WO2019033565A1 - 一种获取发射概率、转移概率以及序列定位的方法和装置 - Google Patents

一种获取发射概率、转移概率以及序列定位的方法和装置 Download PDF

Info

Publication number
WO2019033565A1
WO2019033565A1 PCT/CN2017/108647 CN2017108647W WO2019033565A1 WO 2019033565 A1 WO2019033565 A1 WO 2019033565A1 CN 2017108647 W CN2017108647 W CN 2017108647W WO 2019033565 A1 WO2019033565 A1 WO 2019033565A1
Authority
WO
WIPO (PCT)
Prior art keywords
mrs
terminal
target
information
trajectory data
Prior art date
Application number
PCT/CN2017/108647
Other languages
English (en)
French (fr)
Inventor
朱方舟
袁明轩
曾嘉
饶卫雄
Original Assignee
华为技术有限公司
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 华为技术有限公司 filed Critical 华为技术有限公司
Priority to CN201780087427.1A priority Critical patent/CN110326323B/zh
Priority to EP17921756.7A priority patent/EP3657841A1/en
Publication of WO2019033565A1 publication Critical patent/WO2019033565A1/zh
Priority to US16/791,795 priority patent/US11290975B2/en
Priority to US17/688,087 priority patent/US20220191818A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • H04W64/003—Locating users or terminals or network equipment for network management purposes, e.g. mobility management locating network equipment
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/08—Testing, supervising or monitoring using real traffic
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/10—Scheduling measurement reports ; Arrangements for measurement reports
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/023—Services making use of location information using mutual or relative location information between multiple location based services [LBS] targets or of distance thresholds
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/029—Location-based management or tracking services
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • H04W64/006—Locating users or terminals or network equipment for network management purposes, e.g. mobility management with additional information processing, e.g. for direction or speed determination
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/04—Wireless resource allocation
    • H04W72/044—Wireless resource allocation based on the type of the allocated resource
    • H04W72/0446—Resources in time domain, e.g. slots or frames
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W76/00—Connection management
    • H04W76/10—Connection setup
    • H04W76/11—Allocation or use of connection identifiers

Definitions

  • the present invention relates to the field of telecommunications positioning, and more particularly to a method and apparatus for obtaining a transmission probability for sequence positioning.
  • Telecommunications positioning refers to calculating the location of the mobile device by transmitting data to the pipeline side (eg, a telecommunications carrier) and base station side data by the mobile device.
  • the common telecommunication positioning technologies include the Range-Based method, the Fingerprint method, and the sequence positioning method.
  • the main idea of the sequence positioning method is to describe the positioning process as a matching model from the observed sequence to the hidden sequence, latitude and longitude.
  • the position is the hidden state and the signal strength is taken as the observation.
  • an sequence of observation values is used as an input, and an optimal hidden state sequence is outputted in one-to-one correspondence with the input sequence of observation values as a positioning result.
  • the core of this type of positioning method is to define the emission probability (emission probability) and the transition probability.
  • the probability of transmission is the mapping from the hidden state to the observed value.
  • the transition probability refers to the transition between different hidden states.
  • the main advantage of the sequence location method is that the context information of the location is utilized, so that the prediction result of each location can be constrained by the previous location, and the obtained trajectory is relatively smooth, which can effectively avoid the occurrence of the Range-Based method and the Fingerprint method.
  • the prediction result is "jumping".
  • the acquisition probability of transmission and the probability of transition directly affect the positioning ability of the sequence location method.
  • the probability of transmission is obtained by modeling the mean variance of the signal strength.
  • the emission probability obtained by this method cannot express complex observation information and directly affects Positioning accuracy and reliability of the sequence positioning method.
  • the present application provides a method for obtaining a transmission probability, so that the obtained transmission probability can express complex observation information, and the use of the transmission probability for sequence positioning can improve the positioning accuracy of the sequence positioning method. ,reliability.
  • the present application provides a method for obtaining an emission probability, the method comprising: acquiring a plurality of measurement reports MR of a terminal in a target area and engineering parameters of at least one base station in the target area, where The target area is a predetermined geographical area.
  • the target area may be divided according to the population quantity and the administrative area, for example, the suburb of a certain city is the target area, or the urban area of a certain city is the target area, and the application is not limited.
  • Each of the plurality of MRs includes location information and parameter information, the location information is used to indicate a location of the terminal corresponding to the MR that includes the location information in the target area; Without limitation, generally speaking, other information except the location information in the MR can be attributed to the parameter information;
  • the parameter information includes an environment parameter, where the environment parameter is used to indicate an environment condition of the terminal corresponding to the MR that includes the environment parameter, such as: time period information, weather information, event information (holiday, celebration day, sports meeting, etc.),
  • environment parameter such as: time period information, weather information, event information (holiday, celebration day, sports meeting, etc.).
  • the feature vector obtained by the engineering parameter of the station can express complex observation information; the position information of each MR of the plurality of MRs and the feature vector corresponding to each of the plurality of MRs are processed by a machine learning model, Obtaining a single point positioning model; calculating the plurality of MRs according to the single point positioning model, position information of each MR of the plurality of MRs, and a feature vector corresponding to each of the plurality of MRs a transmission probability of a feature vector corresponding to each of the MRs, wherein the transmission probability includes at least one transmission probability value, the transmission probability value being used to indicate a probability that a certain feature vector corresponds to a certain location information; The MR inputting the single point positioning model and the MR for training
  • the MR used to input the single point positioning model to calculate the transmission probability may be the MR uploaded by different terminals in the target area.
  • a single-point positioning model is obtained by MR training of the target area terminal, and the MR of the terminal in the target area is input into the single-point positioning model, and the spatial model of the single-point positioning model is used to calculate the correspondence between the feature vector and the position information, so that the corresponding The relationship is more reliable.
  • the parameter information of each MR of the multiple MRs includes at least one base station ID, where the base station ID is used to indicate that the terminal corresponding to the MR that includes the base station ID is connected a base station
  • an MR may include information of a plurality of base stations connected to a terminal, the at least one base station includes at least a base station indicated by a base station ID included in the plurality of MRs, and further a base station ID in each MR
  • Corresponding engineering parameters are used to obtain a feature vector; and according to parameter information of each MR of the plurality of MRs and engineering parameters of the at least one base station, each MR of the plurality of MRs is correspondingly obtained
  • the feature vector includes: matching the plurality of MRs with the engineering parameters of the at least one base station according to the base station ID, and obtaining associated engineering parameters of each of the plurality of MRs, wherein any MR association
  • the engineering parameter includes an engineering parameter of the base station indicated by each base station ID in
  • the engineering parameters of the eNB in the feature vector generally depend on the number of base stations indicated by the base station ID in the corresponding MR.
  • the eigenvector may only include at least one base station indicated by the at least one base station ID included in the MR. The engineering parameters of one of them.
  • a possible implementation manner of the first aspect the processing, by using a machine learning model, processing location information of each MR of the plurality of MRs and a feature vector corresponding to each MR of each of the plurality of MRs to obtain a single a point location model, comprising: obtaining, according to position information of each MR of the plurality of MRs and a feature vector corresponding to each of the plurality of MRs, obtaining respective MRs of the plurality of MRs a training set, wherein any training set includes feature vectors and position information corresponding to the same MR; and a training set corresponding to each of the plurality of MRs is input into the machine learning model for training, to obtain the single Point positioning model.
  • the location information according to the single point positioning model, each MR of the plurality of MRs, and a corresponding feature vector of each of the plurality of MRs Calculating a transmission probability of a feature vector corresponding to each of the plurality of MRs, comprising: respectively, position information of each MR of the plurality of MRs and each MR of the plurality of MRs
  • the feature vector is input to the single-point positioning model to obtain a mapping relationship, wherein the mapping relationship is used to indicate a correspondence relationship between the feature vector and the location information; and each MR of the plurality of MRs is calculated according to the mapping relationship.
  • the probability of emission of the feature vector is a mapping relationship, wherein the mapping relationship is used to indicate a correspondence relationship between the feature vector and the location information.
  • the feature vector of the MR input into the single-point positioning model and the location information therein may not be the feature vector corresponding to the MR of the terminal and the location information therein, and may be other terminals in the same target area as the terminal.
  • MR and the method for obtaining the feature vector is the same as the method for obtaining the feature vector corresponding to the MR of the terminal, and details are not described herein again.
  • the machine learning model is a regression model, such as logistic regression, random forest, etc., and the specific model of the regression model is not limited herein.
  • the method for obtaining the transmission probability uses the feature vector obtained by the plurality of parameter information in the MR and the engineering parameter of the corresponding base station as an observation value, and trains the single point positioning model with the feature vector and the position information corresponding to the MR, and uses the single
  • the transmission probability obtained by the spatial model of the point location model can express complex observation information, and the correspondence between the feature vector (observation value) and the position information is more reliable.
  • the present invention provides a method for obtaining a transition probability, which acquires a plurality of trajectory data of a terminal in a target area, wherein the target area is a predetermined geographical area, specifically, according to the population
  • the administrative area is divided into target areas, for example, the suburb of a certain city is the target area, or the urban area of a certain city is the target area. This application does not limit the size, geographical location, etc. of the target area.
  • each of the plurality of trajectory data includes at least two pieces of position information, the position information is used to indicate the trajectory data including the position information a position of the corresponding terminal in the target area, each of the plurality of pieces of position information included in the plurality of trajectory data corresponds to a time stamp; and calculating a transition probability according to the plurality of trajectory data, wherein the The transition probability includes at least one transition probability value, which is used to indicate a certain location information over time The probability of separating information from T to another location.
  • the multiple trajectory data of the terminal in the target area is from a third-party platform, such as a third-party APP drip, a shared bicycle platform, and the like.
  • the multiple trajectory data includes the same environment parameter, where the environment parameter is used to indicate an environment condition of the terminal corresponding to the trajectory data including the environment parameter, such as: time period information, weather information, event information.
  • the environment parameter such as: time period information, weather information, event information.
  • the environmental parameters can be used as the identifier to more accurately support the positioning under different environmental conditions.
  • the calculating the transition probability according to the multiple trajectory data includes: processing the multiple trajectory data, and obtaining at least each trajectory data of the plurality of trajectory data a combined sequence, wherein the combined sequence includes any two location information in the same trajectory data and a time interval of the any two location information, wherein the time interval between the two location information may be based on two location information Calculating a timestamp corresponding to the grid; obtaining a transition probability corresponding to the first preset condition according to the first preset condition and the at least one combined sequence of each of the plurality of trajectory data, wherein the The first preset condition is any one of a plurality of preset conditions, and each of the plurality of preset conditions includes a preset time interval and preset position information, where the preset time interval corresponds to the The time interval T, the preset location information corresponds to the certain location information.
  • the transition probability corresponding to the first preset condition is obtained according to the first preset condition and the at least one combined sequence of each of the plurality of trajectory data
  • the method includes: determining, in all combinations of the plurality of trajectory data, a combined sequence of preset time intervals and preset position information in the first preset condition; A combined sequence of time intervals and preset position information is set, and a transition probability corresponding to the first preset condition is calculated.
  • the number of all combined sequences including the location information A and the time interval T1 is determined to be M, which can be expressed as [location information A, time interval T1, location information X n ], and all combinations of sequences including location A and time interval T1.
  • the position information X 1 to X n are counted, and the probability values including the position information X 1 to X n occupy M are obtained, and the respective probability values corresponding to the position information X 1 to X n constitute the corresponding position information A and the time interval T1. The probability of transition of the condition.
  • the preset time interval is a preset time interval range
  • the preset time interval may be a specific duration or a duration range, for example, the preset time interval is 2 seconds, or the preset time interval is 2 ⁇ 4 seconds, so that the transition probability corresponding to a certain duration range can be obtained, because the frequencies of the MRs obtained by different terminals may be different, and the time intervals in the combined sequence obtained by the trajectory data of different terminals may be different, by setting the preset conditions.
  • the time interval is set to a range of values, and the combined sequences of different time intervals can be fused to maximize the use of existing data.
  • the trajectory data obtained by the third party may have unreliable data, that is, the trajectory data, and the trajectory data is removed from the acquired trajectory data, and the trajectory data after the culling is further processed to obtain the transition probability. It is possible to improve the movement trajectory of the recovery/prediction terminal with this transition probability to be more reliable and smooth.
  • the calculating the transition probability according to the multiple trajectory data further comprising: determining sparse trajectory data in the plurality of trajectory data, wherein the sparse trajectory data is Included trajectory data of a distance between any two adjacent ones of the at least two pieces of position information that is greater than a third threshold; any two of the sparse trajectory data according to map information of the target area Insert one or more location information between adjacent location information.
  • the position information in the acquired trajectory data can be made more dense, and the transition probability obtained from the trajectory data after the interpolation processing and used to recover/predict the movement trajectory of the terminal can make the recovery/predicted movement trajectory smoother, or Get more transition probability for different time intervals.
  • the acquiring the plurality of trajectory data of the terminal in the target area includes: acquiring a plurality of trajectory data of the terminal in the target area during a traffic peak period or a traffic non-peak period.
  • a traffic peak period or a traffic non-peak period there is usually a difference between the terminal movement trajectory during the peak traffic hours and the terminal movement trajectory during the off-peak hours of the traffic. Different transition probabilities are obtained for the traffic peak hours and traffic non-peak hours, which are used to recover/predict the corresponding time periods.
  • the movement trajectory of the terminal can improve the accuracy and reliability of recovering/predicting the movement trajectory of the terminal.
  • the method for obtaining the transition probability provided by the present application, through the terminal movement trajectory data in the target area provided by the third-party platform, and the transition probability calculated according to the movement trajectory data is used to restore or predict the movement trajectory of a terminal in the target area is smoother, Effectively avoid track jumps.
  • the present application provides a method for sequence positioning, acquiring multiple target measurement reports MR of a target terminal in a target area and engineering parameters of at least one base station in the target area, where the target area is a predetermined a geographic area, each target MR of the plurality of target MRs includes parameter information; and the plurality of targets are obtained according to parameter information of each target MR of the plurality of target MRs and engineering parameters of the at least one base station a target feature vector corresponding to each target MR in the MR; a target feature vector corresponding to each target MR of the plurality of target MRs is input into a sequence positioning model to obtain a moving track of the target terminal;
  • the application parameters of the sequence positioning model include a transmission probability and a transition probability, which may be obtained by the method described in the first aspect or any possible implementation of the first aspect, or/and the transition probability may be passed
  • the method described in the second aspect or any of the possible implementations of the second aspect is obtained, and is not described herein.
  • the method for sequence location obtained by the present application obtains a transmission probability according to a plurality of parameter information in the MR and a feature vector obtained from an engineering parameter of the corresponding base station, and can express more complex observation information, thereby further improving the motion trajectory of the sequence location recovery/prediction. ,reliable.
  • the present application provides a method for sequence positioning, the method comprising: acquiring a plurality of target measurement reports MR of a target terminal in a target area and engineering parameters of at least one base station in the target area, where the target The area is a predetermined geographical area, each target MR of the plurality of target MRs includes parameter information; and according to parameter information of each target MR of the plurality of target MRs and engineering parameters of the at least one base station, a target feature vector corresponding to each target MR of the plurality of target MRs; input a target feature vector corresponding to each target MR of the plurality of target MRs into a sequence positioning model, to obtain a moving track of the target terminal ;
  • the application parameters of the sequence positioning model include a transmission probability and a transition probability, and the transition probability is obtained by acquiring a plurality of trajectory data of a terminal in the target area, where each of the plurality of trajectory data of the terminal A track data includes at least two second location information, the second location information being used to indicate that the second location information is included a location of the terminal corresponding to the trajectory data in the target area, each second location information included in the plurality of trajectory data included in the trajectory data of the terminal corresponds to a time stamp;
  • the trajectory data calculates a transition probability, wherein the transition probability includes at least one transition probability value, the transition probability value being used to indicate a probability that a certain second location information passes through the time interval T to another second location information.
  • the calculating the transition probability according to the multiple trajectory data of the terminal includes: processing a plurality of trajectory data of the terminal, and obtaining each of the plurality of trajectory data of the terminal At least one combined sequence of trajectory data, wherein the combined sequence includes any two second location information of the same trajectory data of the terminal and a time interval of the any two second location information; according to the first pre- Setting a condition and at least one combined sequence of each of the plurality of trajectory data of the terminal to obtain a transition probability corresponding to the first preset condition, wherein the first preset condition is a plurality of presets Any one of the plurality of preset conditions, the preset time interval and the preset second position information, wherein the preset time interval corresponds to the time interval T, the preset The second location information corresponds to the certain second location information.
  • the probability includes: determining, in all combinations of the plurality of trajectory data of the terminal, a combination sequence of the preset time interval and the preset second location information in the first preset condition; A combination sequence of preset time intervals in the condition and preset second position information is set, and a transition probability corresponding to the first preset condition is calculated.
  • trajectory data is trajectory data in which at least one second position information deviates from a road in the target area by a distance greater than a first threshold, or trajectory data in which a distance between two adjacent second position information is greater than a second threshold .
  • determining sparse trajectory data in the plurality of trajectory data of the terminal where the sparse trajectory The trajectory data is trajectory data including a distance between any two adjacent second position information of the at least two second position information being greater than a third threshold; according to the map information of the target area, in the One or more second positional information is inserted between any two adjacent second positional information of the sparse trajectory data.
  • the acquiring the plurality of trajectory data of the terminal in the target area includes: acquiring a plurality of trajectory data of the terminal in the target area during a traffic peak period or a traffic non-peak period.
  • the preset time interval is a preset time interval range.
  • the transmission probability is obtained by the method described in the first aspect or any of the possible implementations of the first aspect.
  • the method for sequence positioning uses the transition probability obtained by the third-party real trajectory data for sequence positioning, can improve the smoothness of the recovered/predicted moving trajectory, and obtain a more reliable moving trajectory.
  • the present application provides a transmission probability calculation apparatus, including: a measurement report MR acquisition module, a feature vector module, a regression processing module, and a transmission probability calculation module; and the MR acquisition module is configured to acquire a target a plurality of measurement reports MR of the terminal in the area and engineering parameters of at least one base station in the target area, wherein the target area is a predetermined geographical area, and each MR of the plurality of MRs includes location information and parameter information The location information is used to indicate a location of the terminal corresponding to the MR that includes the location information in the target area; the feature vector module is configured to use each MR of the plurality of MRs acquired by the MR acquisition module The parameter information and the engineering parameters of the at least one base station, the feature vectors corresponding to each of the plurality of MRs are obtained; the regression processing module is configured to use each of the plurality of MRs acquired by the MR acquiring module Position information of the MR and the feature vector module a feature vector corresponding to each of
  • the parameter information of each of the plurality of MRs includes at least one base station ID, where the base station ID is used to indicate a base station to which the terminal corresponding to the MR corresponding to the base station ID is connected
  • the at least one base station includes at least the base station indicated by the base station ID included in the plurality of MRs; the feature vector module is specifically configured to:
  • Matching by the base station ID, the plurality of MRs acquired by the MR acquiring module and the engineering parameters of the at least one base station acquired by the MR acquiring module, to obtain associated engineering parameters of each of the plurality of MRs, where
  • An associated engineering parameter of an MR includes an engineering parameter of a base station indicated by each base station ID in the MR; and an associated engineering parameter and parameter information of each MR of the plurality of MRs acquired by the MR acquiring module,
  • Each of the plurality of MRs corresponds to a feature vector corresponding to each MR, wherein any feature vector includes associated engineering parameters and parameter information of the same MR.
  • the regression processing module is specifically configured to: according to location information of each MR of the plurality of MRs acquired by the MR acquisition module, and multiple MRs obtained by the feature vector module
  • Each of the MRs corresponds to a feature vector
  • a training set corresponding to each of the plurality of MRs is obtained, wherein any training set includes feature vectors and position information corresponding to the same MR;
  • Each of the MRs respectively corresponds to a training set and inputs the machine learning model for training to obtain the single point positioning model.
  • the transmission probability calculation module is specifically configured to: use location information of each MR of the plurality of MRs acquired by the MR acquisition module, and multiple MRs obtained by the feature vector module The corresponding feature vector of each of the MRs is input to the single-point positioning model obtained by the regression processing module to obtain a mapping relationship, wherein the mapping relationship is used to indicate a correspondence between the feature vector and the location information; according to the mapping relationship A transmission probability of a feature vector corresponding to each of the plurality of MRs is calculated.
  • the transmission probability calculation device takes the feature vector obtained by using the plurality of parameter information in the MR and the engineering parameter of the corresponding base station as an observation value, and trains the single point positioning model with the feature vector and the position information corresponding to the MR, and utilizes a single point.
  • the transmission probability obtained by the spatial model of the positioning model can express complex observation information, and the correspondence between the feature vector (observation value) and the position information is more reliable.
  • the present application provides a transition probability calculation device, where the transition probability calculation module includes: a trajectory acquisition module and a transition probability calculation module; and the trajectory acquisition module is configured to acquire a plurality of trajectory data of the terminal in the target area, where The target area is a predetermined geographic area, and each of the plurality of trajectory data includes at least two pieces of position information, where the position information is used to indicate that the terminal corresponding to the trajectory data including the position information is in the a position in the target area, each of the plurality of pieces of position information included in the plurality of pieces of trajectory data corresponds to a time stamp; and the transition probability calculation module is configured to acquire a plurality of trajectory data acquired by the trajectory acquiring module A transition probability is calculated, wherein the transition probability includes at least one transition probability value, the transition probability value being used to indicate a probability that a certain location information passes through the time interval T to another location information.
  • the transition probability calculation module includes: a trajectory acquisition module and a transition probability calculation module; and the trajectory acquisition module is configured to acquire a plurality of trajectory
  • the transition probability calculation module includes: a pre-processing unit and a transition probability calculation unit; the pre-processing unit is configured to process the plurality of trajectory data obtained by the trajectory acquisition module to obtain At least one combined sequence of each of the plurality of trajectory data, wherein the combined sequence includes any two pieces of position information in the same trajectory data and a time interval of the any two pieces of position information;
  • the probability calculation unit is configured to use at least one group of each of the plurality of trajectory data obtained according to the first preset condition and the pre-processing unit Combining the sequence to obtain a transition probability corresponding to the first preset condition, wherein the first preset condition is any one of a plurality of preset conditions, and each of the plurality of preset conditions
  • the preset time interval and the preset position information are corresponding to the time interval T, and the preset position information corresponds to the certain position information.
  • the preset time interval is a preset time interval range.
  • the transition probability calculation unit is configured to: determine that all combinations of the plurality of trajectory data obtained by the pre-processing unit include presets in the first preset condition a combined sequence of the time interval and the preset position information; the statistical combination comprising the preset time interval of the preset condition and the preset position information, and calculating a transition probability corresponding to the first preset condition.
  • the device further includes: a first trajectory processing module, wherein the first trajectory processing module is configured to cancel ⁇ trajectory data in the plurality of trajectory data acquired by the trajectory acquiring module, where The ⁇ trajectory data is trajectory data in which at least one position information deviates from a road in the target area by a distance greater than a first threshold, or trajectory data in which a distance between two adjacent position information is greater than a second threshold .
  • the device further includes: a second trajectory processing module; the second trajectory processing module is configured to determine sparse trajectory data in the plurality of trajectory data acquired by the trajectory acquiring module, where The sparse trajectory data is trajectory data including a distance between any two adjacent ones of the at least two pieces of position information that is greater than a third threshold; according to map information of the target area, One or more pieces of position information are inserted between any two adjacent positional information of the sparse trajectory data.
  • the transition probability calculation device provided by the present application, through the terminal movement trajectory data in the target area provided by the third-party platform, and the transition probability calculated according to the movement trajectory data is used to restore or predict that the movement trajectory of a terminal in the target area is smoother and can be effective. Avoid track jumps.
  • the present application provides a sequence positioning apparatus, where the sequence positioning apparatus includes: a transmission probability calculation module, a transition probability calculation module, and a sequence positioning module;
  • the sequence positioning module includes a target measurement report MR acquisition unit, a target feature vector unit, and a trajectory prediction unit, wherein the target MR unit is configured to acquire a plurality of target measurement reports MR of the target terminal in the target area and the target area An engineering parameter of the at least one base station, wherein the target area is a predetermined geographic area, each target MR of the plurality of target MRs includes parameter information, and the target feature vector unit is configured to acquire multiple according to the target MR unit a parameter information of each target MR in the target MR and an engineering parameter of the at least one base station, and a target feature vector corresponding to each of the plurality of target MRs is obtained, and the trajectory prediction unit is configured to use Deriving a target feature vector corresponding to each target MR of the plurality of target MRs obtained by the target feature vector unit, and obtaining a movement trajectory of the target terminal, where the application parameters of the sequence positioning model include a transmission probability and a transition probability;
  • the transmission probability calculation module is configured to calculate the transmission probability, and the transition probability calculation module is configured to calculate the transition probability;
  • the transmission probability calculation module includes a measurement report MR acquisition unit, a feature vector unit, a regression processing unit, and a transmission probability calculation unit.
  • the MR acquisition unit is configured to acquire a plurality of measurement reports MR and a location of the first terminal in the target area.
  • each MR of the plurality of MRs includes location information and parameter information, where the location information is used to indicate that the first terminal corresponding to the MR that includes the location information is a location in the target area;
  • the feature vector unit is configured to obtain, according to parameter information of each MR of the plurality of MRs acquired by the MR acquiring unit, and engineering parameters of at least one base station, a feature vector corresponding to each MR;
  • the regression processing unit is configured to use each of a plurality of MRs obtained from the plurality of MRs acquired by the MR acquisition unit and each of the plurality of MRs obtained by the feature vector unit
  • Each of the MR corresponding feature vectors obtains a single point positioning model;
  • the transmission probability calculation unit is configured to obtain a single point positioning model according to the regression processing unit, and the MR Fetch unit eligible And taking the parameter information of each of the plurality of MRs and the feature vector unit to obtain a feature vector corresponding to each of the plurality of the plurality of
  • the transition probability calculation module includes a trajectory acquisition unit and a transition probability calculation unit; the trajectory acquisition unit is configured to acquire a plurality of trajectory data of the second terminal in the target area, wherein each of the plurality of trajectory data The trajectory data includes at least two pieces of position information, the position information is used to indicate a position of the second terminal corresponding to the trajectory data including the position information in the target area, and the plurality of trajectory data includes a plurality of pieces of position information Each of the location information corresponds to a time stamp; the transition probability calculation unit is configured to calculate a transition probability according to the plurality of trajectory data acquired by the trajectory acquisition unit, wherein the transition probability includes at least one transition probability value, The transition probability value is used to indicate the probability that a certain location information passes through the time interval T to another location information.
  • the parameter information of each of the plurality of MRs includes at least one base station ID, where the base station ID is used to indicate that the first terminal corresponding to the MR that includes the base station ID is connected
  • the base station the at least one base station includes at least the base station indicated by the base station ID included in the plurality of MRs
  • the feature vector unit is specifically configured to: use the multiple MRs acquired by the MR acquiring unit according to the base station ID, The engineering parameters of the at least one base station acquired by the MR acquiring unit are matched, and the associated engineering parameters of each of the plurality of MRs are obtained, where the associated engineering parameters of any MR include each base station ID in any of the MRs.
  • An engineering parameter of the indicated base station obtaining, according to an associated engineering parameter and parameter information of each MR of each of the plurality of MRs acquired by the MR acquiring unit, a feature vector corresponding to each of the plurality of MRs, wherein A feature vector includes associated engineering parameters and parameter information for the same MR.
  • the regression processing unit is specifically configured to: according to location information of each MR of the plurality of MRs acquired by the MR acquiring unit, and multiple MRs obtained by the feature vector unit
  • Each of the MRs corresponds to a feature vector, and a training set corresponding to each of the plurality of MRs is obtained, wherein any training set includes feature vectors and position information corresponding to the same MR;
  • Each of the MRs respectively corresponds to a training set and inputs the machine learning model for training to obtain the single point positioning model.
  • the transmission probability calculation unit is specifically configured to: acquire location information of each MR of the plurality of MRs acquired by the MR acquisition unit, and multiple MRs obtained by the feature vector unit The corresponding feature vector of each of the MRs is input to the single-point positioning model obtained by the regression processing unit to obtain a mapping relationship, wherein the mapping relationship is used to indicate a correspondence between the feature vector and the location information; according to the mapping relationship A transmission probability of a feature vector corresponding to each of the plurality of MRs is calculated.
  • the transition probability calculation unit includes: a pre-processing sub-unit and a transition probability calculation sub-unit; and the pre-processing sub-unit is configured to perform, by using the trajectory data obtained by the trajectory acquisition unit Processing, obtaining at least one combined sequence of each of the plurality of trajectory data, wherein the combined sequence includes any two location information in the same trajectory data and a time interval of the any two location information;
  • the transition probability calculation subunit is configured to obtain, according to the first preset condition and at least one combination sequence of each of the plurality of trajectory data obtained by the preprocessing subunit, the first preset condition a transition probability, wherein the first preset condition is any one of a plurality of preset conditions, and each of the plurality of preset conditions includes a preset time interval and preset position information,
  • the preset time interval corresponds to the time interval T
  • the preset position information corresponds to the certain position information.
  • the transition probability calculation subunit is specifically configured to: determine that all the combined sequences included in the plurality of trajectory data obtained by the preprocessing subunit include the first preset condition And a combination sequence of the preset time interval and the preset position information; and the combined sequence including the preset time interval and the preset position information in the preset condition, and calculating a transition probability corresponding to the first preset condition.
  • the present application provides a sequence positioning system
  • a sequence positioning system which includes: a positioning device, a fifth aspect, or a transmission probability described in any possible implementation manner of the fifth aspect.
  • a computing device and a transition probability computing device as described in any of the possible implementations of the sixth or sixth aspect;
  • the positioning device comprising a target measurement report MR acquisition module, a target feature vector module, and a trajectory prediction module;
  • the target MR module is configured to acquire a plurality of target measurement reports MR of the target terminal in the target area and engineering parameters of at least one base station in the target area, where the target area is a predetermined geographical area, and the plurality of target MRs
  • Each target MR includes parameter information;
  • the target feature vector module is configured to obtain, according to parameter information of each target MR of the plurality of target MRs acquired by the target MR module, and engineering parameters of the at least one base station a target feature vector corresponding to each of the plurality of target MRs;
  • the trajectory prediction module is configured
  • the present application provides a transmission probability calculation device, including: a memory, a processor; the memory is configured to store a programmable instruction, and the processor invokes the programmable instruction stored in the memory
  • a transmission probability calculation device including: a memory, a processor; the memory is configured to store a programmable instruction, and the processor invokes the programmable instruction stored in the memory
  • the present application provides a transition probability calculation device, including: a memory, a processor; the memory is configured to store a programmable instruction, and the processor invokes the programmable instruction stored in the memory
  • a transition probability calculation device including: a memory, a processor; the memory is configured to store a programmable instruction, and the processor invokes the programmable instruction stored in the memory
  • the present application provides a sequence locating device, including: a memory, a processor; the memory is configured to store a programmable instruction, and the processor invokes the programmable instruction stored in the memory A method as described in the third aspect or the fourth aspect or any of the possible implementations of the fourth aspect.
  • the present application provides a computer readable storage medium comprising instructions which, when executed on a computer, cause the computer to perform any of the first or first aspect of the first aspect or the second aspect or Any of the possible implementations of the second aspect or the method of the third or fourth aspect or any of the possible implementations of the fourth aspect.
  • the present application provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform any of the first aspect or the first aspect or the second aspect or the second Any of the possible implementations of aspects or the method of the third or fourth aspect or any of the possible implementations of the fourth aspect.
  • the present application provides a sequence positioning system comprising the transmission probability calculation device described in the ninth aspect, the transition probability calculation device described in the tenth aspect, and a sequence locating device.
  • the sequence locating device includes a processor and a memory, the memory is configured to store a programmable instruction, and the processor invokes the programmable instruction stored in the memory to perform: acquiring a plurality of target measurements of the target terminal in the target area Reporting MR and engineering parameters of at least one base station in the target area, wherein the target area is a predetermined geographical area, each target MR of the plurality of target MRs includes parameter information; according to the plurality of target MRs a parameter information of each target MR and an engineering parameter of the at least one base station, obtaining a target feature vector corresponding to each of the plurality of target MRs; each target of the plurality of target MRs The respective target feature vectors of the MR are input to the sequence positioning model to obtain a moving trajectory of the target terminal.
  • the method for sequence location obtaineds a transmission probability according to a plurality of parameter information in the MR and a feature vector obtained from an engineering parameter of the corresponding base station, and can express more complex observation information, thereby further improving the motion trajectory of the sequence location recovery/prediction.
  • Reliable or, the transition probability obtained by the third party's real trajectory data is used for sequence positioning, which can improve the smoothness of the recovered/predicted moving trajectory, and obtain a more reliable moving trajectory.
  • FIG. 1 is a schematic structural diagram of a sequence positioning system provided by the present application.
  • FIG. 2 is a schematic diagram of an application scenario of a sequence positioning system provided by the present application.
  • FIG. 3 is a flowchart of a method for acquiring a transmission probability provided by the present application.
  • FIG. 4 is a schematic diagram of calculating a transmission probability provided by the present application.
  • FIG. 5 is a flowchart of a method for acquiring a transition probability provided by the present application.
  • FIG. 6 is a schematic diagram of a map matching and interpolation method provided by the present application.
  • FIG. 7 is a schematic diagram of a method for calculating a transition probability in an offline indexing manner provided by the present application.
  • FIG. 8 is a schematic diagram of a method for calculating a transition probability in an online indexing manner provided by the present application.
  • FIG. 10 is a schematic diagram of a recovery track according to a column positioning method provided by the present application.
  • FIG. 11 is a schematic diagram of a particle filter sequence positioning method provided by the present application.
  • FIG. 13 is a schematic diagram of a transition probability calculation apparatus provided by the present application.
  • FIG. 14 is a schematic diagram of a sequence locating device provided by the present application.
  • 15 is a schematic diagram of a device for calculating a transmission probability and or a transition probability provided by the present application
  • Figure 16 is an experimental setup and results of an actual test experiment provided by the present application.
  • the positioning system includes an emission probability calculation device, a transition probability calculation device, and a sequence positioning device, and the transmission probability device, the transition probability device, and the sequence positioning device are connected through a data link.
  • the transmission probability calculation device is configured to calculate a transmission probability for sequence positioning, the input is a measurement report (MR) data with position annotation, and the transmission probability is output;
  • the transition probability calculation device is used for calculating the sequence
  • the transition probability of the positioning is input as the movement trajectory data, and the output is the transition probability;
  • the sequence locating device is configured to restore the movement trajectory of the to-be-determined terminal/device according to the transmission probability, the transition probability, and the MR data of the terminal/device to be located, and the input is to be determined.
  • a series of MR (no position information) of the target terminal is output as the moving track of the target terminal to be located.
  • the transmission probability device, the transition probability device, and the sequence positioning device may also constitute an overall device, the transmission probability device is a transmission probability calculation module of the whole device, and the transition probability device is a transition probability calculation module of the whole device, the sequence The positioning device is a sequence positioning module of the whole device, and the functions of each module are the same as those of the corresponding device, and data can be transmitted between the modules.
  • the functions implemented by the entire components of the transmission probability calculation module, the transition probability calculation module, and the sequence positioning module or the above three functional modules are implemented by software, or may be implemented by software + hardware.
  • the above positioning system can be deployed on the big data analysis platform.
  • the daily MR is stored on the platform, it is first input to the positioning system, and the positioning system predicts according to the MR extraction feature and then uses our offline trained model.
  • the latitude and longitude position of each MR record is added back to the MR, and the MR with latitude and longitude position can be further analyzed and modeled.
  • the positioning system can be used as a component of the operator's big data analysis platform to upload the mobile device to the MR input positioning system in the operator pipeline in hours or days.
  • the positioning system will get a corresponding input to each MR of the input system.
  • the location information is added to the MR; subsequently, the MRs with the added location information can be used by the operator for user portraits, crowd forecasting, outdoor advertising strategy optimization, and the like.
  • An embodiment of the present application describes a method for obtaining a transmission probability. As shown in FIG. 3, the method includes the following steps:
  • the target area refers to a specific geographical area, such as a suburb, an urban area, a city, a rural area, and the like, and is not limited to an area size, an administrative area, or a geographical position;
  • Each MR includes location information and parameter information, and the location information is used to mark the location of the corresponding terminal within the target area.
  • the parameter information includes an environment parameter, and the environment parameter is used to indicate at least one of a real environment condition, such as time period information, weather information, and event information, when the corresponding terminal generates the MR.
  • a real environment condition such as time period information, weather information, and event information
  • different transmission probabilities obtained for different environmental conditions can more accurately support positioning under different environmental conditions.
  • the environmental parameter corresponding to the peak time period is represented by 1, and the environmental parameter swimming 0 corresponding to the off-peak time period indicates that the emission probability is obtained according to the MR with the environmental parameter of 1 for the peak time period.
  • the positioning of the terminal in the target area is more accurate; similarly, the probability of the transmission is obtained according to the MR whose environment parameter is 0, and the positioning of the terminal in the target area during the off-peak period is more accurate.
  • the information parameter in the MR may include one or more base station IDs, which also indicates that the terminal corresponding to the MR is connected to one or more base stations at the same time; when the parameter information of the MR terminal includes multiple base station IDs, Then, the associated project of a certain MR is a set of engineering parameters of the base station indicated by the plurality of base station IDs.
  • the single-point positioning model is obtained by processing the position information and the feature vector by the machine learning model. Specifically, the position information of each MR and the corresponding feature vector form a set of training sets, and the training set corresponding to each MR is input into the machine.
  • the learning model is trained to obtain a single point positioning model; optionally, the single point positioning model can be directly used to locate the terminal to be located in the target area, and the input is the MR parameter information and the base station engineering parameter of the terminal to be located.
  • the composed feature vector is output as the location information of the terminal to be located.
  • the machine learning model is a regression model, such as linear regression, random forest, and the like.
  • the transmission probability includes at least one transmission probability value, where the transmission probability value is used to indicate that a certain feature vector as the observation corresponds to a certain position of the implicit value.
  • the mapping relationship of information Specifically, the location information of each MR and the corresponding feature vector are input into the single-point positioning model to obtain a mapping relationship, where the mapping relationship is used to indicate a correspondence relationship between the feature vector and the location information; and each MR corresponding to each mapping is calculated according to the mapping relationship. The probability of emission of a feature vector.
  • the location information and feature vectors involved in S104 and the location information and feature vectors that may be involved in S101–S103 are not necessarily obtained based on the same MR, because a large number of MRs need to be acquired, and the same MR is used to reduce data and then acquire. Link, and from the results, if the MR base is large enough, whether to use different MRs has less effect on the results.
  • the MR includes location annotations, such as GPS information carried in the MR, and acquiring data in some other telecommunication networks, such as engineering parameters of the base station, and engineering parameters mainly include engineering
  • the parameters mainly include the ID of the base station, the latitude and longitude position, the antenna hanging height, and the antenna azimuth.
  • the feature related to the positioning is extracted from the data acquired by the feature engineering method as the feature vector of the corresponding MR, and the feature vector may include the engineering parameters of the connected base station, the adjacent base station, the signal strength of the connection, etc.;
  • the engineering parameters of the MR carrying the GPS and the base station are matched according to the base station ID, and the corresponding parameter information is found in the engineering parameters for each base station in the MR; further, some simple characteristic engineering parameters are added to the matched data, For example, the number of base stations connected to each MR, different base stations are located at different latitude and longitude positions; or, the number of sectors, multiple sectors may be located at the same latitude and longitude.
  • a feature vector can be created for each MR, and the feature vector and position label of the same MR form a training set corresponding to the MR for training the single point positioning model.
  • Table 1 shows some of the telecommunication features used to train the positioning model. Some of the original fields in the MR and some of the base station engineering parameters are included. The feature with * indicates that these fields are for the connected base station and the primary base station of the same MR, so the same field will appear multiple times in the feature vector, corresponding to different base stations.
  • environmental information is extracted from the MR as environmental parameters, such as weather information (clear, rain, snow, etc.), time period (peak hours, off-peak hours, weekdays, weekends, statutory holidays), event information (sports meeting, concert) , National Day, etc.; the environmental parameters and the telecommunication features in Table 1 constitute a feature vector, and the corresponding position labeling constitutes a training set.
  • all the training sets obtained in the above steps are input into the machine learning regression model, and the corresponding model is trained to obtain a single point positioning model.
  • machine learning regression models such as linear regression, random forests, etc.
  • the feature vector obtained in the above step or the feature sequence obtained based on other MR samples with position labeling is taken as the observation value, and the position label is taken as the implicit state.
  • the above step is obtained to obtain the single point positioning model, the single point can be analyzed.
  • the model space of the model is located, and the correspondence between the position label and the feature vector is obtained, and then P (feature vector
  • the model space of different models corresponds to different calculation methods.
  • the random forest is taken as an example to introduce the analysis method of model space and the calculation method of emission probability. The details are as follows:
  • each leaf node in the random forest can be regarded as a series of feature vectors (obtained through the splitting features of a series of decision nodes), and a leaf node is regarded as an observation value, so the emission probability is converted into a given label to obtain the leaf node.
  • Probability P leaf node
  • the emission probability value can be obtained by dividing the number of the label sample in the leaf node by the number of the total table label sample of the same label in the label sample, for example, there are 10 dot samples and only one falls on FIG. 4 .
  • the emission probability value is 1/10, which indicates the degree of coincidence of the position of the labeled sample with the observed value (the probability of obtaining similar observations at that position), A higher level of compliance indicates that the predicted position is more accurate.
  • An embodiment of the present application describes a method for obtaining a transition probability. As shown in FIG. 5, the method includes the following steps:
  • trajectory data can be obtained from a third-party platform, such as a drip trip, a traffic data publishing platform, and the like.
  • acquiring the plurality of trajectory data of the one or more terminals in the target area is: acquiring trajectory data of the terminal in the target area during the traffic peak period, or acquiring one or more of the target areas in the traffic off-peak period Trajectory data of the terminal; also, trajectory data of one or more terminals of a preset time period.
  • the transition probability is obtained according to the trajectory data of the terminal acquired in a certain time period, and the sequence positioning applied to the same certain time period has the best effect, and can also be applied to the sequence positioning of the time period similar to the time period.
  • transition probability Calculate a transition probability according to the multiple trajectory data, where the transition probability includes at least one transition probability value, where the transition probability value is used to indicate that a certain location information (starting location) passes through the time interval T to another location information (arrival location). Probability.
  • the acquired multiple trajectory data is processed to obtain a combined sequence of each trajectory data, and any two positional information in the same trajectory data and a time interval of the arbitrary two positional information constitute a combined sequence;
  • one trajectory data There may be one or more combined sequences; two positional information in one combined sequence, starting from one positional information as the starting position and the other being the arriving position, the combined sequence indicating that the starting position has passed a certain time interval to arrive Position; the combined sequence containing the preset condition is selected in all the combined sequences obtained above, and the preset condition is a preset starting position and a time interval, and the combined sequence of the different preset conditions is also different;
  • the arrival position is taken as the object, the number of different arrival positions is counted, and the probability values of all the combined sequences corresponding to the preset conditions are calculated, that is, the transition probability values are calculated, and the set of all probability values is calculated as The transition probability of the condition should be preset.
  • the preset time interval in the preset condition may be a certain time interval or a time interval range, for example, the preset time interval is 2 seconds, or the preset time interval is 2-4 seconds, It is the preset time interval that meets the preset condition when the time interval satisfies the range of 2-4.
  • the ⁇ trajectory data in the plurality of trajectory data is excluded, where the trajectory data refers to trajectory data in which the distance information of the location information deviates from the road in the target area by more than a certain threshold, or There is trajectory data in which the distance between two adjacent positional information is greater than a certain threshold.
  • multiple trajectory data is used to calculate the transition probability, which can improve the reliability of the transition probability or the smoothness of the transition between two adjacent locations.
  • the sparse trajectory data in the trajectory data obtained by interpolation and densification means that the distance between any two adjacent location information is greater than the third Threshold track Trace data
  • interpolation refers to adding one or more location information between adjacent location information according to the map information and the trajectory data, so that the location information in the trajectory data is dense, and the specific location information may be added according to the interval time, for example, A track has only two positions and the time interval between the two positions is 6 seconds.
  • the plurality of trajectory data includes at least one of environmental parameters, such as time period information, weather information, and event information, indicating a corresponding trajectory of the corresponding terminal under which environmental conditions are generated.
  • environmental parameters such as time period information, weather information, and event information, indicating a corresponding trajectory of the corresponding terminal under which environmental conditions are generated.
  • the trajectory data can be classified, and trajectory data including the same environmental parameters are obtained from existing trajectories with different environmental parameters, such as trajectory data during peak hours and trajectory data at rainy days, according to trajectory data including different environmental parameters.
  • Different transition probabilities are obtained, for example, the transition probability corresponding to the rainy environment is obtained for the positioning of the terminal in the target area in the rainy day.
  • the method for obtaining the transition probability described in one embodiment of the present application is to utilize the concept of migration knowledge, and the motion pattern learned from the real trajectory data is used to calculate the transition probability.
  • the transition probability calculation process can be divided into two major processes: trajectory intensification and transition probability learning.
  • the trajectory intensification process is to be able to learn the transition probability at any time interval. Specifically, each trajectory is first mapped to the road network through a map matching algorithm, so that the path through which each trajectory passes can be inferred. Then we interpolate uniformly along the path through which the trajectory passes, so that the time interval between two adjacent points after interpolation is 1 s, so that the transition probability of any time interval in seconds can be learned.
  • a method for implementing map matching and interpolation as shown in FIG. There are many kinds of map matching methods, such as the map matching method for low sampling frequency trajectory. By calculating the matching probability of the original trajectory point to the nearby road segment and the transition probability between the road segments, a sequence sequence with the highest probability can be obtained. After the matching path is obtained, the values are evenly interpolated between the adjacent points of the track until the time interval between the two adjacent points is equal to 1 s.
  • the transition probability learning process mainly learns the transition probability from each position to other locations from the trajectory. There are many specific learning methods.
  • An embodiment of the present application describes a method for learning transition probability in an offline indexing manner. As shown in FIG. 7, the method is divided into two parts, the upper part is an offline index establishment, and the lower part is an online query.
  • a total of three steps are required to establish an offline index.
  • the first step is to process the trajectory data into a table form in Figure 7.
  • a total of three columns (track ID, time stamp, grid ID), each row representing a track point record.
  • the next step is to extract the triplet ⁇ t,i,j> from this table, ⁇ t is the difference between the timestamps of the two records, and i and j are the raster IDs of the two records, respectively.
  • a record of the same trajectory ID can generate a triple, indicating that it can be moved from grid i to grid j within ⁇ t time. In the actual implementation, it is only necessary to extract a record of ⁇ t ⁇ 60s.
  • the third step is to obtain the transition probability matrix by counting the triples generated in the second step.
  • the probability of transition of the piece According to different starting grids i, the transition probability matrix can be obtained (the time interval of the transition is the same), and then different transition time intervals can obtain different transition probability matrices.
  • the trajectory data of the peak hours (such as 7:00-9:00, 17:00-19:00) is used to generate the transition matrix of the peak period, and the trajectory data of other periods is used to generate the transition matrix of the off-peak period.
  • the online query process mainly obtains the probability distribution vector reaching other grids under the condition of the transition time interval and the departure grid when the sequence is located. First, according to whether the current time is a peak time period, the corresponding offline index is selected; then, the corresponding transition probability matrix is selected according to the time interval ⁇ t; finally, according to the departure grid i, the corresponding line in the transition probability matrix is found as the required transfer. Probability vector.
  • One embodiment of the present application describes a method for calculating a transition probability of an online indexing method. As shown in FIG. 8, the method can also be divided into three steps, the first two steps being the same as the offline index.
  • the trajectory data is processed into a three-column table form (track ID, time stamp, grid ID), and then the triplet ⁇ t, i, j> is extracted from the table.
  • use RTree to create a three-dimensional index for all the extracted triples ⁇ t, i, j> (the three elements in the triple are respectively corresponding to the three-dimensional index).
  • RTree's range query (Range Query)
  • Range Query Given the range of ⁇ t and i, such as 1 ⁇ ⁇ t ⁇ 2, 1 ⁇ i ⁇ 1, RTree can return all triples that satisfy this condition, and then from All third elements j are taken out, and the transition probability distribution is obtained according to the value distribution of j.
  • the difference between an online index and an offline index is that it can set the time interval to a range, such as 1-2s set in the above example.
  • An embodiment of the present application describes a method for sequence location, as shown in FIG. 9, the method includes the following steps:
  • each target MR corresponds to a target feature vector, and the target feature vector is used as an observation value, and is used to input a sequence positioning model to obtain a corresponding implicit position;
  • the sequence locating method can be used to recover the user's trajectory. Similar to the method of obtaining the feature vector when calculating the transmission probability, the same processing is performed on the MR of the terminal to be located, and the corresponding feature vector is generated, and a series of feature vectors of the same terminal to be located is added to the previously obtained transmission probability. And the transition probability is input into the sequence positioning method, and the algorithm can predict the trajectory of the terminal to be positioned according to the feature sequence, as shown in FIG.
  • One embodiment of the present application describes a particle filter sequence positioning method, as shown in FIG.
  • the idea of particle filtering is to find a series of particle sequences with a length T (the length is the same as the length of the track to be recovered), so that this sequence best matches the eigenvector of MR.
  • the number of particles is in the range of hundreds of thousands.
  • Each particle initialized will form a sequence of particles through state transition in the following steps, and the initial state will select a random initialization within a reasonable range (such as in the connection).
  • the base station is within a few hundred meters.
  • the importance weight is the probability that the emission probability p(y
  • the second step is sampling. Then according to the current state of each particle x_j ⁇ ((i)), the time interval ⁇ between the two points in the MR T_j samples the next state, where the offline index in the previous embodiment is used, and the state transition probability distribution p(x_(j+1) ⁇ is obtained after inputting ⁇ t_j and x_j ⁇ ((i)) by the online query method described above. (i))
  • the fourth step is resampling. If the distribution of importance weights of all particle sequences at this time satisfies certain conditions, the particles are resampled. Resampling is a process of putting back samples, which are sampled according to the weight. The probability that the larger the weight of the sequence is sampled (and may be sampled multiple times). Replace the previous sequence with the resampled particle sequence (the number of sequences before and after the sample is unchanged). If the current particle sequence length is less than T, then the importance weights of all particle sequences need to be reset to 1/N.
  • the particle sequence with the highest importance weight is used as the predicted trajectory output, and one particle sequence corresponds to a series of states, that is, a series of latitude and longitude positions.
  • V_(t,k) represents the probability of a state sequence in which the final state of the first t sequences is k.
  • V_(t+1,k) the maximum value of V_(t,x)*a_(x,k) is found, where x is the variable, that is, the previous state of the most suitable k is found;
  • A_(x,k) is the probability that we find the transition probability from the x-grid to the k-grid.
  • the transmission probability calculation device 100 includes: an MR acquisition module 110, a feature vector module 120, a regression processing module 130, and a transmission probability calculation module 140;
  • the module 110 is configured to acquire a plurality of measurement reports MR of the terminal in the target area and engineering parameters of at least one base station in the target area, where the target area is a predetermined geographical area, and each MR of the plurality of MRs includes location information and parameter information.
  • the location information is used to indicate the location of the corresponding terminal in the target area.
  • the feature vector module 120 is configured to obtain, according to the parameter information of each MR of the plurality of MRs acquired by the MR acquiring module 110 and the engineering parameters of the at least one base station.
  • Each of the MRs has a corresponding feature vector for each MR;
  • the regression processing module 130 is configured to: according to the position information of each MR of the plurality of MRs acquired by the MR acquisition module 110 and each of the plurality of MRs obtained by the feature vector module 120
  • Each of the MR corresponding feature vectors obtains a single point positioning model;
  • the transmission probability calculation module 140 is configured to perform single point positioning according to the regression processing module 130.
  • the position information and the feature vector module 120 of each MR of the plurality of MRs acquired by the MR acquisition module 110 obtains a feature vector corresponding to each of the plurality of MRs, and calculates each MR of each of the plurality of MRs.
  • the transmission probability of the feature vector wherein the transmission probability includes at least one transmission probability value, and the transmission probability value is used to indicate a probability that a certain feature vector corresponds to a certain location information.
  • the parameter information of each of the plurality of MRs includes at least one base station ID
  • the base station ID is used to indicate a base station to which the terminal corresponding to the MR of the base station ID is connected
  • the at least one base station includes at least a plurality of base station IDs included in the MR The indicated base station
  • the feature vector module 120 is specifically configured to: match the plurality of MRs acquired by the MR acquiring module 110 with the engineering parameters of the at least one base station acquired by the MR acquiring module 110 according to the base station ID, and obtain each MR of the plurality of MRs.
  • the associated engineering parameter of any MR includes an engineering parameter of the base station indicated by each base station ID in any MR; and an associated engineering parameter of each MR of each of the plurality of MRs acquired by the MR acquiring module Parameter information, obtaining a feature vector corresponding to each MR of each of the plurality of MRs, wherein any feature vector includes an associated worker of the same MR Program parameters and parameter information.
  • the regression processing module 130 is specifically configured to: according to the location information of each MR of the plurality of MRs acquired by the MR acquisition module 110 and the feature vector corresponding to each of the plurality of MRs obtained by the feature vector module 120, Obtaining a training set corresponding to each MR of each of the plurality of MRs, wherein any training set includes feature vectors and position information corresponding to the same MR; and inputting a training set corresponding to each MR of the plurality of MRs into the machine learning model Train to get a single point positioning model.
  • the transmission probability calculation module 140 is specifically configured to: the position information of each MR of the plurality of MRs acquired by the MR acquisition module 110 and the feature vector corresponding to each of the plurality of MRs obtained by the feature vector module 120 Entering a single-point positioning model obtained by the regression processing module 130 to obtain a mapping relationship, wherein the mapping relationship is used to indicate a correspondence relationship between the feature vector and the location information; and calculating a feature vector corresponding to each of the plurality of MRs according to the mapping relationship The probability of launch.
  • the transmission probability calculation device described in this embodiment is used to implement the method described in the corresponding embodiment of FIG. 3.
  • a transmission probability calculation apparatus uses a plurality of parameter information in the MR and a feature vector obtained by an engineering parameter of the corresponding base station as an observation value, and trains a single point positioning model by using the feature vector and the position information corresponding to the MR,
  • the transmission probability obtained by the spatial model of the single-point positioning model can express complex observation information, and the correspondence between the feature vector (observation value) and the position information is more reliable.
  • the transition probability calculation device 200 includes: a trajectory acquisition module 210 and a Zhuangyi probability calculation module 220.
  • the trajectory acquisition module 210 is configured to acquire a terminal in the target area.
  • each of the plurality of trajectory data includes at least two pieces of position information, and the position information is used to indicate a position of the corresponding terminal in the target area, the plurality of Each of the plurality of pieces of position information included in the trajectory data corresponds to a time stamp;
  • the transition probability calculation module 220 is configured to calculate a transition probability according to the plurality of trajectory data acquired by the trajectory acquisition module 210, wherein the transition probability includes at least one transition probability The value, the transition probability value is used to indicate the probability that a certain location information passes through the time interval T to another location information.
  • the plurality of trajectory data includes at least one of environmental parameters, such as time period information, weather information, and event information, indicating a corresponding trajectory of the corresponding terminal under which environmental conditions are generated.
  • environmental parameters such as time period information, weather information, and event information, indicating a corresponding trajectory of the corresponding terminal under which environmental conditions are generated.
  • the trajectory data can be classified, and trajectory data including the same environmental parameters are obtained from existing trajectories with different environmental parameters, such as trajectory data during peak hours and trajectory data at rainy days, according to trajectory data including different environmental parameters.
  • Different transition probabilities are obtained, for example, the transition probability corresponding to the rainy environment is obtained for the positioning of the terminal in the target area in the rainy day.
  • the transition probability calculation module 220 includes: a pre-processing unit 221 and a transition probability calculation unit 222; the pre-processing unit 221 is configured to process the plurality of trajectory data obtained by the trajectory acquisition module 210 to obtain each of the plurality of trajectory data.
  • the transition probability calculation unit 222 is configured to perform the pre-processing according to the first preset condition At least one combined sequence of each of the plurality of trajectory data obtained by the unit 221, the transition probability corresponding to the first preset condition is obtained, wherein the first preset condition is any one of a plurality of preset conditions,
  • Each of the plurality of preset conditions includes a preset time interval and preset position information, where the preset time interval corresponds to the time interval T, and the preset position information corresponds to the certain position information.
  • the preset time interval is a preset time interval range.
  • transition probability calculation unit 222 is specifically configured to: determine a combination sequence of the preset time interval and the preset position information in the first preset condition in all the combined sequences included in the plurality of trajectory data obtained by the pre-processing unit 221;
  • the statistic includes a combined sequence of the preset time interval and the preset position information in the preset condition, and the calculation corresponds to the first preset condition. Transfer probability.
  • the transition probability calculation device 200 further includes: a first trajectory processing module 230; the first trajectory processing module 230 is configured to eliminate ⁇ trajectory data in the plurality of trajectory data acquired by the trajectory acquiring module 210, where the ⁇ trajectory data is There is at least one piece of trajectory data in which the distance of the position information from the road in the target area is greater than the first threshold, or there is trajectory data in which the distance between the adjacent two pieces of position information is greater than the second threshold.
  • the transition probability calculation device 200 further includes: a second trajectory processing module 240; the second trajectory processing module 240 is configured to determine sparse trajectory data in the plurality of trajectory data acquired by the trajectory acquiring module 210, where the sparse trajectory data is The trajectory data of the distance between any two adjacent position information included in the at least two pieces of position information is greater than the third threshold value; according to the map information of the target area, any two adjacent position information of the sparse trajectory data Insert one or more location information.
  • a second trajectory processing module 240 is configured to determine sparse trajectory data in the plurality of trajectory data acquired by the trajectory acquiring module 210, where the sparse trajectory data is The trajectory data of the distance between any two adjacent position information included in the at least two pieces of position information is greater than the third threshold value; according to the map information of the target area, any two adjacent position information of the sparse trajectory data Insert one or more location information.
  • transition probability calculation device described in this embodiment is used to implement the method described in the corresponding embodiment of FIG. 5 .
  • the transition probability calculation device described in this embodiment is used to implement the method described in the corresponding embodiment of FIG. 5 .
  • the transition probability calculation device described in this embodiment is used to implement the method described in the corresponding embodiment of FIG. 5 .
  • the transition probability calculation apparatus provided by an embodiment of the present application is configured to recover or predict the movement trajectory of a terminal in the target area by using the transition trajectory data in the target area provided by the third-party platform and the transition probability calculated according to the movement trajectory data. , can effectively avoid the phenomenon of track jump.
  • the sequence positioning apparatus 300 includes a transmission probability calculation module 310, a transition probability calculation module 320, and a sequence positioning module 330.
  • the application parameters of the sequence positioning model 330 include The transmission probability calculation module 310 is configured to calculate the transmission probability, the transition probability calculation module 320 is used to calculate the transition probability, and the sequence positioning module 330 is configured to obtain the movement trajectory of the target terminal.
  • the transmission probability calculation module 310 includes a measurement report MR acquisition unit 311, a feature vector unit 312, a regression processing unit 313, and a transmission probability calculation unit 314.
  • the MR acquisition unit 311 is configured to acquire multiple first terminals in the target area.
  • each MR of the plurality of MRs includes location information and parameter information, and the location information is used to indicate that the corresponding first terminal is a location in the target area;
  • the feature vector unit 312 is configured to obtain, according to the parameter information of each MR of the plurality of MRs acquired by the MR acquiring unit 311 and the engineering parameters of the at least one base station, respectively, corresponding to each MR of the plurality of MRs The feature vector;
  • the regression processing unit 313 is configured to obtain a single according to the position information of each of the plurality of MRs acquired by the MR acquiring unit 311 and the feature vector corresponding to each of the plurality of MRs obtained by the feature vector unit 312.
  • the transmission probability calculation unit 314 is configured to acquire a plurality of M of 311 according to the single point location model obtained by the regression processing unit 313 and the MR acquisition unit
  • the parameter information and feature vector unit 312 of each MR in R obtains a feature vector corresponding to each of the plurality of MRs, and calculates a transmission probability of a feature vector corresponding to each of the plurality of MRs, wherein the emission probability
  • the probability includes at least one transmission probability value, and the transmission probability value is used to indicate a probability that a certain feature vector corresponds to a certain location information.
  • the transmission probability calculation module 310 described in this embodiment is functionally consistent with the transmission probability calculation apparatus described in the embodiment corresponding to FIG. 12. For a specific description of the transmission probability calculation module 310, reference may be made to the purchase description of the embodiment corresponding to FIG. 12, where No longer.
  • the transition probability calculation module 320 includes a trajectory acquisition unit 321, a first trajectory processing unit 322, a second trajectory processing unit 323, and a transition probability calculation unit 324.
  • the trajectory acquisition unit 321 is configured to acquire multiple trajectories of the second terminal in the target area.
  • each of the plurality of trajectory data includes at least two pieces of position information
  • the position information is used to indicate a position of the corresponding second terminal in the target area
  • the plurality of pieces of trajectory data are included in the plurality of pieces of position information
  • Each position information corresponds to a time stamp
  • the first trajectory processing unit 322 is configured to cull the sinus trajectory data in the plurality of trajectory data acquired by the trajectory acquiring unit 321 , wherein the ⁇ trajectory data is that there is at least one position information deviating from the road in the target area The distance is greater than the first threshold of the trajectory data, or the distance between the adjacent two positional information is greater than the second threshold Track data.
  • the second trajectory processing unit 323 is configured to determine the sparse trajectory data in the plurality of trajectory data acquired by the trajectory acquiring unit 321 , wherein the sparse trajectory data is between any two adjacent position information included in the at least two pieces of position information The distance data is greater than the third threshold value; and one or more pieces of position information are inserted between any two adjacent position information of the sparse track data according to the map information of the target area.
  • the transition probability calculation unit 324 calculates a transition probability for the plurality of trajectory data processed by the first trajectory processing unit 322 and/or the second trajectory processing unit 323, wherein the transition probability includes at least one transition probability value, and the transition probability value is used for A probability of indicating that a certain location information has passed the time interval T to another location information.
  • the plurality of trajectory data obtained by the trajectory acquiring unit 321 is not processed by the first trajectory processing unit 322 and the second trajectory processing unit 323, and the transition probability calculating unit 324 calculates the transfer according to the plurality of trajectory data obtained by the trajectory acquiring unit 321 Probability.
  • the transition probability calculation module 320 described in this embodiment is functionally consistent with the transition probability calculation apparatus described in the embodiment corresponding to FIG. 13. For a detailed description of the transition probability calculation module 320, refer to the description of the embodiment corresponding to FIG. 13, where No longer.
  • the sequence positioning module 330 includes a target measurement report MR acquisition unit 331, a target feature vector unit 332, and a trajectory prediction unit 333.
  • the target MR unit 331 is configured to acquire a plurality of target measurement reports MR of the target terminal in the target area and at least within the target area.
  • each target MR of the plurality of target MRs includes parameter information
  • the target feature vector unit 332 is configured to each of the plurality of target MRs acquired according to the target MR unit 331
  • the parameter information of the target MR and the engineering parameters of the at least one base station obtain a target feature vector corresponding to each of the plurality of target MRs
  • the trajectory prediction unit 333 is used for the plurality of target MRs obtained according to the target feature vector unit 332.
  • the target feature vector corresponding to each target MR in each of them obtains the movement trajectory of the target terminal.
  • the sequence positioning apparatus obtains a transmission probability according to a plurality of parameter information in the MR and a feature vector obtained by an engineering parameter of the corresponding base station, and can express more complex observation information, thereby further improving the movement of the sequence positioning recovery/prediction.
  • the trajectory is more accurate and reliable; or the transition probability obtained by the third-party real trajectory data is used for sequence positioning, which can improve the smoothness of the recovered/predicted moving trajectory, and obtain a more reliable moving trajectory.
  • the device 400 includes a memory 410, a processor 420, an input/output port 430, and a power source 440.
  • the memory 410 is configured to store a programmable instruction
  • the processor 420 can execute the method for acquiring the transmission probability described in the corresponding embodiment of FIG. 3 and the method for obtaining the transition probability described in the corresponding embodiment of FIG. 5 by calling the programmable instruction stored in the memory 410.
  • the processor 420 can execute the method for acquiring the transmission probability described in the corresponding embodiment of FIG. 3 and the method for obtaining the transition probability described in the corresponding embodiment of FIG. 5 by calling the programmable instruction stored in the memory 410.
  • the specific method refer to the corresponding embodiment. Description, no longer repeat here.
  • the input/output port 430 is used by the processor 420 to perform data interaction with a device or device outside the device 400. Specifically, the processor 420 acquires MR and/or trajectory data from the outside through the input/output port 430, and passes the input/output. Port 430 outputs the calculated result;
  • Power source 440 is used to provide the required power to device 400.
  • Figure 16 shows the experimental setup and results of the actual test experiment of the solution of the present application; the sequence positioning method based on machine learning and feature engineering described in the embodiment of the present application is compared with the current industry mainstream single point location method Fingerprint positioning, Range-based positioning and other sequence positioning have a significant improvement in accuracy; the median error of the sequence positioning method described in the embodiment of the present application can reach 22 meters in the road test data, and the accuracy is improved by more than 20%. the above.
  • transition probability There are two main calculations for the transition probability in the prior art, one is to directly transfer from the current grid average probability to the adjacent grid; the other is to assume the equation of the motion pattern, and the transition probability is derived according to the equation.
  • the method described in the embodiment of the present application calculates the transition probability for each time granularity by real data, which is more realistic and practical than the existing technology. And the probability is more subtle.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Position Fixing By Use Of Radio Waves (AREA)

Abstract

本申请提供了一种获得发射概率的方法，该方法包括：获取目标区域内终端的多个测量报告MR和所述目标区域内至少一个基站的工程参数，根据所述多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量；通过回归模型处理所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到单点定位模型；根据所述单点定位模型、所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率。

Description

一种获取发射概率、转移概率以及序列定位的方法和装置 技术领域
本发明涉及电信定位领域，尤其涉及获得用于序列定位的发射概率的方法和装置。
背景技术
电信定位是指通过移动设备发送到管道侧(例如电信运营商)的数据和基站侧数据计算出该移动设备的位置。目前，常见的电信定位技术有Range-Based方法、Fingerprint方法、序列定位方法，其中序列定位方法(或序列方法)的主要思想是将定位过程描述成从观测序列到隐含序列的匹配模型，经纬度位置作为隐状态，信号强度作为观测值。在定位的时候，以一个观测值序列作为输入，输出一个最佳的隐状态序列与输入的观测值序列一一对应来作为定位结果。这类定位方法的核心是要定义发射概率(emission probability)和转移概率(transition probability)发射概率是指从隐状态到观测值的映射，转移概率是指不同隐状态之间的转换。
序列定位方法的主要优点是利用了位置的上下文信息，使得每个位置的预测结果都能够受到前一个位置的约束，进而得到的轨迹比较平滑，可以有效避免出现如Range-Based方法和Fingerprint方法得到的预测结果“乱跳”的情况。发射概率和转移概率的获取直接影响序列定位方法的定位能力，现有技术中通过对信号强度的均值方差进行建模得到发射概率，通过该方法得到的发射概率无法表述复杂的观测信息，直接影响序列定位方法的定位精度、可靠性。
发明内容
针对现有技术中存在的技术问题，本申请提供了一种获得发射概率的方法，使得获得的发射概率能够表述复杂的观测信息，将该发射概率用于序列定位能够提高序列定位方法的定位精度、可靠性。
第一方面，本申请提供了一种获得发射概率(emission probability)的方法，该方法包括：获取目标区域内终端的多个测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，具体地，可以根据人口数量、行政区域划分目标区域，例如某一城市的郊区为目标区域，或某一城市的市区为目标区域，本申请并不限制目标区域的面积大小、地理位置等，一般情况而言，根据某一区域的终端的MR得到的发射概率应用于该某一区域；需要说明的是，获取的目标区域内终端的多个MR可以是获取目标区域内的一个终端的多个MR；也可以是获取目标区域内的多个终端的多个MR，且不限定一个终端的MR的数量，多个终端的MR可以是一个也可以多个。所述多个MR中的每个MR包含位置信息和参数信息,所述位置信息用于指示包含所述位置信息的MR对应的终端在所述目标区域内的位置；本申请对MR的参数信息不做限定，一般是指MR中除位置信息之外的其他信息均可以归于参数信息；
可选的，参数信息包括环境参数，环境参数用于指示包含环境参数的MR对应的终端所处环境状况，例如：时间段信息、天气信息、事件信息(节假日、庆典日、运动会等)，针对不同环境状况得到的不同的发射概率，将环境参数作为特征可以更准确地支撑不同环境状况下的定位。
根据所述多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量，参数信息包括很多信息，以参数信息及相应基 站的工程参数得到的特征向量能够表达复杂的观测信息；通过机器学习模型处理所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到单点定位模型；根据所述单点定位模型、所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率,其中，所述发射概率包括至少一个发射概率值，所述发射概率值用于指示某一特征向量对应某一位置信息的概率；需要注意的，输入单点定位模型的MR和用于训练单点定位模型的MR可以不限于相同终端的，也可以是目标区域终端之外的其他终端的MR，即用于得到单点定位模型的MR和用于输入单点定位模型计算发射概率的MR可以是目标区域内不同的终端上传的MR。通过目标区域终端的MR训练得到单点定位模型，再将该目标区域内的终端的MR输入该单点定位模型，利用单点定位模型的空间模型统计特征向量与位置信息的对应关系，使得对应关系更加可靠。
第一方面的一种可能的实现方式，所述多个MR中的每个MR的参数信息包含至少一个基站ID，所述基站ID用于指示包含所述基站ID的MR对应的终端所连接的基站,在实际中，一个MR可能包含对应终端连接的多个基站的信息，所述至少一个基站至少包括所述多个MR包含的基站ID所指示的基站，进而每个MR中的基站ID都有对应的工程参数用于得到特征向量；所述根据所述多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量，包括：根据基站ID将所述多个MR与所述至少一个基站的工程参数进行匹配，得到所述多个MR中每个MR各自的关联工程参数，其中，任一MR的关联工程参数包括所述任一MR中每个基站ID所指示基站的工程参数；根据所述多个MR中每个MR各自的关联工程参数和参数信息，得到所述多个MR中每个MR各自对应的特征向量，其中，任一特征向量包括同一MR的关联工程参数和参数信息。需要注意的，特征向量中包含几个基站的工程参数一般取决于对应MR中基站ID指示的基站数量，可选的，特征向量可以仅包含对应MR中包含的至少一个基站ID指示的至少一个基站中的一个的工程参数。
第一方面的一种可能的实现方式，所述通过机器学习模型处理所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到单点定位模型，包括：根据所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到所述多个MR中的每个MR各自对应的训练集合，其中，任一训练集合包括同一MR对应的特征向量和位置信息；将所述多个MR中的每个MR各自对应的训练集输入所述机器学习模型进行训练，得到所述单点定位模型。
第一方面的一种可能的实现方式，所述根据所述单点定位模型、所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率，包括：将所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量输入所述单点定位模型得到映射关系，其中，所述映射关系用于指示特征向量与位置信息的对应关系；根据所述映射关系计算所述多个MR中的每个MR各自对应的特征向量的发射概率。可选的，输入单点定位模型的MR的特征向量和其中的位置信息，可以不是终端的MR对应的特征向量和其中的位置信息，可能是与终端所在相同目标区域内的其他多个终端的MR，且特征向量的获得方法与终端的MR对应的特征向量的获得方法相同，此处不再赘述。
第一方面的一种可能的实现方式，所述机器学习模型为回归模型，例如逻辑回归、随机森林等，此处对回归模型的具体模型不做限定。
本申请提供的获得发射概率的方法，以MR中多种参数信息、相应基站的工程参数得到的特征向量作为观测值，在以上述MR对应的特征向量和位置信息训练单点定位模型，利用单点定位模型的空间模型得到的发射概率，能够表达复杂的观测信息，且特征向量(观测值)与位置信息之间的对应关系更可靠。
第二方面，本发明提供了一种获得转移概率(transition probability)的方法，获取目标区域内终端的多个轨迹数据，其中，所述目标区域为预定的地理区域，具体地，可以根据人口数量、行政区域划分目标区域，例如某一城市的郊区为目标区域，或某一城市的市区为目标区域，本申请并不限制目标区域的面积大小、地理位置等，一般情况而言，根据某一区域的终端的MR得到的发射概率应用于该某一区域；所述多个轨迹数据中的每一轨迹数据包含至少两个位置信息，该位置信息用于指示包含所述位置信息的轨迹数据对应的终端在所述目标区域内的位置，所述多个轨迹数据包含的多个位置信息中的每一位置信息对应一个时间戳；根据所述多个轨迹数据计算转移概率,其中，所述转移概率包括至少一个转移概率值，所述转移概率值用于指示某一位置信息经时间间隔T到另一位置信息的概率。可选的，目标区域内终端的多个轨迹数据来自第三方平台，例如第三方APP滴滴出行、共享单车平台等。
可选的，所述多个轨迹数据包含相同的环境参数，所述环境参数用于指示包含所述环境参数的轨迹数据对应的终端所处环境状况，例如：时间段信息、天气信息、事件信息中的至少一个，针对不同环境状况得到的不同的转移概率，将环境参数作为标识可以更准确地支撑不同环境状况下的定位。
第二方面的一种可能的实现方式，所述根据所述多个轨迹数据计算转移概率包括：对所述多个轨迹数据进行处理，得到所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，所述组合序列包括同一轨迹数据中的任意两个位置信息及所述任意两个位置信息的时间间隔，其中，两个位置信息之间的时间间隔可以根据两个位置信息格子对应的时间戳计算得到；根据第一预设条件和所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率，其中，所述第一预设条件为多个预设条件中的任一个，所述多个预设条件中的每一预设条件包括预设时间间隔和预设位置信息，所述预设时间间隔对应所述时间间隔T，所述预设位置信息对应所述某一位置信息。
第二方面的一种可能的实现方式，所述根据第一预设条件和所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率包括：确定所述多个轨迹数据包含的全部组合序列中包含所述第一预设条件中的预设时间间隔和预设位置信息的组合序列；统计所述包含所述预设条件中的预设时间间隔和预设位置信息的组合序列，计算对应所述第一预设条件的转移概率。举例，确定包含位置信息A、时间间隔T1的全部组合序列的数量为M，可表示为[位置信息A，时间间隔T1，位置信息Xn]，在包含位置A、时间间隔T1的全部组合序列中分别统计位置信息X1～Xn个数，进而得到包含位置信息X1～Xn占M的概率值，对应位置信息X1～Xn的各个概率值构成对应位置信息A、时间间隔T1条件的转移概率。可选的，所述预设时间间隔为预设的时间间隔范围，预设时间间隔可以是特定时长，也可以是一个时长范围，例如预设时间间隔为2秒，或预设时间间隔为2～4秒，这样可以获得某一时长范围对应的转移概率，因为不同终端获得MR的频率有可能不同，导致不同终端的轨迹数据得到的组合序列中的时间间隔可能不同，通过将预设条件中时间间隔设定为一个取值范围，可以将不同时间间隔的组合序列进行融合，最大化利用现有数据。
第二方面的一种可能的实现方式，在所述根据所述多个轨迹数据计算转移概率之前，还包括：剔除所述多个轨迹数据中的瑕疵轨迹数据，其中，所述瑕疵轨迹数据为存在至少一个 位置信息偏离所述目标区域内道路的距离大于第一阈值的轨迹数据，或，存在相邻的两个位置信息之间的距离大于第二阈值的轨迹数据。在实际中，通过第三方获得轨迹数据可能存在不可靠的数据，即瑕疵轨迹数据，通过从获取的多个轨迹数据中剔除瑕疵轨迹数据，在对剔除后的轨迹数据继续进行处理得到转移概率，可以提高以此转移概率恢复/预测终端的移动轨迹更加可靠、平滑。
第二方面的一种可能的实现方式，在所述根据所述多个轨迹数据计算转移概率之前，还包括：确定所述多个轨迹数据中的稀疏轨迹数据，其中，所述稀疏轨迹数据为包含的所述至少两个位置信息中的任意两个相邻的位置信息之间的距离大于第三阈值的轨迹数据；根据所述目标区域的地图信息，在所述稀疏轨迹数据的任意两个相邻的位置信息之间插入一个或多个位置信息。通过插值处理，可以使得获取的轨迹数据中的位置信息更加稠密，根据插值处理后的轨迹数据得到的转移概率并用于恢复/预测终端的移动轨迹可以使得恢复/预测的移动轨迹更加平滑，也可以得到更多不同的时间间隔的转移概率。第二方面的一种可能的实现方式，所述获取目标区域内终端的多个轨迹数据包括：获取交通高峰时段或交通非高峰时段的所述目标区域内终端的多个轨迹数据。在实际中，交通高峰时段内的终端移动轨迹与交通非高峰时段内的终端移动轨迹通常存在区别，针对交通高峰时段和交通非高峰时段得到不同的转移概率，用于恢复/预测相应时段内的终端的移动轨迹，能够提高恢复/预测终端移动轨迹的准确性、可靠性。
本申请提供的获得转移概率的方法，通过第三方平台提供的目标区域内终端移动轨迹数据，根据移动轨迹数据计算的转移概率用于恢复或预测目标区域内某终端的移动轨迹会更平滑，可以有效避免轨迹跳跃现象。
第三方面，本申请提供了一种序列定位的方法，获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息；根据所述多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量；将所述多个目标MR中的每个目标MR各自对应的目标特征向量输入序列定位模型，得到所述目标终端的移动轨迹；其中，所述序列定位模型的应用参数包括发射概率和转移概率，该发射概率可以通过为第一方面或第一方面的任一可能的实现方式中描述的方法获得，或者/和，该转移概率可以通过为第二方面或第二方面的任一可能的实现方式中描述的方法获得，此处不在赘述。
本申请提供的序列定位的方法，根据MR中的多种参数信息及相应基站的工程参数得到的特征向量获得发射概率能够表达更复杂的观测信息，进一步提高序列定位恢复/预测的移动轨迹更加准确、可靠。
第四方面，本申请提供了一种序列定位的方法，该方法包括:获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息；根据所述多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量；将所述多个目标MR中的每个目标MR各自对应的目标特征向量输入序列定位模型，得到所述目标终端的移动轨迹；
其中，所述序列定位模型的应用参数包括发射概率和转移概率，所述转移概率通过以下方法获得：获取目标区域内终端的多个轨迹数据，其中，所述终端的多个轨迹数据中的每一轨迹数据包含至少两个第二位置信息，所述第二位置信息用于指示包含所述第二位置信息的 轨迹数据对应的终端在所述目标区域内的位置，所述终端的多个轨迹数据包含的多个第二位置信息中的每一第二位置信息对应一个时间戳；根据所述终端的多个轨迹数据计算转移概率,其中，所述转移概率包括至少一个转移概率值，所述转移概率值用于指示某一第二位置信息经时间间隔T到另一第二位置信息的概率。
第四方面的一个可能的实现方式，所述根据所述终端的多个轨迹数据计算转移概率包括：对所述终端的多个轨迹数据进行处理，得到所述终端的多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，所述组合序列包括所述终端的同一轨迹数据中的任意两个第二位置信息及所述任意两个第二位置信息的时间间隔；根据第一预设条件和所述终端的多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率，其中，所述第一预设条件为多个预设条件中的任一个，所述多个预设条件中的每一预设条件包括预设时间间隔和预设第二位置信息，所述预设时间间隔对应所述时间间隔T，所述预设第二位置信息对应所述某一第二位置信息。
第四方面的一个可能的实现方式，所述根据第一预设条件和所述终端的多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率包括：确定所述终端的多个轨迹数据包含的全部组合序列中包含所述第一预设条件中的预设时间间隔和预设第二位置信息的组合序列；统计所述包含所述预设条件中的预设时间间隔和预设第二位置信息的组合序列，计算对应所述第一预设条件的转移概率。
第四方面的一个可能的实现方式，在所述根据所述终端的多个轨迹数据计算转移概率之前，还包括：剔除所述终端的多个轨迹数据中的瑕疵轨迹数据，其中，所述瑕疵轨迹数据为存在至少一个第二位置信息偏离所述目标区域内道路的距离大于第一阈值的轨迹数据，或，存在相邻的两个第二位置信息之间的距离大于第二阈值的轨迹数据。
第四方面的一个可能的实现方式，在所述根据所述终端的多个轨迹数据计算转移概率之前，还包括：确定所述终端的多个轨迹数据中的稀疏轨迹数据，其中，所述稀疏轨迹数据为包含的所述至少两个第二位置信息中的任意两个相邻的第二位置信息之间的距离大于第三阈值的轨迹数据；根据所述目标区域的地图信息，在所述稀疏轨迹数据的任意两个相邻的第二位置信息之间插入一个或多个第二位置信息。
第四方面的一个可能的实现方式，所述获取目标区域内终端的多个轨迹数据包括：获取交通高峰时段或交通非高峰时段的所述目标区域内终端的多个轨迹数据。
第四方面的一个可能的实现方式，所述预设时间间隔为预设的时间间隔范围。
第四方面的一个可能的实现方式，所述发射概率通过第一方面或第一方面的任一可能的实现方式中描述的方法获得。
本申请提供的序列定位的方法，以第三方真实的轨迹数据获得的转移概率用于序列定位，能够提高恢复/预测的移动轨迹的平滑性，得到移动轨迹更加可靠。
第五方面，本申请提供了一种发射概率计算装置，该发射概率计算装置包括：测量报告MR获取模块、特征向量模块、回归处理模块、发射概率计算模块；所述MR获取模块用于获取目标区域内终端的多个测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个MR中的每个MR包含位置信息和参数信息,所述位置信息用于指示包含所述位置信息的MR对应的终端在所述目标区域内的位置；所述特征向量模块用于根据所述MR获取模块获取的多个MR中的每个MR的参数信息和至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量；所述回归处理模块用于根据所述MR获取模块获取的多个MR中的每个MR的位置信息和所述特征向量模块得到的 多个MR中的每个MR各自对应的特征向量，得到单点定位模型；所述发射概率计算模块用于根据所述回归处理模块得到的单点定位模型、所述MR获取模块获取的多个MR中的每个MR的位置信息和所述特征向量模块得到多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率,其中，所述发射概率包括至少一个发射概率值，所述发射概率值用于指示某一特征向量对应某一位置信息的概率。
第五方面的一个可能的实现方式，所述多个MR中的每个MR的参数信息包含至少一个基站ID，所述基站ID用于指示包含所述基站ID的MR对应的终端所连接的基站,所述至少一个基站至少包括所述多个MR包含的基站ID所指示的基站；所述特征向量模块具体用于:
根据基站ID将所述MR获取模块获取的多个MR与所述MR获取模块获取的至少一个基站的工程参数进行匹配，得到所述多个MR中每个MR各自的关联工程参数，其中，任一MR的关联工程参数包括所述任一MR中每个基站ID所指示基站的工程参数；根据所述MR获取模块获取的多个MR中每个MR各自的关联工程参数和参数信息，得到所述多个MR中每个MR各自对应的特征向量，其中，任一特征向量包括同一MR的关联工程参数和参数信息。
第五方面的一个可能的实现方式，所述回归处理模块具体用于：根据所述MR获取模块获取的多个MR中的每个MR的位置信息和所述特征向量模块得到的多个MR中的每个MR各自对应的特征向量，得到所述多个MR中的每个MR各自对应的训练集合，其中，任一训练集合包括同一MR对应的特征向量和位置信息；将所述多个MR中的每个MR各自对应的训练集输入所述机器学习模型进行训练，得到所述单点定位模型。
第五方面的一个可能的实现方式，所述发射概率计算模块具体用于：将所述MR获取模块获取的多个MR中的每个MR的位置信息和所述特征向量模块得到的多个MR中的每个MR各自对应的特征向量输入所述回归处理模块得到的单点定位模型，得到映射关系，其中，所述映射关系用于指示特征向量与位置信息的对应关系；根据所述映射关系计算所述多个MR中的每个MR各自对应的特征向量的发射概率。
本申请提供的发射概率计算装置，以MR中多种参数信息、相应基站的工程参数得到的特征向量作为观测值，在以上述MR对应的特征向量和位置信息训练单点定位模型，利用单点定位模型的空间模型得到的发射概率，能够表达复杂的观测信息，且特征向量(观测值)与位置信息之间的对应关系更可靠。
第六方面，本申请提供了一种转移概率计算装置，该转移概率计算装置包括：轨迹获取模块、转移概率计算模块；所述轨迹获取模块用于获取目标区域内终端的多个轨迹数据，其中，所述目标区域为预定的地理区域，所述多个轨迹数据中的每一轨迹数据包含至少两个位置信息，所述位置信息用于指示包含所述位置信息的轨迹数据对应的终端在所述目标区域内的位置，所述多个轨迹数据包含的多个位置信息中的每一位置信息对应一个时间戳；所述转移概率计算模块用于根据所述轨迹获取模块获取的多个轨迹数据计算转移概率,其中，所述转移概率包括至少一个转移概率值，所述转移概率值用于指示某一位置信息经时间间隔T到另一位置信息的概率。可选的，获取交通高峰时段或交通非高峰时段的所述目标区域内终端的多个轨迹数据。
第六方面的一个可能的实现方式，所述转移概率计算模块包括：预处理单元和转移概率计算单元；所述预处理单元用于对所述轨迹获取模块获得的多个轨迹数据进行处理，得到所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，所述组合序列包括同一轨迹数据中的任意两个位置信息及所述任意两个位置信息的时间间隔；所述转移概率计算单元用于根据第一预设条件和所述预处理单元得到的多个轨迹数据中的每一轨迹数据的至少一个组 合序列，得到对应所述第一预设条件的转移概率，其中，所述第一预设条件为多个预设条件中的任一个，所述多个预设条件中的每一预设条件包括预设时间间隔和预设位置信息，所述预设时间间隔对应所述时间间隔T，所述预设位置信息对应所述某一位置信息。可选的，所述预设时间间隔为预设的时间间隔范围。
第六方面的一个可能的实现方式，所述转移概率计算单元具体用于：确定所述预处理单元得到的多个轨迹数据包含的全部组合序列中包含所述第一预设条件中的预设时间间隔和预设位置信息的组合序列；统计所述包含所述预设条件中的预设时间间隔和预设位置信息的组合序列，计算对应所述第一预设条件的转移概率。
第六方面的一个可能的实现方式，所述装置还包括：第一轨迹处理模块；所述第一轨迹处理模块用于剔除所述轨迹获取模块获取的多个轨迹数据中的瑕疵轨迹数据，其中，所述瑕疵轨迹数据为存在至少一个位置信息偏离所述目标区域内道路的距离大于第一阈值的轨迹数据，或，存在相邻的两个位置信息之间的距离大于第二阈值的轨迹数据。
第六方面的一个可能的实现方式，所述装置还包括：第二轨迹处理模块；所述第二轨迹处理模块用于确定所述轨迹获取模块获取的多个轨迹数据中的稀疏轨迹数据，其中，所述稀疏轨迹数据为包含的所述至少两个位置信息中的任意两个相邻的位置信息之间的距离大于第三阈值的轨迹数据；根据所述目标区域的地图信息，在所述稀疏轨迹数据的任意两个相邻的位置信息之间插入一个或多个位置信息。
本申请提供的转移概率计算装置，通过第三方平台提供的目标区域内终端移动轨迹数据，根据移动轨迹数据计算的转移概率用于恢复或预测目标区域内某终端的移动轨迹会更平滑，可以有效避免轨迹跳跃现象。
第七方面，本申请提供了一种序列定位装置，该序列定位装置包括：发射概率计算模块、转移概率计算模块、序列定位模块；
所述序列定位模块包括目标测量报告MR获取单元、目标特征向量单元、轨迹预测单元；其中，所述目标MR单元用于获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息，所述目标特征向量单元用于根据所述目标MR单元获取的多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量，所述轨迹预测单元用于根据所述目标特征向量单元得到的多个目标MR中的每个目标MR各自对应的目标特征向量，得到所述目标终端的移动轨迹，所述序列定位模型的应用参数包括发射概率和转移概率；
所述发射概率计算模块用于计算所述发射概率，所述转移概率计算模块用于计算所述转移概率；
其中，所述发射概率计算模块包括测量报告MR获取单元、特征向量单元、回归处理单元、发射概率计算单元；所述MR获取单元用于获取目标区域内第一终端的多个测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述多个MR中的每个MR包含位置信息和参数信息,所述位置信息用于指示包含所述位置信息的MR对应的第一终端在所述目标区域内的位置；所述特征向量单元用于根据所述MR获取单元获取的多个MR中的每个MR的参数信息和至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量；所述回归处理单元用于根据所述MR获取单元获取的多个MR中的每个MR的位置信息和所述特征向量单元得到的多个MR中的每个MR各自对应的特征向量，得到单点定位模型；所述发射概率计算单元用于根据所述回归处理单元得到的单点定位模型、所述MR获取单元获 取的多个MR中的每个MR的参数信息和所述特征向量单元得到多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率,其中，所述发射概率包括至少一个发射概率值，所述发射概率值用于指示某一特征向量对应某一位置信息的概率；
或者，所述转移概率计算模块包括轨迹获取单元、转移概率计算单元；所述轨迹获取单元用于获取目标区域内第二终端的多个轨迹数据，其中，所述多个轨迹数据中的每一轨迹数据包含至少两个位置信息，所述位置信息用于指示包含所述位置信息的轨迹数据对应的第二终端在所述目标区域内的位置，所述多个轨迹数据包含的多个位置信息中的每一位置信息对应一个时间戳；所述转移概率计算单元用于根据所述轨迹获取单元获取的多个轨迹数据计算转移概率,其中，所述转移概率包括至少一个转移概率值，所述转移概率值用于指示某一位置信息经时间间隔T到另一位置信息的概率。
第七方面的一个可能的实现方式，所述多个MR中的每个MR的参数信息包含至少一个基站ID，所述基站ID用于指示包含所述基站ID的MR对应的第一终端所连接的基站,所述至少一个基站至少包括所述多个MR包含的基站ID所指示的基站；所述特征向量单元具体用于:根据基站ID将所述MR获取单元获取的多个MR与所述MR获取单元获取的至少一个基站的工程参数进行匹配，得到所述多个MR中每个MR各自的关联工程参数，其中，任一MR的关联工程参数包括所述任一MR中每个基站ID所指示基站的工程参数；根据所述MR获取单元获取的多个MR中每个MR各自的关联工程参数和参数信息，得到所述多个MR中每个MR各自对应的特征向量，其中，任一特征向量包括同一MR的关联工程参数和参数信息。
第七方面的一个可能的实现方式，所述回归处理单元具体用于：根据所述MR获取单元获取的多个MR中的每个MR的位置信息和所述特征向量单元得到的多个MR中的每个MR各自对应的特征向量，得到所述多个MR中的每个MR各自对应的训练集合，其中，任一训练集合包括同一MR对应的特征向量和位置信息；将所述多个MR中的每个MR各自对应的训练集输入所述机器学习模型进行训练，得到所述单点定位模型。
第七方面的一个可能的实现方式，所述发射概率计算单元具体用于：将所述MR获取单元获取的多个MR中的每个MR的位置信息和所述特征向量单元得到的多个MR中的每个MR各自对应的特征向量输入所述回归处理单元得到的单点定位模型，得到映射关系，其中，所述映射关系用于指示特征向量与位置信息的对应关系；根据所述映射关系计算所述多个MR中的每个MR各自对应的特征向量的发射概率。
第七方面的一个可能的实现方式，所述转移概率计算单元包括：预处理子单元和转移概率计算子单元；所述预处理子单元用于对所述轨迹获取单元获得的多个轨迹数据进行处理，得到所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，所述组合序列包括同一轨迹数据中的任意两个位置信息及所述任意两个位置信息的时间间隔；所述转移概率计算子单元用于根据第一预设条件和所述预处理子单元得到的多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率，其中，所述第一预设条件为多个预设条件中的任一个，所述多个预设条件中的每一预设条件包括预设时间间隔和预设位置信息，所述预设时间间隔对应所述时间间隔T，所述预设位置信息对应所述某一位置信息。
第七方面的一个可能的实现方式，所述转移概率计算子单元具体用于：确定所述预处理子单元得到的多个轨迹数据包含的全部组合序列中包含所述第一预设条件中的预设时间间隔和预设位置信息的组合序列；统计所述包含所述预设条件中的预设时间间隔和预设位置信息的组合序列，计算对应所述第一预设条件的转移概率。
第八方面，本申请提供了一种序列定位系统，本申请提供了一种序列定位系统，该系统包括：定位装置、第五方面或第五方面的任一可能的实现方式中描述的发射概率计算装置和第六方面或第六方面的任一可能的实现方式中描述的转移概率计算装置；所述定位装置包括目标测量报告MR获取模块、目标特征向量模块和轨迹预测模块；其中，所述目标MR模块用于获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息；所述目标特征向量模块用于根据所述目标MR模块获取的多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量；所述轨迹预测模块用于根据输入所述序列定位模块的所述目标特征向量模块得到的多个目标MR中的每个目标MR各自对应的目标特征向量，得到到所述目标终端的移动轨迹。所述发射概率计算装置向所述轨迹预测模块输入发射概率，所述转移概率计算装置相所述轨迹预测模块输入转移概率。
第九方面，本申请提供了一种发射概率计算装置，该发射概率计算装置包括：存储器、处理器；上述存储器用于存储可编程序指令，上述处理器调用上述存储器中存储的可编程序指令可以实现第一方面或第一方面的任一可能的实现方式中描述的方法。
第十方面，本申请提供了一种转移概率计算装置，该转移概率计算装置包括：存储器、处理器；上述存储器用于存储可编程序指令，上述处理器调用上述存储器中存储的可编程序指令可以实现第二方面或第二方面的任一可能的实现方式中描述的方法。
第十一方面，本申请提供了一种序列定位装置，该序列定位装置包括：存储器、处理器；上述存储器用于存储可编程序指令，上述处理器调用上述存储器中存储的可编程序指令可以实现第三方面或第四方面或第四方面的任一可能的实现方式中描述的方法。
第十二方面，本申请提供了一种计算机可读存储介质，包括指令，当其在计算机上运行时，使得计算机执行第一方面或第一方面的任一可能的实现方式或第二方面或第二方面的任一可能的实现方式或第三方面或第四方面或第四方面的任一可能实现方式中描述的方法。
第十三方面，本申请提供了一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行第一方面或第一方面的任一可能的实现方式或第二方面或第二方面的任一可能的实现方式或第三方面或第四方面或第四方面的任一可能实现方式中描述的方法。
第十四方面，本申请提供的一种序列定位系统，该序列定位系统包括第九方面描述的发射概率计算装置、第十方面描述的转移概率计算装置和序列定位装置。所述序列定位装置包括处理器和存储器，所述存储器用于存储可编程序指令，所述处理器调用所述存储器中存储的可编程序指令执行：获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息；根据所述多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量；将所述多个目标MR中的每个目标MR各自对应的目标特征向量输入序列定位模型，得到所述目标终端的移动轨迹。
本申请提供的序列定位的方法，根据MR中的多种参数信息及相应基站的工程参数得到的特征向量获得发射概率能够表达更复杂的观测信息，进一步提高序列定位恢复/预测的移动轨迹更加准确、可靠；或者，以第三方真实的轨迹数据获得的转移概率用于序列定位，能够提高恢复/预测的移动轨迹的平滑性，得到移动轨迹更加可靠。
附图说明
图1是本申请提供的一种序列定位系统的结构示意图；
图2是本申请提供的一种序列定位系统应用场景的示意图；
图3是本申请提供的一种获取发射概率方法的流程图；
图4是本申请提供的一种计算发射概率的示意图；
图5是本申请提供的一种获取转移概率方法的流程图；
图6是本申请提供的一种地图匹配和插值方法的示意图；
图7是本申请提供的一种离线索引方式计算转移概率方法的示意图；
图8是本申请提供的一种在线索引方式计算转移概率方法的示意图；
图9是本申请提供的一种序列定位方法的流程图；
图10是本申请提供的一种根据列定位方法恢复轨迹的示意图；
图11是本申请提供的一种粒子滤波序列定位方法的示意图；
图12是本申请提供的一种发射概率计算装置的示意图；
图13是本申请提供的一种转移概率计算装置的示意图；
图14是本申请提供的一种序列定位装置的示意图；
图15是本申请提供的一种计算发射概率和或转移概率设备的示意图；
图16是本申请提供的实际测试实验的实验设置及结果。
具体实施方式
下面将结合附图，对本申请实施例中的技术方案进行描述。
本申请一个实施例描述了一种定位系统，该系统用于对电信用户进行定位，恢复电信用户移动的轨迹。如图1所示，该定位系统包括发射概率(emission probability)计算装置、转移概率(transition probability)计算装置和序列定位装置，发射概率装置、转移概率装置、序列定位装置之间通过数据链路进行数据通信；其中，发射概率计算装置用于计算用于序列定位的发射概率,输入为带位置标注的测量报告(Measurement Report，MR)数据，输出发射概率；转移概率计算装置用于计算用于序列定位的转移概率，输入为移动轨迹数据，输出为转移概率；序列定位装置用于根据发射概率、转移概率以及待定位终端/设备的MR数据,恢复该待定终端/设备的移动轨迹，输入为待定位目标终端的一串MR(无位置信息)，输出为待定位目标终端移动轨迹，需要注意的是，也可以用于单点定位，即输入为一个MR，输出一个对应位置；下面会结合附图在进一步描述发射概率装置、转移概率装置和序列定位装置的具体结构及功能，在此不再赘述。需要注意的是，发射概率装置、转移概率装置、序列定位装置也可以组成一个整体设备，发射概率装置是该整体设备的发射概率计算模块，转移概率装置是该整体设备的转移概率计算模块，序列定位装置是该整体设备的序列定位模块，各模块的功能与对应装置的功能相同，各模块之间可以进行数据传递。可选的,由发射概率计算模块、转移概率计算模块和序列定位模块各个模块或上述三个功能模块构成的整体所实现的功能通过软件实现，也可以软件+硬件实现。
通常上，可以将上述定位系统部署在大数据分析平台上，当每天的MR存储到平台上时，首先会输入到定位系统，定位系统会根据MR抽取特征然后用我们离线训练好的模型预测出每一条MR记录的经纬度位置并添加回MR中，带经纬度位置的MR可以进行进一步的分析建模。
本申请一个实施例描述了一种应用场景，如图2所示，在该应用场景中，上述实施例描 述的定位系统可以作为运营商大数据分析平台的一个组件，以小时或日为单位将移动设备上传运营商管道中的MR输入定位系统中，定位系统会对输入其中的每一个MR得到一个相应的位置信息并添加到MR中；随后，这些已添加位置信息的MR可以用于运营商进行用户画像、人流预测、户外广告投放策略优化等用途。
本申请一个实施例描述了一种获得发射概率的方法，如图3所示，该方法包括以下步骤：
S101，获取目标区域内终端的多个测量报告MR和基站的工程参数；目标区域是指特定的地理区域，例如郊区、市区、城市、农村等，不限于面积大小、行政区域、地理位置；每个MR中包含位置信息和参数信息,位置信息用于标注对应的终端在该目标区域内的位置。
参数信息包括环境参数，环境参数用于指示对应的终端生成MR时所处真实的环境状况，例如时间段信息、天气信息、事件信息中的至少一个。将环境参数作为特征，针对不同环境状况得到的不同的发射概率，可以更准确地支撑不同环境状况下的定位。在实际实现中，以高峰时段和非高峰时段为例，高峰时段对应的环境参数用1表示，非高峰时段对应的环境参数泳0表示，根据环境参数为1的MR得到发射概率用于高峰时段目标区域内终端的定位更准确；同样地，根据环境参数为0的MR得到发射概率用于非高峰时段目标区域内终端的定位更准确。
S102，根据参数信息和工程参数得到每个MR各自对应的特征向量；具体地，根据参数信息中的基站ID，将MR与基站工程参数进行匹配，得到每个MR的关联工程参数；将每个MR的关联工程参数与其信息参数组合成其对应的特征向量。在具体实现上，MR中的信息参数可能包含一个或多个基站ID，这也表示该MR对应的终端同一时刻与一个或多个基站连接；当MR终端的参数信息包含多个基站ID时，那么某一MR的关联工程则为多个基站ID指示的基站的工程参数的集合。
S103，通过机器学习模型处理位置信息和特征向量得到单点定位模型；具体地，将每一MR的位置信息和对应的特征向量构成一组训练集合，并将每一MR对应的训练集合输入机器学习模型进行训练，得到单点定位模型；可选的，该单点定位模型可直接用于对目标区域的待定位终端进行定位，其输入为该待定位终端的MR中参数信息和基站工程参数组成的特征向量，输出为该待定位终端的位置信息。可选的，机器学习模型为回归模型，例如线性回归、随机森林等。
S104，根据单点定位模型、位置信息和特征向量计算发射概率；发射概率包括至少一个发射概率值，发射概率值用于指示作为观测值的某一特征向量对应作为隐含值的某一的位置信息的映射关系。具体地，将每个MR的位置信息和对应的特征向量输入所述单点定位模型得到映射关系，映射关系用于指示特征向量与位置信息的对应关系；根据映射关系计算每个MR各自对应的特征向量的发射概率。需要注意的是，S104涉及的位置信息和特征向量与S101–S103中可能涉及的位置信息和特征向量不一定是基于相同的MR得到，因为需要获取大量的MR，使用相同的MR减少数据再获取环节，且从结果来看，在MR基数足够大的情况下，是否使用不同的MR对结果影响较小。
一个计算发射概率的可能的实现方式，如下：
通过训练一个单点定位模型，随后使用这个单点定位模型的模型空间进行发射概率的计算，具体的包括：
首先，获取预设的地理区域内多个移动终端的MR，MR包括位置标注，例如MR中携带的GPS信息，以及获取一些其他电信网络中的数据，例如基站的工程参数，工程参数主要包括工程参数中主要包括了基站的ID、经纬度位置、天线挂高、天线方位角等信息。
然后，通过特征工程的方法从上述获取的数据中抽取与定位相关的特征作为相应MR的特征向量，特征向量可能包括连接基站、临区基站的工程参数、连接的信号强度等；具体的，将携带GPS的MR与基站的工程参数按照基站ID进行匹配，为MR中每一基站在工程参数中找到对应的参数信息；进一步地，在匹配后的数据上再加上一些简单的特征工程参数，例如每个MR连接的基站数量，不同基站位于不同的经纬度位置；或者，扇区数量，多个扇区可能会位于同一个经纬度为位置。这样，就可以为每一个MR创建一个特征向量，同一MR的特征向量和位置标注则组成相应该MR的训练集合，用于训练单点定位模型。
表1给出了一些用于训练定位模型的电信特征。包括了MR中的一些原始字段和基站工程参数中的一些字段。带*号的特征表示这些字段针对的是同一条MR的连接基站和临区基站，因此同一个字段会在特征向量中出现多次，对应不同的基站。
表1定位模型使用的特征列表
特征名称 说明
RNCID* RNC设备的ID
CellID* 小区ID
RSCP* 接收信号码功率
Ec/No* 信噪比
RSSI* 接收信号强度指示
天线挂高* 天线高度
天线方位角* 天线朝向
机械下倾角* 机械下倾角
电子下倾角* 电子下倾角
扇区经度* 天线所在位置的经度
扇区纬度* 天线所在位置的纬度
基站类型* 基站类型(宏站、室分)
基站生产厂商* 基站设备生产厂商(华为、诺西等)
设备连接扇区数量 该MR连接的扇区数量
设备连接基站数量 连接的不同位置的扇区的数量
进一步地，从MR中提取环境信息作为环境参数，例如天气信息(晴、雨、雪等)、时段(高峰时段、非高峰时段、工作日、周末、法定假日)、事件信息(运动会、演唱会、国庆等)；将环境参数和表1中的电信特征组成特征向量，与相应的位置标注组成训练集合。
再则，将上面步骤得到的全部训练集合输入机器学习回归模型，进行训练得到相对应模型——单点定位模型。机器学习回归模型可以有多种，比如线性回归，随机森林等；通常对于较大区域的单点定位模型训练，我们会将区域进行分块，对每块都训练一个单点定位模型，这样可以学习到不同区域的不同特征，比如城区和郊区训练两个模型。
最后，将上面步骤得到的特征向量或者基于其他带有位置标注的MR样本得到的特征序列作为观测值，位置标注作为隐含状态，输入上面步骤得到单点定位模型，则可以通过分析该单点定位模型的模型空间，得到位置标注和特征向量之间的对应关系，进而得得到P(特征向量|位置)。
不同模型的模型空间对应着不同的计算方式，下面以随机森林为例介绍模型空间的分析方法以及发射概率的计算方式，具体介绍如下:
当一个样本(包括特征向量和位置标注)输入到随机森林之后，会不断根据比较样本的特征值与决策节点的分裂特征值的大小，从而选择左孩子节点或者右孩子节点，直到最后到达叶子节点。因此，随机森林中的每一个叶子节点都可以看成一系列特征向量(通过一系列决策节点的分裂特征得到)，一个叶子节点看作一个观测值，于是发射概率就转变为给定标签得到叶子节点的概率P(叶子节点|标注)。如图4所示，将标注样本输入到训练好的树模型中，每个标注样本都能在树中找到自己对应的一个叶子节点(如图4第三部分所示)。于是，发射概率值就可以由该标注样本落在该叶子节点中的数量除以该标注样本中相同标签的总表标注样本数得到，比如圆点样本一共有10个而只有一个落在图4第三幅图的最左侧的的叶子节点中，那么发射概率值就是1/10，它表示标注样本的位置与观测值的符合程度(在该位置上可以得到类似的观测值的概率)，符合程度越高说明预测的位置越准。
本申请一个实施例描述了一种获得转移概率的方法，如图5所示，该方法包括以下步骤：
S201，获取目标区域内一个或多个终端的多个轨迹数据,目标区域是指特定的地理区域，例如郊区、市区、城市、农村等，不限于面积大小、行政区域、地理位置；每一轨迹数据包含至少两个位置信息，位置信息用于标注终端在该目标区域内的位置，每一位置信息对应一个时间戳，时间戳用于指示终端生成对应的位置信息的时刻。在具体实现中，可以从第三方平台获取轨迹数据，例如滴滴出行、交通数据公布平台等。
可选的，获取目标区域内一个或多个终端的多个轨迹数据为：获取交通高峰时段的该目标区域内终端的轨迹数据，或者，获取交通非高峰时段的该目标区域内一个或多个终端的轨迹数据；也可以，预设的时间段的一个或多个终端的轨迹数据。当然，根据某一时间段获取的终端的轨迹数据得到转移概率，应用于相同的某一时间段的序列定位的效果最佳，也可以应用于与该时间段类似的时间段的序列定位的。
S202,根据上述多个轨迹数据计算转移概率，转移概率包括至少一个转移概率值，转移概率值用于指示某一位置信息(起始位置)经时间间隔T到另一位置信息(到达位置)的概率。具体地，对获取的多个轨迹数据进行处理，得到每一轨迹数据的组合序列，同一轨迹数据中的任意两个位置信息及该任意两个位置信息的时间间隔构成一个组合序列；一个轨迹数据可能有一个或多个组合序列；一个组合序列中的两个位置信息，起一个位置信息作为起始位置，另一个则为到达位置，该组合序列则表示起始位置经过某一时间间隔至到达位置；在上面得到所有组合序列中筛选出包含预设条件的组合序列，该预设条件为预设起始位置和时间间隔，预设条件不同筛选出的组合序列也不同；在筛选出的符合预设条件的组合序列中，以到达位置为对象，统计不同到达位置的数量并计算相应占符合预设条件的全部组合序列的概率值，即转移概率值，计算得到全部概率值的集合为对应该预设条件的转移概率。可选的，预设条件中的预设时间间隔可以是某一时间间隔，也可以是时间间隔范围，例如：预设时间间隔为2秒，或者，预设时间间隔为2—4秒，也就是时间间隔满足2-4范围则符合预设条件的预设时间间隔。
可选的，在根据上多个轨迹数据计算转移概率之前，剔除多个轨迹数据中的瑕疵轨迹数据，瑕疵轨迹数据是指存在位置信息偏离目标区域内道路的距离大于一定阈值的轨迹数据，或，存在相邻的两个位置信息之间的距离大于一定阈值的轨迹数据。将剔除瑕疵轨迹数据后多个轨迹数据用于计算转移概率，可以提高转移概率的可靠程度或两相邻位置过渡的平滑程度。
可选的，在根据上多个轨迹数据计算转移概率之前，通过插值稠密化获得的轨迹数据中的稀疏轨迹数据，稀疏轨迹数据是指任意两个相邻的位置信息之间的距离大于第三阈值的轨 迹数据，插值是指根据地图信息和轨迹数据在相邻的位置信息之间添加一个或多个位置信息，使得轨迹数据中的位置信息稠密，具体的可以根据间隔时间进行添加位置信息，例如某一轨迹仅有两个位置且两位置之时间间隔6秒，要实现时间间隔为3秒，则在两个位置之间插入一个位置，可以使得该轨迹的位置之间的时间间隔为3秒；基于上一例子，要实现时间间隔为1秒，则可以两个位置之间插入5个位置，即每个1秒插入一个位置，可以使得该轨迹的位置之间的时间间隔为1秒。插入的位置的具体地理位置信息则根据地图信息和轨迹可以比较准确的确定。
可选的，多条轨迹数据包含指示对应的终端在怎样的环境状况下移动生成的相应轨迹的环境参数，例如时间段信息、天气信息、事件信息中的至少一个。根据环境参数可以将轨迹数据进行分类，从环境参数不同的现有轨迹中得到包含相同环境参数的轨迹数据，例如高峰时段的轨迹数据、雨天时候的轨迹数据，根据包含不同环境参数的轨迹数据可以得到不同的转移概率，例如得到对应雨天环境的转移概率用于雨天时目标区域内终端的定位。本申请一个实施例描述的获得转移概率的方法是利用迁移知识的概念，将从真实轨迹数据中学习到的运动模式用于计算转移概率。
一个计算转移概率的可能的实现方式，如下：
首先，获取一批在定位区域中的第三方轨迹数据，随后将数据中偏差比较大的轨迹去掉，偏差大表现在轨迹中的点距离道路比较远或者轨迹中频繁出现点瞬间跳跃到很远地方的情况。然后，对不够稠密的轨迹数据使用地图匹配加插值的方法进行密集化，使得我们能够得到比较细粒度(轨迹相邻两点的时间间隔尽量小)的转移概率。最后需要对轨迹点坐标进行离散化，对整块定位区域均匀划分矩形栅格(栅格大小在20m*20m左右)，这样每一个坐标点就可以唯一对应一个栅格ID。
转移概率计算过程可以分为轨迹密集化和转移概率学习两大过程。
轨迹密集化过程是为了能够学习到任意时间间隔的转移概率，具体的，首先把每条轨迹通过地图匹配算法映射到路网上，这样每条轨迹经过的路段都可以被推测出来。随后我们沿着轨迹经过的道路均匀地进行插值，使得插值之后相邻两点间的时间间隔为1s，这样以秒为粒度任意时间间隔的转移概率都可以被学习出来。
如图6所示的一种地图匹配和插值的实现方法。地图匹配方法有很多种，比如使用的是针对低采样频率轨迹的地图匹配方法，通过计算原始轨迹点到附近路段的匹配概率和路段之间的转移概率，就能得到一个概率最大的路段序列，得到匹配的路径之后，再在轨迹相邻点之间均匀插值直到满足相邻两点之间的时间间隔等于1s。
转移概率学习过程主要是从轨迹中学习到从每个位置到其他位置的转移概率，具体的学习方式有很多种。
本申请一个实施例描述了一种离线索引方式学习转移概率的方法，如图7所示，该方式分为两个部分，上半部分是离线索引建立，下半部分是在线查询。
建立离线索引一共需要三步，第一步将轨迹数据处理成图7中表格形式，一共为3列(轨迹ID，时间戳，栅格ID)，每一行代表一个轨迹点的记录。接下来第二步是从这个表格中抽取出三元组<Δt,i,j>，Δt为两条记录的时间戳之差，i和j分别为两条记录的栅格ID。相同轨迹ID的记录两两可以生成一个三元组，表示可以在Δt时间内从栅格i移动到栅格j。在实际实现中，只需要抽取Δt<60s的记录即可。第三步通过统计第二步生成的三元组得到转移概率矩阵。比如从栅格1出发1s内到达其他栅格的概率，只需要找出所有满足Δt＝1,i＝1的三元组<1,1,j>，然后统计不同j出现的频率可以得到转移概率向量，即满足Δt＝1,i＝1条 件的转移概率。根据不同的出发栅格i可以得到转移概率矩阵(转移的时间间隔相同)，然后不同的转移时间间隔又可以得到不同的转移概率矩阵。同时，为了更符合现实情况，我们区分了高峰时段和非高峰时段的转移概率。使用高峰时段(比如7:00-9:00,17:00-19:00)的轨迹数据来生成高峰时段的转移矩阵，用其他时段的轨迹数据来生成非高峰时段的转移矩阵。
在线查询过程主要是在序列定位时，给定转移时间间隔和出发栅格的条件下，获得到达其他栅格的概率分布向量。首先，根据当前的时间是否是高峰时段选择对应的离线索引；随后，根据时间间隔Δt选取对应的转移概率矩阵；最后，根据出发栅格i，找到转移概率矩阵中的对应行即为要求的转移概率向量。
本申请一个实施例描述了一种在线索引方式的转移概率计算方法，如图8所示，该方式也可以分为三步，前两步和离线索引一样。首先将轨迹数据处理成三列的表格形式(轨迹ID，时间戳，栅格ID)，然后从这个表格中抽取出三元组<Δt,i,j>。接下来用RTree对所有抽取出来的三元组<Δt,i,j>建立三维索引(三元组中的三个元素分别对应索引的三维)。
在线查询时，利用RTree的范围查询(Range Query)，给定Δt和i的范围，比如1≤Δt≤2,1≤i≤1，RTree能返回所有满足这个条件的三元组，之后再从中取出所有第三个元素j，根据j的值分布得到转移概率分布。在线索引与离线索引不同之处是它可以把时间间隔设置为一个范围，比如上面例子里设置的1-2s。
本申请一个实施例描述了一种序列定位的方法,如图9所示，该方法包括以下步骤:
S301,获取目标区域内目标终端的多个目标MR和基站的工程参数，每个目标MR包含参数信息，参数信息包含环境参数，具体可以参见上面相应实施例的描述；
S302,根据目标MR的参数信息和基站的工程参数得到目标特征向量；每个目标MR对应一个目标特征向量，目标特征向量作为观测值，用于输入序列定位模型得到相应的隐含的位置；
S303,将得到的目标特征向量输入序列定位模型得到该目标终端的移动轨迹。该序列定位模型应用到的发射概率和转移概率通过上面实施例描述的获得发射概率的方法和获得转移概率的方法计算，此处不再赘述。需要注意的是，获得发射概率、转移概率的目标区域与序列定位的目标区域为同一个地理区域，同样的，时间段上也为同一时间段，这样的效果更佳。
当发射概率和转移概率都准备好之后，就可以使用序列定位方法恢复用户的轨迹。与计算发射概率时得到特征向量的方法一样，这里也需要对待定位的终端的MR做同样的处理，生成对应的特征向量，将同一个待定位终端的一串特征向量加上之前得到的发射概率和转移概率输入到序列定位方法中，算法就能够根据特征序列预测出待定位终端移动的轨迹，如图10所示。
序列定位方法有多种方式，本申请一个实施例描述了一种粒子滤波序列定位方法，如图11所示。粒子滤波的思想是找到一串粒子序列长度为T(长度和要恢复的轨迹长度相同)，使这个序列和MR的特征向量最相符。
第一步，在状态空间内初始化粒子，生成一个粒子集P＝{p^((1)),p^((2)),…,p^((N))}，每个粒子对应一个状态和一个重要性权重<x_1^((i)),w_1^((i))>(上标i表示第几个粒子，下标1表示这个粒子对应轨迹中的第一个点)，通常粒子的个数在几百几千这个范围，初始化的每个粒子都会在后面的步骤中通过状态转移形成一个粒子序列，并且初始的状态会选择一个在合理的范围内进行随机初始化(比如在连接基站方圆几百米之内)。重要性权重即为发射概率p(y|x)给定状态得到观测值的概率，根据之前发射概率的计算方法，输入粒子的状态(相当于标注)，得到对应的发射概率值。
第二步是采样。随后根据每个粒子的当前状态x_j^((i))、MR中前后两点的时间间隔Δ t_j采样下一状态，这里使用到了前面实施例中的离线索引，通过上面描述的在线查询方法输入Δt_j和x_j^((i))之后得到状态转移概率分布p(x_(j+1)^((i))|x_j^((i)))。从这个分布中采样一个状态作为第i个粒子j+1时刻的状态x_(j+1)^((i))。
第三步是决策。根据x_(j+1)^((i))和w_j^((i))计算对应的重要性权重w_(j+1)^((i))＝w_j^((i))p(y_(j+1)|x_(j+1)^((i)))。对所有粒子序列的重要性权重进行归一化可以得到一个重要性的分布。做完第二和第三步之后，粒子集中的所有粒子序列的长度都增加了1。
第四步是重采样。如果此时所有粒子序列的重要性权重的分布满足一定条件，则对粒子进行重采样。重采样就是一种有放回采样的过程，按照权重的大小来进行采样，权重越大的粒子序列被采样到的概率就越大(并且可能会被采样到多次)。用重采样之后的粒子序列替换之前的(采样前后序列的数量不变)。如果当前粒子序列长度小于T，则需要把所有粒子序列的重要性权重都重置成1/N。
重复二三四步，直到粒子序列长度等于T。此时用重要性权重最大的粒子序列来作为预测轨迹输出，一个粒子序列对应着一串状态，也就是一串经纬度位置。
本申请一个实施例描述了一种Viterbi序列定位的方法。该方法利用了动态规划的思想，不断地更新V_(t,k)矩阵，它表示前t个序列的最终状态为k的状态序列的概率。每次计算V_(t+1,k)时，都要找V_(t,x)*a_(x,k)的最大值，其中x是变量，也就是找到最合适的k的前一个状态；a_(x,k)是我们求出的转移概率，从x栅格移动到k栅格的概率。随后
Figure PCTCN2017108647-appb-000001
Figure PCTCN2017108647-appb-000002
其中b代表求出的发射概率，这样更新完所有的V矩阵值之后就可以找到V_(T,k)这一行中的最大值，接着回溯找前一个状态(满足前面公式中最大值得状态)，直到得到一个状态转移序列。
本申请一个实施例描述了一种发射概率计算装置，如图12所示，发射概率计算装置100包括：MR获取模块110、特征向量模块120、回归处理模块130和发射概率计算模块140；MR获取模块110用于获取目标区域内终端的多个测量报告MR和目标区域内至少一个基站的工程参数，其中，目标区域为预定的地理区域，多个MR中的每个MR包含位置信息和参数信息,位置信息用于指示对应的终端在目标区域内的位置；特征向量模块120用于根据MR获取模块110获取的多个MR中的每个MR的参数信息和至少一个基站的工程参数，得到多个MR中的每个MR各自对应的特征向量；回归处理模块130用于根据MR获取模块110获取的多个MR中的每个MR的位置信息和特征向量模块120得到的多个MR中的每个MR各自对应的特征向量，得到单点定位模型；发射概率计算模块140用于根据回归处理模块130得到的单点定位模型、MR获取模块110获取的多个MR中的每个MR的位置信息和特征向量模块120得到多个MR中的每个MR各自对应的特征向量，计算多个MR中的每个MR各自对应的特征向量的发射概率,其中，发射概率包括至少一个发射概率值，发射概率值用于指示某一特征向量对应某一位置信息的概率。
进一步地，多个MR中的每个MR的参数信息包含至少一个基站ID，基站ID用于指示包含基站ID的MR对应的终端所连接的基站,至少一个基站至少包括多个MR包含的基站ID所指示的基站；特征向量模块120具体用于:根据基站ID将MR获取模块110获取的多个MR与MR获取模块110获取的至少一个基站的工程参数进行匹配，得到多个MR中每个MR各自的关联工程参数，其中，任一MR的关联工程参数包括任一MR中每个基站ID所指示基站的工程参数；根据MR获取模块获取的多个MR中每个MR各自的关联工程参数和参数信息，得到多个MR中每个MR各自对应的特征向量，其中，任一特征向量包括同一MR的关联工 程参数和参数信息。
进一步地，回归处理模块130具体用于：根据MR获取模块110获取的多个MR中的每个MR的位置信息和特征向量模块120得到的多个MR中的每个MR各自对应的特征向量，得到多个MR中的每个MR各自对应的训练集合，其中，任一训练集合包括同一MR对应的特征向量和位置信息；将多个MR中的每个MR各自对应的训练集输入机器学习模型进行训练，得到单点定位模型。
进一步地，发射概率计算模块140具体用于：将MR获取模块110获取的多个MR中的每个MR的位置信息和特征向量模块120得到的多个MR中的每个MR各自对应的特征向量输入回归处理模块130得到的单点定位模型，得到映射关系，其中，映射关系用于指示特征向量与位置信息的对应关系；根据述映射关系计算多个MR中的每个MR各自对应的特征向量的发射概率。
本实施例描述的发射概率计算装置用于实现图3对应实施例描述的方法，更为具体的描述可以参见图3对应的实施例，此处不再赘述。
本申请一个实施例提供的发射概率计算装置，以MR中多种参数信息、相应基站的工程参数得到的特征向量作为观测值，在以上述MR对应的特征向量和位置信息训练单点定位模型，利用单点定位模型的空间模型得到的发射概率，能够表达复杂的观测信息，且特征向量(观测值)与位置信息之间的对应关系更可靠。
本申请一个实施例描述了一种转移概率计算装置，如图13所示，转移概率计算装置200包括：轨迹获取模块210、庄毅概率计算模块220；轨迹获取模块210用于获取目标区域内终端的多个轨迹数据，其中，目标区域为预定的地理区域，多个轨迹数据中的每一轨迹数据包含至少两个位置信息，位置信息用于指示对应的终端在目标区域内的位置，多个轨迹数据包含的多个位置信息中的每一位置信息对应一个时间戳；转移概率计算模块220用于根据轨迹获取模块210获取的多个轨迹数据计算转移概率,其中，转移概率包括至少一个转移概率值，转移概率值用于指示某一位置信息经时间间隔T到另一位置信息的概率。可选的，获取交通高峰时段或交通非高峰时段的目标区域内终端的多个轨迹数据。
可选的，多条轨迹数据包含指示对应的终端在怎样的环境状况下移动生成的相应轨迹的环境参数，例如时间段信息、天气信息、事件信息中的至少一个。根据环境参数可以将轨迹数据进行分类，从环境参数不同的现有轨迹中得到包含相同环境参数的轨迹数据，例如高峰时段的轨迹数据、雨天时候的轨迹数据，根据包含不同环境参数的轨迹数据可以得到不同的转移概率，例如得到对应雨天环境的转移概率用于雨天时目标区域内终端的定位。
进一步地，转移概率计算模块220包括：预处理单元221和转移概率计算单元222；预处理单元221用于对轨迹获取模块210获得的多个轨迹数据进行处理，得到多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，组合序列包括同一轨迹数据中的任意两个位置信息及该任意两个位置信息的时间间隔；转移概率计算单元222用于根据第一预设条件和预处理单元221得到的多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应该第一预设条件的转移概率，其中，第一预设条件为多个预设条件中的任一个，多个预设条件中的每一预设条件包括预设时间间隔和预设位置信息，预设时间间隔对应上述时间间隔T，预设位置信息对应上述某一位置信息。可选的，预设时间间隔为预设的时间间隔范围。
进一步地，转移概率计算单元222具体用于：确定预处理单元221得到的多个轨迹数据包含的全部组合序列中包含第一预设条件中的预设时间间隔和预设位置信息的组合序列；统计包含预设条件中的预设时间间隔和预设位置信息的组合序列，计算对应该第一预设条件的 转移概率。
可选的，转移概率计算装置200还包括：第一轨迹处理模块230；第一轨迹处理模块230用于剔除轨迹获取模块210获取的多个轨迹数据中的瑕疵轨迹数据，其中，瑕疵轨迹数据为存在至少一个位置信息偏离目标区域内道路的距离大于第一阈值的轨迹数据，或，存在相邻的两个位置信息之间的距离大于第二阈值的轨迹数据。
可选的，转移概率计算装置200还包括：第二轨迹处理模块240；第二轨迹处理模块240用于确定轨迹获取模块210获取的多个轨迹数据中的稀疏轨迹数据，其中，稀疏轨迹数据为包含的至少两个位置信息中的任意两个相邻的位置信息之间的距离大于第三阈值的轨迹数据；根据目标区域的地图信息，在稀疏轨迹数据的任意两个相邻的位置信息之间插入一个或多个位置信息。
本实施例描述的转移概率计算装置用于实现图5对应实施例描述的方法，更为具体的描述可以参见图5对应的实施例，此处不再赘述
本申请一个实施例提供的转移概率计算装置，通过第三方平台提供的目标区域内终端移动轨迹数据，根据移动轨迹数据计算的转移概率用于恢复或预测目标区域内某终端的移动轨迹会更平滑，可以有效避免轨迹跳跃现象。
本申请一个实施例提供了一种序列定位装置，如图14所示，序列定位装置300包括：发射概率计算模块310、转移概率计算模块320和序列定位模块330，序列定位模型330的应用参数包括发射概率和转移概率；发射概率计算模块310用于计算发射概率，转移概率计算模块320用于计算转移概率，序列定位模块330用于得到目标终端的移动轨迹。
具体的，发射概率计算模块310包括测量报告MR获取单元311、特征向量单元312、回归处理单元313、发射概率计算单元314；其中，MR获取单元311用于获取目标区域内第一终端的多个测量报告MR和目标区域内至少一个基站的工程参数，其中，目标区域为预定的地理区域，多个MR中的每个MR包含位置信息和参数信息,位置信息用于指示对应的第一终端在目标区域内的位置；特征向量单元312用于根据MR获取单元311获取的多个MR中的每个MR的参数信息和至少一个基站的工程参数，得到多个MR中的每个MR各自对应的特征向量；回归处理单元313用于根据MR获取单元311获取的多个MR中的每个MR的位置信息和特征向量单元312得到的多个MR中的每个MR各自对应的特征向量，得到单点定位模型；发射概率计算单元314用于根据回归处理单元313得到的单点定位模型、MR获取单元获取311的多个MR中的每个MR的参数信息和特征向量单元312得到多个MR中的每个MR各自对应的特征向量，计算多个MR中的每个MR各自对应的特征向量的发射概率,其中，发射概率包括至少一个发射概率值，发射概率值用于指示某一特征向量对应某一位置信息的概率。本实施例描述的发射概率计算模块310与图12对应的实施例描述的发射概率计算装置的功能上一致，关于发射概率计算模块310具体描述可以参见图12对应的实施例的买描述，此处不再赘述。
转移概率计算模块320包括轨迹获取单元321、第一轨迹处理单元322、第二轨迹处理单元323和转移概率计算单元324；其中，轨迹获取单元321用于获取目标区域内第二终端的多个轨迹数据，其中，多个轨迹数据中的每一轨迹数据包含至少两个位置信息，位置信息用于指示对应的第二终端在目标区域内的位置，多个轨迹数据包含的多个位置信息中的每一位置信息对应一个时间戳；第一轨迹处理单元322用于剔除轨迹获取单元321获取的多个轨迹数据中的瑕疵轨迹数据，其中，瑕疵轨迹数据为存在至少一个位置信息偏离目标区域内道路的距离大于第一阈值的轨迹数据，或，存在相邻的两个位置信息之间的距离大于第二阈值的 轨迹数据。第二轨迹处理单元323用于确定轨迹获取单元321获取的多个轨迹数据中的稀疏轨迹数据，其中，稀疏轨迹数据为包含的至少两个位置信息中的任意两个相邻的位置信息之间的距离大于第三阈值的轨迹数据；根据目标区域的地图信息，在稀疏轨迹数据的任意两个相邻的位置信息之间插入一个或多个位置信息。转移概率计算单元324用于第一轨迹处理单元322和/或第二轨迹处理单元323处理后得的多个轨迹数据计算转移概率,其中，转移概率包括至少一个转移概率值，转移概率值用于指示某一位置信息经时间间隔T到另一位置信息的概率。可选的，对轨迹获取单元321获得的多个轨迹数据不经过第一轨迹处理单元322和第二轨迹处理单元323处理，转移概率计算单元324根据轨迹获取单元321获得的多个轨迹数据计算转移概率。本实施例描述的转移概率计算模块320与图13对应的实施例描述的转移概率计算装置的功能上一致，关于转移概率计算模块320的具体描述可以参见图13对应的实施例的描述，此处不再赘述。
序列定位模块330包括目标测量报告MR获取单元331、目标特征向量单元332、轨迹预测单元333；其中，目标MR单元331用于获取目标区域内目标终端的多个目标测量报告MR和目标区域内至少一个基站的工程参数，上述目标区域为预定的地理区域，上述多个目标MR中每个目标MR包含参数信息，目标特征向量单元332用于根据目标MR单元331获取的多个目标MR中的每个目标MR的参数信息和至少一个基站的工程参数，得到多个目标MR中的每个目标MR各自对应的目标特征向量，轨迹预测单元333用于根据目标特征向量单元332得到的多个目标MR中的每个目标MR各自对应的目标特征向量，得到目标终端的移动轨迹。本申请的一个实施例提供的序列定位装置，根据MR中的多种参数信息及相应基站的工程参数得到的特征向量获得发射概率能够表达更复杂的观测信息，进一步提高序列定位恢复/预测的移动轨迹更加准确、可靠；或者，以第三方真实的轨迹数据获得的转移概率用于序列定位，能够提高恢复/预测的移动轨迹的平滑性，得到移动轨迹更加可靠。
本申请一个实施例提供设备，如图15所示，设备400包括：存储器410、处理器420、输入/输出端口430和电源440。
存储器410用于存储可编程序指令；
处理器420调用存储器410中存储的可编程序指令可以执行图3对应实施例描述的获取发射概率的方法和/或图5对应实施例描述的获取转移概率的方法；具体方法参见相应实施例的描述，此处不再赘述。
输入/输出端口430用于处理器420与设备400外的设别或装置进行数据交互，具体的，处理器420通过输入/输出端口430从外界获取MR和/或轨迹数据，并通过输入/输出端口430将计算的结果输出；
电源440用于为设备400提供所需的电力。
图16展示了本申请方案实际测试实验的实验设置及结果；本申请实施例描述的基于机器学习和特征工程的序列定位方法，相比当前业界主流的单点定位方法指纹库(Fingerprint)定位、距离相关(Range-Based)定位及其他序列定位，精度上均有了大幅的提升；本申请实施例描述的序列定位方法的中位误差在路测数据上可以达到22米，精度提高超过20％以上。
相对现有技术中对于观测值的定义仅考虑了单一的信号强度，本发明将这个扩展到了任意维度的特征组合。
现有技术中对于转移概率的计算主要有两种，一种是直接从当前栅格平均概率转移到临近的栅格；另一种是假定运动模式的方程，根据该方程推算出转移概率。本申请实施例描述的方法通过真实数据计算出对于每个时间粒度的转移概率，相比现有的技术更佳切合实际并 且概率更精细。
最后应说明的是：以上各实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述各实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。

Claims (28)

  1. 一种获得发射概率的方法，所述发射概率用于序列定位，其特征在于，包括：
    获取目标区域内终端的多个测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个MR中的每个MR包含位置信息和参数信息,所述位置信息用于指示包含所述位置信息的MR对应的终端在所述目标区域内的位置，所述参数信息包括环境参数，所述环境参数用于指示包含所述环境参数的MR对应的终端所处环境状况；
    根据所述多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量；
    通过回归模型处理所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到单点定位模型；
    根据所述单点定位模型、所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率,其中，所述发射概率包括至少一个发射概率值，所述发射概率值用于指示某一特征向量对应某一位置信息的概率。
  2. 如权利要求1所述的方法，其特征在于，所述多个MR中的每个MR的参数信息包含至少一个基站ID，所述基站ID用于指示包含所述基站ID的MR对应的终端所连接的基站,所述至少一个基站至少包括所述多个MR包含的基站ID所指示的基站；
    所述根据所述多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述多个MR中的每个MR各自对应的特征向量，包括：
    根据基站ID将所述多个MR与所述至少一个基站的工程参数进行匹配，得到所述多个MR中每个MR各自的关联工程参数，其中，任一MR的关联工程参数包括所述任一MR中每个基站ID所指示基站的工程参数；
    根据所述多个MR中每个MR各自的关联工程参数和参数信息，得到所述多个MR中每个MR各自对应的特征向量，其中，任一特征向量包括同一MR的关联工程参数和参数信息。
  3. 如权利要求1所述的方法，其特征在于，所述通过回归模型处理所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到单点定位模型，包括：
    根据所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，得到所述多个MR中的每个MR各自对应的训练集合，其中，任一训练集合包括同一MR对应的特征向量和位置信息；
    将所述多个MR中的每个MR各自对应的训练集输入所述机器学习模型进行训练，得到所述单点定位模型。
  4. 如权利要求1所述的方法，其特征在于，所述根据所述单点定位模型、所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量，计算所述多个MR中的每个MR各自对应的特征向量的发射概率，包括：
    将所述多个MR中的每个MR的位置信息和所述多个MR中的每个MR各自对应的特征向量输入所述单点定位模型，得到映射关系，其中，所述映射关系用于指示特征向量与位置信息的对应关系；
    根据所述映射关系计算所述多个MR中的每个MR各自对应的特征向量的发射概率。
  5. 如权利要求1-4任选一所述的方法，其特征在于，所述环境参数包括时间段信息、天气信息、事件信息中的至少一个。
  6. 一种获得转移概率的方法，其特征在于，包括：
    从第三方平台获取目标区域内终端的多个轨迹数据，其中，所述目标区域为预定的地理区域，所述多个轨迹数据包含相同的环境参数，所述环境参数用于指示包含所述环境参数的轨迹数据对应的终端所处环境状况，所述多个轨迹数据中的每一轨迹数据包含至少两个位置信息，所述位置信息用于指示包含所述位置信息的轨迹数据对应的终端在所述目标区域内的位置，所述多个轨迹数据包含的多个位置信息中的每一位置信息对应一个时间戳；
    根据所述多个轨迹数据计算转移概率,其中，所述转移概率包括至少一个转移概率值，所述转移概率值用于指示某一位置信息经时间间隔T到另一位置信息的概率。
  7. 如权利要求6所述的方法，其特征在于，所述根据所述多个轨迹数据计算转移概率包括：
    对所述多个轨迹数据进行处理，得到所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，所述组合序列包括同一轨迹数据中的任意两个位置信息及所述任意两个位置信息的时间间隔；
    根据第一预设条件和所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率，其中，所述第一预设条件为多个预设条件中的任一个，所述多个预设条件中的每一预设条件包括预设时间间隔和预设位置信息，所述预设时间间隔对应所述时间间隔T，所述预设位置信息对应所述某一位置信息。
  8. 如权利要求7所述的方法，其特征在于，所述根据第一预设条件和所述多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率包括：
    确定所述多个轨迹数据包含的全部组合序列中包含所述第一预设条件中的预设时间间隔和预设位置信息的组合序列；
    统计所述包含所述预设条件中的预设时间间隔和预设位置信息的组合序列，计算对应所述第一预设条件的转移概率。
  9. 如权利要求6-8任选一所述的方法，其特征在于，在所述根据所述多个轨迹数据计算转移概率之前，还包括：
    剔除所述多个轨迹数据中的瑕疵轨迹数据，其中，所述瑕疵轨迹数据为存在至少一个位置信息偏离所述目标区域内道路的距离大于第一阈值的轨迹数据，或，存在相邻的两个位置信息之间的距离大于第二阈值的轨迹数据。
  10. 如权利要求6-8任选一所述的方法，其特征在于，在所述根据所述多个轨迹数据计算转移概率之前，还包括：
    确定所述多个轨迹数据中的稀疏轨迹数据，其中，所述稀疏轨迹数据为包含的所述至少两个位置信息中的任意两个相邻的位置信息之间的距离大于第三阈值的轨迹数据；
    根据所述目标区域的地图信息，在所述稀疏轨迹数据的任意两个相邻的位置信息之间插入一个或多个位置信息。
  11. 如权利要求6-8任选一所述的方法，其特征在于，所述获取目标区域内终端的多个轨迹数据包括：
    获取交通高峰时段或交通非高峰时段的所述目标区域内终端的多个轨迹数据。
  12. 如权利要求6-11任选一所述的方法，其特征在于，所述环境参数包括时间段信息、天气信息、事件信息中的至少一个。
  13. 如权利要求7所述的方法，其特征在于，所述预设时间间隔为预设的时间间隔范围。
  14. 一种序列定位方法，其特征在于，包括：
    获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息，所述多个目标MR中每个目标MR包含的目标参数信息包括目标环境参数，所述目标环境参数用于指示包含所述目标环境参数的目标MR对应的目标终端所处环境状况；
    根据所述多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量；
    将所述多个目标MR中的每个目标MR各自对应的目标特征向量输入序列定位模型，得到所述目标终端的移动轨迹；
    其中，所述序列定位模型的应用参数包括发射概率和转移概率，所述发射概率通过以下方法获得：获取目标区域内第一终端的多个测量报告MR，所述第一终端的多个MR中的每个MR包含位第一置信息和参数信息,所述第一位置信息用于指示包含所述第一位置信息的MR对应的第一终端在所述目标区域内的位置,所述第一终端的多个MR中的每个MR包含的参数信息包括第一环境参数，所述第一环境参数用于指示包含所述第一环境参数的MR对应的第一终端所处环境状况；根据所述第一终端的多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述第一终端的多个MR中的每个MR各自对应的特征向量；通过机器学习模型处理所述第一终端的多个MR中的每个MR的第一位置信息和所述第一终端的多个MR中的每个MR各自对应的特征向量，得到单点定位模型；根据所述单点定位模型、所述第一终端的多个MR中的每个MR的第一位置信息和所述第一终端的多个MR中的每个MR各自对应的特征向量，计算所述第一终端的多个MR中的每个MR各自对应的特征向量的发射概率，所述发射概率包括至少一个发射概率值，所述发射概率值用于指示某一特征向量对应某一第一位置信息的概率，所述机器学习模型为回归模型。
  15. 如权利要求14所述的方法，其特征在于，所述第一终端的多个MR中的每个MR的参数信息包含至少一个基站ID，所述基站ID用于指示包含所述基站ID的MR对应的第一终端所连接的基站,所述至少一个基站至少包括所述第一终端的多个MR包含的全部基站ID所指示的基站；
    所述根据所述第一终端的多个MR中的每个MR的参数信息和所述至少一个基站的工程参数，得到所述第一终端的多个MR中的每个MR各自对应的特征向量，包括：
    根据基站ID将所述第一终端的多个MR与所述至少一个基站的工程参数进行匹配，得到所述第一终端的多个MR中每个MR各自的关联工程参数，其中，任一第一终端的MR各自的关联工程参数包括所述任一第一终端的MR中每个基站ID所指示基站的工程参数；
    根据所述第一终端的多个MR中每个MR的关联工程参数和参数信息，得到所述第一终端的多个MR中每个MR各自对应的特征向量，其中，任一特征向量包括所述第一终端的同一MR的关联工程参数和参数信息。
  16. 如权利要求14所述的方法，其特征在于，所述通过机器学习模型处理所述第一终端的多个MR中的每个MR的第一位置信息和所述第一终端的多个MR中的每个MR各自对应的特征向量，得到单点定位模型，包括：
    根据所述第一终端的多个MR中的每个MR的第一位置信息和所述第一终端的多个MR中的每个MR各自对应的特征向量，得到所述第一终端的多个MR中的每个MR各自对应的训练集合，其中，任一训练集合包括所述第一终端的同一MR对应的特征向量和第一位置信 息；
    将所述第一终端的多个MR中的每个MR各自对应的训练集输入所述机器学习模型进行训练，得到所述单点定位模型。
  17. 如权利要求14所述的方法，其特征在于，所述根据所述单点定位模型、所述第一终端的多个MR中的每个MR的第一位置信息和所述第一终端的多个MR中的每个MR各自对应的特征向量，计算所述第一终端的多个MR中的每个MR各自对应的特征向量的发射概率，包括：
    将所述第一终端的多个MR中的每个MR的第一位置信息和所述第一终端的多个MR中的每个MR各自对应的特征向量输入所述单点定位模型，得到映射关系，其中，所述映射关系用于指示特征向量与第一位置信息的对应关系；
    根据所述映射关系计算所述第一终端的多个MR中的每个MR各自对应的特征向量的发射概率。
  18. 如权利要求14-17任选一所述的方法，其特征在于，所述第一环境参数包括时间段信息、天气信息、事件信息中的至少一个。
  19. 如权利要求14-17任选一所述的方法，其特征在于，所述转移概率通过以下方法获得：
    从第三方平台获取目标区域内第二终端的多个轨迹数据，其中，所述多个轨迹数据包含相同的第二环境参数，所述第二环境参数用于指示包含所述第二环境参数的轨迹数据对应的第二终端所处环境状况，所述第二终端的多个轨迹数据中的每一轨迹数据包含至少两个第二位置信息，所述第二位置信息用于指示包含所述第二位置信息的轨迹数据对应的第二终端在所述目标区域内的位置，所述第二终端的多个轨迹数据包含的多个位置信息中的每一第二位置信息对应一个时间戳；
    根据所述第二终端的多个轨迹数据计算转移概率,其中，所述转移概率包括至少一个转移概率值，所述转移概率值用于指示某一第二位置信息经时间间隔T到另一第二位置信息的概率。
  20. 如权利要求19所述的方法，其特征在于，所述根据所述第二终端的多个轨迹数据计算转移概率包括：
    对所述第二终端的多个轨迹数据进行处理，得到所述第二终端的多个轨迹数据中的每一轨迹数据的至少一个组合序列，其中，所述组合序列包括所述第二终端的同一轨迹数据中的任意两个第二位置信息及所述任意两个第二位置信息的时间间隔；
    根据第一预设条件和所述第二终端的多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率，其中，所述第一预设条件为多个预设条件中的任一个，所述多个预设条件中的每一预设条件包括预设时间间隔和预设第二位置信息，所述预设时间间隔对应所述时间间隔T，所述预设第二位置信息对应所述某一第二位置信息。
  21. 如权利要求20所述的方法，其特征在于，所述根据第一预设条件和所述第二终端的多个轨迹数据中的每一轨迹数据的至少一个组合序列，得到对应所述第一预设条件的转移概率包括：
    确定所述第二终端的多个轨迹数据包含的全部组合序列中包含所述第一预设条件中的预设时间间隔和预设第二位置信息的组合序列；
    统计所述包含所述预设条件中的预设时间间隔和预设第二位置信息的组合序列，计算对应所述第一预设条件的转移概率。
  22. 如权利要求19-21任选一所述的方法，其特征在于，所述第二环境参数包括时间段信息、天气信息、事件信息中的至少一个。
  23. 如权利要求14-22所述的方法，其特征在于，所述目标环境参数包括时间段信息、天气信息、事件信息中的至少一个。
  24. 一种发射概率计算装置，其特征在于，包括：存储器、处理器；
    所述存储器用于存储可编程序指令；
    所述处理器调用所述存储器中存储的可编程序指令执行权利要求1-5任选一所述的方法。
  25. 一种转移概率计算装置，其特征在于，包括：存储器、处理器；
    所述存储器用于存储可编程序指令；
    所述处理器调用所述存储器中存储的可编程序指令执行权利要求6-13任选一所述的方法。
  26. 一种序列定位装置，其特征在于，包括：存储器、处理器；
    所述存储器用于存储可编程序指令；
    所述处理器调用所述存储器中存储的可编程序指令执行权利要求14-23任选一所述的方法。
  27. 一种序列定位系统，其特征在于，包括：序列定位装置、如权利要求24所述的发射概率计算装置、如权利要求25所述的转移概率计算装置；
    所述序列定位装置包括目标测量报告MR获取模块、目标特征向量模块和轨迹预测模块；其中，所述目标MR模块用于获取目标区域内目标终端的多个目标测量报告MR和所述目标区域内至少一个基站的工程参数，其中，所述目标区域为预定的地理区域，所述多个目标MR中每个目标MR包含参数信息，所述多个目标MR中每个目标MR包含参数信息包括目标环境参数，所述目标环境参数用于指示包含所述目标环境参数的目标MR对应的目标终端所处环境状况；所述目标特征向量模块用于根据所述目标MR模块获取的多个目标MR中的每个目标MR的参数信息和所述至少一个基站的工程参数，得到所述多个目标MR中的每个目标MR各自对应的目标特征向量；所述轨迹预测模块用于根据所述目标特征向量模块得到的多个目标MR中的每个目标MR各自对应的目标特征向量，得到到所述目标终端的移动轨迹，所述序列定位装置的应用参数包括发射概率和转移概率；
    所述发射概率计算装置向所述轨迹预测模块输入发射概率，所述转移概率计算装置向所述轨迹预测模块输入转移概率。
  28. 如权利要求27所述的系统，其特征在于，所述目标环境参数包括时间段信息、天气信息、事件信息中的至少一个。
PCT/CN2017/108647 2017-08-15 2017-10-31 一种获取发射概率、转移概率以及序列定位的方法和装置 WO2019033565A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
CN201780087427.1A CN110326323B (zh) 2017-08-15 2017-10-31 一种获取发射概率、转移概率以及序列定位的方法和装置
EP17921756.7A EP3657841A1 (en) 2017-08-15 2017-10-31 Method and device for obtaining emission probability, transfer probability and sequence positioning
US16/791,795 US11290975B2 (en) 2017-08-15 2020-02-14 Method and apparatus for obtaining emission probability, method and apparatus for obtaining transition probability, and sequence positioning method and apparatus
US17/688,087 US20220191818A1 (en) 2017-08-15 2022-03-07 Method and Apparatus for Obtaining Emission Probability, Method and Apparatus for Obtaining Transition Probability, and Sequence Positioning Method and Apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710698562.1A CN109413683B (zh) 2017-08-15 2017-08-15 一种获取发射概率、转移概率以及序列定位的方法和装置
CN201710698562.1 2017-08-15

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/791,795 Continuation US11290975B2 (en) 2017-08-15 2020-02-14 Method and apparatus for obtaining emission probability, method and apparatus for obtaining transition probability, and sequence positioning method and apparatus

Publications (1)

Publication Number Publication Date
WO2019033565A1 true WO2019033565A1 (zh) 2019-02-21

Family

ID=65361759

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/108647 WO2019033565A1 (zh) 2017-08-15 2017-10-31 一种获取发射概率、转移概率以及序列定位的方法和装置

Country Status (4)

Country Link
US (2) US11290975B2 (zh)
EP (1) EP3657841A1 (zh)
CN (2) CN109413683B (zh)
WO (1) WO2019033565A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109413683B (zh) * 2017-08-15 2021-09-21 华为技术有限公司 一种获取发射概率、转移概率以及序列定位的方法和装置
CN111867049B (zh) * 2019-04-25 2021-11-19 华为技术服务有限公司 定位方法、装置及存储介质
CN110809284B (zh) * 2019-09-25 2022-09-16 福建新大陆软件工程有限公司 基于mr数据的定位方法、系统、设备、可读存储介质
CN115297506B (zh) * 2022-10-10 2022-12-20 中通服建设有限公司 一种基于大数据的地铁线路网络智能测评方法和系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060240842A1 (en) * 1998-09-22 2006-10-26 Polaris Wireless, Inc. Estimating the Location of a Wireless Terminal Based on Non-Uniform Probabilities of Movement
CN103634810A (zh) * 2013-12-24 2014-03-12 山东润谱通信工程有限公司 一种室内无线网络覆盖问题区域定位的方法
CN104202761A (zh) * 2014-09-15 2014-12-10 南通大学 信道状态转移概率估计方法

Family Cites Families (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4053234B2 (ja) * 2000-12-11 2008-02-27 株式会社日立グローバルストレージテクノロジーズ ディスク装置
FI113409B (fi) * 2002-05-31 2004-04-15 Ekahau Oy Sekvenssiperusteinen paikannustekniikka
EP1943777B1 (en) * 2005-10-31 2016-07-20 LG Electronics Inc. Method for processing control information in a wireless mobile communication system
US8209121B1 (en) * 2007-10-10 2012-06-26 Google Inc. Registration of location data to street maps using hidden markov models, and application thereof
US8554247B2 (en) * 2008-11-13 2013-10-08 Glopos Fzc Method and system for refining accuracy of location positioning
CN103037507A (zh) * 2012-12-17 2013-04-10 浙江鸿程计算机系统有限公司 一种基于Cell-ID定位技术的地图匹配方法
US9225789B2 (en) * 2013-10-10 2015-12-29 Pushd, Inc. Automated mobile positional social media method and system
ES2458621B1 (es) * 2013-10-15 2015-02-10 Aoife Solutions, S.L. Sistema de control descentralizado de redes inalámbricas
CN104853432A (zh) * 2014-02-18 2015-08-19 电信科学技术研究院 Wlan接入点的位置确定方法及用户设备、网络侧设备
US9125019B1 (en) * 2014-05-01 2015-09-01 Glopos Fzc Positioning arrangement, method, mobile device and computer program
IL239503B (en) * 2014-06-19 2018-08-30 Cellwize Wireless Tech Ltd A method and system for analyzing data collected in a cellular network
US9173064B1 (en) * 2014-10-06 2015-10-27 Polaris Wireless, Inc. Estimating proximity to a mobile station by manipulating an interfering signal
CN105303837A (zh) * 2015-11-24 2016-02-03 东南大学 一种检测驾驶人跟驰行为特性参数的方法及系统
CN106595633B (zh) * 2016-11-25 2019-07-19 北京邮电大学 室内定位方法及装置
CN106709606B (zh) * 2016-12-29 2020-10-30 平安科技（深圳）有限公司 个性化场景预测方法及装置
CN107231615A (zh) * 2017-06-27 2017-10-03 深圳市优网精蜂网络有限公司 一种基于网络指纹的定位方法及系统
CN109413683B (zh) * 2017-08-15 2021-09-21 华为技术有限公司 一种获取发射概率、转移概率以及序列定位的方法和装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060240842A1 (en) * 1998-09-22 2006-10-26 Polaris Wireless, Inc. Estimating the Location of a Wireless Terminal Based on Non-Uniform Probabilities of Movement
CN103634810A (zh) * 2013-12-24 2014-03-12 山东润谱通信工程有限公司 一种室内无线网络覆盖问题区域定位的方法
CN104202761A (zh) * 2014-09-15 2014-12-10 南通大学 信道状态转移概率估计方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3657841A4 *

Also Published As

Publication number Publication date
CN110326323A (zh) 2019-10-11
CN109413683A (zh) 2019-03-01
US20220191818A1 (en) 2022-06-16
CN109413683B (zh) 2021-09-21
US20200187149A1 (en) 2020-06-11
EP3657841A4 (en) 2020-05-27
EP3657841A1 (en) 2020-05-27
US11290975B2 (en) 2022-03-29
CN110326323B (zh) 2021-12-14

Similar Documents

Publication Publication Date Title
US11290975B2 (en) Method and apparatus for obtaining emission probability, method and apparatus for obtaining transition probability, and sequence positioning method and apparatus
CN103068035B (zh) 一种无线网络定位方法、装置及系统
KR102116824B1 (ko) 딥러닝 기반 측위 시스템 및 이의 구축 방법
CN106851571B (zh) 一种基于决策树的快速KNN室内WiFi定位方法
CN110972062B (zh) 一种基于手机信令数据的基站位置参数校准方法和系统
CN105682025A (zh) 基于移动信令数据的用户驻留地识别方法
CN107688906B (zh) 多方法融合的输电线路气象要素降尺度分析系统及方法
Zaidi et al. Mobility estimation for wireless networks based on an autoregressive model
CN108061556B (zh) 一种基于隐马尔可夫模型的地图匹配方法
CN109041218B (zh) 一种预测用户位置的方法和智能硬件
CN110109055B (zh) 一种基于rssi测距的室内定位方法
CN111148030A (zh) 指纹数据库的更新方法、装置、服务器及存储介质
CN109874104B (zh) 用户位置定位方法、装置、设备及介质
JP2011158459A (ja) パターン分類を使用した、無線端末が屋内にあるかどうかの推定
CN103929719A (zh) 定位信息的优化方法和优化装置
CN111787488B (zh) 一种用户定位方法及系统
CN108770057B (zh) 预定指纹库的确定方法及指纹定位方法、装置、存储介质
CN108259097B (zh) 基于mr数据的修正无线传播模型的方法及装置
CN114236480A (zh) 一种机载平台传感器系统误差配准算法
CN107750051B (zh) 无线传播模型的优化方法及其装置
CN114916059A (zh) 基于区间随机对数阴影模型的WiFi指纹稀疏地图扩建方法
CN115942231A (zh) 一种基于rss的5g室外定位方法
CN114222238B (zh) 定位方法、装置和计算机可读存储介质
CN111356152B (zh) 基站位置矫正方法、装置、设备及介质
CN116634378B (zh) 一种出行道路轨迹识别方法、装置、设备及存储介质

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: 17921756

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2017921756

Country of ref document: EP

Effective date: 20200221