的地球表面距离,请参见公式8。score(0)=score(1)=0。公式9可以从正面反映轨迹当前异常程度,取决于前后两点距离和当前支持度,该值越大,轨迹越异常。
本公开实施例提供的轨迹补全方法AE-AUG,步骤简明,实际应用时能快速求出一条路径让两不相邻网格连通。该实施例基于大量车辆历史GPS记录,生成历史轨迹数据,结合AE-AEG补全算法、异常检测算法、Bing Maps Tile System地图网格计算算法,实现了响应速度快、整体检测准确率高的车辆轨迹监测方法。
实施例二:
本实施例提供了一种车辆行驶轨迹监测系统,参见图7所示,包括:检测子系统61;网格子系统62,其包括网格管理器组件,该组件可以具有网格映射Mapping以及轨迹补全 Augmenting功能;数据库63,设置为存储维护网格轨迹库;以及实时数据采集子系统64。
检测子系统61设置为检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标,并通过网格子系统62将起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标(作为例子,转换算法可以采用上述映射算法),根据起点网格坐标和终点网格坐标从数据库63的网格轨迹库中查找(作为例子,可以采用上述反向查找方法)出包含起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
实时数据采集子系统64设置为对车辆在行程过程中当前所在位置的地理坐标进行采集并通过网格子系统62转换为采样点网格坐标后发给检测子系统61。采样规则可以根据应用场景灵活设定。
检测子系统61还设置为获取车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,并将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合中历史轨迹信息,根据当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值,并将得到的所述支持度值与预设支持度阈值进行比较,根据比较结果确定当前采样点网格坐标是否异常,且在比较结果为支持度值小于所述预设支持度阈值时,判断当前采样点网格坐标异常,并将当前采样点网格坐标对应的轨迹集合更新为包含起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
本实施例中检测子系统61、网格子系统62、实时数据采集子系统64实现各自功能,以执行,例如,实施例一中的方法。例如,检测子系统61还可以采用实施例一所示的方式进行轨迹距离和/或异常值的计算。本实施例对上述各系统的一种组网结构进行示例说明,参见图8所示,包括Hadoop平台和Web服务端,分别执行离线处理阶段和在线处理阶段任务。结合实际数据,各节点说明如下。
(1)历史车辆71指在过去任意时刻某一区域(例如深圳市)的出租车(当然也可以包括其他运营或私家车辆)群,本实施例可以仅关注其产生的历史位置数据。
(2)地理位置数据包收集器72(GPS Packet Collector):数据合并节点,收集合并该区域所有出租车辆上传的历史GPS数据记录。
(3)HDFS(分布式文件系统)73:Hadoop平台依赖的数据存储节点。由于出租车产生的数据总量较大,例如深圳市的出租车每天新增数据约40G文件,记录条数达500万余条。受限于存储空间,实际中最多可保存半年数据。
(4)数据清洗器74(Data Cleaner):由于网络不稳定、设备老化等原因,导致原始数 据存在大量异常情况,如字段缺失、记录上报延时、GPS坐标漂移、车辆间歇性失联等,大大降低数据质量。针对此类问题,清洗器在尽可能保证数据完整性的前提下,去掉异常记录。
(5)地理位置包接收器75(GPS Packet Receiver)是实时数据采集子系统64中的一部分,是实时数据中转节点,配置成在线接收检测车辆发送的GPS数据包。
(6)数据格式化器76(Data Formatter),是实时数据采集子系统64中的一部分,配置成提取检测车辆数据包车牌号、经纬度、上报时间、载客状态等字段,并格式化为语义对象,方便数据传输和分析。
(7)网格管理器77(Grid Manager),是网格子系统62中的一部分,是Hadoop平台和Web服务端公共组件,提供网格操作相关功能,如求网格邻域N(gi)、网格在轨迹中的索引位置pos(t,gi),同时可以包含网格映射(Mapping)和轨迹补全(Augmenting)两子组件。
(8)网格映射组件78(Mapping),是网格子系统62中的一部分,通过网格映射函数ρ,将经纬度坐标映射为网格坐标,实现轨迹点离散化。
(9)轨迹补全组件79(Augmenting),网格子系统62中的一部分,通过轨迹补全函数aug,在不相邻的两个网格插入网格,使两者连通。
(10)数据库管理器710(Database Manager):离线阶段时将Grid Manager生成的网格轨迹数据存入轨迹数据库,在线阶段时为检测引擎提供轨迹查询服务。
(11)轨迹数据库711(Trajectories Database),也即数据库:包含正序和逆序两部分数据。正序指通过轨迹编号索引所有轨迹点,逆序指通过网格坐标索引所有经过该点的轨迹。假设轨迹t1=<g1,g2,g3>,t2=<g1,g2,g4>,t3=<g2,g3,g4>,此部分为正序数据,对应的逆序数据为g1:<(t1,1),(t2,1)>,g2:<(t1,2),(t2,2),(t3,1)>,g3:<(t1,3),(t3,2)>,g4:<(t2,3),(t3,3)>。正序逆序数据互相冗余,目的是加快轨迹检索速度。
(12)检测引擎712(Detection Engine),检测子系统61的一部分:作为检测子系统61中的核心组件,实现检测算法。同时与Grid Manager和Database Manager进行交互,输入轨迹点和历史轨迹集合,检测该点是否异常。
(13)Web控制器713(Web Controller),检测子系统61中的一部分:与终端设备进行交互,发布检测结果。
基于上述组网系统,本实施例以车辆的整个监测过程为例进行说明,参见图9所示,包括以下步骤。
在S801,输入结构化实时车辆状态记录。出租车车载终端设备将车辆当前行驶状态信息上传至检测服务器,服务器抽取有效检测字段,并将数据结构化,最后将结构化数据传送到检测引擎。执行步骤S802。
在S802,检测引擎接收到新的轨迹点信息时,根据记录是否包含终点经纬度坐标,判断车辆是否开始新的运营轨迹,也即开始新的行程。若是,执行步骤S803,否则执行步骤S808。
在S803,出租车开始了一条新的运营轨迹,即新载客人,从记录中取出起点、终点经纬度坐标。执行步骤S804。
在S804,检测引擎向网格管理器发送命令,将起点、终点坐标映射为对应网格点。执行步骤S805。
在S805,检测引擎向数据库管理器发送检索命令,传递起、终点网格,求出历史轨迹集合。
求解时,由于轨迹数量众多,本实施例不采用轨迹库全遍历方法,即,逐条判断是否同时经过起、终点,而是通过网格逆向索引出各自落入的全部轨迹ID,通过两者交集求得。该集合由所有恰好或经过起点、终点网格的轨迹构成。然后,执行步骤S806。
在S806,设置初始检测结果。起始坐标点默认为正常,支持度为1,异常值、轨迹距离为0。然后,执行步骤S807。
在S807,发布检测结果。此处所有检测结果统一发送到出租车模拟检测子系统61控制台Web页面,以图表和地图形式展示。
在S808,读取车辆上一次检测结果状态信息。出租车仍在载客运营行驶过程中,当前轨迹点为在线轨迹中最新一点,且本次检测受到上一次检测结果影响,故需要读取上一次状态信息。执行步骤S809。
S809:读取记录经纬度坐标点,检测引擎向网格管理器发送命令,将坐标点映射为对应网格。执行步骤S810。
S810:判读最新坐标点对应的网格与上一坐标点对象的网格是否相同。若相同,执行步骤S811。否则执行步骤S812。
S811:最新轨迹点与上一轨迹点落入同一网格,所有检测状态保持不变。若上一轨迹点正常,当前轨迹点正常,否则异常。前后两点含有相同的支持度,异常值和异常距离。执行步骤S807。
S812:最新接收的轨迹网格与上一轨迹网格不同,需要重新计算当前网格支持度,需读取上一状态轨迹集合。执行步骤S813。
S813:记录过滤前轨迹集合包含的轨迹总数,即上一状态轨迹集合轨迹数count(Ti-1)。 执行步骤S813。
S814:根据hasPath函数过滤当前工作轨迹集合。hasPath函数两输入参数分别是被过滤的候选轨迹集合和作为过滤条件的参考轨迹,此处对应上一状态轨迹集合和当前最新接收的轨迹点,找出包含当前轨迹点的所有轨迹。执行步骤S815。
S815:由步骤S814得出过滤后的轨迹集合,记录过滤后集合包含的轨迹数目count(Tsi)。执行步骤S816。
S816:求出当前轨迹点的支持度support。计算如下:support=count(Tsi)/count(Ts,i-1)。执行步骤S817。
S817:判断当前轨迹点支持度support是否低于设定阈值。若否,则判定当前轨迹点是正常的,执行步骤S818;否则为异常的,执行步骤S822。
S818:根据步骤S817的判定标准,最新接收的轨迹点是正常的。正常指车辆从上一个轨迹点gi行驶到当前轨迹点gj的历史轨迹数目至少超过了设定参考值,即从gi到gj的走法属于常规路线。执行步骤S819。
S819:计算当前轨迹点所对应的异常值score。其中score(i-1)是上一状态异常值,与当前支持度呈负相关,与上一轨迹点到当前轨迹点的距离呈正相关。执行步骤S820。
S820:计算当前轨迹点对应的轨迹距离。轨迹距离为所有异常点和异常点到正常点之间的球面距离之和。只有当前和上一状态轨迹点均为正常时,异常距离才保持不变;否则需要累加dist(pi,pi-1)。执行步骤S821。
S821:设置当前轨迹点检测结果。根据上述计算的支持度、异常值、异常距离,重新设置当前轨迹点的检测结果。执行步骤S807。
S822:根据步骤S817的判定标准,最新接收的轨迹点是异常的。异常指车辆从上一个轨迹点gi行驶到当前轨迹点gj的历史轨迹数目不超过设定的参考值,即从gi到gj的走法属于非常规路线。执行步骤S823。
S823:重置轨迹集合到初始轨迹集合。由于上一个状态轨迹集合经过当前轨迹点过滤后,轨迹数过少已低于预设值,若不将轨迹集合重置到初始状态,则之后接收的所有轨迹点都会判定为异常的。执行步骤S819。
为了便于理解,下面结合一个在线检测例子进行说明。
如图10所示,假设有3组常规路线,即从起点S形式至终点D大部分出租车司机载客的首选线路,箭头方向代表行车方向,灰色网格代表常规线路所占区域。假定有40个司机沿H2线路行驶,30个司机沿H3线路,30个司机沿H1线路。H4线路为目标测试轨迹t,各黑点为服务器实际接收到的GPS坐标点,黑点编号代表服务器接收数据的先后顺序。
总体观察,测试轨迹t中大部分点落在常规线路所占网格,只有<g7,g8>落入新网格中。另一方面,虽然<g4,g5,g6>落入常规线路网格,但其出现方向与红色轨迹相反。
当检测开始时,服务器依次接收到<g1,g2,g3,g4>点,此段有H1历史轨迹支持,4点检测均为正常。当接收到<g5,g6>时,3组常规线路中只有H3轨迹经过该网格区域,但两者行进方向相反,即没有历史轨迹符合该走法,故<g5,g6>两点检测是异常的。而<g7,g8>完全落入新网格,不在任何历史轨迹所经区域上,故<g7,g8>也是异常点。当接收到g9时,虽然没有落入常规线路网格中,但其网格邻域在H2轨迹集上,该点正常。当接收到<g10,g11>时,测试轨迹回落到常规网格,且方向与H2轨迹集一致,因此<g10,g11>为正常点。
在检测过程中,历史轨迹工作集和支持度变化如下表2,假设异常判别阈值为0.1。
表2
网格点编号 工作集轨迹数 是否重置工作集 支持度 是否异常
1(S) 100 1.0
2 30 0.3
3 30 1.0
4 30 1.0
5 0 0.0
6 0 0.0
7 0 0.0
8 0 0.0
9 40 0.4
10 40 1.0
11(D) 40 1.0
如上表2所示,当接收到起点S时,历史轨迹工作集处于初始状态,T1=100。当第i个点到来,对上一状态工作集Ti-1进行过滤并求出支持度。每当支持度低于异常阈值时,需要将当前工作集重置为T1状态。最终输出正常点集为<g1,g2,g3,g4,g9,g10,g11>,异常点集为<g5,g6,g7,g8>。
本公开实施例提供的轨迹补全方法AE-AUG可以简单、快速地求出一条路径让两不相邻网格连通。本公开实施例提供的系统可以基于大量车辆历史GPS记录,生成历史轨迹数据,结 合AE-AEG补全算法、异常检测算法、Bing Maps Tile System地图网格计算算法实现对车辆行驶轨迹快速、可靠的检测,避免司机恶意绕行,提升用户体验满意度。
以上内容是结合实施方式对本公开实施例所作的说明,不能认定本公开只局限于这些说明。对于本公开所属技术领域的普通技术人员来说,在不脱离本公开构思的前提下,还可以做出若干推演、修改或替换,都应当视为属于本公开的保护范围。

Claims (22)

  1. 一种车辆行驶轨迹监测方法,包括:
    检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标;
    将所述行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标;并在网格轨迹库中查找出包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;
    对所述车辆在所述行程过程中当前所在位置的地理坐标进行采样并转换为采样点网格坐标;
    获取所述车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,并将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合中历史轨迹信息;
    根据所述当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值;
    将得到的所述支持度值与预设支持度阈值进行比较,根据比较结果确定所述当前采样点网格坐标是否异常。
  2. 如权利要求1所述的车辆行驶轨迹监测方法,其中,检测到车辆开始新的行程之前,还包括设置网格轨迹库中的各历史轨迹的过程,包括:
    获取各车辆的历史行程以及在各历史行程中的历史位置地理坐标;
    将各车辆在一个历史行程中的各历史位置地理坐标换成对应的网格坐标,并将该历史行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标;
    将每一车辆在所述历史行程中的起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标进行轨迹补全处理得到各车辆完成所述历史行程的历史轨迹,并得到各历史轨迹的行车数量。
  3. 如权利要求2所述的车辆行驶轨迹监测方法,其中,将在所述历史行程中的起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标进行轨迹补全处理包括:
    将所述起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标映射到网格系统对应的各网格中;
    以所述起点网格坐标对应的网格为起点,所述终点网格坐标对应的网格为终点,按照 各历史位置的网格坐标获取的时序,依次找到相邻两个网格;
    如果相邻两个网格在同一行或列,则以相邻两个网格在同一行或列之间的网格作为补全网格将两个网格连接;
    否则,确定所述相邻两个网格之间的网格组成的内矩形,然后以网格内矩形的短边为边长确定正方形,并以所述相邻网格中靠近起点的网格为起点取正方形对角上的各网格作为补全网格,然后再取对角线上最后一个网格所在行或列到所述相邻网格中另一个网格之间的所有网格作为补全网格。
  4. 如权利要求2所述的车辆行驶轨迹监测方法,其中,在网格轨迹库中查找出包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合包括:
    在所述网格轨迹库中查找出所有包含所述起点网格坐标的历史轨迹作为起点轨迹集合,并查找出所有包含所述终点网格坐标的历史轨迹作为终点轨迹集合;
    取所述起点轨迹集合和所述终点轨迹集合的交集得到包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
  5. 如权利要求2所述的车辆行驶轨迹监测方法,其中,将所述上一采样点网格坐标对应的轨迹集合中不包含当前采样点网格坐标的历史轨迹清除包括:
    判断上一采样点网格坐标对应的轨迹集合中的某一历史轨迹对应的网格集合中是否包含所述当前采样点网格坐标对应的网格,如是,判断该历史轨迹包含当前采样点网格坐标;否则,判断该历史轨迹对应的网格集合中是否包含所述当前采样点网格坐标对应的网格邻域中的某一相邻网格,且该相邻网格满足pos(N(gk-1))<pos(N(gk)),如是,判断该历史轨迹包含当前采样点网格坐标,否则,判断该历史轨迹不包含当前采样点网格坐标。
  6. 如权利要求2所述的车辆行驶轨迹监测方法,其中,根据所述当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值包括:
    将当前采样点网格坐标对应的轨迹集合中的历史轨迹数量除以所述上一采样点网格坐标对应的轨迹集合中的历史轨迹数量,得到支持度值;
    或,
    将当前采样点网格坐标对应的轨迹集合中的各历史轨迹的行车数量和除以所述上一采样点网格坐标对应的轨迹集合中的各历史轨迹行车数量和,得到支持度值。
  7. 如权利要求1-6任一项所述的车辆行驶轨迹监测方法,其中,判断车辆是否开始新的行程包括:
    接收车辆当前发送的实时位置上报信息,所述实时位置上报信息中包含当前所在位置的地理坐标以及当前行程的起点地理坐标和终点的地理坐标;
    判断所述车辆当前实时位置上报信息中的起点地理坐标和终点的地理坐标与上一次发送的实时位置上报信息中的起点地理坐标和终点的地理坐标是否相同,如否,则判断所述车辆开始新的行程;否则,判断所述车辆在执行原行程。
  8. 如权利要求1-6任一项所述的车辆行驶轨迹监测方法,其中,所述比较结果为所述支持度值小于所述预设支持度阈值时,判断所述当前采样点网格坐标异常,将所述当前采样点网格坐标对应的轨迹集合更新为包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
  9. 如权利要求1-6任一项所述的车辆行驶轨迹监测方法,其中,在获取到所述当前采样点网格坐标后,还包括采用以下计算公式计算所述当前采样点网格坐标对应的轨迹距离;
    Figure PCTCN2017114305-appb-100001
    所述pi-1、pi分别为上一采样点和当前采样点;所述RE为地球半径;所述t1=cos(ai-1)×cos(ai)×cos(bi-1)×cos(bi),所述t2=cos(ai-1)×sin(ai)×cos(bi-1)×sin(bi),所述t3=sin(ai-1)×sin(bi-1),所述
    Figure PCTCN2017114305-appb-100002
    所述
    Figure PCTCN2017114305-appb-100003
    所述xi-1和yi-1为所述pi-1的经度和纬度,所述xi和yi述为所述pi的经度和纬度。
  10. 如权利要求9所述的车辆行驶轨迹监测方法,其中,在获取到所述当前采样点网格坐标后,还包括采用以下计算公式计算所述当前采样点网格坐标对应的异常值;
    Figure PCTCN2017114305-appb-100004
    所述
    Figure PCTCN2017114305-appb-100005
    所述x=support(i)*dist(pi-1,pi);所述λ为温度常量参数,所述θ为所述支持度阈值,所述dist(pi,pi-1)为采样点pi,pi-1的地球表面距离。
  11. 一种车辆行驶轨迹监测系统,其中,包括检测子系统、网格子系统、数据库以及实时数据采集子系统;
    检测子系统,设置为检测到车辆开始新的行程时,获取该行程之起点地理坐标和终点的地理坐标,并通过所述网格子系统将所述起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标,根据所述起点网格坐标和终点网格坐标从所述数据库的网格轨迹库中查找出包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合;
    所述实时数据采集子系统设置为对所述车辆在所述行程过程中当前所在位置的地理坐标进行采集并通过所述网格子系统转换为采样点网格坐标后发给所述检测子系统;
    所述检测子系统还设置为获取所述车辆上一采样点网格坐标对应的轨迹集合中历史轨迹信息,并将该轨迹集合中不包含当前采样点网格坐标的历史轨迹清除,得到当前采样点网格坐标对应的轨迹集合中历史轨迹信息,根据所述当前采样点网格坐标对应的轨迹集合以及上一采样点网格坐标对应的轨迹集合计算当前采样点网格坐标的支持度值,并将得到的所述支持度值与预设支持度阈值进行比较,根据比较结果确定所述当前采样点网格坐标是否异常。
  12. 如权利要求11所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为在所述比较结果为所述支持度值小于所述预设支持度阈值时,判断所述当前采样点网格坐标异常,并将所述当前采样点网格坐标对应的轨迹集合更新为包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
  13. 如权利要求11所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为:
    获取各车辆的历史行程以及在各历史行程中的历史位置地理坐标;
    将各车辆在一个历史行程中的各历史位置地理坐标换成对应的网格坐标,并将该历史行程的起点地理坐标和终端地理坐标分别转换为起点网格坐标和终点网格坐标;
    将每一车辆在所述历史行程中的起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标进行轨迹补全处理得到各车辆完成所述历史行程的历史轨迹,并得到各历史轨迹的行车数量。
  14. 如权利要求13所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为:
    将所述起点网格坐标、各个历史位置的网格坐标、以及终点网格坐标映射到网格系统 对应的各网格中;
    以所述起点网格坐标对应的网格为起点,所述终点网格坐标对应的网格为终点,按照各历史位置的网格坐标获取的时序,依次找到相邻两个网格;
    如果相邻两个网格在同一行或列,则以相邻两个网格在同一行或列之间的网格作为补全网格将两个网格连接;
    否则,确定所述相邻两个网格之间的网格组成的内矩形,然后以网格内矩形的短边为边长确定正方形,并以所述相邻网格中靠近起点的网格为起点取正方形对角上的各网格作为补全网格,然后再取对角线上最后一个网格所在行或列到所述相邻网格中另一个网格之间的所有网格作为补全网格。
  15. 如权利要求13所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为:
    在所述网格轨迹库中查找出所有包含所述起点网格坐标的历史轨迹作为起点轨迹集合,并查找出所有包含所述终点网格坐标的历史轨迹作为终点轨迹集合;
    取所述起点轨迹集合和所述终点轨迹集合的交集得到包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
  16. 如权利要求13所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为:
    判断上一采样点网格坐标对应的轨迹集合中的某一历史轨迹对应的网格集合中是否包含所述当前采样点网格坐标对应的网格,如是,判断该历史轨迹包含当前采样点网格坐标;否则,判断该历史轨迹对应的网格集合中是否包含所述当前采样点网格坐标对应的网格邻域中的某一相邻网格,且该相邻网格满足pos(N(gk-1))<pos(N(gk)),如是,判断该历史轨迹包含当前采样点网格坐标,否则,判断该历史轨迹不包含当前采样点网格坐标。
  17. 如权利要求13所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为:
    将当前采样点网格坐标对应的轨迹集合中的历史轨迹数量除以所述上一采样点网格坐标对应的轨迹集合中的历史轨迹数量,得到支持度值;
    或,
    将当前采样点网格坐标对应的轨迹集合中的各历史轨迹的行车数量和除以所述上一 采样点网格坐标对应的轨迹集合中的各历史轨迹行车数量和,得到支持度值。
  18. 如权利要求11-17任一项所述的车辆行驶轨迹监测系统,其中,所述检测子系统还设置为:
    接收车辆当前发送的实时位置上报信息,所述实时位置上报信息中包含当前所在位置的地理坐标以及当前行程的起点地理坐标和终点的地理坐标;
    判断所述车辆当前实时位置上报信息中的起点地理坐标和终点的地理坐标与上一次发送的实时位置上报信息中的起点地理坐标和终点的地理坐标是否相同,如否,则判断所述车辆开始新的行程;否则,判断所述车辆在执行原行程。
  19. 如权利要求11-17任一项所述的车辆行驶轨迹监测系统,其中,所述比较结果为所述支持度值小于所述预设支持度阈值时,所述检测子系统判断所述当前采样点网格坐标异常,将所述当前采样点网格坐标对应的轨迹集合更新为包含所述起点网格坐标和终点网格坐标的所有历史轨迹的轨迹集合。
  20. 如权利要求11-17任一项所述的车辆行驶轨迹监测系统,其中,在获取到所述当前采样点网格坐标后,所述检测子系统采用以下计算公式计算所述当前采样点网格坐标对应的轨迹距离;
    Figure PCTCN2017114305-appb-100006
    所述pi-1、pi分别为上一采样点和当前采样点;所述RE为地球半径;所述t1=cos(ai-1)×cos(ai)×cos(bi-1)×cos(bi),所述t2=cos(ai-1)×sin(ai)×cos(bi-1)×sin(bi),所述t3=sin(ai-1)×sin(bi-1),所述
    Figure PCTCN2017114305-appb-100007
    所述
    Figure PCTCN2017114305-appb-100008
    所述xi-1和yi-1为所述pi-1的经度和纬度,所述xi和yi述为所述pi的经度和纬度。
  21. 如权利要求20所述的车辆行驶轨迹监测系统,其中,在获取到所述当前采样点网格坐标后,所述检测子系统采用以下计算公式计算所述当前采样点网格坐标对应的异常值;
    Figure PCTCN2017114305-appb-100009
    所述
    Figure PCTCN2017114305-appb-100010
    所述x=support(i)*dist(pi-1,pi);所述λ为温度常量参数,所 述θ为所述支持度阈值,所述dist(pi,pi-1)为采样点pi,pi-1的地球表面距离。
  22. 一种计算机存储介质,所述计算机存储介质存储有执行指令,所述执行指令设置为执行权利要求1至10中任一项所述的车辆行驶轨迹监测方法。
PCT/CN2017/114305 2016-12-01 2017-12-01 车辆行驶轨迹监测方法及系统 WO2018099480A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201611092075.2A CN108133611A (zh) 2016-12-01 2016-12-01 车辆行驶轨迹监测方法及系统
CN201611092075.2 2016-12-01

Publications (1)

Publication Number Publication Date
WO2018099480A1 true WO2018099480A1 (zh) 2018-06-07

Family

ID=62241220

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/114305 WO2018099480A1 (zh) 2016-12-01 2017-12-01 车辆行驶轨迹监测方法及系统

Country Status (2)

Country Link
CN (1) CN108133611A (zh)
WO (1) WO2018099480A1 (zh)

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109035768A (zh) * 2018-07-25 2018-12-18 北京交通大学 一种出租车绕路行为的识别方法
CN109670656A (zh) * 2019-02-27 2019-04-23 重庆邮电大学 一种基于4g网络的无人机最优通信路线规划方法
CN109726737A (zh) * 2018-11-27 2019-05-07 武汉极意网络科技有限公司 基于轨迹的异常行为检测方法及装置
CN109933066A (zh) * 2019-02-22 2019-06-25 上海岚豹智能科技有限公司 一种机器人移动轨迹回放的方法及设备
CN110276950A (zh) * 2019-06-24 2019-09-24 华南理工大学 一种基于卡口视频数据的城市交通出行链重构方法
CN111190891A (zh) * 2019-12-27 2020-05-22 武汉长江通信产业集团股份有限公司 一种多语义轨迹数据分段存储方法
CN111383294A (zh) * 2018-12-27 2020-07-07 杭州海康威视数字技术股份有限公司 一种安防雷达系统中防区的绘制方法及装置
CN111784728A (zh) * 2020-06-29 2020-10-16 杭州海康威视数字技术股份有限公司 轨迹处理方法、装置、设备及存储介质
CN111858620A (zh) * 2019-04-29 2020-10-30 北京四维图新科技股份有限公司 地图数据更新方法及装置
CN112035873A (zh) * 2020-08-18 2020-12-04 合肥市大数据资产运营有限公司 一种时空轨迹数据脱敏方法
CN112070247A (zh) * 2020-09-01 2020-12-11 上海钧正网络科技有限公司 一种运维网格的管理方法、装置、介质及电子设备
CN112153573A (zh) * 2020-09-28 2020-12-29 平安国际融资租赁有限公司 基于位置轨迹的切分方法、装置、计算机设备及存储介质
CN112255658A (zh) * 2020-10-28 2021-01-22 湖南聚沙信息技术有限公司 基于gps定位点统计车辆行驶圈数的方法、装置及介质
CN112329815A (zh) * 2020-09-30 2021-02-05 华南师范大学 模型训练方法、行驶轨迹异常性检测方法、装置和介质
CN112929816A (zh) * 2021-02-01 2021-06-08 北京嘀嘀无限科技发展有限公司 车辆异常行为识别方法、装置、介质以及计算机程序产品
CN113096209A (zh) * 2021-03-29 2021-07-09 惠州华阳通用电子有限公司 一种车载影像轨迹线的显示方法
CN113329398A (zh) * 2021-05-28 2021-08-31 南京邮电大学 一种基于整数规划的敏感上下车模式的隐藏方法
CN113742536A (zh) * 2021-09-17 2021-12-03 上海交通大学 一种面向轨迹流数据的连续范围查询的方法和系统
CN113848071A (zh) * 2021-09-24 2021-12-28 上汽通用五菱汽车股份有限公司 定位故障测试方法、装置、车辆及计算机可读存储介质
CN113868362A (zh) * 2021-11-30 2021-12-31 亿海蓝(北京)数据技术股份公司 航路轨迹构建方法、系统、船舶和船舶管理系统
CN113984077A (zh) * 2021-10-26 2022-01-28 武汉大学 一种电力巡检导航路径的获取方法和系统
CN114074674A (zh) * 2020-08-14 2022-02-22 上海汽车工业(集团)总公司 一种引导车辆历史轨迹曲线的获取方法及相关装置
CN114166241A (zh) * 2021-10-27 2022-03-11 北京中交兴路车联网科技有限公司 一种货运车载终端异常检测方法、装置、存储介质及终端
CN114550449A (zh) * 2022-02-16 2022-05-27 腾讯云计算(北京)有限责任公司 车辆轨迹补全方法、装置、计算机可读介质及电子设备
CN114812584A (zh) * 2021-01-19 2022-07-29 北京四维图新科技股份有限公司 一种路径规划方法、系统、存储介质及设备
CN115342817A (zh) * 2022-10-20 2022-11-15 北京理工大学 无人履带车辆轨迹跟踪状态监测方法、系统、设备、介质
US11536582B2 (en) * 2017-08-10 2022-12-27 Beijing Didi Infinity Technology And Development Co., Ltd. System and method for estimating travel time and distance
CN116011694A (zh) * 2023-03-24 2023-04-25 北京快成科技有限公司 车辆行驶路线异常的识别方法、存储介质及装置
CN117150431A (zh) * 2023-08-31 2023-12-01 南京领行科技股份有限公司 行驶轨迹融合方法、装置、服务器以及介质
CN117891892A (zh) * 2024-03-14 2024-04-16 南京师范大学 一种基于历史巡护数据的护林员轨迹优化算法

Families Citing this family (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109218985B (zh) * 2018-08-08 2020-09-22 上海中交水运设计研究有限公司 一种基于模式相似度的船舶作业行为异常检测方法及系统
CN110874660B (zh) * 2018-08-29 2024-03-15 中兴通讯股份有限公司 道路测试路线规划方法、装置以及存储介质
CN110751531A (zh) * 2018-11-13 2020-02-04 北京嘀嘀无限科技发展有限公司 轨迹识别方法、装置及电子设备
CN110992072A (zh) * 2018-11-30 2020-04-10 北京嘀嘀无限科技发展有限公司 异常订单预测方法和系统
CN109766777B (zh) * 2018-12-18 2021-08-13 东软集团股份有限公司 异常轨迹的检测方法、装置、存储介质及电子设备
CN111444294B (zh) * 2019-01-17 2023-10-10 杭州海康威视系统技术有限公司 一种轨迹补全方法、装置及电子设备
CN111581306B (zh) * 2019-02-15 2023-04-14 阿里巴巴集团控股有限公司 一种行车轨迹模拟方法及装置
CN111695713B (zh) * 2019-03-12 2023-09-15 北京嘀嘀无限科技发展有限公司 一种历史轨迹数据处理方法及装置
CN111752254A (zh) * 2019-03-28 2020-10-09 深圳市商汤科技有限公司 轨迹跟踪控制器的测试方法、装置、介质及设备
CN111721312B (zh) * 2019-05-28 2022-05-06 腾讯科技(深圳)有限公司 一种运行轨迹生成方法和装置
CN110188242B (zh) * 2019-05-30 2020-09-04 北京三快在线科技有限公司 无人驾驶设备定位方法、装置、无人驾驶设备和存储介质
CN110428500B (zh) * 2019-07-29 2022-07-12 腾讯科技(深圳)有限公司 轨迹数据处理方法、装置、存储介质以及设备
CN110609881A (zh) * 2019-08-28 2019-12-24 中山大学 车辆轨迹偏离检测方法、系统和存储介质
CN111024096B (zh) * 2019-11-08 2022-06-03 安徽中科美络信息技术有限公司 一种车辆定位数据上传方法和车载终端
CN110851551B (zh) * 2019-11-12 2022-05-06 芜湖盟博科技有限公司 一种实时自动分析车辆轨迹数据和查询方法
CN111811642B (zh) * 2020-07-22 2022-05-03 三一重机有限公司 一种挖掘机发动机的故障检测方法及故障检测装置
CN112052873B (zh) * 2020-07-30 2022-10-21 北京邮电大学 基于通用转换器的异常轨迹检测方法及装置
CN111652201B (zh) * 2020-08-10 2020-10-27 中国人民解放军国防科技大学 基于深度视频事件补全的视频数据异常识别方法和装置
CN112258879B (zh) * 2020-10-16 2021-11-02 安徽亿力停车场投资有限公司 一种利用随机矩阵优化城市停车构造方法
CN112633165A (zh) * 2020-12-23 2021-04-09 远光软件股份有限公司 基于车辆车厢的采样监督方法、系统、存储介质及电子设备
CN112598767B (zh) * 2020-12-29 2024-05-10 厦门市美亚柏科信息股份有限公司 基于时空大数据的轨迹行为分析方法、终端设备及存储介质
CN112699942B (zh) * 2020-12-30 2024-08-02 东软睿驰汽车技术(沈阳)有限公司 一种运营车辆识别方法、装置、设备及存储介质
CN113487858B (zh) * 2021-06-25 2022-05-10 大连海事大学 一种面向路网数据的异常轨迹检测方法
CN116029736B (zh) * 2023-01-05 2023-09-29 浙江警察学院 一种网约车异常轨迹实时检测和安全预警方法、系统
CN118555543A (zh) * 2023-02-27 2024-08-27 腾讯云计算(北京)有限责任公司 轨迹异常检测方法、装置、计算机设备和存储介质
CN116819582B (zh) * 2023-08-28 2023-11-10 中国联合网络通信有限公司湖南省分公司 一种巡查打卡控制方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5774824A (en) * 1995-08-24 1998-06-30 The Penn State Research Foundation Map-matching navigation system
CN101991935A (zh) * 2009-08-12 2011-03-30 英业达股份有限公司 设定运动参数进行运动装置的控制方法及其装置
CN103512581A (zh) * 2012-06-28 2014-01-15 北京搜狗科技发展有限公司 一种路径规划方法和装置
CN103605362A (zh) * 2013-09-11 2014-02-26 天津工业大学 基于车辆轨迹多特征的运动模式学习及异常检测方法
CN104567906A (zh) * 2015-01-14 2015-04-29 合肥革绿信息科技有限公司 一种基于北斗的城市路网车辆路径规划方法及装置

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4305300B2 (ja) * 2004-07-07 2009-07-29 トヨタ自動車株式会社 車両追跡装置
CN101383090A (zh) * 2008-10-24 2009-03-11 北京航空航天大学 一种并行路网结构下的浮动车信息处理方法
CN101894460B (zh) * 2009-05-22 2014-07-09 触动多媒体技术(上海)有限公司 空出租车位置信息控制方法
CN103903433B (zh) * 2012-12-27 2016-09-14 南京中兴新软件有限责任公司 一种道路交通状态的实时动态判别方法及装置
CN103325245B (zh) * 2013-06-25 2015-01-07 中国科学院信息工程研究所 一种黑名单车辆的时空行驶轨迹预测方法
CN104462190B (zh) * 2014-10-24 2018-01-05 中国电子科技集团公司第二十八研究所 一种基于海量空间轨迹挖掘的在线的位置预测方法
CN104331466B (zh) * 2014-10-31 2018-01-19 南京邮电大学 基于时空邻近搜索的移动轨迹序列模式快速挖掘方法
CN104598992A (zh) * 2015-01-12 2015-05-06 北京中交兴路车联网科技有限公司 一种车辆常跑路线分析方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5774824A (en) * 1995-08-24 1998-06-30 The Penn State Research Foundation Map-matching navigation system
CN101991935A (zh) * 2009-08-12 2011-03-30 英业达股份有限公司 设定运动参数进行运动装置的控制方法及其装置
CN103512581A (zh) * 2012-06-28 2014-01-15 北京搜狗科技发展有限公司 一种路径规划方法和装置
CN103605362A (zh) * 2013-09-11 2014-02-26 天津工业大学 基于车辆轨迹多特征的运动模式学习及异常检测方法
CN104567906A (zh) * 2015-01-14 2015-04-29 合肥革绿信息科技有限公司 一种基于北斗的城市路网车辆路径规划方法及装置

Cited By (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11536582B2 (en) * 2017-08-10 2022-12-27 Beijing Didi Infinity Technology And Development Co., Ltd. System and method for estimating travel time and distance
CN109035768A (zh) * 2018-07-25 2018-12-18 北京交通大学 一种出租车绕路行为的识别方法
CN109726737A (zh) * 2018-11-27 2019-05-07 武汉极意网络科技有限公司 基于轨迹的异常行为检测方法及装置
CN109726737B (zh) * 2018-11-27 2020-11-10 武汉极意网络科技有限公司 基于轨迹的异常行为检测方法及装置
CN111383294A (zh) * 2018-12-27 2020-07-07 杭州海康威视数字技术股份有限公司 一种安防雷达系统中防区的绘制方法及装置
CN111383294B (zh) * 2018-12-27 2023-10-17 杭州海康威视数字技术股份有限公司 一种安防雷达系统中防区的绘制方法及装置
CN109933066A (zh) * 2019-02-22 2019-06-25 上海岚豹智能科技有限公司 一种机器人移动轨迹回放的方法及设备
CN109670656A (zh) * 2019-02-27 2019-04-23 重庆邮电大学 一种基于4g网络的无人机最优通信路线规划方法
CN109670656B (zh) * 2019-02-27 2023-04-07 重庆邮电大学 一种基于4g网络的无人机最优通信路线规划方法
CN111858620A (zh) * 2019-04-29 2020-10-30 北京四维图新科技股份有限公司 地图数据更新方法及装置
CN111858620B (zh) * 2019-04-29 2024-03-01 北京四维图新科技股份有限公司 地图数据更新方法及装置
CN110276950A (zh) * 2019-06-24 2019-09-24 华南理工大学 一种基于卡口视频数据的城市交通出行链重构方法
CN111190891A (zh) * 2019-12-27 2020-05-22 武汉长江通信产业集团股份有限公司 一种多语义轨迹数据分段存储方法
CN111784728A (zh) * 2020-06-29 2020-10-16 杭州海康威视数字技术股份有限公司 轨迹处理方法、装置、设备及存储介质
CN111784728B (zh) * 2020-06-29 2023-08-22 杭州海康威视数字技术股份有限公司 轨迹处理方法、装置、设备及存储介质
CN114074674B (zh) * 2020-08-14 2024-04-09 上海汽车工业(集团)总公司 一种引导车辆历史轨迹曲线的获取方法及相关装置
CN114074674A (zh) * 2020-08-14 2022-02-22 上海汽车工业(集团)总公司 一种引导车辆历史轨迹曲线的获取方法及相关装置
CN112035873A (zh) * 2020-08-18 2020-12-04 合肥市大数据资产运营有限公司 一种时空轨迹数据脱敏方法
CN112070247A (zh) * 2020-09-01 2020-12-11 上海钧正网络科技有限公司 一种运维网格的管理方法、装置、介质及电子设备
CN112153573A (zh) * 2020-09-28 2020-12-29 平安国际融资租赁有限公司 基于位置轨迹的切分方法、装置、计算机设备及存储介质
CN112153573B (zh) * 2020-09-28 2023-04-07 平安国际融资租赁有限公司 基于位置轨迹的切分方法、装置、计算机设备及存储介质
CN112329815B (zh) * 2020-09-30 2022-07-22 华南师范大学 模型训练方法、行驶轨迹异常性检测方法、装置和介质
CN112329815A (zh) * 2020-09-30 2021-02-05 华南师范大学 模型训练方法、行驶轨迹异常性检测方法、装置和介质
CN112255658B (zh) * 2020-10-28 2024-03-26 湖南聚沙信息技术有限公司 基于gps定位点统计车辆行驶圈数的方法、装置及介质
CN112255658A (zh) * 2020-10-28 2021-01-22 湖南聚沙信息技术有限公司 基于gps定位点统计车辆行驶圈数的方法、装置及介质
CN114812584A (zh) * 2021-01-19 2022-07-29 北京四维图新科技股份有限公司 一种路径规划方法、系统、存储介质及设备
CN112929816B (zh) * 2021-02-01 2023-06-30 北京嘀嘀无限科技发展有限公司 车辆异常行为识别方法、装置、介质以及计算机程序产品
CN112929816A (zh) * 2021-02-01 2021-06-08 北京嘀嘀无限科技发展有限公司 车辆异常行为识别方法、装置、介质以及计算机程序产品
CN113096209B (zh) * 2021-03-29 2024-06-04 惠州华阳通用电子有限公司 一种车载影像轨迹线的显示方法
CN113096209A (zh) * 2021-03-29 2021-07-09 惠州华阳通用电子有限公司 一种车载影像轨迹线的显示方法
CN113329398A (zh) * 2021-05-28 2021-08-31 南京邮电大学 一种基于整数规划的敏感上下车模式的隐藏方法
CN113742536A (zh) * 2021-09-17 2021-12-03 上海交通大学 一种面向轨迹流数据的连续范围查询的方法和系统
CN113742536B (zh) * 2021-09-17 2024-04-19 上海交通大学 一种面向轨迹流数据的连续范围查询的方法和系统
CN113848071B (zh) * 2021-09-24 2024-05-28 上汽通用五菱汽车股份有限公司 定位故障测试方法、装置、车辆及计算机可读存储介质
CN113848071A (zh) * 2021-09-24 2021-12-28 上汽通用五菱汽车股份有限公司 定位故障测试方法、装置、车辆及计算机可读存储介质
CN113984077A (zh) * 2021-10-26 2022-01-28 武汉大学 一种电力巡检导航路径的获取方法和系统
CN114166241A (zh) * 2021-10-27 2022-03-11 北京中交兴路车联网科技有限公司 一种货运车载终端异常检测方法、装置、存储介质及终端
CN113868362A (zh) * 2021-11-30 2021-12-31 亿海蓝(北京)数据技术股份公司 航路轨迹构建方法、系统、船舶和船舶管理系统
CN114550449A (zh) * 2022-02-16 2022-05-27 腾讯云计算(北京)有限责任公司 车辆轨迹补全方法、装置、计算机可读介质及电子设备
CN114550449B (zh) * 2022-02-16 2023-03-14 腾讯云计算(北京)有限责任公司 车辆轨迹补全方法、装置、计算机可读介质及电子设备
CN115342817B (zh) * 2022-10-20 2023-02-03 北京理工大学 无人履带车辆轨迹跟踪状态监测方法、系统、设备、介质
CN115342817A (zh) * 2022-10-20 2022-11-15 北京理工大学 无人履带车辆轨迹跟踪状态监测方法、系统、设备、介质
CN116011694A (zh) * 2023-03-24 2023-04-25 北京快成科技有限公司 车辆行驶路线异常的识别方法、存储介质及装置
CN117150431A (zh) * 2023-08-31 2023-12-01 南京领行科技股份有限公司 行驶轨迹融合方法、装置、服务器以及介质
CN117150431B (zh) * 2023-08-31 2024-06-11 南京领行科技股份有限公司 行驶轨迹融合方法、装置、服务器以及介质
CN117891892A (zh) * 2024-03-14 2024-04-16 南京师范大学 一种基于历史巡护数据的护林员轨迹优化算法
CN117891892B (zh) * 2024-03-14 2024-05-31 南京师范大学 一种基于历史巡护数据的护林员轨迹优化方法

Also Published As

Publication number Publication date
CN108133611A (zh) 2018-06-08

Similar Documents

Publication Publication Date Title
WO2018099480A1 (zh) 车辆行驶轨迹监测方法及系统
KR102205096B1 (ko) 거래 위험 검출 방법 및 장치
He et al. RoadRunner: improving the precision of road network inference from GPS trajectories
He et al. Mapping to cells: A simple method to extract traffic dynamics from probe vehicle data
CN107241512B (zh) 基于手机数据的城际交通出行方式判断方法和设备
CN105261217B (zh) 一种基于密度聚类算法的城市交通拥堵状态检测方法
Zheng et al. Probabilistic range queries for uncertain trajectories on road networks
US20190329788A1 (en) Road condition status prediction method, device, and server, and storage medium
WO2022143549A1 (zh) 一种基于收费数据的高速公路路况的监测方法及装置
US20160102987A1 (en) Method for inferring type of road segment
CN107885795B (zh) 一种卡口数据的数据校验方法、系统和装置
Hong et al. Recognizing network trip patterns using a spatio-temporal vehicle trajectory clustering algorithm
Li et al. Automatic generation and validation of road maps from GPS trajectory data sets
CN107808552A (zh) 航班行为检测方法和装置
JP5697810B2 (ja) プローブデータ処理装置及びプローブデータ処理方法及びプログラム及びプローブデータ処理システム
CN113276874B (zh) 一种车辆行驶轨迹处理方法及相关装置
Mao et al. Online clustering of streaming trajectories
US11537767B2 (en) Automated control through a traffic model
Chen et al. Local path searching based map matching algorithm for floating car data
RU2664034C1 (ru) Способ и система создания информации о трафике, которая будет использована в картографическом приложении, выполняемом на электронном устройстве
Hu et al. A comprehensive survey of recommendation system based on taxi GPS trajectory
CN111949840B (zh) 一种基于物联网数据的拓扑图结构的构建方法、装置
Wang et al. Large-scale freeway network traffic monitoring: A map-matching algorithm based on low-logging frequency GPS probe data
Felemban et al. An Interactive Analysis Platform for Bus Movement: A Case Study of One of the World’s Largest Annual Gathering
Efentakis et al. Crowdsourcing turning-restrictions from map-matched trajectories

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17875202

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17875202

Country of ref document: EP

Kind code of ref document: A1