WO2020215783A1 - 定位方法、装置及存储介质 - Google Patents

定位方法、装置及存储介质 Download PDF

Info

Publication number
WO2020215783A1
WO2020215783A1 PCT/CN2019/129333 CN2019129333W WO2020215783A1 WO 2020215783 A1 WO2020215783 A1 WO 2020215783A1 CN 2019129333 W CN2019129333 W CN 2019129333W WO 2020215783 A1 WO2020215783 A1 WO 2020215783A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
information
historical position
positioning
historical
Prior art date
Application number
PCT/CN2019/129333
Other languages
English (en)
French (fr)
Inventor
李汐
毛娜
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
Publication of WO2020215783A1 publication Critical patent/WO2020215783A1/zh

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
    • H04W64/006—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
    • 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
    • H04W4/029—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

Definitions

  • This application relates to the field of communication technology, and in particular to a positioning method, device and storage medium.
  • a measurement report is reported by the mobile terminal periodically or triggered by an event, and contains the wireless environment information of the mobile terminal at a certain moment, and it has a corresponding relationship with the current geographic location of the mobile terminal. Therefore, positioning the mobile terminal based on the measurement report is currently a common method for positioning the mobile terminal.
  • the MR reported by the mobile terminal is first obtained, and the MR within a preset time period is selected based on the user identification carried by the MR or the user process call detail record (CDR) identification, And sort according to the timestamp information carried by the MR, and then calculate the center of gravity of the triangle formed by the primary serving cell and the strongest two neighboring cells of each MR within the preset time period. The triangle center of gravity of the cell and the time difference between every two adjacent MRs are calculated to calculate the moving direction and speed of the mobile terminal. Finally, according to the direction and speed of the mobile terminal, the current MR cell level intensity is smoothed, and finally based on the smoothed cell The level strength and the pre-established fingerprint library determine the location of the mobile terminal.
  • CDR user process call detail record
  • the embodiments of the present application provide a positioning method, device and storage medium to solve the problem of low positioning accuracy of terminal equipment in the prior art.
  • the first aspect of the present application provides a positioning method, including: obtaining a measurement report MR data set, the MR data set including: multiple pieces of MR data; for any piece of MR data in the MR data set, according to the MR data Determine the characteristic information of the MR data with at least one piece of MR data that is temporally adjacent to the MR data, and the characteristic information of the MR data is information used to describe the location of the terminal device corresponding to the MR data; The characteristic information of the MR data is input into the positioning model to obtain the position information of the terminal device corresponding to the MR data.
  • the positioning model takes the characteristic information of each historical position MR data in the historical position MR data set as input, and
  • the historical position MR data corresponding to the position information of the terminal device is a model obtained by output training, and the historical position MR data refers to historical MR data carrying position information.
  • the positioning device processes the MR data based on at least one piece of MR data that has a time sequence relationship with the MR data during the positioning process, and the characteristic information of the obtained MR data can more accurately characterize the position information of the terminal device. , Thereby improving the positioning accuracy.
  • the method before the determining the characteristic information of the MR data based on the MR data and at least one piece of MR data temporally adjacent to the MR data, the method further include:
  • all MR data in the MR data set are sorted to obtain a sorted MR data set, after the sorting process MR data with the same user ID in the MR data set is arranged together according to the time stamp information;
  • the positioning device when the positioning device processes the acquired MR data set, the MR data with the same user ID is arranged together according to the time stamp information, so that the positioning device can sort the MR data in the MR data set after processing.
  • the data is feature extraction
  • the time sequence of each MR data report in the MR data set can be taken into account, laying the foundation for subsequent extraction of reasonable and accurate feature information; by combining the data at each moment or each time period
  • the integration of MR data can make the feature information in each moment or time period more complete, thereby improving the accuracy of the feature information corresponding to the MR data.
  • the method further includes:
  • the combined MR data set is split to obtain multiple MR data sub-sets.
  • Each MR data sub-set includes: multiple pieces of MR data sorted according to timestamp information and with the same user ID, each MR The total duration of all MR data in the data subset is less than the preset second duration, and the time difference corresponding to the time stamp information of two adjacent MR data in each MR data subset is less than the preset third duration, so The total duration is the time difference between the last piece of MR data in the MR data subset and the time stamp information of the first piece of MR data.
  • the merged MR data set is split according to the preset call extraction strategy to obtain multiple MR data subsets, which can improve the performance of feature information when extracting features of MR data. Extraction efficiency and accuracy.
  • the method further includes:
  • the processed MR data is obtained by filtering all the MR data in each MR data subset Collecting to provide positioning accuracy during subsequent positioning.
  • the determining the characteristic information of the MR data according to the MR data and at least one piece of MR data temporally adjacent to the MR data includes:
  • the first feature information set is used as the feature information of the MR data.
  • the first feature information set is obtained based on the MR call sequence corresponding to the MR data, it takes into account the timing relationship between the MR data in the process of feature extraction, which enhances the MR data.
  • the rationality and accuracy of the feature information improves the positioning accuracy.
  • the determining the characteristic information of the MR data according to the MR data that is temporally adjacent to the MR data further includes:
  • the MR data and the network engineering parameter information corresponding to the MR data extract the second feature information set of the MR data, and the network engineering parameter information corresponding to the MR data is used to indicate the terminal equipment that reports the MR data Attribute information of the cell to which it belongs;
  • the using the first feature information set as the feature information of the MR data includes:
  • the fusion feature information of the MR data is used as the feature information of the MR data.
  • the positioning model is specifically based on a machine learning method, taking the characteristic information of each historical position MR data in the historical position MR data set as input, and taking the historical position MR data
  • the location label of is used as a model obtained by output training; the feature information of the historical location MR data is obtained by merging, splitting, and filtering the historical location MR data in the historical location MR data set.
  • the location label of the historical location MR data is obtained by rasterizing a preset area, and the preset area includes: an area determined by the location of the site to which the terminal device reporting the historical location MR data belongs.
  • the time relationship between the historical position MR data is considered, and the dimension of the characteristic information of each historical position MR data is enriched, so that the positioning model can be more effective
  • the relationship between the characteristic information of the MR data and the position information of the terminal equipment corresponding to the MR data is depicted, thereby improving the subsequent positioning accuracy.
  • the positioning model is specifically based on the feature information obtained based on the time sequence relationship between all historical position MR data in the historical position MR data set and the position information corresponding to the historical position MR data
  • the time sequence relationship is the time sequence corresponding to the time stamp information of the historical position MR data.
  • the characteristic information of each historical position MR data is enriched based on the time series relationship between the historical position MR data in the corresponding historical position MR call sequence and other historical position MR data.
  • the dimension of, so that the positioning model can more effectively describe the relationship between the characteristic information of the MR data and the position information of the terminal device corresponding to the MR data, which also improves the subsequent positioning accuracy.
  • the second aspect of the present application provides a positioning method, including: obtaining a measurement report MR data set, the MR data set including: multiple pieces of MR data; for any piece of MR data in the MR data set, according to the MR data Determine characteristic information of the MR data with at least one piece of MR data that is temporally adjacent to the MR data, where the characteristic information of the MR data is information used to describe the location of the terminal device corresponding to the MR data; The characteristic information of the MR data, the target fingerprint file matching the characteristic information is determined from the fingerprint characteristic database, and the fingerprint characteristic database stores the fingerprint file used to describe the association relationship between the characteristic information and the location of the terminal device; The target fingerprint file and the characteristic information determine the location information of the terminal device corresponding to the MR data.
  • the time dimension is introduced in the feature extraction stage, and the time sequence relationship between the MR data in the corresponding MR call sequence and other MR data is taken into consideration, which enriches the dimension of feature information and improves the positioning accuracy.
  • the fingerprint feature library includes: a fingerprint file corresponding to the current area where each piece of historical position MR data in the historical position MR data set is located, and the fingerprint file is used to describe the current area
  • the characteristic information of each grid and the position information of the grid; the characteristic information of the historical position MR data is processed by merging, splitting, and filtering the historical position MR data in the historical position MR data set Extracted.
  • the time relationship between the historical position MR data is considered, and the dimension of the feature information of each historical position MR data is enriched, so that the fingerprint feature library obtained can be more
  • the relationship between the characteristic information of the MR data and the position information of the terminal equipment corresponding to the MR data is effectively depicted, thereby improving the subsequent positioning accuracy.
  • a third aspect of the present application provides a positioning device, including: an acquisition module, a processing module, and a positioning module;
  • the acquiring module is configured to acquire a measurement report MR data set, and the MR data set includes: multiple pieces of MR data;
  • the processing module is configured to determine characteristic information of the MR data according to the MR data and at least one MR data adjacent to the MR data in time for any piece of MR data in the MR data set ,
  • the characteristic information of the MR data is information used to describe the location of the terminal device corresponding to the MR data;
  • the positioning module is used to input the characteristic information of the MR data into a positioning model to obtain the position information of the terminal device corresponding to the MR data.
  • the positioning model is based on the historical position MR data set of each historical position MR
  • the characteristic information of the data is input, and the position information of the terminal device corresponding to the historical position MR data is used as the output trained model, and the historical position MR data refers to historical MR data carrying position information.
  • the processing module is further configured to determine the characteristic information of the MR data based on the MR data and at least one piece of MR data adjacent to the MR data in time Previously, according to the user identification information and time stamp information carried by each MR data in the MR data set, all MR data in the MR data set were sorted to obtain a sorted MR data set.
  • the MR data with the same user identification in the processed MR data set are arranged together according to the time stamp information, and the MR data set after the sorting process has the same user identification information and the time difference is less than the preset first Multiple pieces of MR data of duration are merged to obtain a merged MR data set.
  • the processing module is further used for sorting and processing multiple pieces of the MR data set that have the same user identification information and the time difference is less than the preset first duration Merge the MR data to obtain the merged MR data set, split the merged MR data set to obtain multiple MR data subsets;
  • each MR data subset includes: multiple pieces of MR data sorted by timestamp information and with the same user ID, the total duration of all MR data in each MR data subset is less than the preset second duration, and each The time difference corresponding to the time stamp information of two adjacent MR data in the MR data subset is less than the preset third time length, and the total time length is the last MR data and the first MR data in the MR data subset The time difference corresponding to the timestamp information.
  • the processing module is further configured to perform split processing on the combined MR data set to obtain multiple MR data sub-sets, and perform processing on all MR data in each MR data sub-set Filter processing to obtain a processed MR data set.
  • the processing module has at least one piece of MR data temporally located before the MR data, the MR data, and at least temporally located after the MR data.
  • a piece of MR data forms the MR call sequence corresponding to the MR data.
  • feature extraction is performed on the MR data to obtain the first feature information set of the MR data, and The first feature information set is used as feature information of the MR data.
  • the processing module further has a second feature information set for extracting the MR data according to the MR data and the network engineering parameter information corresponding to the MR data, and the first feature information Fusing the set and the second feature information set to obtain the fusion feature information of the MR data, and use the fusion feature information of the MR data as the feature information of the MR data;
  • the network parameter information corresponding to the MR data is used to indicate the attribute information of the cell to which the terminal device reporting the MR data belongs.
  • the positioning model is specifically based on the machine learning device taking the characteristic information of each historical position MR data in the historical position MR data set as input, and taking the historical position MR data
  • the location label of is used as a model obtained by output training; the feature information of the historical location MR data is obtained by merging, splitting, and filtering the historical location MR data in the historical location MR data set.
  • the location label of the historical location MR data is obtained by rasterizing a preset area, and the preset area includes: an area determined by the location of the site to which the terminal device reporting the historical location MR data belongs.
  • the positioning model is specifically based on feature information obtained based on the time series relationship between all historical position MR data in the historical position MR data set and the position information corresponding to the historical position MR data
  • the time sequence relationship is the time sequence corresponding to the time stamp information of the historical position MR data.
  • a fourth aspect of the present application provides a positioning device, including: an acquisition module, a processing module, and a positioning module;
  • the acquiring module is configured to acquire a measurement report MR data set, and the MR data set includes: multiple pieces of MR data;
  • the processing module is configured to determine characteristic information of the MR data according to the MR data and at least one MR data adjacent to the MR data in time for any piece of MR data in the MR data set , And according to the characteristic information of the MR data, a target fingerprint file matching the characteristic information is determined from the fingerprint characteristic database, and the characteristic information of the MR data is information used to describe the location of the terminal device corresponding to the MR data,
  • the fingerprint feature database stores fingerprint files for describing the association relationship between feature information and the location of the terminal device;
  • the positioning module is configured to determine the location information of the terminal device corresponding to the MR data based on the target fingerprint file and the characteristic information.
  • the fingerprint feature library includes: a fingerprint file corresponding to the current area where each piece of historical position MR data in the historical position MR data set is located, and the fingerprint file is used to describe the current area
  • the characteristic information of each grid and the position information of the grid; the characteristic information of the historical position MR data is processed by merging, splitting, and filtering the historical position MR data in the historical position MR data set Extracted.
  • a fifth aspect of the present application provides a positioning device, including a processor, a memory, and a computer program that is stored on the memory and can run on the processor.
  • the processor executes the program, the above-mentioned first aspect and The methods described in the first aspect of various possible designs.
  • a sixth aspect of the present application provides a positioning device, including: a processor and a memory, the memory is used to store computer program code, and the processor is used to call the computer program code to execute the second aspect and the second aspect described above The methods described in the various possible designs.
  • the seventh aspect of the present application provides a storage medium that stores instructions in the storage medium, which when run on a computer, causes the computer to execute the methods described in the first aspect and various possible designs of the first aspect .
  • the eighth aspect of the present application provides a storage medium that stores instructions in the storage medium, which when run on a computer, causes the computer to execute the methods described in the above second aspect and various possible designs of the second aspect .
  • the ninth aspect of the present application provides a program product containing instructions, which when run on a computer, causes the computer to execute the methods described in the first aspect and various possible designs of the first aspect.
  • the tenth aspect of the present application provides a program product containing instructions, which when run on a computer, causes the computer to execute the methods described in the second aspect and various possible designs of the second aspect.
  • the eleventh aspect of the present application provides a chip that includes a memory and a processor.
  • the memory stores code and data.
  • the memory is coupled to the processor.
  • the processor runs the code in the memory so that the chip is used to execute the first Aspects and the methods described in the various possible designs of the first aspect.
  • a twelfth aspect of the present application provides a chip.
  • the chip includes a memory and a processor.
  • the memory stores code and data.
  • the memory is coupled to the processor.
  • the processor runs the code in the memory so that the chip is used to execute the second Aspect and the methods described in the various possible designs of the second aspect.
  • a thirteenth aspect of this application provides a communication system, including: positioning equipment and training equipment;
  • the positioning device is the device described in the third aspect and various possible designs of the third aspect
  • the training device is a device for training the positioning model in the third aspect and various possible designs of the third aspect.
  • the positioning method, device, and storage medium obtained by the embodiments of the present application obtain an MR data set including multiple pieces of MR data.
  • the MR data is temporally related to the MR data.
  • the adjacent MR data determines the feature information used to describe the location of the terminal device corresponding to the MR data, and finally inputs the feature information of the MR data into the positioning model to obtain the location information of the terminal device corresponding to the MR data. It takes the characteristic information of each historical position MR data in the historical position MR data set as the input, takes the position information of the terminal equipment corresponding to the historical position MR data as the output training model, and considers the MR when determining the characteristic information of the MR data.
  • the influence of the time sequence relationship of the data on the geographical location of the terminal device improves the positioning accuracy and solves the problem of low positioning accuracy of the terminal device in the prior art.
  • FIG. 1 is a schematic structural diagram of a positioning system provided by an embodiment of this application.
  • Embodiment 1 of a positioning method provided by an embodiment of this application;
  • FIG. 3 is a schematic flowchart of Embodiment 2 of a positioning method provided by an embodiment of this application;
  • FIG. 5 is a schematic flowchart of the implementation manner (1) of the positioning model in the positioning method provided by the embodiment of the application;
  • FIG. 6 is a schematic flowchart of the implementation manner (2) of the positioning model in the positioning method provided by the embodiment of the application;
  • FIG. 7 is a schematic flowchart of Embodiment 4 of a positioning method provided by an embodiment of this application.
  • FIG. 8 is a schematic diagram of a method for generating a fingerprint feature database in a positioning method provided by an embodiment of the application.
  • FIG. 9 is a schematic flowchart of a training method provided by an embodiment of the application.
  • FIG. 10 is a schematic structural diagram of Embodiment 1 of a positioning device according to an embodiment of this application.
  • FIG. 11 is a schematic flowchart of Embodiment 2 of a positioning device provided by an embodiment of this application;
  • FIG. 12 is a schematic structural diagram of Embodiment 3 of a positioning device provided by an embodiment of this application.
  • FIG. 13 is a schematic structural diagram of a training device provided by an embodiment of the application.
  • FIG. 14 is a schematic structural diagram of Embodiment 4 of a positioning device provided by an embodiment of this application.
  • 15 is a schematic structural diagram of another training device provided by an embodiment of the application.
  • FIG. 16 is a schematic structural diagram of an embodiment of a positioning system provided by an embodiment of this application.
  • the network engineering parameters are engineering parameters used to describe the attribute information of the station (base station) in the wireless network planning.
  • the network engineering parameters may include: the latitude and longitude of the station antenna position, antenna directivity, gain, and azimuth , Downtilt angle, hanging height, feeder model, site type (indoor, outdoor), level value of each cell (for example, reference signal receiving power (RSRP)), etc.
  • RSRP reference signal receiving power
  • a user refers to a person who carries a terminal (Terminal, that is, a communication terminal, including but not limited to a mobile phone, etc.) that can communicate through a wireless network.
  • Terminal that is, a communication terminal, including but not limited to a mobile phone, etc.
  • the fingerprint of a certain geographic location can be interpreted as the signal level value and other information contained in the measurement report corresponding to the geographic location.
  • the fingerprint database is a reference database for fingerprint matching and positioning.
  • the "positioning model” is also called the “model”, which can receive input data and generate a prediction output based on the received input data and current model parameters.
  • the positioning model may be a regression model, artificial neural network (ANN), deep neural network (DNN), support vector machine (SVM), or other machine learning models.
  • Measurement report (MR) data refers to data sent every preset time on the service channel, and can include: timestamp information, user identification information, latitude and longitude (optional), delay information, cell identification information, and signal Level value information, interference information, etc.; in the embodiment of this application, MR data mainly refers to the information that the terminal device feeds back to the network device side, and it carries the signal level value and the cell identifier received by the terminal device from the serving cell and neighboring cells. Information, user identification information, time stamp information, etc.
  • the MR data set composed of the multiple pieces of MR data is used to locate the position of the terminal device corresponding to any piece of MR data in the MR data set.
  • Historical location MR data refers to the MR data that carries the location information of the terminal device.
  • the historical location MR data is mainly through assisted global positioning system (AGPS), minimization of drive tests (MDT), Data obtained through drive test (DT) and other methods.
  • AGPS assisted global positioning system
  • MDT minimization of drive tests
  • DT Drive test
  • the historical position MR data set composed of the multiple pieces of historical position MR data is used to train a positioning model with the function of locating the position of the terminal device.
  • the position of the terminal device corresponding to any piece of MR data in the aforementioned MR data set is obtained based on the positioning of the model obtained by training the historical position MR data set.
  • multiple MR data subsets are obtained by performing data processing on the MR data set.
  • Each MR data subset may include multiple samples, and each sample may include user identification information and time stamp information.
  • preprocessing may include one or more of data sorting, data merging, data splitting, data filtering, and the like. For example, sort all the MR data in the MR data set according to the time stamp information of each MR data, merge multiple MR data with the same user identification information and the time difference less than the preset first duration, etc. .
  • feature information of MR data in the embodiments of the present application is used to describe the attribute information of the cell where the terminal device reporting the MR data is located, and can characterize the location information when the terminal device reports the MR data.
  • the characteristic information of the MR data may include: signal level values received by each cell, cell identification information, cell type information, cell location information, etc. These characteristic information may be used to describe the location information of the terminal device.
  • Fig. 1 is a schematic structural diagram of a positioning system provided by an embodiment of the application.
  • the positioning system may include: a training device 11, a positioning device 12, a network device 13, at least one terminal device 14, and a data storage device 15.
  • at least one terminal device 14 may report MR data to the network device 13 periodically or triggered by an event after accessing the wireless network, and the network device 13 may report the received MR data
  • the data is stored in the data storage device 15.
  • the above-mentioned MR data may be historical location MR data that carries location information, or MR data that does not carry location information.
  • the historical location MR data carrying location information can be obtained by at least one terminal device 14 in specific application scenarios such as network assessment, network planning and optimization, virtual drive testing, etc., through key road monitoring, user behavior identification, value area identification, etc.
  • the MR data that does not carry location information may be the MR data reported by the terminal device in daily applications.
  • the data storage device 15 can store a large amount of historical position MR data used to train the positioning model and MR data that does not carry terminal position information, and the training device 11 is used to set MR data based on the historical position in the data storage device 15.
  • the program code of the model training method is executed to train the positioning model; the positioning device 12 is used to execute the program code of the data processing method based on the MR data set in the data storage device 15 that does not carry terminal location information, using the processed MR data set and training
  • the obtained positioning model obtains the position information of the terminal device.
  • the data storage device 15 in this embodiment may be a database for storing data, it may be an independent device, or it may be integrated in a data platform.
  • the training device 11 can send the trained positioning model to the positioning device 12.
  • the positioning device 12 executes the method of locating the location of the terminal device based on the MR data set and the positioning model.
  • the specific positioning method please refer to the relevant in the following embodiments Description, not repeat them here.
  • the positioning device 12 is implemented by one or more servers. Optionally, it cooperates with other computing devices, such as data storage, routers, load balancers, etc.; the positioning device 12 can be arranged on a physical site , Or distributed on multiple physical sites.
  • the positioning device 12 can use the data stored in the data storage device 15 or call the program code in the data storage device 15 to implement the positioning method described in the embodiment of the present application.
  • the positioning device 12 may For any MR data, determine the characteristic information of the MR data according to the MR data adjacent to the MR data in time, and then input the characteristic information of the MR data into the positioning model to obtain the terminal equipment corresponding to the MR data Location information.
  • Fig. 1 is only a schematic diagram of a system architecture provided by an embodiment of the present application, and the positional relationship between the devices, devices, modules, etc. shown in Fig. 1 does not constitute any limitation.
  • the data storage device 15 may be an external memory relative to the positioning device 12, and in other cases, the data storage device 15 may also be placed in the positioning device 12.
  • the training device 11 and the positioning device 12 in the embodiment of the present application may be the same device or different devices.
  • the training device 11 and/or the positioning device 12 can be a terminal, such as a mobile phone terminal, a tablet computer, a notebook computer, AR/VR, a vehicle-mounted terminal, etc., or a network device 13, such as a wireless access device, a core network device, or It may be a server or a virtual machine, etc., and may also be a distributed computer system composed of one or more servers and/or computers, etc., which is not limited in the embodiment of the present application.
  • the aforementioned terminal device 14 may also be referred to as user equipment (UE), mobile station (mobile station, MS), mobile terminal (mobile terminal), terminal (terminal), etc., for example, terminal
  • UE user equipment
  • MS mobile station
  • mobile terminal mobile terminal
  • terminal terminal
  • the device 14 may be a smart phone, a tablet computer, a personal computer, a desktop computer, an on-board unit (OBU), a virtual reality device, an artificial intelligence device (such as a robot, etc.), or a smart wearable device, etc., and this embodiment does not limited.
  • OBU on-board unit
  • the network device 13 may include various forms of macro base stations, micro base stations (also referred to as small stations), relay stations, access points, and so on.
  • the network device 13 may be a base transceiver station (BTS) in GSM or CDMA, a base station (nodeB, NB) in WCDMA, or an evolved base station (evolutional node B, eNB) in LTE. Or e-NodeB), and may be the corresponding device gNB in the 5G network, the base station in the future communication system, or the access node in the WiFi system.
  • BTS base transceiver station
  • NB base station
  • evolutional node B evolutional node B
  • e-NodeB evolved base station
  • the embodiment of the present application does not limit the specific technology and specific device form adopted by the network device 13.
  • the positioning system shown in Figure 1 may be a communication system, which may be a global system of mobile communication (GSM) system, code division multiple access (CDMA) system, Wideband code division multiple access (WCDMA) system, general packet radio service (GPRS), long term evolution (LTE) system, advanced long term evolution (LTE advanced, LTE- A), LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), universal mobile telecommunication system (UMTS), and other applications of orthogonal frequency division duplex
  • GSM global system of mobile communication
  • CDMA code division multiple access
  • WCDMA Wideband code division multiple access
  • GPRS general packet radio service
  • LTE long term evolution
  • LTE advanced, LTE- A advanced long term evolution
  • TDD LTE time division duplex
  • UMTS universal mobile telecommunication system
  • OFDM orthogonal frequency division duplex
  • OFDM orthogonal frequency division multiplexing
  • NR new radio
  • the product implementation form of this application is a program code contained in machine learning and deep learning platform software and deployed on a server (or a computing cloud or a mobile terminal and other hardware with computing capabilities).
  • the program code of this application can be stored in the positioning device and the training device. When running, the program code runs on the host memory and/or GPU memory of the server.
  • the positioning device can determine the location of the terminal device based on the MR reported by the terminal device, and then recommend the user's services around the location based on the location, such as food, entertainment and other product recommendations . Therefore, accurate positioning of the terminal equipment held by the user is the key to targeted product recommendation.
  • a terminal device when a terminal device accesses a wireless network, it can periodically or event trigger report measurement report MR data. Because MR data contains the wireless environment information of the terminal device at a certain moment, it has a corresponding relationship with the current geographic location of the terminal device. , Locating users based on measurement reports is a common method for locating terminal devices at this stage.
  • the MR-based positioning method can use fingerprints for positioning.
  • the information contained in the measurement report corresponding to a certain geographic location is called the “fingerprint” of the geographic location.
  • a fingerprint database is established in advance based on the fingerprint.
  • the application embodiment provides a positioning method, device, and storage medium.
  • the MR data is compared with the MR data in time.
  • the adjacent MR data determines the characteristic information used to describe the location of the terminal equipment corresponding to the MR data, and finally inputs the characteristic information of the MR data into the positioning model to obtain the position information of the terminal equipment corresponding to the MR data.
  • the model takes the characteristic information of each historical position MR data in the historical position MR data set as the input, takes the position information of the terminal equipment corresponding to the historical position MR data as the output training model, and considers the characteristic information of the MR data
  • the time sequence of MR data affects the geographic location of the terminal device, which improves the positioning accuracy.
  • “multiple” refers to two or more.
  • “And/or” describes the association relationship of the associated objects, indicating that there can be three types of relationships, for example, A and/or B, which can mean: A alone exists, A and B exist at the same time, and B exists alone.
  • the character “/” generally indicates that the associated objects are in an "or” relationship.
  • FIG. 2 is a schematic flowchart of Embodiment 1 of a positioning method provided by an embodiment of this application.
  • the positioning method may be executed by a positioning device, or may be executed by a processor in the positioning device.
  • a positioning device executes the method for description.
  • the positioning method may include the following steps:
  • Step 21 Obtain a measurement report MR data set.
  • the MR data set includes multiple pieces of MR data.
  • the positioning device can execute the positioning method under the trigger of an external application. Specifically, the positioning device first obtains the MR data set of at least one terminal device. Specifically, the positioning device may directly obtain the MR data set reported by the terminal device from the data storage device, or may obtain the MR data set of the terminal device from the network device, and different MR data sets contain different numbers of MR data.
  • the embodiment of the present application does not limit the acquisition method of the MR data set and the number of MR data included in each MR data set, which can be determined according to actual conditions.
  • the MR data set may include multiple pieces of MR data, and each piece of MR data carries multi-dimensional features such as user identification information, time stamp information, and cell information.
  • the MR data in the MR data set may come from the same terminal device or multiple terminal devices. The embodiment of the present application does not limit the specific source of each piece of MR data in the MR data set.
  • MR data refers to MR data that does not carry location information.
  • historical position MR data refers to historical MR data carrying position information.
  • Step 22 For any piece of MR data in the MR data set, determine the characteristic information of the MR data according to the MR data and at least one piece of MR data adjacent to the MR data in time, and the characteristic information of the MR data is Information used to describe the location of the terminal equipment corresponding to the MR data.
  • the MR data set may be a data set of the original MR data set obtained by the positioning device after data preprocessing.
  • the preprocessing of the original MR data set may include one or more of data sorting, merging, splitting, and filtering.
  • the positioning device may perform the same processing on each piece of MR data in the MR data set. Therefore, the embodiment of the present application may take any piece of MR data in the MR data set as an example.
  • the positioning device introduces timing information between the MR data during the positioning process, that is, according to the MR data and the MR data adjacent to the MR data in time Data to extract the characteristic information of the MR data.
  • the characteristic information of the MR data may be used to characterize the location of the terminal device corresponding to the MR data.
  • the characteristic information of the MR data may include, but is not limited to, cell level information and cell identification. Information, cell location information, cell site type and other information.
  • the positioning device may process the MR data based on the chronological order of the MR data in the MR data set to deeply dig the relationship between the feature information of the MR data and the terminal position information, thereby improving the positioning accuracy.
  • Step 23 Input the characteristic information of the MR data into the positioning model to obtain the position information of the terminal device corresponding to the MR data.
  • the positioning model takes the characteristic information of each historical position MR data in the historical position MR data set as input. Take the position information of the terminal equipment corresponding to the historical position MR data as the output training model.
  • the historical position MR data refers to historical MR data carrying position information.
  • using the positioning model to predict the location information of the terminal device is the goal of the embodiment of this application.
  • the positioning device uses the feature information of the MR data obtained above and the positioning model trained by the training device to predict the MR data corresponding to the terminal device.
  • the location specifically, the positioning device uses the feature information of the MR data obtained above as the input of the positioning model, and uses the positioning model to locate the MR data, thereby predicting the location information of the terminal device corresponding to the MR data.
  • the predicted MR data corresponds to the location information of the terminal device, that is, the current longitude and latitude information of the user carrying the terminal device.
  • the positioning device can also filter and smooth the positioning results after obtaining the positioning results, so as to obtain the final positioning results and provide the required location services for external applications.
  • the filtering processing method in this step includes, but is not limited to, various filtering methods such as mean filtering and Kalman filtering.
  • the positioning model may be obtained by the training device using the characteristic information of each piece of historical position MR data in the historical position MR data set and the position information of the terminal device corresponding to the historical position MR data.
  • the training device may perform, for example, the processing in step 22 on the acquired historical position MR data set, so as to obtain the characteristic information of each historical position MR data, and then use the characteristic information of each historical position MR data as the training network Use the terminal location information carried in each historical location MR data as the output of the training network to train the positioning model.
  • the training network may be a deep neural network, for example, a recurrent neural network (RNN), a long short-term memory network (long short-term memory, LSTM), etc.
  • RNN recurrent neural network
  • LSTM long short-term memory
  • the embodiments of this application are not used for training.
  • the network of the positioning model is defined.
  • the positioning method provided by the embodiment of the present application obtains an MR data set including multiple pieces of MR data. For any piece of MR data in the MR data set, according to the MR data and at least one piece adjacent to the MR data in time MR data, determine the characteristic information used to describe the location of the terminal equipment corresponding to the MR data, and input the characteristic information of the MR data into the positioning model to obtain the position information of the terminal equipment corresponding to the MR data.
  • the positioning model is based on the historical position
  • the characteristic information of each historical position MR data in the MR data set is used as input, and the position information of the terminal device corresponding to the historical position MR data is used as the output trained model.
  • the positioning device processes the MR data based on at least one piece of MR data that has a time sequence relationship with the MR data during the positioning process, and the characteristic information of the obtained MR data can more accurately characterize the position information of the terminal device. Thereby improving the positioning accuracy.
  • FIG. 3 is a schematic flowchart of Embodiment 2 of the positioning method provided in an embodiment of this application. As shown in FIG. 3, in this embodiment, before the above step 22, the method may further include some or all of the following steps:
  • Step 31 According to the user identification information and time stamp information carried in each MR data in the MR data set, sort all the MR data in the MR data set to obtain a sorted MR data set.
  • the MR data with the same user identifier in the sorted MR data set are arranged together according to the time stamp information.
  • each piece of MR data since each piece of MR data carries user identification information and time stamp information, when the positioning device processes the acquired MR data set, it may first be based on the user identification information carried in each piece of MR data Divide the MR data with the same user identification information together, and then sort the MR data of the same user according to the timestamp information, so that the MR data with the same user identification are arranged together according to the timestamp information, and then get sorted MR data set, so that the positioning device can perform feature extraction on the MR data in the sorted MR data set, and the time sequence of each MR data report in the MR data set is taken into account during feature extraction, which is the subsequent extraction Lay the foundation for reasonable and accurate feature information.
  • the movement track of the terminal device can be obtained, which further improves the positioning accuracy.
  • Step 32 Combine multiple pieces of MR data with the same user identification information and a time difference less than the preset first duration in the sorted MR data set to obtain a combined MR data set.
  • the preset first duration generally refers to a relatively short period of time, such as 1 s or 2 s, etc.
  • the embodiment of the present application does not limit the specific value of the first duration.
  • the MR data set after the sorting process is merged with the same time or the time difference is less than 1s to obtain the merged MR data set.
  • the timestamp information is determined as the average value of the timestamp information of the multiple pieces of MR data participating in the merge; if multiple pieces of MR data are participating in the merge If the primary serving cell of the corresponding terminal equipment is the same, the primary serving cell remains unchanged.
  • the signal level value of the primary serving cell is the average of the signal level values of all primary serving cells, and its corresponding neighboring cells are deduplicated and combined; if participating The combined multiple pieces of MR data correspond to different primary serving cells of the terminal equipment, and the cell with the largest signal level value is used as the primary serving cell, and all the remaining cells undergo adjacent cell deduplication and combination.
  • the neighboring cells are deduplicated and combined, the signal level values of the same cell are averaged, and the neighboring cells are re-determined according to the signal level value.
  • the feature information at each time or time period can be made more complete, thereby improving the accuracy of the feature information corresponding to the MR data.
  • the method may further include the following steps:
  • Step 33 Split the combined MR data set to obtain multiple MR data sub-sets.
  • Each MR data sub-set includes: multiple pieces of MR data sorted by timestamp information and with the same user ID, each The total duration of all MR data in each MR data subset is less than the preset second duration, and the time difference corresponding to the timestamp information of two adjacent MR data in each MR data subset is less than the preset third duration .
  • the total duration is the time difference between the last piece of MR data in the MR data subset and the time stamp information of the first piece of MR data.
  • the time dimension corresponding to the set is relatively long. Therefore, in order to improve the feature information when extracting features from MR data
  • the extraction efficiency and accuracy of the MR data set can be split according to the preset call extraction strategy to obtain multiple MR data subsets.
  • the call extraction strategy is: all MR data in each MR data subset have the same user identification, and the total duration of all MR data in each MR data subset is less than the preset second duration, and each The time difference corresponding to the time stamp information of two adjacent MR data in the MR data subset is less than the preset third time length.
  • the MR data subset may also be referred to as a call.
  • the call refers to a collection of multiple pieces of MR data that have the same user identification information and meet a preset time constraint rule.
  • the total duration of all MR data in each MR data subset does not exceed 180s, and the time interval between two adjacent MR data does not exceed 30s. It is worth noting that the second duration and the third duration in this embodiment are both preset values, which can be determined according to actual conditions, and the embodiments of the present application do not limit them.
  • the method may further include the following steps:
  • Step 34 Perform filtering processing on all MR data in each MR data subset to obtain a processed MR data set.
  • the signal level value received by the terminal equipment from each cell may have great fluctuations.
  • a stable signal level can be obtained.
  • the average value is obtained by filtering all MR data in each MR data subset to obtain the processed MR data set, thereby providing positioning accuracy during subsequent positioning.
  • this embodiment may use multiple filtering methods to perform filtering processing on the MR data in each of the foregoing MR data subsets, for example, common filtering methods such as weighted filtering, Kalman filtering, and synovial average filtering.
  • common filtering methods such as weighted filtering, Kalman filtering, and synovial average filtering.
  • the embodiment of the present application does not limit the specific filtering method, which may be limited according to actual conditions.
  • processing steps for the MR data set in the embodiment of the present application may include one or more of steps 31 to 34, and the specific steps included may be determined according to actual conditions, and will not be repeated here.
  • the positioning method provided in the embodiment of the present application sorts all the MR data in the MR data set according to the user identification information and time stamp information carried by each MR data in the MR data set to obtain the sorted MR data set , So that the MR data with the same user ID in the sorted MR data set are arranged together according to the time stamp information, and then the sorted MR data set with the same user ID information and the time difference is less than the preset first One-time-long multiple pieces of MR data are merged to obtain a merged MR data set, and then the merged MR data set is split to obtain multiple MR data subsets, and finally each MR data subset is Filter processing is performed on all MR data of MR data to obtain a processed MR data set.
  • the accuracy of the subsequently extracted feature information can be improved by sorting, merging, splitting, and filtering the MR data set.
  • FIG. 4 is a schematic flowchart of Embodiment 3 of the positioning method provided in an embodiment of this application.
  • the above step 22 can be implemented through the following steps:
  • Step 41 Form an MR call sequence corresponding to the MR data according to at least one piece of MR data temporally before the MR data, the MR data, and at least one MR data temporally after the MR data.
  • the convolutional neural network can convolve data of the same nature to extract the correlation between the data, when the positioning device extracts the characteristic information of the MR data, after introducing the time dimension, the time At least one piece of MR data before the MR data, the MR data, and at least one piece of MR data temporally after the MR data form an MR call sequence corresponding to the MR data, and then feature information of the MR data is extracted based on the MR call sequence.
  • the first M MRs and the last N MRs within a certain time window of the current MR are combined to form the MR call sequence corresponding to the MR data, where M and N are both positive integers.
  • Step 42 Perform feature extraction on the MR data based on the convolutional neural network and the MR call sequence to obtain the first feature information set of the MR data.
  • the information contained in the MR data is roughly divided into two types of data, signal level value and cell information, it is possible to construct feature maps separately based on different indicators, such as signal level value and cell information. Extract features.
  • one piece of MR data when constructing a feature map, one piece of MR data generates an n*1 vector, where n represents the length of the vector, which can be represented by n features of a single piece of MR data.
  • n represents the length of the vector, which can be represented by n features of a single piece of MR data.
  • an MR call sequence including t pieces of MR data it can form an n*t vector as a feature map.
  • the indicators used include, but are not limited to, cell identification vector, signal level value, Fourier transform of level value, combination of cell and level value, etc.
  • the cell identity vector refers to the cell vector obtained by encoding the cell identity in the MR data of the current area (similar to the word2vec model).
  • the index used by each piece of MR data can be determined according to the actual situation, which is not limited in this embodiment.
  • the feature map can be constructed based on the call sequence first, and the convolutional neural network can be used to automatically mine and extract the level, cell, and other manifestations of MR.
  • the characteristic information of the data, the characteristic information of the MR can be obtained after combining the characteristics of different indicators.
  • connection layer can also be implemented by convolution.
  • the first feature information set may include dimensional features related to signal level values and cell information, including but not limited to signal level values of cells, cell identifiers, and the like.
  • Step 43 Use the first feature information set as feature information of the MR data.
  • the time sequence relationship between the MR data is taken into consideration during the feature extraction process to enhance The rationality and accuracy of the feature information of MR data is improved, and the positioning accuracy is improved.
  • the above step 22 may further include the following step 40.
  • this step 40 may be located before or after the above step 41.
  • the embodiment shown in FIG. 4 takes step 40 after step 42 for example:
  • Step 40 Extract the second feature information set of the MR data according to the MR data and the network parameter information corresponding to the MR data.
  • the network parameter information corresponding to the MR data is used to indicate the cell to which the terminal device reporting the MR data belongs Attribute information.
  • network engineering parameter information can be used to describe site (base station) attribute information
  • its corresponding network engineering parameters can include: latitude and longitude of the site antenna position, antenna directivity, gain, azimuth, Downtilt, hanging height, feeder model, site type (indoor, outdoor), etc. Therefore, in this embodiment, for the MR data, first determine the signal level value of the primary serving cell of the terminal device corresponding to the MR data, The identification information of the primary serving cell, the signal level value of each neighboring cell, and the identification information of each neighboring cell, etc., and then the signal level value of all cells is calculated to determine the cell that characterizes the cell where the MR data corresponds to the terminal device.
  • Level value and cell identification and then query the network parameters information corresponding to the MR data based on the cell identification, determine the longitude, latitude, height, direction angle, downtilt and other information of the cell, and then determine the information used to characterize the MR data
  • the second feature information collection The second feature information collection.
  • first feature information set and the second feature information set in the embodiment of the present application represent different feature information sets obtained in two ways, and do not indicate a sequence.
  • step 40 can also be performed after any one or more of all the steps included in the embodiment shown in FIG. 3, and when step 40 is performed directly after step 31, step 32, or step 34,
  • the second feature information set of the extracted MR data is the feature information of a single piece of MR data.
  • the method may further include:
  • feature enhancement is performed on the feature information in the second feature representation set.
  • the extracted features in the second feature information set may be feature-enhanced based on the SdA model.
  • the SDA model is first used to train the features, and then the feature transformation enhancement model is obtained through learning, and then the features in the second feature information set are processed by the feature transformation enhancement model to obtain the enhanced second feature information set.
  • the training process of the feature conversion enhancement model can be explained as follows: For the feature X in the second feature information set, that is, the feature X of a single piece of MR data is first processed by the encoder (Encoder) and decoder (Decoder) to obtain the feature X' , Train the parameters of the current network layer by layer through unsupervised learning, so that the error between X and X'is minimized; when the network training formed by stacking all autoencoders is completed, use the previous autoencoder in a multilayer neural network The middle layer then fine-tunes the network weights of the multilayer neural network in a supervised learning manner to obtain the feature transformation enhancement model.
  • step 43 can be replaced with the following steps 431 and 432:
  • Step 431 Perform fusion on the first feature information set and the second feature information set to obtain the fusion feature information of the MR data.
  • the feature information in the first feature information set and the second feature information set can be fused.
  • the feature fusion operation can combine the feature information in the two sets Features are directly connected to complete, or can be converted through a fully connected layer.
  • a fully connected layer and a hidden layer based on a neural network connect the first feature information set and the second feature information set, thereby expanding the dimension of the feature information of the MR data, which improves the subsequent determination of the corresponding MR data The accuracy of the location of the terminal device.
  • Step 432 Use the fusion feature information of the MR data as the feature information of the MR data.
  • the rationality and accuracy of the feature information of the MR data are further enhanced, and the positioning accuracy is improved.
  • the first feature information set and the enhanced first feature information set are Two feature information sets are fused to obtain updated fused feature information.
  • the updated fusion feature information can be used as the feature information of the MR data.
  • the positioning method provided in the embodiment of the present application forms an MR call sequence corresponding to the MR data according to at least one piece of MR data temporally before the MR data, the MR data, and at least one MR data temporally after the MR data, based on
  • the convolutional neural network and the MR call sequence perform feature extraction on the MR data to obtain the first feature information set of the MR data, and extract the second feature information of the MR data according to the network parameter information corresponding to the MR data
  • the first feature information set and the second feature information set are further merged to obtain the fusion feature information of the MR data, and finally the fusion feature information of the MR data is used as the feature information of the MR data.
  • This technical solution not only considers the feature information of a single piece of MR data, but also considers the feature information extracted based on the MR data corresponding to the MR call sequence, which further enhances the rationality and accuracy of the MR data, and determines accurate terminals for the subsequent The location laid the foundation.
  • constructing and training a positioning model is the core of the positioning method proposed in this application, which can use training equipment and acquired historical position MR data for offline training.
  • the training device may first process the acquired historical position MR data set (for example, one or more of multiple processing such as sorting, merging, splitting, filtering, etc.) to obtain the characteristics of each historical position MR data Information, and then use the characteristic information of each historical location MR data as the input of the model, because the historical location MR data subset obtained by sorting according to the user identification information and the time stamp information (each historical location MR data set can be split to get more A subset of historical position MR data) can form the movement trajectory of the user to which the terminal device belongs.
  • multiple processing such as sorting, merging, splitting, filtering, etc.
  • the positioning model can be constructed based on the movement trajectory, that is, the characteristic information of the aforementioned historical position MR data is used as the input of the model, and the historical position
  • the position information of the terminal equipment corresponding to the MR data is used as the output of the model, and the parameters of the model are trained to obtain the positioning model in the target area corresponding to each historical position MR data.
  • the training device can run on the open source Tensorflow machine learning platform, and extract the feature information of each historical position MR data in the historical position MR data set based on the machine learning method to train the positioning model, specifically , Can run on servers with NVIDIA GPU cards, where NVIDIA GPU cards provide computing acceleration capabilities through the CUDA programming interface to accelerate the feature extraction process and the positioning model construction process.
  • the positioning model can be used to predict the location of the terminal device corresponding to the MR data based on the characteristic information of the processed MR data, which can be obtained by offline training.
  • the training process of the positioning model may include the following two implementation modes:
  • the positioning model is specifically based on a machine learning method that takes the characteristic information of each historical position MR data in the historical position MR data set as input, and uses the position label of the historical position MR data as an output training model;
  • the characteristic information of the historical position MR data is extracted after merging, splitting, and filtering the historical position MR data in the historical position MR data set.
  • the location label of the historical location MR data is obtained by rasterizing a preset area, and the preset area includes: an area determined by the location of the site to which the terminal device reporting the historical location MR data belongs.
  • the preset area may be the primary serving cell where the terminal device reporting the historical location MR data is located, or an area determined by the neighboring cell, and the embodiment of the present application does not limit the specific implementation of the preset area.
  • the embodiment of the present application is based on a machine learning method and uses the time sequence information of the adjacent historical position MR data in the historical position MR data set to extract the feature information during the positioning process, which enriches the dimension of the feature information , Increase the degree of discrimination between MR data in different locations, build a positioning model based on machine learning methods (such as random forest, DNN, etc.), optimize the relationship between features, and improve positioning accuracy.
  • the training method may be referred to as a "time-space feature model positioning method".
  • the positioning model is trained based on the random forest, DNN and other machine learning methods.
  • the better callability of the MR data refers to a high proportion of the number of MR data contained in each MR data subset is not less than 8.
  • the positioning model trained based on the historical position MR data corresponding to the call sequence may be called a time series positioning model
  • the positioning model trained based on the historical position MR data corresponding to the position label may be called a label positioning model . That is, the positioning obtained through this implementation (1) is a label positioning model, and the positioning model obtained through the following implementation (2) is a time series positioning model.
  • FIG. 5 is a schematic flowchart of the implementation manner (1) of the positioning model in the positioning method provided by the embodiment of the application.
  • the training method can be executed by the training device in the positioning system shown in FIG. 1 above.
  • the training method may include the following steps:
  • Step 51 The training device obtains a historical position MR data set, the historical position MR data set includes: multiple pieces of historical position MR data, and the historical position MR data refers to historical MR data carrying position information.
  • Step 52 For any piece of historical position MR data in the historical position MR data set, the training device determines the historical position according to the historical position MR data and at least one historical position MR data adjacent to the historical position MR data in time Characteristic information of location MR data.
  • the training device may merge, split, and filter the historical position MR data in the historical position MR data set, and then extract the characteristic information of each historical position MR data.
  • the merging, splitting, and filtering process of the historical position MR data by the training device is the same as the merging and splitting of the MR data in the MR data set by the positioning device in the embodiment shown in FIG.
  • the classification and filtering processes are similar. For details, please refer to the introduction in the foregoing embodiment, which will not be repeated here.
  • Step 53 The training device rasterizes the preset area corresponding to the historical position MR data to obtain the position label of the historical position MR data.
  • the training device first determines a preset area based on the location of the site to which the terminal device reporting the historical location MR data belongs.
  • the preset area may be the coverage of the main serving cell or the neighboring cell.
  • the training device when it builds and trains a positioning model based on a machine learning method, it can perform classification tasks or regression tasks. Since machine learning methods, such as random forest and DNN models, require location labels, the training device can rasterize the preset area when doing classification tasks, and the grid where the latitude and longitude of the historical location MR data is located The ID is determined as the location label of the piece of historical location MR data; when the training device performs a regression task, the longitude and latitude of the historic location MR is used as the location label of the historical location MR data.
  • machine learning methods such as random forest and DNN models
  • Step 54 The training device takes the characteristic information of each historical position MR data in the historical position MR data set as input, and uses the position label corresponding to the historical position MR data as an output training model to obtain the positioning model.
  • the positioning model has a corresponding model file, and the model file can determine whether the MR data can use the positioning model for position prediction during the positioning process.
  • Step 55 The training device stores the model file corresponding to the positioning model.
  • the training device may save the generated model file in a storage device, for example, a model storage database, for subsequent direct use in positioning.
  • a storage device for example, a model storage database
  • the training device uses the historical position MR data and at least one historical position adjacent to the historical position MR data in time.
  • MR data determine the characteristic information of the historical position MR data, and then rasterize the preset area corresponding to the historical position MR data to obtain the position label of the historical position MR data, and finally combine each historical position in the historical position MR data set
  • the feature information of the position MR data is used as input, and the position label corresponding to the historical position MR data is used as the output training model to obtain the positioning model.
  • this technical solution takes into account the time relationship between the historical position MR data, and enriches the dimension of the characteristic information of each historical position MR data, so that the positioning model can more effectively describe the MR data The relationship between the characteristic information of the MR data and the position information of the terminal device corresponding to the MR data, thereby improving the subsequent positioning accuracy.
  • the positioning model is specifically obtained based on the association relationship between the feature information obtained from the time series relationship between all historical position MR data in the historical position MR data set and the position information corresponding to the historical position MR data, and the time series relationship is The time sequence corresponding to the time stamp information of the historical location MR data.
  • the embodiment of the present application is also based on a machine learning method.
  • the time sequence information of the adjacent historical position MR data in the historical position MR data set is used to construct the positioning model.
  • the position information of the terminal device corresponding to the historical position MR data is not only related to the characteristic information of the historical position MR data, but also related to the historical position MR data before and after the historical position MR data in time.
  • the feature information is related to the location.
  • the training device considers the time series relationship between the historical location MR data to construct a positioning model, which improves the positioning accuracy.
  • the training method may be referred to as the "time series model positioning method" according to the feature information used in the implementation manner.
  • this training method is especially suitable for the poor call continuity of historical location MR data.
  • the time series characteristics of each historical location MR data are not obvious, but the amount of historical location MR data is sufficient to form a call sequence for training Position the scene of the model.
  • FIG. 6 is a schematic flowchart of the implementation manner (2) of the positioning model in the positioning method provided by the embodiment of the application.
  • the training method can be executed by the training device in the positioning system shown in FIG. 1 above.
  • the training method may include the following steps:
  • Step 61 The training device obtains a historical position MR data set, the historical position MR data set includes: multiple pieces of historical position MR data, and the historical position MR data refers to historical MR data carrying position information.
  • Step 62 For any piece of historical position MR data in the historical position MR data set, the training device determines the historical position according to the historical position MR data and at least one historical position MR data adjacent to the historical position MR data in time Characteristic information of location MR data.
  • Step 63 The training device trains the positioning model based on the historical position MR call sequence corresponding to the historical position MR data.
  • the historical position MR call sequence corresponds to the historical position MR data subset to which the historical position MR data belongs. Therefore, when constructing the positioning model, the time sequence information between the characteristic information of the historical position MR data can be considered, and Training the positioning model based on recursive Bayesian estimation RBE or neural network model.
  • the position information of the terminal device corresponding to the historical position MR data is only the same as the position information of the terminal device at the previous moment.
  • the location of the terminal device corresponding to the historical location MR data can be predicted by an iterative method, where the distribution of feature information at a certain time at a certain time is estimated, A particle filter algorithm or random forest can be used to simulate this distribution.
  • a neural network model with temporal representation capability can be used to correspond to the characteristics of the terminal device at multiple times for the historical location MR data Information and location information are modeled, and the parameters of the neural network model are trained according to the feature information and location information.
  • Step 64 The training device stores the model file corresponding to the positioning model.
  • step 61, step 62, and step 64 in this embodiment can be introduced in the embodiment shown in FIG. 5, and will not be repeated here.
  • the positioning model based on the time series relationship between the historical position MR data in the corresponding historical position MR call sequence and other historical position MR data, the dimension of the feature information of each historical position MR data is enriched , So that the positioning model can more effectively characterize the relationship between the characteristic information of the MR data and the position information of the terminal device corresponding to the MR data, which also improves the subsequent positioning accuracy.
  • the embodiment of the present application considers the time sequence of the adjacent historical position MR data in the historical position MR data set.
  • the positioning model may be based on recursive Bayesian estimation (RBE) or Neural network models (including but not limited to RNN, LSTM, or temporal convolutional network (TCN)) are constructed.
  • step 23 in the embodiment shown in FIG. 2 can be implemented in the following manner. That is, the positioning device first queries the storage device based on the characteristic information of the MR data whether the storage device uses the above According to the model file generated by the realization method (1) or the realization method (2), the specific positioning method is determined according to the query result.
  • model file generated by an implementation method obtain the model file corresponding to the implementation method from the storage device, generate the positioning model corresponding to the model file, and input the characteristic information of the MR data obtained above into the positioning model , Using the positioning model to predict the location information of the terminal equipment corresponding to the MR data.
  • the model file generated based on the implementation method (2) is obtained from the storage device, the positioning model is determined based on the model file, and the characteristic information of the MR data obtained above is input into the In the positioning model, the positioning model is used to predict the position information of the terminal device corresponding to the MR data.
  • the feature information of the MR data determined by the positioning device can be either the fusion feature information or the first fusion feature information obtained after processing and extracting the steps of the embodiment shown in FIG. 3 and the embodiment shown in FIG. 4
  • a feature information set may also be processed through part of the steps of the embodiment shown in FIG. 3, and the second feature information set obtained by performing feature extraction on a single piece of MR data using the embodiment shown in FIG. 4.
  • the positioning model obtained by the positioning device through implementation (1) or implementation (2) can predict and obtain an accurate MR.
  • the data corresponds to the location information of the terminal device.
  • the positioning device using the positioning model obtained by the implementation (2) can also predict and obtain an accurate corresponding MR data Location information of the terminal device.
  • FIG. 7 is a schematic flowchart of Embodiment 4 of the positioning method provided in an embodiment of this application.
  • the execution subject of this method may be the positioning device in the positioning system shown in Figure 1 above.
  • the positioning method may include the following steps:
  • Step 71 Obtain an MR data set, where the MR data set includes multiple pieces of MR data.
  • Step 72 For any piece of MR data in the MR data set, determine the characteristic information of the MR data according to the MR data and at least one piece of MR data adjacent to the MR data in time.
  • the characteristic information of the MR data is used It describes the location of the terminal equipment corresponding to the MR data.
  • Step 71 in this embodiment is consistent with step 21 in the embodiment shown in FIG. 2 above
  • step 72 is consistent with step 22 in the embodiment shown in FIG. 2 above.
  • step 71 and step 72 refer to the above figure. The descriptions in the embodiments shown in 2 to 4 will not be repeated here.
  • Step 73 According to the characteristic information of the MR data, a target fingerprint file matching the characteristic information is determined from the fingerprint characteristic database, and the fingerprint characteristic database stores a fingerprint file for describing the association relationship between the characteristic information and the location of the terminal device.
  • the fingerprint feature database stores a fingerprint file used to describe the association relationship between feature information and the location of the terminal device.
  • the fingerprint file is the feature of each historical location MR data in the historical location MR data set by the training device. Information and the location information of the corresponding terminal equipment are processed and correlated.
  • the positioning device can use the feature information of the MR data obtained in step 72 to query the storage device, and determine whether there is a target matching the feature information of the MR data in the fingerprint feature database of the storage device. If the fingerprint file exists, the target fingerprint file is determined; if it does not exist, the location fails.
  • Step 74 Based on the target fingerprint file and the characteristic information, determine the location information of the terminal device corresponding to the MR data.
  • the positioning device obtains the target fingerprint file in the fingerprint feature library, and searches for a grid that matches the feature information in an area corresponding to the target fingerprint file based on the feature information of the MR data. Specifically, when searching for a grid that matches the feature information, it is necessary to calculate the degree of matching between the feature information and all grids in the region, for example, using Euclidean distance to measure.
  • weights can be introduced when calculating the matching degree, and then the weighted Euclidean distance method is used to measure the matching degree, and, The smaller the Euclidean distance, the higher the matching degree.
  • the weight value of each parameter information can be obtained by an optimization method.
  • the latitude and longitude of the grid with the highest matching degree is used as the location information of the terminal device corresponding to the MR data.
  • the determined location information can also be filtered and smoothed to make the positioning result of the solution more reasonable.
  • the filtering processing method in this step also includes, but is not limited to, various filtering methods such as mean filtering and Kalman filtering.
  • the positioning method of the embodiment of the present application obtains the measurement report MR data set, and for any piece of MR data in the MR data set, it is determined to use the MR data according to the MR data and at least one piece of MR data adjacent to the MR data in time.
  • a target fingerprint file matching the characteristic information is determined from the fingerprint characteristic database, and the fingerprint characteristic database stores the characteristic information for describing
  • the fingerprint file associated with the location of the terminal device is finally determined based on the target fingerprint file and characteristic information to determine the location information of the terminal device corresponding to the MR data.
  • This technical solution introduces the time dimension in the feature extraction stage, takes into account the time sequence relationship between the MR data in the corresponding MR call sequence and other MR data, enriches the dimension of feature information, and improves the positioning accuracy.
  • the fingerprint feature database includes: a fingerprint file corresponding to the current area where each piece of historical position MR data in the historical position MR data set is located, and the fingerprint file is used to describe each grid in the current area
  • the feature information of and the location information of the grid; the feature information of the historical location MR data is extracted by merging, splitting, and filtering the historical location MR data in the historical location MR data set.
  • the embodiment of the present application may be based on a machine learning method.
  • a fingerprint feature library feature information is extracted based on the time sequence relationship between MR data, and the discrimination of each fingerprint information is enhanced in a fingerprint positioning scheme (such as feature library positioning) , Thereby improving positioning accuracy.
  • the method based on the fingerprint feature database can be called "time-space feature database positioning method".
  • the method of this embodiment is particularly suitable for the historical location MR data set to characterize the user's conversational ability, but the data volume of the historical location MR data is not enough to form a call sequence for training a positioning model, and calculation In resource-constrained scenarios.
  • the better callability of the MR data refers to a higher proportion of the number of MR data contained in each MR data subset is not less than 8.
  • FIG. 8 is a schematic diagram of a method for generating a fingerprint feature database in a positioning method provided in an embodiment of the application. This method can be executed by the training device in the positioning system shown in FIG. 1 above. Exemplarily, as shown in FIG. 8, the method may include the following steps:
  • Step 81 Obtain a historical position MR data set.
  • the historical position MR data set includes: multiple pieces of historical position MR data, and the historical position MR data refers to historical MR data carrying position information.
  • Step 82 For any piece of historical position MR data in the historical position MR data set, the training device determines the historical position based on the historical position MR data and at least one historical position MR data adjacent to the historical position MR data in time Characteristic information of location MR data.
  • Step 81 in this embodiment is consistent with step 51 in the embodiment shown in FIG. 5, and step 82 is consistent with step 52 in the embodiment shown in FIG. 5.
  • step 81 and step 82 refer to the above figure.
  • the records in the embodiment shown in 5 will not be repeated here.
  • Step 83 Perform rasterization processing on the current region corresponding to the historical position MR data, and determine all the grids included in the current region.
  • the current area refers to the area where the historical location MR and the MR data to be positioned are located.
  • Step 84 Count and calculate the feature information in each grid to form a fingerprint file, and build a fingerprint feature database based on the fingerprint files of all grids included in the current area.
  • the dimensional value of each feature information in the feature file of each grid includes, but is not limited to, the average value, maximum value, minimum value and other statistical values that can describe the data distribution.
  • the fingerprint file may include the following content: grid number, grid latitude and longitude, and statistical values of each feature information on the grid.
  • Each grid forms a fingerprint file, and all fingerprint files on the grid form a fingerprint feature library.
  • Step 85 Store the constructed fingerprint feature library.
  • the generated fingerprint feature library is saved in a storage device for subsequent direct use in positioning.
  • the training device by acquiring the historical position MR data set, for any piece of historical position MR data in the historical position MR data set, the training device is based on the historical position MR data and temporally adjacent to the historical position MR data At least one piece of historical position MR data of the historical position, the characteristic information of the historical position MR data is determined, the current region corresponding to the historical position MR data is rasterized, all the grids included in the current region are determined, and each grid is counted and calculated The feature information in the grid forms a fingerprint file, a fingerprint feature database is constructed based on the fingerprint files of all grids included in the current area, and the fingerprint feature database obtained is finally stored.
  • This technical solution based on processing the historical MR data set, considers the time relationship between the historical position MR data, and enriches the dimension of the feature information of each historical position MR data, so that the obtained fingerprint feature library can be more effective
  • the relationship between the characteristic information of the MR data and the position information of the terminal equipment corresponding to the MR data is depicted, thereby improving the subsequent positioning accuracy.
  • FIG. 9 is a schematic flowchart of a training method provided in an embodiment of this application. As shown in Figure 9, the method may include the following steps:
  • Step 91 Obtain a historical location MR data set.
  • Step 92 Determine whether the historical position MR data set meets the scene condition for training the first positioning model; if yes, go to step 93, if not, go to step 94.
  • the scenario condition for training the first type of positioning model is that the amount of data in the historical position MR set is greater than the amount of data required to train the first type of positioning model, and the first type of positioning model is based on all the data in the historical position MR data set.
  • the characteristic information obtained from the time series relationship between the historical position MR data and the position information corresponding to the historical position MR data are obtained.
  • Step 93 Based on the association relationship between the feature information obtained from the time sequence relationship between all the historical position MR data in the historical position MR data set and the position information corresponding to the historical position MR data, a first positioning model is obtained.
  • the first positioning model is the positioning model obtained through training in the embodiment shown in FIG. 6.
  • the implementation principle of this step please refer to the record in the embodiment shown in FIG. 6 for details, which will not be repeated here.
  • Step 94 Determine whether the historical position MR data set meets the scene condition for training the second positioning model; if yes, execute step 95, if not, execute step 96.
  • the scenario condition for training the second positioning model is that the amount of data in the historical position MR set is less than the amount of data required to train the second positioning model, but the historical position MR data set characterizes the user's conversational ability better, but its It can support model training such as random forest and DNN.
  • Step 95 Based on the machine learning method, the feature information of each historical position MR data in the historical position MR data set is taken as input, and the position label of the historical position MR data is taken as output, and the second positioning model is trained.
  • the second positioning model is the positioning model trained in the embodiment shown in FIG. 5.
  • the second positioning model is the positioning model trained in the embodiment shown in FIG. 5.
  • the record in the embodiment shown in FIG. 5 for details. Repeat it again.
  • Step 96 Based on each historical position MR data in the historical position MR data set, a fingerprint feature library including multiple fingerprint files is determined, and each fingerprint file is used to describe the feature information of each grid in the current area and the grid. Grid location information.
  • the fingerprint feature library is the fingerprint feature library generated in the embodiment shown in FIG. 8.
  • the fingerprint feature library is the fingerprint feature library generated in the embodiment shown in FIG. 8.
  • the record in the embodiment shown in FIG. 8 for details, which will not be repeated here. .
  • the priority of different models or fingerprint libraries in this embodiment is: first positioning model>second positioning model>fingerprint feature library.
  • the solution considers the time sequence relationship of MR data in the feature extraction stage of MR data, so that the positioning model or fingerprint feature library obtained by training can better reflect the relationship between MR data and geographic location.
  • the positioning model construction stage or the fingerprint feature database generation stage according to the actual live network conditions, determine the applicable conditions of the current scene, select and construct a suitable positioning model or fingerprint feature database, so that the positioning method of this application has more advantages. Strong adaptability further improves the positioning accuracy and ensures the positioning rate of MR data.
  • the positioning device predicts the location information of the terminal device corresponding to the MR data based on the feature information of any piece of MR data in the acquired MR data set and the positioning model/fingerprint feature library, Similarly, the first positioning model is greater than the second positioning model, and the second positioning model is greater than the fingerprint feature library.
  • the specific positioning process please refer to the relevant records in the embodiments described in Figures 2 to 7. I won't repeat them here.
  • FIG. 10 is a schematic structural diagram of Embodiment 1 of a positioning device provided by an embodiment of this application.
  • the device can be integrated in the positioning device or realized by the positioning device.
  • the device may include: an acquisition module 101, a processing module 102, and a positioning module 103.
  • the obtaining module 101 is configured to obtain a measurement report MR data set, and the MR data set includes: multiple pieces of MR data;
  • the processing module 102 is configured to determine characteristic information of the MR data according to the MR data and at least one MR data adjacent to the MR data in time for any piece of MR data in the MR data set ,
  • the characteristic information of the MR data is information used to describe the location of the terminal device corresponding to the MR data;
  • the positioning module 103 is configured to input the characteristic information of the MR data into a positioning model to obtain the position information of the terminal device corresponding to the MR data.
  • the positioning model takes the characteristic information of each piece of historical position MR data in the historical position MR data set as input, and takes the position information of the terminal device corresponding to the historical position MR data as the model obtained by output training, and the historical position MR data refers to historical MR data that carries location information.
  • the processing module 102 is further configured to determine the MR data based on the MR data and at least one piece of MR data adjacent to the MR data in time. Before the characteristic information of the data, sort all the MR data in the MR data set according to the user identification information and time stamp information carried by each MR data in the MR data set to obtain the sorted MR data set ;
  • the MR data with the same user identification in the sorted MR data set are arranged together according to the time stamp information, and the MR data set after the sorting process has the same user identification information and the time difference is less than Multiple pieces of MR data of the preset first duration are merged to obtain a merged MR data set.
  • the processing module 102 is further configured to perform multiple items in the MR data set after sorting processing that have the same user identification information and the time difference is less than the preset first duration. Merge the MR data to obtain the merged MR data set, split the merged MR data set to obtain multiple MR data subsets;
  • each MR data subset includes: multiple pieces of MR data sorted by timestamp information and with the same user ID, the total duration of all MR data in each MR data subset is less than the preset second duration, and each The time difference corresponding to the time stamp information of two adjacent MR data in the MR data subset is less than the preset third time length, and the total time length is the last MR data and the first MR data in the MR data subset The time difference corresponding to the timestamp information.
  • the processing module 102 is further configured to perform split processing on the combined MR data set to obtain multiple MR data sub-sets, and then perform a split processing on each MR data sub-set All MR data in the MR data is filtered to obtain a processed MR data set.
  • the processing module 102 has at least one piece of MR data located before the MR data in time, the MR data, and the MR data in time. At least one piece of MR data after the data forms the MR call sequence corresponding to the MR data, and based on the convolutional neural network and the MR call sequence, feature extraction is performed on the MR data to obtain the first feature information of the MR data Set, and use the first feature information set as feature information of the MR data.
  • the processing module 102 further has a second feature information set for extracting the MR data according to the MR data and the network parameter information corresponding to the MR data, and Fusing the first feature information set and the second feature information set to obtain the fusion feature information of the MR data, and use the fusion feature information of the MR data as the feature information of the MR data;
  • the network parameter information corresponding to the MR data is used to indicate the attribute information of the cell to which the terminal device reporting the MR data belongs.
  • the device of this embodiment can be used to implement the implementation solutions of the method embodiments shown in FIGS. 2 to 4 and part of the implementation solutions in the embodiment shown in FIG. 9.
  • the specific implementation methods and technical effects are similar, and details are not described herein again.
  • the positioning model is specifically based on the machine learning device taking the characteristic information of each historical position MR data in the historical position MR data set as input, and taking the The position label of the historical position MR data is used as the model obtained by output training;
  • the characteristic information of the historical position MR data is obtained by merging, splitting, and filtering the historical position MR data in the historical position MR data set.
  • the location label of the historical location MR data is obtained by rasterizing a preset area, and the preset area includes: an area determined by the location of the site to which the terminal device reporting the historical location MR data belongs.
  • the training process of the positioning model can be referred to the record in the embodiment shown in FIG. 5, and the specific implementation and technical effects are similar, and will not be repeated here.
  • the positioning model is specifically based on the feature information and the corresponding historical position MR obtained based on the time series relationship between all historical position MR data in the historical position MR data set.
  • the time sequence relationship is the time sequence corresponding to the time stamp information of the historical position MR data.
  • the training process of the positioning model can be referred to the record in the embodiment shown in FIG. 6, and the specific implementation and technical effects are similar, and will not be repeated here.
  • FIG. 11 is a schematic flowchart of Embodiment 2 of a positioning device provided by an embodiment of this application.
  • the device can also be integrated in the positioning device, or realized by the positioning device.
  • the device may include: an acquisition module 111, a processing module 112, and a positioning module 113.
  • the obtaining module 111 is configured to obtain a measurement report MR data set, and the MR data set includes: multiple pieces of MR data;
  • the processing module 112 is configured to determine characteristic information of the MR data based on the MR data and at least one MR data adjacent to the MR data in time for any piece of MR data in the MR data set , And according to the characteristic information of the MR data, a target fingerprint file matching the characteristic information is determined from the fingerprint characteristic database.
  • the feature information of the MR data is information used to describe the location of the terminal device corresponding to the MR data
  • the fingerprint feature database stores a fingerprint file used to describe the association relationship between the feature information and the location of the terminal device
  • the positioning module 113 is configured to determine the location information of the terminal device corresponding to the MR data based on the target fingerprint file and the characteristic information.
  • the device in this embodiment can be used to implement the implementation scheme of the method embodiment shown in FIG. 7 and some implementation schemes in the embodiment shown in FIG. 9. The specific implementation manner and technical effect are similar, and details are not repeated here.
  • the fingerprint feature database includes: a fingerprint file corresponding to the current area where each piece of historical position MR data in the historical position MR data set is located, and the fingerprint file is used to describe the current area
  • the feature information of each grid in the grid and the location information of the grid is processed by merging, splitting, and filtering the historical position MR data in the historical position MR data set After extraction.
  • the method for generating the fingerprint feature database can be referred to the record in the embodiment shown in FIG. 8.
  • the specific implementation and technical effects are similar, and will not be repeated here.
  • modules of the above device is only a division of logical functions, and may be fully or partially integrated into a physical entity in actual implementation, or may be physically separated.
  • modules can all be implemented in the form of software called by processing elements; they can also be implemented in the form of hardware; some modules can be implemented in the form of calling software by processing elements, and some of the modules can be implemented in the form of hardware.
  • the processing module may be a separately established processing element, or it may be integrated in a chip of the above-mentioned device for implementation.
  • it may also be stored in the memory of the above-mentioned device in the form of program code, and a certain processing element of the above-mentioned device Call and execute the functions of the above processing module.
  • the implementation of other modules is similar.
  • each step of the above method or each of the above modules can be completed by hardware integrated logic circuits in the processor element or instructions in the form of software.
  • the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuit (ASIC), or one or more microprocessors (digital signal processor, DSP), or, one or more field programmable gate arrays (FPGA), etc.
  • ASIC application specific integrated circuit
  • DSP digital signal processor
  • FPGA field programmable gate arrays
  • the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call program codes.
  • CPU central processing unit
  • these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).
  • SOC system-on-a-chip
  • the above embodiments it may be implemented in whole or in part by software, hardware, firmware or any combination thereof.
  • software it can be implemented in the form of a computer program product in whole or in part.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in a readable storage medium, or transmitted from one readable storage medium to another readable storage medium.
  • the computer instructions may be transmitted from a website, computer, server, or data center through a wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means to transmit to another website, computer, server or data center.
  • the readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center integrated with one or more available media.
  • the usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, and a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)).
  • FIG. 12 is a schematic structural diagram of Embodiment 3 of a positioning device provided by an embodiment of this application.
  • the positioning device can be integrated in the positioning device.
  • the positioning device may include: a processor 121, a memory 122, a communication interface 123, and a system bus 124.
  • the memory 122 and the communication interface 123 are connected to the processor 121 through the system bus 124.
  • the memory 122 is used to store computer-executed instructions
  • the communication interface 123 is used to communicate with other devices
  • the processor 121 executes the computer-executed instructions when the implementation is shown in Figures 2 to 4
  • FIG. 13 is a schematic structural diagram of a training device provided by an embodiment of the application.
  • the training device may include: a processor 131, a memory 132, a communication interface 133, and a system bus 134.
  • the memory 132 and the communication interface 133 are connected to the processor 131 through the system bus 134.
  • the memory 132 is used to store computer-executed instructions
  • the communication interface 133 is used to communicate with other devices
  • the processor 131 executes the computer-executed instructions as shown in Figure 5 or Figure 6 The implementation scheme of the method embodiment shown.
  • FIG. 14 is a schematic structural diagram of Embodiment 4 of a positioning device provided by an embodiment of this application.
  • the device can be integrated in the positioning device.
  • the device may include: a processor 141, a memory 142, a communication interface 143, and a system bus 144.
  • the memory 142 and the communication interface 143 are connected to the processor 141 through the system bus 144 and To complete mutual communication, the memory 142 is used to store computer-executed instructions, the communication interface 143 is used to communicate with other devices, and the processor 141 executes the computer-executed instructions to implement the method shown in FIG. 7
  • FIG. 7 The implementation scheme of the example and part of the implementation scheme in the embodiment shown in FIG. 9.
  • FIG. 15 is a schematic structural diagram of another training device provided by an embodiment of the application.
  • the training device may include: a processor 151, a memory 152, a communication interface 153, and a system bus 154.
  • the memory 152 and the communication interface 153 are connected to the processor 151 through the system bus 154.
  • the memory 152 is used to store computer execution instructions
  • the communication interface 153 is used to communicate with other devices
  • the processor 151 implements the method shown in FIG. 8 when executing the computer execution instructions Implementation scheme of the embodiment.
  • system bus mentioned in FIGS. 12 to 15 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus.
  • PCI peripheral component interconnect
  • EISA extended industry standard architecture
  • the system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
  • the communication interface is used to realize the communication between the database access device and other devices (such as client, read-write library and read-only library).
  • the memory may include random access memory (RAM), and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
  • the above-mentioned processor may be a general-purpose processor, including a central processing unit CPU, a network processor (NP), etc.; it may also be a digital signal processor DSP, an application specific integrated circuit ASIC, and a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • a general-purpose processor including a central processing unit CPU, a network processor (NP), etc.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA field programmable gate array
  • the foregoing memory may be a read only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), and the memory may store programs and data.
  • ROM read only memory
  • RAM random access memory
  • the foregoing processor and memory may also be integrated in an application specific integrated circuit, and the integrated circuit may also include a communication interface.
  • the application specific integrated circuit can be a processing chip or a processing circuit.
  • the communication interface may be a communication interface including wireless transmission and reception, or an interface for digital signals input after processing the received wireless signals through other processing circuits, and may also be a software or hardware interface for communicating with other modules.
  • the aforementioned positioning device or training device may also include an artificial intelligence processor, which may be a neural network processor (network processing unit, NPU), a tensor processing unit (TPU), or graphics Processors (graphics processing unit, GPU) and other processors suitable for large-scale XOR operations.
  • the artificial intelligence processor can be mounted on the host CPU (Host CPU) as a coprocessor, and the host CPU assigns tasks to it.
  • the artificial intelligence processor can implement one or more operations involved in the above-mentioned positioning model training method. For example, taking the NPU as an example, the core part of the NPU is an arithmetic circuit, and the controller controls the arithmetic circuit to extract matrix data in the memory and perform multiplication and addition operations.
  • an embodiment of the present application provides a storage medium that stores instructions in the storage medium, and when the instructions run on a computer, the computer executes the implementation of the method embodiments shown in FIGS. 2 to 6 above. Solution, and the implementation solution in the embodiment shown in FIG. 9; or
  • the computer When the instructions run on the computer, the computer is caused to execute the implementation solutions of the method embodiments shown in FIG. 7 and FIG. 8 and the implementation solutions in the embodiment shown in FIG. 9.
  • An embodiment of the present application further provides a program product, the program product includes a computer program, the computer program is stored in a storage medium, at least one processor can read the computer program from the storage medium, and the at least one When the processor executes the computer program, the implementation solution of the method embodiment shown in FIG. 2 to FIG. 6 and the implementation solution of the embodiment shown in FIG. 9 can be implemented; or
  • the at least one processor executes the computer program
  • the implementation solutions of the method embodiments shown in FIG. 7 and FIG. 8 and the implementation solutions in the embodiment shown in FIG. 9 can be implemented.
  • FIG. 16 is a schematic structural diagram of an embodiment of a positioning system provided by an embodiment of this application.
  • the positioning system may include: a positioning device 161 and a training device 162.
  • the training device 162 can communicate with the positioning device 161, and send the trained positioning model or the generated fingerprint feature library to the positioning device 161, and the positioning device 161 can use the received positioning model or fingerprint feature library to predict the terminal corresponding to the MR data Location information of the device.
  • At least one refers to one or more, and “multiple” refers to two or more.
  • “And/or” describes the association relationship of the associated objects, indicating that there can be three relationships, for example, A and/or B, which can mean: A alone exists, both A and B exist, and B exists alone, where A, B can be singular or plural.
  • the character “/” generally indicates that the associated objects before and after are in an “or” relationship; in the formula, the character “/” indicates that the associated objects before and after are in a “division” relationship.
  • “The following at least one item (a)” or similar expressions refers to any combination of these items, including any combination of a single item (a) or plural items (a).
  • at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple One.
  • the size of the sequence numbers of the foregoing processes does not mean the order of execution.
  • the execution order of each process should be determined by its function and internal logic, and should not be implemented in this application.
  • the implementation process of the example constitutes any limitation.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Position Fixing By Use Of Radio Waves (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请实施例提供了一种定位方法、装置及存储介质，其中，该方法包括：获取包括多条MR数据的MR数据集合，对于该MR数据集合中的任意一条MR数据，根据该MR数据在时间上与该MR数据相邻的MR数据，确定用于描述该MR数据对应终端设备所在位置的特征信息，将该MR数据的特征信息输入到定位模型中，得到该MR数据对应终端设备的位置信息。该技术方案中，由于该定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，且确定该MR数据的特征信息时考虑了MR数据的时间先后关系对终端设备所在地理位置的影响，提升了定位精度。

Description

定位方法、装置及存储介质
本申请要求于2019年04月25提交中国专利局、申请号为201910339092.9、申请名称为“定位方法、装置及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及通信技术领域，尤其涉及一种定位方法、装置及存储介质。
背景技术
测量报告(measurement report，MR)是移动终端周期性或事件触发上报的，包含移动终端某时刻的无线环境信息，其和移动终端当前所在地理位置存在对应关系。因而，基于测量报告对移动终端进行定位，是当前对移动终端定位的常见手段。
现有技术中基于MR的移动终端定位方法中，首先获取移动终端上报的MR，基于MR携带的用户标识或用户流程呼叫详细记录(call detail record，CDR)标识选取预设时间段内的MR，并按照MR携带的时间戳信息进行排序，然后计算预设时间段内每个MR的主服务小区和最强的两个邻区所形成三角形的重心，根据每两个相邻MR的主/邻小区的三角形重心和每两个相邻MR的时间差计算移动终端的移动方向和速度，最后根据移动终端的方向和速度，对当前MR的小区电平强度进行平滑处理，最终根据平滑处理后的小区电平强度和预先建立的指纹库确定出移动终端的位置。
然而，上述移动终端的定位方法在定位过程中只考虑了每条MR的小区电平强度对移动终端定位结果的影响，存在终端设备定位精度低的问题。
发明内容
本申请实施例提供一种定位方法、装置及存储介质，以解决现有技术中终端设备定位精度低的问题。
本申请第一方面提供一种定位方法，包括：获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；对于所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；将所述MR数据的特征信息输入到定位模型中，得到所述MR数据对应终端设备的位置信息，所述定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以所述历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，所述历史位置MR数据是指携带位置信息的历史MR数据。
在本实施例中，定位设备在定位过程中基于与该MR数据具有时间先后关系的至少一条MR数据对该MR数据进行处理，得到的MR数据的特征信息可以更加准确的表征终端 设备的位置信息，从而提升了定位精度。
在第一方面的一种可能设计中，在所述根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息之前，所述方法还包括：
根据所述MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对所述MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合，所述排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起；
将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合。
在本实施例中，定位设备对获取到的MR数据集合进行处理时，使具有相同用户标识的MR数据按照时间戳信息排列在一起，这样定位设备可以对排序处理后的MR数据集合中的MR数据进行特征提取时，可以将MR数据集合中每个MR数据上报的时间顺序考虑在内，为后续提取到合理、准确的特征信息奠定了基础；通过将每个时刻或者每个时间段内的MR数据进行整合，可以使得每个时刻或时间段内的特征信息更完整，从而提高了MR数据对应特征信息的准确性。
在第一方面的上述可能设计中，在所述将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合之后，所述方法还包括：
对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合，每个MR数据子集合包括：按照时间戳信息排序且具有相同用户标识的多条MR数据，每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，且每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长，所述总时长为所述MR数据子集合中最后一条MR数据与第一条MR数据的时间戳信息对应的时间差值。
在本实施例中，按照预设的通话提取策略对上述合并处理后的MR数据集合进行拆分处理，以得到多个MR数据子集合，能够在对MR数据进行特征提取时，提高特征信息的提取效率和准确度。
在第一方面的上述可能设计中，在所述对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合之后，所述方法还包括：
对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合。
本实施例中，为了减少信号电平值的波动对定位精度的影响，得到稳定的信号电平值，通过对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合，从而在后续定位时提供定位精度。
在第一方面的另一种可能设计中，所述根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，包括：
根据时间上位于所述MR数据之前的至少一条MR数据、所述MR数据以及时间上位于所述MR数据之后的至少一条MR数据，形成所述MR数据对应的MR通话序列；
基于卷积神经网络和所述MR通话序列，对所述MR数据进行特征提取，得到所述MR数据的第一特征信息集合；
将所述第一特征信息集合作为所述MR数据的特征信息。
在本实施例中，由于该第一特征信息集合是基于该MR数据对应的MR通话序列得到 的，其在特征提取的过程中将MR数据之间的时序关系考虑在内，增强了MR数据的特征信息的合理性和准确度，提高了定位精度。
在第一方面的这种可能设计中，所述根据所述MR数据在时间上与所述MR数据相邻的MR数据，确定所述MR数据的特征信息，还包括：
根据所述MR数据和所述MR数据对应的网络工参信息，提取所述MR数据的第二特征信息集合，所述MR数据对应的网络工参信息用于表示上报所述MR数据的终端设备所属小区的属性信息；
相应的，所述将所述第一特征信息集合作为所述MR数据的特征信息，包括：
对所述第一特征信息集合和所述第二特征信息集合进行融合，得到所述MR数据的融合特征信息；
将所述MR数据的融合特征信息作为所述MR数据的特征信息。
在本实施例中，既考虑了单条MR数据的特征信息，也考虑了基于该MR数据对应MR通话序列提取到的特征信息，其进一步增强了MR数据的合理性和准确度，为后续确定出准确的终端位置奠定了基础。
在第一方面的再一种可能设计中，所述定位模型具体是基于机器学习方法以所述历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以所述历史位置MR数据的位置标签作为输出训练得到的模型；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
可选的，所述历史位置MR数据的位置标签是对预设区域进行栅格化得到的，所述预设区域包括：上报所述历史位置MR数据的终端设备所属站点的位置确定的区域。
在本实施例中，在对历史MR数据集合进行处理时，考虑了历史位置MR数据之间的时间关系，丰富了每个历史位置MR数据的特征信息的维度，使得该定位模型可以更有效地刻画MR数据的特征信息与该MR数据对应终端设备的位置信息之间的关系，从而提升了后续的定位精度。
在第一方面的又一种可能设计中，所述定位模型具体是基于所述历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的，所述时序关系为所述历史位置MR数据的时间戳信息对应的时间先后顺序。
在本实施例中，在定位模型的训练过程中，基于历史位置MR数据在对应历史位置MR通话序列中与其他历史位置MR数据之间的时序关系，丰富了每个历史位置MR数据的特征信息的维度，使得该定位模型可以更有效地刻画MR数据的特征信息与该MR数据对应终端设备的位置信息之间的关系，同样提升了后续的定位精度。
本申请第二方面提供一种定位方法，包括：获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；对于所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；根据MR数据的特征信息，从指纹特征库中确定出与所述特征信息匹配的目标指纹文件，所述指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件；基于所述目标指纹文件和所述特征信息，确定出所述MR数据对应终端设备的位置信息。
在本实施例中，在特征提取阶段引入了时间维度，考虑了该MR数据在对应MR通话序列与其他MR数据的时序关系，丰富了特征信息的维度，提升了定位精度。
在第二方面的一种可能设计中，所述指纹特征库包括：历史位置MR数据集合中每条历史位置MR数据所在当前区域对应的指纹文件，所述指纹文件用于描述所述当前区域中每个栅格的特征信息与所述栅格的位置信息；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
在本实施例中，基于对历史MR数据集合进行处理时，考虑了历史位置MR数据之间的时间关系，丰富了每个历史位置MR数据的特征信息的维度，使得得到的指纹特征库可以更有效地刻画MR数据的特征信息与该MR数据对应终端设备的位置信息之间的关系，从而提升了后续的定位精度。
本申请第三方面提供一种定位装置，包括：获取模块、处理模块和定位模块；
所述获取模块，用于获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
所述处理模块，用于针对所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；
所述定位模块，用于将所述MR数据的特征信息输入到定位模型中，得到所述MR数据对应终端设备的位置信息，所述定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以所述历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，所述历史位置MR数据是指携带位置信息的历史MR数据。
在第三方面的一种可能设计中，所述处理模块，还用于在根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息之前，根据所述MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对所述MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合，所述排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起，以及将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合。
在第三方面的上述可能设计中，所述处理模块，还用于在将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合之后，对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合；
其中，每个MR数据子集合包括：按照时间戳信息排序且具有相同用户标识的多条MR数据，每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，且每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长，所述总时长为所述MR数据子集合中最后一条MR数据与第一条MR数据的时间戳信息对应的时间差值。
可选的，所述处理模块，还用于在对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合之后，对每个MR数据子集合中的所有MR数据进行滤波处理， 得到处理后的MR数据集合。
在第三方面的另一种可能设计中，所述处理模块，具有用于根据时间上位于所述MR数据之前的至少一条MR数据、所述MR数据以及时间上位于所述MR数据之后的至少一条MR数据，形成所述MR数据对应的MR通话序列，基于卷积神经网络和所述MR通话序列，对所述MR数据进行特征提取，得到所述MR数据的第一特征信息集合，以及将所述第一特征信息集合作为所述MR数据的特征信息。
可选的，所述处理模块，还具有用于根据所述MR数据和所述MR数据对应的网络工参信息，提取所述MR数据的第二特征信息集合，以及对所述第一特征信息集合和所述第二特征信息集合进行融合，得到所述MR数据的融合特征信息，将所述MR数据的融合特征信息作为所述MR数据的特征信息；
其中，所述MR数据对应的网络工参信息用于表示上报所述MR数据的终端设备所属小区的属性信息。
在第三方面的再一种可能设计中，所述定位模型具体是基于机器学习装置以所述历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以所述历史位置MR数据的位置标签作为输出训练得到的模型；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
可选的，所述历史位置MR数据的位置标签是对预设区域进行栅格化得到的，所述预设区域包括：上报所述历史位置MR数据的终端设备所属站点的位置确定的区域。
在第三方面的又一种可能设计中，所述定位模型具体是基于所述历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的，所述时序关系为所述历史位置MR数据的时间戳信息对应的时间先后顺序。
本申请第四方面提供一种定位装置，包括：获取模块、处理模块和定位模块；
所述获取模块，用于获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
所述处理模块，用于针对所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，以及根据MR数据的特征信息，从指纹特征库中确定出与所述特征信息匹配的目标指纹文件，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息，所述指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件；
所述定位模块，用于基于所述目标指纹文件和所述特征信息，确定出所述MR数据对应终端设备的位置信息。
在第四方面的一种可能设计中，所述指纹特征库包括：历史位置MR数据集合中每条历史位置MR数据所在当前区域对应的指纹文件，所述指纹文件用于描述所述当前区域中每个栅格的特征信息与所述栅格的位置信息；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
关于第三方面和第四方面中各可能设计未详尽的有益技术效果可以参见第一方面和第二方面中的记载，此处不再赘述。
本申请第五方面提供一种定位装置，包括处理器、存储器及存储在所述存储器上并可在处理器上运行的计算机程序，所述处理器执行所述程序时实现如上述第一方面以及第一方面各种可能的设计中所述的方法。
本申请第六方面提供一种定位装置，包括：处理器和存储器，所述存储器用于存储计算机程序代码，所述处理器用于调用所述计算机程序代码，执行如上述第二方面以及第二方面各种可能的设计中所述的方法。
本申请第七方面提供一种存储介质，所述存储介质中存储有指令，当其在计算机上运行时，使得计算机执行如上述第一方面以及第一方面各种可能的设计中所述的方法。
本申请第八方面提供一种存储介质，所述存储介质中存储有指令，当其在计算机上运行时，使得计算机执行如上述第二方面以及第二方面各种可能的设计中所述的方法。
本申请第九方面提供一种包含指令的程序产品，当其在计算机上运行时，使得计算机执行上述第一方面以及第一方面各种可能的设计中所述的方法。
本申请第十方面提供一种包含指令的程序产品，当其在计算机上运行时，使得计算机执行上述第二方面以及第二方面各种可能的设计中所述的方法。
本申请第十一方面提供一种芯片，所述芯片包括存储器、处理器，存储器中存储代码和数据，存储器与所述处理器耦合，处理器运行存储器中的代码使得芯片用于执行上述第一方面以及第一方面各种可能的设计中所述的方法。
本申请第十二方面提供一种芯片，所述芯片包括存储器、处理器，存储器中存储代码和数据，存储器与所述处理器耦合，处理器运行存储器中的代码使得芯片用于执行上述第二方面以及第二方面各种可能的设计中所述的方法。
本申请第十三方面提供一种通信系统，包括：定位设备和训练设备；
所述定位设备为上述第三方面以及第三方面各种可能设计中所述的装置，所述训练设备为用于训练上述第三方面以及第三方面各种可能设计中定位模型的设备。
本申请实施例提供的定位方法、装置及存储介质，通过获取包括多条MR数据的MR数据集合，对于该MR数据集合中的任意一条MR数据，根据该MR数据在时间上与该MR数据相邻的MR数据，确定用于描述该MR数据对应终端设备所在位置的特征信息，最后将该MR数据的特征信息输入到定位模型中，得到该MR数据对应终端设备的位置信息，由于该定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，且确定该MR数据的特征信息时考虑了MR数据的时间先后关系对终端设备所在地理位置的影响，提升了定位精度，解决了现有技术中终端设备定位精度低的问题。
附图说明
图1为本申请实施例提供的一种定位系统的结构示意图；
图2为本申请实施例提供的定位方法实施例一的流程示意图；
图3为本申请实施例提供的定位方法实施例二的流程示意图；
图4为本申请实施例提供的定位方法实施例三的流程示意图；
图5为本申请实施例提供的定位方法中定位模型的实现方式(1)的流程示意图；
图6为本申请实施例提供的定位方法中定位模型的实现方式(2)的流程示意图；
图7为本申请实施例提供的定位方法实施例四的流程示意图；
图8为本申请实施例提供的定位方法中指纹特征库的生成方法示意图；
图9为本申请实施例提供的训练方法的流程示意图；
图10为本申请实施例提供的定位装置实施例一的结构示意图；
图11为本申请实施例提供的定位装置实施例二的流程示意图；
图12为本申请实施例提供的定位装置实施例三的结构示意图；
图13为本申请实施例提供的一种训练设备的结构示意图；
图14为本申请实施例提供的定位装置实施例四的结构示意图；
图15为本申请实施例提供的另一种训练设备的结构示意图；
图16为本申请实施例提供的定位系统实施例的结构示意图。
具体实施方式
以下，对本申请实施例中的部分用语进行解释说明，以便于本领域技术人员理解：
在本申请实施例中，网络工程参数是在无线网络规划中用于描述站点(基站)属性信息的工程参数，该网络工程参数可以包括：站点天线位置的经纬度、天线方向性、增益、方位角、下倾角、挂高、馈线型号、站点类型(室内、室外)、各小区的电平值(例如，参考信号接收功率(reference signal receiving power，RSRP))等。
用户指的是携带可通过无线网络进行通信的终端(Terminal，即通信终端，包括但不限于手机等)的人。
值得说明的是，本申请实施例描述的无线网络以及业务场景是为了更加清楚的说明本申请实施例的技术方案，并不构成对于本申请实施例提供的技术方案的限定，本领域普通技术人员可知，随着网络架构的演变和新业务场景的出现，本申请实施例提供的技术方案对于类似的技术问题，同样适用。
在本申请实施例中，将某一地理位置的指纹可以解释为该地理位置对应的测量报告所包含的信号电平值等信息。相应的，指纹库是用于进行指纹匹配定位的基准数据库。
本申请实施例中，“定位模型”也称“模型”，可以接收输入数据，并根据接收的输入数据和当前的模型参数生成预测输出。该定位模型可以是回归模型、神经网络(artificial neural network，ANN)、深度神经网络(deep neural network，DNN)、支持向量机(support vector machine，SVM)或其他的机器学习模型等。
测量报告(measurement report，MR)数据指在业务信道上每间隔预设时间发送的一次数据，可以包括：时间戳信息、用户标识信息、经纬度(可选)、时延信息、小区标识信息、信号电平值信息、干扰信息等内容；在本申请实施例中，MR数据主要指终端设备反馈给网络设备侧的信息，携带有终端设备从服务小区、邻区接收的信号电平值、小区标识信息、用户标识信息和时间戳信息等。
在本申请实施例中，该多条MR数据组成的MR数据集合用于定位该MR数据集合中任意一条MR数据对应终端设备的位置。
历史位置MR数据指携带终端设备所在位置信息的MR数据，该历史位置MR数据主要是通过辅助全球卫星定位系统(assisted global positioning system，AGPS)、最小化路测(minimization of drive tests，MDT)、路测(drive test，DT)等方式获取到的数据。
在本申请实施例中，该多条历史位置MR数据组成的历史位置MR数据集合用于训练具有定位终端设备所在位置功能的定位模型。
应理解，上述MR数据集合中任意一条MR数据对应终端设备的位置是基于通过对历史位置MR数据集合训练得到的模型进行定位得到的。
本申请实施例中，通过对MR数据集合进行数据处理得到多个MR数据子集合，每个MR数据子集合可以包括多个样本，每个样本中可以包括用户标识信息和时间戳信息。其中，预处理可以包括数据排序、数据合并、数据拆分、数据滤波等中的一种或多种。例如，按照每条MR数据的时间戳信息对MR数据集合中的所有MR数据进行排序处理，将具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并等等。
需要说明的是，上述仅示例性地给出了一些数据处理的形式，本申请实施例还可以包括其他的处理形式，对此，本申请实施例不作限定。
本申请实施例中“MR数据的特征信息”用于描述上报该MR数据的终端设备所在小区的属性信息，可以表征终端设备上报该MR数据时的位置信息。
例如，该MR数据的特征信息可以包括：各小区接收到的信号电平值、小区标识信息、小区类型信息、小区位置信息等，这些特征信息可以用于描述终端设备的位置信息。
下面结合图1介绍本申请实施例设计的一种系统结构。图1为本申请实施例提供的一种定位系统的结构示意图。如图1所示，该定位系统可以包括：训练设备11、定位设备12、网络设备13、至少一个终端设备14、数据存储设备15。示例性的，在图1所示的定位系统中，至少一个终端设备14在接入无线网络后可以周期性或经事件触发向网络设备13上报MR数据，该网络设备13可以将接收到的MR数据存储在数据存储设备15中。
示例性的，在本实施例中，上述MR数据可以是携带位置信息的历史位置MR数据，也可以是不携带位置信息的MR数据。其中，携带位置信息的历史位置MR数据可以是至少一个终端设备14在网络评估、网规网优、虚拟路测等具体应用场景中，通过重点道路监控、用户行为识别、价值区域识别等方式获取到的，不携带位置信息的MR数据可以是日常应用中终端设备上报的MR数据。
可以理解的是，本申请实施例并不限定上述MR数据的获取方式，其可以根据实际情况确定，此处不再赘述。
在本实施例中，数据存储设备15可以存储大量用于训练定位模型的历史位置MR数据和未携带终端位置信息的MR数据，训练设备11用于基于数据存储设备15中的历史位置MR数据集合执行模型训练方法的程序代码，以训练定位模型；定位设备12用于基于数据存储设备15中未携带终端位置信息的MR数据集合执行数据处理方法的程序代码，利用处理后的MR数据集合和训练得到的定位模型得到终端设备的位置信息。
可选的，本实施例中的数据存储设备15可以是用于存储数据的数据库，其可以是一个独立的设备，也可以集成在数据平台中。
关于训练设备11训练定位模型的方法可以参见下述实施例中的相关描述，此处不再赘述。训练设备11可以将训练出的定位模型发送至定位设备12，由定位设备12基于MR数据集合和定位模型执行定位终端设备所在位置的方法，关于具体的定位方法可以参见下述实施例中的相关描述，此处不再赘述。
在本实施例中，定位设备12由一个或多个服务器实现，可选的，与其它计算设备配 合，例如：数据存储、路由器、负载均衡器等设备；定位设备12可以布置在一个物理站点上，或者分布在多个物理站点上。定位设备12可以使用数据存储设备15中存储的数据，或者调用数据存储设备15中的程序代码实现本申请实施例所述的定位方法，具体地，定位设备12对于获取到的MR数据集合中的任意一个MR数据，根据该MR数据在时间上与MR数据相邻的MR数据，确定该MR数据的特征信息，再将该MR数据的特征信息输入到定位模型中，得到该MR数据对应终端设备的位置信息。
需要说明的是，附图1仅是本申请实施例提供的一种系统架构的示意图，图1中所示设备、器件、模块等之间的位置关系不构成任何限制，例如，在附图1中，数据存储设备15相对定位设备12可以是外部存储器，在其它情况下，也可以将数据存储设备15置于定位设备12中。
还需要说明的是，本申请实施例中训练设备11和定位设备12可以是同一设备，或者不同设备。训练设备11和/或定位设备12可以是终端，如手机终端，平板电脑，笔记本电脑，AR/VR，车载终端等，也可以是网络设备13，例如，无线接入设备、核心网设备，还可以是服务器或者虚拟机等，还可以是一个或多个服务器和/或计算机等组成的分布式计算机系统等，本申请实施例不作限定。
在本申请实施例中，上述终端设备14也可称之为用户设备(user equipment，UE)、移动台(mobile station，MS)、移动终端(mobile terminal)、终端(terminal)等，例如，终端设备14可以是智能手机、平板电脑、个人计算机、台式计算机、车载单元(on board unit，OBU)、虚拟现实设备、人工智能设备(例如机器人等)或智能可穿戴设备等，本申请实施例不作限定。
网络设备13可以包括各种形式的宏基站，微基站(也称为小站)，中继站，接入点等。例如，该网络设备13可以是GSM或CDMA中的基站(base transceiver station，BTS)，也可以是WCDMA中的基站(nodeB，NB)，还可以是LTE中的演进型基站(evolutional node B，eNB或e-NodeB)，以及可以是5G网络中对应的设备gNB、未来通信系统中的基站或WiFi系统中的接入节点等。本申请的实施例对网络设备13所采用的具体技术和具体设备形态不做限定。
可以理解的是，图1所示的定位系统可以是通信系统，该通信系统可以为全球移动通讯(global system of mobile communication，GSM)系统、码分多址(code division multiple access，CDMA)系统、宽带码分多址(wideband code division multiple access，WCDMA)系统、通用分组无线业务(general packet radio service，GPRS)、长期演进(long term evolution，LTE)系统、高级的长期演进(LTE advanced，LTE-A)、LTE频分双工(frequency division duplex，FDD)系统、LTE时分双工(time division duplex，TDD)、通用移动通信系统(universal mobile telecommunication system，UMTS)，及其他应用正交频分复用(orthogonal frequency division multiplexing，OFDM)技术的无线通信系统，以及第五代移动通信(5th generation mobile networks，5G)的新空口(new radio，NR)系统，即5G NR。本申请实施例描述的系统架构进行限定。
值得说明的是，本申请的产品实现形态是包含在机器学习、深度学习平台软件中，并部署在服务器(也可以是计算云或移动终端等具有计算能力的硬件)上的程序代码。在图1所示的系统结构图中，本申请的程序代码可以存储在定位设备和训练设备内部。运行时， 程序代码运行于服务器的主机内存和/或GPU内存。
下面首先针对本申请实施例适用场景进行简要说明。
示例性的，随着无线网络的大规模建设，网络结构的日趋复杂，如何打造一个网络性能、网络能力、服务指标均非常优异的精品网对网络运维优化提出了挑战。由于精品网的规划和优化、电信商业数据的挖掘等都需要精确识别用户所处网络的无线环境，因此，如何对用户持有的终端设备进行精准定位成为无线网络运维和优化过程中的关键。
再比如，随着网络设备的迅速发展，定位设备可以基于终端设备上报的MR确定出终端设备所在的位置后，基于该位置为用户定向推荐该位置周边的服务，例如，美食、娱乐等产品推荐。所以，对用户持有的终端设备进行精准定位是定向产品推荐的关键。
通常情况下，终端设备接入无线网络时可以周期性或事件触发上报测量报告MR数据，由于MR数据包含终端设备某时刻的无线环境信息，其和该终端设备当前所在地理位置存在对应关系，因而，基于测量报告对用户进行定位是现阶段对终端设备进行定位的常见手段。
现有技术中，基于MR的定位方法中可以通过指纹进行定位，某一地理位置对应的测量报告包含的信息称为该地理位置的“指纹”，根据该指纹预先建立指纹库，当对用户进行定位时，取该用户的终端设备上报的测量报告，以小区为单位，将用户上传的该小区的电平信息作为度量值，计算其与指纹库中的信息的相似性，最相似的指纹对应的地理位置，称为该用户此时的地理位置。
同理，基于背景技术中的介绍，移动终端的定位方法在定位过程中只考虑了每条MR的小区电平强度对移动终端定位结果的影响，存在定位精度低的问题，基于该问题，本申请实施例提供了一种定位方法、装置及存储介质，通过获取包括多条MR数据的MR数据集合，对于该MR数据集合中的任意一条MR数据，根据该MR数据在时间上与该MR数据相邻的MR数据，确定用于描述该MR数据对应终端设备所在位置的特征信息，最后将该MR数据的特征信息输入到定位模型中，得到该MR数据对应终端设备的位置信息，由于该定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，且确定该MR数据的特征信息时考虑了MR数据的时间先后关系对终端设备所在地理位置的影响，提升了定位精度。
本申请实施例中，“多个”是指两个或两个以上。“和/或”，描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。字符“/”一般表示前后关联对象是一种“或”的关系。
下面，通过具体实施例对本申请的技术方案进行详细说明。需要说明的是，下面这几个具体的实施例可以相互结合，对于相同或相似的概念或过程可能在某些实施例中不再赘述。
图2为本申请实施例提供的定位方法实施例一的流程示意图。该定位方法可以由定位设备执行，也可以由定位设备中的处理器执行。本实施例中以定位设备执行该方法进行说明。如图2所示，该定位方法可以包括如下步骤：
步骤21：获取测量报告MR数据集合，该MR数据集合包括：多条MR数据。
在本实施例中，定位设备可以在外部应用的触发下执行该定位方法。具体的，定位设 备首先获取至少一个终端设备的MR数据集合。具体的，定位设备可以直接从数据存储设备获取终端设备上报的MR数据集合，也可以从网络设备获取终端设备的MR数据集合，而且，不同的MR数据集合其包含的MR数据的条数不同。
值得说明的是，本申请实施例并不对MR数据集合的获取方式和每个MR数据集合包含的MR数据条数进行限定，其可以根据实际情况确定。
示例性的，在本实施例中，该MR数据集合可以包括多条MR数据，每条MR数据均携带用户标识信息、时间戳信息、小区信息等多维度特征。其中，该MR数据集合中的MR数据可以来自同一个终端设备，也可以来自多个终端设备，本申请实施例并不限定MR数据集合中每条MR数据的具体来源。
值得说明的是，未经特殊说明，在本申请的实施例中，MR数据是指不携带位置信息的MR数据。同理，下述历史位置MR数据指的是携带位置信息的历史MR数据。
步骤22：对于该MR数据集合中的任意一条MR数据，根据该MR数据和在时间上与该MR数据相邻的至少一条MR数据，确定该MR数据的特征信息，该MR数据的特征信息是用于描述该MR数据对应终端设备所在位置的信息。
可选的，该MR数据集合可以是定位设备获取到的原始MR数据集合在经过数据预处理后的数据集合。其中，对原始MR数据集合的预处理可以包括数据排序、合并、拆分和滤波等中的一种或多种。
在本申请实施例中，定位设备可以对该MR数据集合中的每条MR数据执行相同的处理，因而，本申请实施例可以以对该MR数据集合中的任意一条MR数据进行举例说明。
示例性的，对于该MR数据集合中的任意一条MR数据，定位设备在定位过程中引入MR数据之间的时序信息，也即，根据该MR数据和在时间上与该MR数据相邻的MR数据来提取该MR数据的特征信息。
其中，在本实施例中，该MR数据的特征信息可以用于表征该MR数据对应终端设备所在的位置，例如，该MR数据的特征信息可以包括但不局限于包括小区电平信息、小区标识信息、小区位置信息、小区站点类型等信息。
本申请实施例中，定位设备可以基于MR数据集合中各MR数据的时间先后顺序，通过对MR数据进行处理，深度挖掘MR数据的特征信息与终端位置信息之间的关系，提升了定位精度。
步骤23：将该MR数据的特征信息输入到定位模型中，得到该MR数据对应终端设备的位置信息，该定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以历史位置MR数据对应终端设备的位置信息为输出训练得到的模型。
其中，该历史位置MR数据是指携带位置信息的历史MR数据。
在申请的实施例中，利用定位模型预测终端设备的位置信息是本申请实施例的目标，定位设备利用上述得到的MR数据的特征信息和训练设备训练好的定位模型预测MR数据对应终端设备的位置，具体的，定位设备将上述得到的MR数据的特征信息作为定位模型的输入，利用该定位模型对该MR数据进行定位，从而预测该MR数据对应终端设备的位置信息。
示例性的，在本实施例中，预测得到的该MR数据对应终端设备的位置信息，即携带终端设备的用户当前的经纬度信息。
值得说明的是，为了使的本方案的定位结果更合理，定位设备在得到定位结果后，还可以对定位结果做滤波平滑处理，从而得到最终的定位结果，为外部应用提供所需的位置服务。示例性的，本步骤的滤波处理方法包括但不限于如均值滤波、卡尔曼滤波等多种不同的滤波方法。
在本实施例中，该定位模型可以是训练设备利用历史位置MR数据集合中每条历史位置MR数据的特征信息和该历史位置MR数据对应终端设备的位置信息得到的。具体的，训练设备可以对获取到的历史位置MR数据集合进行例如步骤22中的处理过程，从而得到每条历史位置MR数据的特征信息，再以每条历史位置MR数据的特征信息作为训练网络的输入，以每条历史位置MR数据携带的终端位置信息作为训练网络的输出来训练该定位模型。
示例性的，该训练网络可以是深度神经网络，例如，循环神经网络(recurrent neural network，RNN)，长短期记忆网络(long short-term memory，LSTM)等，本申请实施例并不对用于训练定位模型的网络进行限定。
关于训练设备训练得到定位模型的具体实现方式可以参见下述实施例中的描述，此处不再赘述。
本申请实施例提供的定位方法，通过获取包括多条MR数据的MR数据集合，对于该MR数据集合中的任意一条MR数据，根据该MR数据和在时间上与该MR数据相邻的至少一条MR数据，确定用于描述该MR数据对应终端设备所在位置的特征信息，将该MR数据的特征信息输入到定位模型中，得到该MR数据对应终端设备的位置信息，该定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以该历史位置MR数据对应终端设备的位置信息为输出训练得到的模型。该技术方案中，定位设备在定位过程中基于与该MR数据具有时间先后关系的至少一条MR数据对该MR数据进行处理，得到的MR数据的特征信息可以更加准确的表征终端设备的位置信息，从而提升了定位精度。
示例性的，在上述实施例的基础上，图3为本申请实施例提供的定位方法实施例二的流程示意图。如图3所示，在本实施例中，在上述步骤22之前，该方法还可以包括如下部分或全部步骤：
步骤31：根据该MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对该MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合。
其中，排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起。
在本申请的实施例中，由于每条MR数据均携带用户标识信息和时间戳信息，因而，定位设备对获取到的MR数据集合进行处理时，可以首先基于每条MR数据携带的用户标识信息将具有相同用户标识信息的MR数据划分到一起，再对同一个用户的MR数据按照时间戳信息进行排序，以使具有相同用户标识的MR数据按照时间戳信息排列在一起，进而得到排序处理后的MR数据集合，这样定位设备可以对排序处理后的MR数据集合中的MR数据进行特征提取，且在特征提取时将MR数据集合中每个MR数据上报的时间顺序考虑在内，为后续提取到合理、准确的特征信息奠定了基础。
可选的，通过将同一个用户的终端设备上报的MR划分到一起，当对该终端设备的每 条MR数据进行位置定位时，可以得到该终端设备的移动轨迹，进一步提升了定位精度。
步骤32：将排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合。
在本实施例中，该预设的第一时长一般指较短的时间段，例如，1s或2s等，本申请实施例并不对第一时长的具体取值进行限定。
在本实施例中，对于上述排序处理后的MR数据集合，通过将排序处理后的MR数据集合中相同时间或时间差值小于1s的MR数据进行合并，得到合并处理后的MR数据集合。
示例性的，在对多条MR数据进行合并时，对于合并得到的MR数据，其时间戳信息确定为参与合并的多条MR数据的时间戳信息的平均值；若参与合并的多条MR数据对应终端设备的主服务小区相同，则保持主服务小区不变，该主服务小区的信号电平值取所有主服务小区的信号电平值的均值，其对应的邻区去重合并；若参与合并的多条MR数据对应终端设备的主服务小区不相同，则将信号电平值最大的小区作为主服务小区，其余所有小区进行邻区去重合并。
值得说明的是，邻区去重合并时，相同小区的信号电平值取均值，并按信号电平值的大小重新确定邻区。
在本实施例中，通过将每个时刻或者每个时间段内的MR数据进行整合，可以使得每个时刻或时间段内的特征信息更完整，从而提高了MR数据对应特征信息的准确性。
示例性的，在本申请实施例的一种可能实现方式中，如图3所示，在步骤32之后，该方法还可以包括如下步骤：
步骤33：对上述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合，每个MR数据子集合包括：按照时间戳信息排序且具有相同用户标识的多条MR数据，每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，且每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长。
其中，该总时长为MR数据子集合中最后一条MR数据与第一条MR数据的时间戳信息对应的时间差值。
可选的，在本实施例中，由于合并处理后的MR数据集合包括的MR数据条数较多，该集合对应的时间维度较长，因而，为了对MR数据进行特征提取时，提高特征信息的提取效率和准确度，可以按照预设的通话提取策略对上述合并处理后的MR数据集合进行拆分处理，以得到多个MR数据子集合。其中，该通话提取策略为：每个MR数据子集合中的所有MR数据具有相同的用户标识，且每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，以及每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长。
示例性的，作为一种示例，该MR数据子集合也可以称为一段通话，该通话是指具有相同的用户标识信息且满足预设的时间约束规则的多条MR数据的集合。
示例性的，每个MR数据子集合中所有MR数据的总时长不超过180s，且相邻两条MR数据之间的时间间隔不超过30s。值得说明的是，本实施例中的第二时长、第三时长均是预设的取值，其可以根据实际情况确定，本申请实施例并不对其进行限定。
进一步的，在本申请实施例的另一种可能实现方式中，如图3所示，在步骤33之后， 该方法还可以包括如下步骤：
步骤34：对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合。
在实际应用中，终端设备从各个小区接收到的信号电平值可能会有很大的波动性，本实施例中，为了减少信号电平值的波动对定位精度的影响，得到稳定的信号电平值，通过对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合，从而在后续定位时提供定位精度。
示例性的，本实施例可以采用多种滤波方法对上述每个MR数据子集合中的MR数据进行滤波处理，例如，加权滤波、卡尔曼滤波、滑膜平均滤波等常见滤波方法。本申请实施例并不限定滤波的具体方式，其可以根据实际情况进行限定。
值得说明的是，本申请实施例中对于MR数据集合的处理步骤可以包括步骤31至步骤34中的一个或多个，具体包括的步骤可以根据实际情况确定，此处不再赘述。
本申请实施例提供的定位方法，通过根据MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合，使得排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起，然后将排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合，再对合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合，最后对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合。该技术方案中，通过对MR数据集合进行排序、合并、拆分和滤波处理，能够提高后续提取到的特征信息的准确性。
示例性的，在上述实施例的基础上，图4为本申请实施例提供的定位方法实施例三的流程示意图。如图4所示，在本实施例中，上述步骤22可以通过如下步骤实现：
步骤41：根据时间上位于该MR数据之前的至少一条MR数据、该MR数据以及时间上位于该MR数据之后的至少一条MR数据，形成该MR数据对应的MR通话序列。
在本实施例中，由于卷积神经网络可以对同性质数据进行卷积以提取数据之间的相互关联，所以，定位设备提取MR数据的特征信息时，引入时间维度后，可以利用时间上位于该MR数据之前的至少一条MR数据、该MR数据以及时间上位于该MR数据之后的至少一条MR数据形成该MR数据对应的MR通话序列，进而基于该MR通话序列提取该MR数据的特征信息。
示例性的，对于当前MR数据，联合当前MR一定时间窗内的前M条MR、后N条MR构成该MR数据对应的MR通话序列，其中，M、N均为正整数。
步骤42：基于卷积神经网络和该MR通话序列，对该MR数据进行特征提取，得到该MR数据的第一特征信息集合。
在本实施例中，由于MR数据包含的信息大体分为信号电平值和小区信息两类数据，因此，可以基于不同的指标，例如，信号电平值和小区信息，分别构造特征图用以提取特征。
示例性的，构造特征图时，1条MR数据产生n*1的向量，其中，n表示向量的长度，可以用单条MR数据的n个特征表示。对于包括t条MR数据的MR通话序列，其可以组 成n*t的向量作为特征图。
可选的，对于每条MR数据，其使用的指标包括但不限于小区标识向量、信号电平值、电平值的傅里叶变换、小区和电平值的组合等。其中，小区标识向量指对当前区域的MR数据中的小区标识进行编码(类似word2vec模型)得到的小区向量。在实际应用中，可以根据实际情况确定每条MR数据使用的指标，本实施例并不对其进行限定。
在本实施例中，基于每条MR数据对应的通话序列对该MR数据进行特征提取时，可以首先基于该通话序列构成特征图，并基于卷积神经网络自动挖掘并提取电平和小区等体现MR数据的特征信息，不同指标上的特征联合后即可得到该MR的特征信息。
值得说明的是，不同的指标使用不同的卷积神经网络来提取特征，对于分别提取出来的特征构造密集层而得到该MR通话序列上的第一特征信息集合；其中，该密集层可以是全连接层，也可以通过卷积来实现。
示例性的，在本实施例中，该第一特征信息集合可以包含与信号电平值、小区信息相关的维度特征，包括但不限于各小区的信号电平值、小区标识等。
步骤43：将该第一特征信息集合作为该MR数据的特征信息。
可选的，在本实施例中，由于该第一特征信息集合是基于该MR数据对应的MR通话序列得到的，其在特征提取的过程中将MR数据之间的时序关系考虑在内，增强了MR数据的特征信息的合理性和准确度，提高了定位精度。
可选的，在本申请实施例的一种可能设计中，如图4所示，上述步骤22还可以包括如下步骤40，在实际应用中，该步骤40可以位于上述步骤41之前或者在上述步骤42之后，示例性的，图4所示的实施例以步骤40位于步骤42之后进行举例说明：
步骤40：根据该MR数据和该MR数据对应的网络工参信息，提取该MR数据的第二特征信息集合，该MR数据对应的网络工参信息用于表示上报该MR数据的终端设备所属小区的属性信息。
可选的，在本实施例中，由于网络工参信息可以用于描述站点(基站)属性信息，其对应的网络工程参数可以包括：站点天线位置的经纬度、天线方向性、增益、方位角、下倾角、挂高、馈线型号、站点类型(室内、室外)等，因而，在本实施例中，对于该MR数据，首先确定出该MR数据对应终端设备的主服务小区的信号电平值、主服务小区标识信息、各个邻区的信号电平值和各邻区标识信息等，再通过对所有小区的信号电平值进行运算，进而确定出用于表征该MR数据对应终端设备所在小区的电平值和小区标识，进而基于该小区标识查询该MR数据对应的网络工参信息，确定出该小区所在的经纬度、高度、方向角、下倾角等信息，进而确定出用于表征该MR数据的第二特征信息集合。
可以理解的是，本申请实施例中的第一特征信息集合和第二特征信息集合表示通过两种方式得到的不同特征信息集合，并不表示先后顺序。
值得说明的是，该步骤40还可以在上述图3所示实施例包括的所有步骤中的任意一个或多个步骤之后执行，当步骤40直接位于步骤31、步骤32或者步骤34之后执行时，提取到的MR数据的第二特征信息集合即为单条MR数据的特征信息。
示例性的，在本申请的实施例的一种设计中，在该步骤40之后，该方法还可以包括：
基于栈式降噪自动编码器(stack denosing autoencoders，SDA)模型，对该第二特征表征集合中的特征信息进行特征增强。
在本实施例中，为增强单条MR数据的特征表征能力，可以基于SdA模型对提取得到的该第二特征信息集合中的特征进行特征增强。
示例性的，首先采用SDA模型对特征进行训练，再通过学习获得了特征转换增强模型，然后该第二特征信息集合中的特征经过该特征转换增强模型的处理后得到增强后的第二特征信息集合。
对于特征转换增强模型的训练过程可以解释如下：对于第二特征信息集合中的特征X，即单条MR数据的特征X首先经过编码器(Encoder)和解码器(Decoder)的处理后得到特征X’，通过无监督学习方式逐层训练当前网络的参数，使得X和X’的误差最小；当所有的自动编码器堆叠形成的网络训练完成后，在一个多层神经网络中使用之前自动编码器的中间层，然后以监督学习的方式对多层神经网络的网络权值进行微调，从而得到该特征转换增强模型。
相应的，如图4所示，上述步骤43可以替换为如下步骤431和步骤432：
步骤431：对该第一特征信息集合和该第二特征信息集合进行融合，得到该MR数据的融合特征信息。
在本实施例中，为了提高MR数据对应特征信息的完整性，可以对第一特征信息集合和第二特征信息集合中的特征信息进行融合，例如，该特征融合操作可以将两个集合中的特征直接进行连接完成，也可以通过全连接层进行转换完成。
示例性的，基于神经网络的全连接层和隐藏层对第一特征信息集合和第二特征信息集合进行连接处理，进而扩展该MR数据的特征信息的维度，其提高了后续确定该MR数据对应终端设备的位置的精度。
步骤432：将该MR数据的融合特征信息作为该MR数据的特征信息。
可选的，在本实施例中，通过将MR数据的融合特征信息，进一步增强了MR数据的特征信息的合理性和准确度，提高了定位精度。
值得说明的是，在本实施例中，若采用SDA模型对第二特征信息集合进行特征增强，并得到了增强后的第二特征信息集合，则对上述第一特征信息集合和增强后的第二特征信息集合进行融合，得到更新后的融合特征信息。相应的，可以将更新后的融合特征信息作为该MR数据的特征信息。
本申请实施例提供的定位方法，根据时间上位于该MR数据之前的至少一条MR数据、该MR数据以及时间上位于该MR数据之后的至少一条MR数据，形成MR数据对应的MR通话序列，基于卷积神经网络和该MR通话序列，对该MR数据进行特征提取，得到该MR数据的第一特征信息集合，以及根据该MR数据对应的网络工参信息，提取该MR数据的第二特征信息集合，进而对该第一特征信息集合和该第二特征信息集合进行融合，得到该MR数据的融合特征信息，最后将该MR数据的融合特征信息作为MR数据的特征信息。该技术方案既考虑了单条MR数据的特征信息，也考虑了基于该MR数据对应MR通话序列提取到的特征信息，其进一步增强了MR数据的合理性和准确度，为后续确定出准确的终端位置奠定了基础。
在本申请的上述任意一种实施例中，构建并训练定位模型是本申请提出的定位方法的核心，其可以利用训练设备和获取到的历史位置MR数据进行离线训练。
具体的，训练设备可以首先对获取到的历史位置MR数据集合进行处理(例如，排序、合并、拆分、滤波等多种处理中的一种或多种)得到每条历史位置MR数据的特征信息，再将每条历史位置MR数据的特征信息作为模型的输入，由于按照用户标识信息和时间戳信息排序得到的历史位置MR数据子集合(每个历史位置MR数据集合拆分后可以得到多个历史位置MR数据子集合)可以形成终端设备所属用户的移动轨迹，因而，基于该移动轨迹可以构建该定位模型，也即，使用上述历史位置MR数据的特征信息作为模型的输入，以历史位置MR数据对应终端设备的位置信息作为模型的输出，训练该模型的参数，从而得到每条历史位置MR数据对应目标区域内的定位模型。
本申请的实施例中，该训练设备可以运行在开源的Tensorflow机器学习平台中，基于机器学习方法提取历史位置MR数据集合中每个历史位置MR数据的特征信息，用以训练定位模型，具体的，可以运行在带有NVIDIA GPU卡的服务器上，其中，NVIDIA GPU卡通过CUDA编程接口提供计算加速能力，对特征提取过程和定位模型构建过程进行加速。
下面介绍本申请实施例中涉及的定位模型的训练方法，应理解，该定位模型可以用于基于处理得到的MR数据的特征信息预测该MR数据对应终端设备的位置，其可以离线训练得到。具体的，该定位模型的训练过程可以包括如下两种实现方式：
实现方式(1)：
示例性的，该定位模型具体是基于机器学习方法以该历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以该历史位置MR数据的位置标签作为输出训练得到的模型；该历史位置MR数据的特征信息是通过对该历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
其中，该历史位置MR数据的位置标签是对预设区域进行栅格化得到的，该预设区域包括：上报该历史位置MR数据的终端设备所属站点的位置确定的区域。
示例性的，该预设区域可以是上报该历史位置MR数据的终端设备所在的主服务小区、邻区确定的区域，本申请实施例并不对预设区域的具体实现进行限定。
应理解，在本实施例中，训练设备对历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理的具体操作与上述图3所示实施例中定位设备对MR数据集合的处理过程类似，此处不再赘述。
在该实现方式(1)中，本申请实施例基于机器学习方法，在定位过程中利用历史位置MR数据集合中相邻历史位置MR数据的时间先后信息来提取特征信息，丰富了特征信息的维度，增加不同位置下MR数据之间的区分度，基于机器学习方法(如随机森林、DNN等)构建定位模型，优化了特征之间关联关系，提升了定位精度。可选的，根据该种实现方式利用的主要特征信息，该训练方法可称为“时间-空间特征模型定位法”。
在实际现网中，通过对获取到的历史位置MR数据集合进行分析可知，若该历史位置MR数据集合表征用户的通话性较好，但该历史位置MR数据的数据量不足以组成通话序列用来训练定位模型，但其可以支撑随机森林、DNN等模型训练时，则基于该随机森林、DNN等机器学习方法来训练定位模型。
其中，MR数据的通话性较好指的是每个MR数据子集合中含有的MR数据的条数不少于8条的比例较高。
可以理解的是，在本实施例中，基于历史位置MR数据对应通话序列训练得到的定位 模型可以称为时序定位模型，基于历史位置MR数据对应位置标签训练得到的定位模型可以称为标签定位模型。也即，通过该实现方式(1)得到的定位为标签定位模型，通过下述实现方式(2)得到的定位模型为时序定位模型。
示例性的，图5为本申请实施例提供的定位方法中定位模型的实现方式(1)的流程示意图。该训练方法可以由上述图1所示定位系统中的训练设备执行。示例性的，如图5所示，该训练方法可以包括如下步骤：
步骤51：训练设备获取历史位置MR数据集合，该历史位置MR数据集合包括：多条历史位置MR数据，该历史位置MR数据是指携带位置信息的历史MR数据。
步骤52：对于该历史位置MR数据集合中的任意一条历史位置MR数据，训练设备根据该历史位置MR数据和在时间上与该历史位置MR数据相邻的至少一条历史位置MR数据，确定该历史位置MR数据的特征信息。
可选的，在本实施例中，训练设备可以通过对该历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后，再提取得到每条历史位置MR数据的特征信息。
值得说明的是，在本实施例中，训练设备对历史位置MR数据的合并、拆分、滤波处理过程与上述图3所示实施例中定位设备对MR数据集合中的MR数据的合并、拆分、滤波处理过程类似，具体可以参见上述实施例中的介绍，此处不再赘述。
步骤53：训练设备对该历史位置MR数据对应的预设区域进行栅格化得到该历史位置MR数据的位置标签。
可选的，在本实施例中，训练设备首先根据上报该历史位置MR数据的终端设备所属站点的位置，确定出预设区域，该预设区域可以是主服务小区或邻区的覆盖范围，其次对该预设区域进行栅格化得到该历史位置MR数据的位置标签。
示例性的，训练设备基于机器学习方法构建并训练定位模型时，可以做分类任务，也可以做回归任务。由于机器学习方法，例如，随机森林和DNN等模型均需要有位置标签，因而，训练设备做分类任务时，可以对预设区域进行栅格化处理，且历史位置MR数据的经纬度所在的栅格ID确定为该条历史位置MR数据的位置标签；训练设备做回归任务时，将历史位置MR的经纬度作为该条历史位置MR数据位置标签。
步骤54：训练设备将历史位置MR数据集合中每个历史位置MR数据的特征信息作为输入，以对应历史位置MR数据的位置标签作为输出训练模型，得到该定位模型。
其中，该定位模型具有对应的模型文件，该模型文件可以在定位过程中确定MR数据是否可以利用该定位模型进行位置预测。
步骤55：训练设备存储该定位模型对应的模型文件。
示例性的，训练设备可以将生成的模型文件保存到存储设备中，例如，模型存储数据库中，以备后续在定位时直接使用。
在本申请实施例中，对于获取到的历史位置MR数据集合中的任意一条历史位置MR数据，训练设备根据该历史位置MR数据和在时间上与该历史位置MR数据相邻的至少一条历史位置MR数据，确定该历史位置MR数据的特征信息，再对该历史位置MR数据对应的预设区域进行栅格化得到该历史位置MR数据的位置标签，最后将历史位置MR数据集合中每个历史位置MR数据的特征信息作为输入，以对应历史位置MR数据的位置标签作为输出训练模型，得到该定位模型。该技术方案在对历史MR数据集合进行处理时，考 虑了历史位置MR数据之间的时间关系，丰富了每个历史位置MR数据的特征信息的维度，使得该定位模型可以更有效地刻画MR数据的特征信息与该MR数据对应终端设备的位置信息之间的关系，从而提升了后续的定位精度。
实现方式(2)：
示例性的，该定位模型具体是基于该历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的，该时序关系为历史位置MR数据的时间戳信息对应的时间先后顺序。
应理解，在本实施例中，训练设备对历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理的具体操作与上述图3所示实施例中定位设备对MR数据集合的处理过程类似，此处不再赘述。
在该实现方式(2)中，本申请实施例同样基于机器学习方法，在定位过程中，利用历史位置MR数据集合中相邻历史位置MR数据的时间先后信息来构建定位模型。对于某一历史位置MR数据，该历史位置MR数据对应终端设备的位置信息不仅与该历史位置MR数据的特征信息有关，还与时间上在该历史位置MR数据前、后的历史位置MR数据的特征信息及位置有关，基于此事实，在模型训练过程中，训练设备考虑了历史位置MR数据之间的时序关系构建定位模型，提升了定位精度。可选的，根据该实现方式利用的特征信息，该训练方法可称为“时序模型定位法”。
在实际现网中，该训练方法尤其适用于历史位置MR数据的通话连续性不佳，每个历史位置MR数据的时序特征不明显，但是历史位置MR数据的数据量足以组成通话序列用来训练定位模型的场景。
示例性的，图6为本申请实施例提供的定位方法中定位模型的实现方式(2)的流程示意图。该训练方法可以由上述图1所示定位系统中的训练设备执行。示例性的，如图6所示，该训练方法可以包括如下步骤：
步骤61：训练设备获取历史位置MR数据集合，该历史位置MR数据集合包括：多条历史位置MR数据，该历史位置MR数据是指携带位置信息的历史MR数据。
步骤62：对于该历史位置MR数据集合中的任意一条历史位置MR数据，训练设备根据该历史位置MR数据和在时间上与该历史位置MR数据相邻的至少一条历史位置MR数据，确定该历史位置MR数据的特征信息。
步骤63：训练设备基于历史位置MR数据对应的历史位置MR通话序列，训练该定位模型。
在本实施例中，该历史位置MR通话序列与该历史位置MR数据所属历史位置MR数据子集合对应，因而，构建定位模型时，可以考虑历史位置MR数据的特征信息之间的时序信息，并基于递归贝叶斯估计RBE或神经网络模型训练定位模型。
示例性的，在基于RBE的方案中，对于该历史位置MR数据集合中的任意一条历史位置MR数据，假设该历史位置MR数据对应终端设备的位置信息只与该终端设备前一时刻的位置信息有关，则可以基于该马可夫(Markov)性质和贝叶斯定理，通过迭代方法对该历史位置MR数据对应终端设备的位置进行预测，其中对某一时刻某位置上的特征信息分布进行估计时，可采用粒子滤波算法或随机森林等对该分布进行模拟表达。
示例性的，在基于神经网络模型的方案中，可以采用带有时序表示能力的神经网络模 型(包括但不限于RNN、LSTM或TCN)对该历史位置MR数据对应终端设备在多个时刻的特征信息和位置信息进行建模，并根据特征信息和位置信息训练该神经网络模型的参数。
步骤64：训练设备存储该定位模型对应的模型文件。
值得说明的是，本实施例中的步骤61、步骤62以及步骤64中的实现方案可以上述图5所示实施例中的介绍，此处不再赘述。
示例性的，在定位模型的训练过程中，基于历史位置MR数据在对应历史位置MR通话序列中与其他历史位置MR数据之间的时序关系，丰富了每个历史位置MR数据的特征信息的维度，使得该定位模型可以更有效地刻画MR数据的特征信息与该MR数据对应终端设备的位置信息之间的关系，同样提升了后续的定位精度。
示例性的，本申请实施例在构建定位模型时考虑了历史位置MR数据集合中相邻历史位置MR数据的时间先后关系，该定位模型可以基于递归贝叶斯估计(recursive bayesian estimation，RBE)或神经网络模型(包括但不限于RNN、LSTM或时间卷积网络(temporal convolutional network，TCN)等方式进行构建。
相应的，在上述两种实现方式中，上述图2所示实施例中的步骤23可以通过如下方式实现，也即，定位设备首先基于该MR数据的特征信息，查询存储设备中是否存在采用上述实现方式(1)或实现方式(2)生成的模型文件，并根据查询结果，确定具体的定位方法。
示例性的，若两种实现方式生成的模型文件均不存在，则定位失败。
若存在一种实现方式生成的模型文件，则从该存储设备中获取该种实现方式对应的模型文件，生成该模型文件对应的定位模型，将上述得到的该MR数据的特征信息输入该定位模型中，利用该定位模型预测该MR数据对应终端设备的位置信息。
若两种实现方式生成的模型文件均存在，则从存储设备中获取基于实现方式(2)生成的模型文件，并基于该模型文件确定定位模型，将上述得到的该MR数据的特征信息输入该定位模型中，利用该定位模型预测该MR数据对应终端设备的位置信息。
可以理解的是，由上述分析可知，定位设备确定的该MR数据的特征信息既可以是经过图3所示实施例和图4所示实施例的各步骤处理并提取得到的融合特征信息或第一特征信息集合，也可以经过图3所示实施例的部分步骤处理，且利用图4所示实施例对单条MR数据进行特征提取得到的第二特征信息集合。
因而，当定位设备确定的MR数据的特征信息是融合特征信息或第一特征信息集合时，定位设备通过实现方式(1)或实现方式(2)得到的定位模型均可以预测得到准确的该MR数据对应终端设备的位置信息。
当定位设备确定的MR数据的特征信息是通过对单条MR数据进行特征提取得到的第二特征信息集合时，定位设备利用实现方式(2)得到的定位模型也可以预测得到准确的该MR数据对应终端设备的位置信息。
值得说明的是，基于定位模型预测MR数据对应终端设备的位置信息的实现原理类似，本申请实施例不再逐一进行介绍。
示例性的，图7为本申请实施例提供的定位方法实施例四的流程示意图。该方法的执 行主体可以是上述图1所示定位系统中的定位设备。如图7所示，该定位方法可以包括如下步骤：
步骤71：获取MR数据集合，该MR数据集合包括：多条MR数据。
步骤72：对于该MR数据集合中的任意一条MR数据，根据该MR数据和在时间上与该MR数据相邻的至少一条MR数据，确定该MR数据的特征信息，该MR数据的特征信息用于描述该MR数据对应终端设备所在位置的信息。
本实施例中的步骤71与上述图2所示实施例中的步骤21一致，步骤72与上述图2所示实施例中的步骤22一致，关于步骤71和步骤72的实现原理可参见上述图2至图4所示实施例中的记载，此处不再赘述。
步骤73：根据MR数据的特征信息，从指纹特征库中确定出与该特征信息匹配的目标指纹文件，该指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件。
在本实施例中，指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件，该指纹文件是训练设备通过对历史位置MR数据集合中每条历史位置MR数据的特征信息和对应终端设备的位置信息进行处理和关联得到的。
因而，在本实施例中，定位设备可以利用步骤72得到的该MR数据的特征信息，查询存储设备，逐一判断该存储设备的指纹特征库中是否存在与该MR数据的特征信息相匹配的目标指纹文件，若存在，则确定出该目标指纹文件，若不存在，则定位失败。
步骤74：基于该目标指纹文件和该特征信息，确定出该MR数据对应终端设备的位置信息。
在本实施例中，定位设备获取指纹特征库中的该目标指纹文件，并基于该MR数据的特征信息在该目标指纹文件对应的区域内查找与该特征信息相匹配的栅格。具体的，在查找与该特征信息匹配的栅格时，需要计算该特征信息与该区域内所有栅格的匹配度，例如，使用欧氏距离来衡量。
在本实施例中，由于该MR数据的特征信息中各参数信息的重要程度不同，因而，在计算上述匹配度时，可以引入权重，进而使用加权欧氏距离的方法来衡量匹配度，并且，欧氏距离越小，匹配度越高。对于各参数信息的权重值可以通过寻优的方法获得。
在本实施例中，将匹配度最高的栅格的经纬度作为该MR数据对应终端设备的位置信息。
可选的，在本实施例中，同样可以对确定的位置信息做滤波平滑处理，以使本方案的定位结果更合理。同理，本步骤的滤波处理方法同样包括但不限于如均值滤波、卡尔曼滤波等多种不同的滤波方法。
本申请实施例的定位方法，通过获取测量报告MR数据集合，并对于MR数据集合中的任意一条MR数据，根据该MR数据和在时间上与该MR数据相邻的至少一条MR数据，确定用于描述所述MR数据对应终端设备所在位置的特征信息，根据MR数据的特征信息，从指纹特征库中确定出与该特征信息匹配的目标指纹文件，该指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件，最后基于该目标指纹文件和特征信息，确定出该MR数据对应终端设备的位置信息。该技术方案在特征提取阶段引入了时间维度，考虑了该MR数据在对应MR通话序列与其他MR数据的时序关系，丰富了特征 信息的维度，提升了定位精度。
示例性的，在本实施例中，上述指纹特征库包括：历史位置MR数据集合中每条历史位置MR数据所在当前区域对应的指纹文件，该指纹文件用于描述该当前区域中每个栅格的特征信息与该栅格的位置信息；该历史位置MR数据的特征信息是通过对该历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
具体的，本申请实施例可以基于机器学习方法，在生成指纹特征库时，基于MR数据之间的时序关系提取特征信息，在指纹定位方案(如特征库定位)中增强各指纹信息的区分度，从而提升定位精度。基于该指纹特征库的方法可称为“时间-空间特征库定位法”。
在实际现网中，本实施例的方法尤其适用于该历史位置MR数据集合表征用户的通话性较好，但该历史位置MR数据的数据量不足以组成通话序列用来训练定位模型，且计算资源受限的场景中。MR数据的通话性较好指的是每个MR数据子集合中含有的MR数据的条数不少于8条的比例较高。
示例性的，图8为本申请实施例提供的定位方法中指纹特征库的生成方法示意图。该方法可以由上述图1所示定位系统中的训练设备执行。示例性的，如图8所示，该方法可以包括如下步骤：
步骤81：获取历史位置MR数据集合，该历史位置MR数据集合包括：多条历史位置MR数据，该历史位置MR数据是指携带位置信息的历史MR数据。
步骤82：对于该历史位置MR数据集合中的任意一条历史位置MR数据，训练设备根据该历史位置MR数据和在时间上与该历史位置MR数据相邻的至少一条历史位置MR数据，确定该历史位置MR数据的特征信息。
本实施例中的步骤81与上述图5所示实施例中的步骤51一致，步骤82与上述图5所示实施例中的步骤52一致，关于步骤81和步骤82的实现原理可参见上述图5所示实施例中的记载，此处不再赘述。
步骤83：对该历史位置MR数据对应的当前区域进行栅格化处理，确定出该当前区域包括的所有栅格。
在本实施例中，当前区域指的是该历史位置MR和需要进行定位的MR数据所在的区域。
步骤84：统计并计算每个栅格里的特征信息形成指纹文件，基于该当前区域包括的所有栅格的指纹文件构建指纹特征库。
可选的，在本实施例中，每个栅格的特征文件中各特征信息的维度值，包括但不限于平均值、最大值、最小值等可以刻画数据分布的统计值。
示例性的，该指纹文件可以包括以下内容：栅格编号、栅格经纬度、栅格上各特征信息的统计值。每个栅格形成一个指纹文件，所有栅格上的指纹文件形成指纹特征库。
步骤85：存储构建得到的指纹特征库。
示例性的，将生成的指纹特征库保存到存储设备中，以备后续在定位时直接使用。
在本实施例中，通过获取历史位置MR数据集合，对于该历史位置MR数据集合中的任意一条历史位置MR数据，训练设备根据该历史位置MR数据和在时间上与该历史位置MR数据相邻的至少一条历史位置MR数据，确定该历史位置MR数据的特征信息，对该历史位置MR数据对应的当前区域进行栅格化处理，确定出该当前区域包括的所有栅格， 统计并计算每个栅格里的特征信息形成指纹文件，基于该当前区域包括的所有栅格的指纹文件构建指纹特征库，最后存储构建得到的指纹特征库。该技术方案，基于对历史MR数据集合进行处理时，考虑了历史位置MR数据之间的时间关系，丰富了每个历史位置MR数据的特征信息的维度，使得得到的指纹特征库可以更有效地刻画MR数据的特征信息与该MR数据对应终端设备的位置信息之间的关系，从而提升了后续的定位精度。
可选的，基于上述图5、图6和图8所示的实施例，图9为本申请实施例提供的训练方法的流程示意图。如图9所示，该方法可以包括如下步骤：
步骤91：获取历史位置MR数据集合。
步骤92：判断该历史位置MR数据集是否满足训练第一种定位模型的场景条件；若是，执行步骤93，若否，执行步骤94。
其中，训练第一种定位模型的场景条件为该历史位置MR集合的数据量大于训练第一种定位模型所需的数据量，且该第一种定位模型是基于该历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的。
步骤93：基于该历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系，得到第一种定位模型。
该第一种定位模型即为上述图6所示实施例中训练得到的定位模型，关于该步骤的实现原理，具体可以参见上述图6所示实施例中的记载，此处不再赘述。
步骤94：判断该历史位置MR数据集是否满足训练第二种定位模型的场景条件；若是，执行步骤95，若否，执行步骤96。
其中，训练第二种定位模型的场景条件为该历史位置MR集合的数据量小于训练第二种定位模型所需的数据量，但该历史位置MR数据集合表征用户的通话性较好，但其可以支撑随机森林、DNN等模型训练。
步骤95：基于机器学习方法以该历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以该历史位置MR数据的位置标签作为输出，训练第二种定位模型。
可选的，该第二种定位模型即为上述图5所示实施例中训练得到的定位模型，关于该步骤的实现原理，具体可以参见上述图5所示实施例中的记载，此处不再赘述。
步骤96：基于历史位置MR数据集合中每条历史位置MR数据，确定出包括多个指纹文件的指纹特征库，每个指纹文件用于描述该当前区域中每个栅格的特征信息与该栅格的位置信息。
可选的，该指纹特征库即为上述图8所示实施例中生成的指纹特征库，关于该步骤的实现原理，具体可以参见上述图8所示实施例中的记载，此处不再赘述。
值得说明的是，在本申请的实施例中，不同模型或指纹库在本实施例中的优先级为：第一种定位模型>第二种定位模型>指纹特征库。
在本申请的实施例中，该方案在MR数据的特征提取阶段考虑了MR数据的时序关系，使得训练得到的定位模型或者指纹特征库能更好地反应MR数据与地理位置之间的关系，其次，在定位模型构建阶段或指纹特征库生成阶段，根据在实际现网中的情况，判定当前场景的适用条件，选择并构建合适的定位模型或指纹特征库，使得本申请的定位方法具有更强的自适应性，进一步提升了定位精度，且保证了MR数据的定位率。
相应的，基于图9所示的方法，定位设备基于获取到的MR数据集合中任意一条MR数据的特征信息和定位模型/指纹特征库，预测该MR数据对应终端设备的位置信息的过程中，同样以第一种定位模型大于第二种定位模型，第二种定位模型大于指纹特征库的顺序进行，关于具体的定位过程，可参见上述图2至图7所述实施例中的相关记载，此处不再赘述。
图10为本申请实施例提供的定位装置实施例一的结构示意图。该装置可以集成在定位设备中，也可以通过定位设备实现。如图10所示，该装置可以包括：获取模块101、处理模块102和定位模块103。
其中，该获取模块101，用于获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
该处理模块102，用于针对所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；
该定位模块103，用于将所述MR数据的特征信息输入到定位模型中，得到所述MR数据对应终端设备的位置信息。
其中，所述定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以所述历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，所述历史位置MR数据是指携带位置信息的历史MR数据。
示例性的，在本实施例的一种可能设计中，该处理模块102，还用于在根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息之前，根据所述MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对所述MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合；
其中，所述排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起，以及将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合。
示例性的，在本实施例中，该处理模块102，还用于在将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合之后，对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合；
其中，每个MR数据子集合包括：按照时间戳信息排序且具有相同用户标识的多条MR数据，每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，且每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长，所述总时长为所述MR数据子集合中最后一条MR数据与第一条MR数据的时间戳信息对应的时间差值。
示例性的，在本实施例中，该处理模块102，还用于在对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合之后，对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合。
示例性的，在本实施例的另一种可能设计中，该处理模块102，具有用于根据时间上位于所述MR数据之前的至少一条MR数据、所述MR数据以及时间上位于所述MR数据 之后的至少一条MR数据，形成所述MR数据对应的MR通话序列，基于卷积神经网络和所述MR通话序列，对所述MR数据进行特征提取，得到所述MR数据的第一特征信息集合，以及将所述第一特征信息集合作为所述MR数据的特征信息。
示例性的，在本实施例中，该处理模块102，还具有用于根据所述MR数据和所述MR数据对应的网络工参信息，提取所述MR数据的第二特征信息集合，以及对所述第一特征信息集合和所述第二特征信息集合进行融合，得到所述MR数据的融合特征信息，将所述MR数据的融合特征信息作为所述MR数据的特征信息；
其中，所述MR数据对应的网络工参信息用于表示上报所述MR数据的终端设备所属小区的属性信息。
本实施例的装置可用于执行图2至图4所示方法实施例的实现方案，以及图9所示实施例中的部分实现方案，具体实现方式和技术效果类似，这里不再赘述。
示例性的，在本申请实施例再一种可能设计中，所述定位模型具体是基于机器学习装置以所述历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以所述历史位置MR数据的位置标签作为输出训练得到的模型；
其中，所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
可选的，所述历史位置MR数据的位置标签是对预设区域进行栅格化得到的，所述预设区域包括：上报所述历史位置MR数据的终端设备所属站点的位置确定的区域。
在本实施例中，该定位模型的训练过程可以参见上述图5所示实施例中的记载，关于具体实现方式和技术效果类似，这里不再赘述。
示例性的，在本申请实施例又一种可能设计中，所述定位模型具体是基于所述历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的，所述时序关系为所述历史位置MR数据的时间戳信息对应的时间先后顺序。
在本实施例中，该定位模型的训练过程可以参见上述图6所示实施例中的记载，关于具体实现方式和技术效果类似，这里不再赘述。
图11为本申请实施例提供的定位装置实施例二的流程示意图。该装置同样可以集成在定位设备中，也可以通过定位设备实现。如图11所示，该装置可以包括：获取模块111、处理模块112和定位模块113。
其中，该获取模块111，用于获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
该处理模块112，用于针对所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，以及根据MR数据的特征信息，从指纹特征库中确定出与所述特征信息匹配的目标指纹文件。
其中，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息，所述指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件；
该定位模块113，用于基于所述目标指纹文件和所述特征信息，确定出所述MR数据 对应终端设备的位置信息。
本实施例的装置可用于执行图7所示方法实施例的实现方案，以及图9所示实施例中的部分实现方案，具体实现方式和技术效果类似，这里不再赘述。
示例性的，在申请的一种实施例中，该指纹特征库包括：历史位置MR数据集合中每条历史位置MR数据所在当前区域对应的指纹文件，所述指纹文件用于描述所述当前区域中每个栅格的特征信息与所述栅格的位置信息；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
在本实施例中，该指纹特征库的生成方法可以参见上述图8所示实施例中的记载，关于具体实现方式和技术效果类似，这里不再赘述。
需要说明的是，应理解以上装置的各个模块的划分仅仅是一种逻辑功能的划分，实际实现时可以全部或部分集成到一个物理实体上，也可以物理上分开。且这些模块可以全部以软件通过处理元件调用的形式实现；也可以全部以硬件的形式实现；还可以部分模块通过处理元件调用软件的形式实现，部分模块通过硬件的形式实现。
例如，处理模块可以为单独设立的处理元件，也可以集成在上述装置的某一个芯片中实现，此外，也可以以程序代码的形式存储于上述装置的存储器中，由上述装置的某一个处理元件调用并执行以上处理模块的功能。其它模块的实现与之类似。
此外这些模块全部或部分可以集成在一起，也可以独立实现。这里所述的处理元件可以是一种集成电路，具有信号的处理能力。在实现过程中，上述方法的各步骤或以上各个模块可以通过处理器元件中的硬件的集成逻辑电路或者软件形式的指令完成。
例如，以上这些模块可以是被配置成实施以上方法的一个或多个集成电路，例如：一个或多个特定集成电路(application specific integrated circuit，ASIC)，或，一个或多个微处理器(digital signal processor，DSP)，或，一个或者多个现场可编程门阵列(field programmable gate array，FPGA)等。再如，当以上某个模块通过处理元件调度程序代码的形式实现时，该处理元件可以是通用处理器，例如中央处理器(central processing unit，CPU)或其它可以调用程序代码的处理器。再如，这些模块可以集成在一起，以片上系统(system-on-a-chip，SOC)的形式实现。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行所述计算机程序指令时，全部或部分地产生按照本申请实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在可读存储介质中，或者从一个可读存储介质向另一个可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质(例如固态硬盘solid state disk(SSD))等。
图12为本申请实施例提供的定位装置实施例三的结构示意图。该定位装置可以集成在定位设备中。如图12所示，该定位装置可以包括：处理器121、存储器122、通信接口123和系统总线124，所述存储器122和所述通信接口123通过所述系统总线124与所述处理器121连接并完成相互间的通信，所述存储器122用于存储计算机执行指令，所述通信接口123用于和其他设备进行通信，所述处理器121执行所述计算机执行指令时实现如图2至图4所示方法实施例的实现方案，以及图9所示实施例中的部分实现方案。
图13为本申请实施例提供的一种训练设备的结构示意图。如图13所示，该训练设备可以包括：处理器131、存储器132、通信接口133和系统总线134，所述存储器132和所述通信接口133通过所述系统总线134与所述处理器131连接并完成相互间的通信，所述存储器132用于存储计算机执行指令，所述通信接口133用于和其他设备进行通信，所述处理器131执行所述计算机执行指令时实现如图5或图6所示方法实施例的实现方案。
图14为本申请实施例提供的定位装置实施例四的结构示意图。该装置可以集成在定位设备中。如图14所示，该装置可以包括：处理器141、存储器142、通信接口143和系统总线144，所述存储器142和所述通信接口143通过所述系统总线144与所述处理器141连接并完成相互间的通信，所述存储器142用于存储计算机执行指令，所述通信接口143用于和其他设备进行通信，所述处理器141执行所述计算机执行指令时实现如图7所示方法实施例的实现方案，以及图9所示实施例中的部分实现方案。
图15为本申请实施例提供的另一种训练设备的结构示意图。如图15所示，该训练设备可以包括：处理器151、存储器152、通信接口153和系统总线154，所述存储器152和所述通信接口153通过所述系统总线154与所述处理器151连接并完成相互间的通信，所述存储器152用于存储计算机执行指令，所述通信接口153用于和其他设备进行通信，所述处理器151执行所述计算机执行指令时实现如图8所示方法实施例的实现方案。
值得说明的是，上述图12至图15中提到的系统总线可以是外设部件互连标准(peripheral component interconnect，PCI)总线或扩展工业标准结构(extended industry standard architecture，EISA)总线等。所述系统总线可以分为地址总线、数据总线、控制总线等。为便于表示，图中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。通信接口用于实现数据库访问装置与其他设备(例如客户端、读写库和只读库)之间的通信。存储器可能包含随机存取存储器(random access memory，RAM)，也可能还包括非易失性存储器(non-volatile memory)，例如至少一个磁盘存储器。
可选的，上述的处理器可以是通用处理器，包括中央处理器CPU、网络处理器(network processor，NP)等；还可以是数字信号处理器DSP、专用集成电路ASIC、现场可编程门阵列FPGA或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
可选的，上述的存储器可以是只读存储器(read only memory，ROM)，静态存储设备，动态存储设备或者随机存取存储器(random access memory，RAM)，该存储器可以存储程序和数据。
可选的，上述处理器、存储器也可以集成在专用集成电路中，集成电路中还可以包括通信接口。专用集成电路可以是处理芯片，也可以是处理电路。其中，通信接口可以是包括无线收发的通信接口，也可以是经过其他处理电路对接收的无线信号进行处理后而输入的数字信号的接口，还可以是和其他模块进行通信的软件或硬件接口。
可选的，上述定位装置或训练设备还可以包括人工智能处理器，人工智能处理器可以是神经网络处理器(network processing unit，NPU)，张量处理器(tensor processing unit，TPU)，或者图形处理器(graphics processing unit，GPU)等一切适合用于大规模异或运算处理的处理器。人工智能处理器可以作为协处理器挂载到主CPU(Host CPU)上，由主CPU为其分配任务。人工智能处理器可以实现上述定位模型的训练方法中涉及的一种或多种运算。例如，以NPU为例，NPU的核心部分为运算电路，通过控制器控制运算电路提取存储器中的矩阵数据并进行乘加运算。
可选的，本申请实施例提供一种存储介质，所述存储介质中存储有指令，当所述指令在计算机上运行时，使得计算机执行如上述图2至图6所示方法实施例的实现方案，以及图9所示实施例中的实现方案；或者
当所述指令在计算机上运行时，使得计算机执行如上述图7和图8所示方法实施例的实现方案以及图9所示实施例中的实现方案。
本申请实施例还提供一种程序产品，所述程序产品包括计算机程序，所述计算机程序存储在存储介质中，至少一个处理器可以从所述存储介质读取所述计算机程序，所述至少一个处理器执行所述计算机程序时可实现上述图2至图6所示方法实施例的实现方案以及图9所示实施例中的实现方案；或者
所述至少一个处理器执行所述计算机程序时可实现上述图7和图8所示方法实施例的实现方案以及图9所示实施例中的实现方案。
图16为本申请实施例提供的定位系统实施例的结构示意图。如图16所示，该定位系统可以包括：定位设备161和训练设备162。该训练设备162可以与该定位设备161通信，并将训练得到的定位模型或者生成的指纹特征库发送给定位设备161，定位设备161可以利用接收到的定位模型或者指纹特征库预测MR数据对应终端设备的位置信息。
其中，该定位设备161可以是上述图2至图4或图7或图9所示实施例的定位装置；该训练设备162可以为上述图5或图6或图8中的定位设备。
在本实施例中，关于定位设备161和训练设备162的具体实现方式可参见上述实施例中的记载，此处不再赘述。
本申请中，“至少一个”是指一个或者多个，“多个”是指两个或两个以上。“和/或”，描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B的情况，其中A，B可以是单数或者复数。字符“/”一般表示前后关联对象是一种“或”的关系；在公式中，字符“/”，表示前后关联对象是一种“相除”的关系。“以下至少一项(个)”或其类似表达，是指的这些项中的任意组合，包括单项(个)或复数项(个)的任意组合。例如，a，b，或c中的至少一项(个)，可以表示：a，b，c，a-b，a-c，b-c，或a-b-c，其中，a，b，c可以是单个，也可以是多个。
可以理解的是，在本申请的实施例中涉及的各种数字编号仅为描述方便进行的区分，并不用来限制本申请的实施例的范围。
可以理解的是，在本申请的实施例中，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请的实施例的实施过程构成任何限定。

Claims (26)

  1. 一种定位方法，其特征在于，包括：
    获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
    对于所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；
    将所述MR数据的特征信息输入到定位模型中，得到所述MR数据对应终端设备的位置信息，所述定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以所述历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，所述历史位置MR数据是指携带位置信息的历史MR数据。
  2. 根据权利要求1所述的方法，其特征在于，在所述根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息之前，所述方法还包括：
    根据所述MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对所述MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合，所述排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起；
    将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合。
  3. 根据权利要求2所述的方法，其特征在于，在所述将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合之后，所述方法还包括：
    对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合，每个MR数据子集合包括：按照时间戳信息排序且具有相同用户标识的多条MR数据，每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，且每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长，所述总时长为所述MR数据子集合中最后一条MR数据与第一条MR数据的时间戳信息对应的时间差值。
  4. 根据权利要求3所述的方法，其特征在于，在所述对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合之后，所述方法还包括：
    对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，所述根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，包括：
    根据时间上位于所述MR数据之前的至少一条MR数据、所述MR数据以及时间上位于所述MR数据之后的至少一条MR数据，形成所述MR数据对应的MR通话序列；
    基于卷积神经网络和所述MR通话序列，对所述MR数据进行特征提取，得到所述MR数据的第一特征信息集合；
    将所述第一特征信息集合作为所述MR数据的特征信息。
  6. 根据权利要求5所述的方法，其特征在于，所述根据所述MR数据在时间上与所述MR数据相邻的MR数据，确定所述MR数据的特征信息，还包括：
    根据所述MR数据和所述MR数据对应的网络工参信息，提取所述MR数据的第二特 征信息集合，所述MR数据对应的网络工参信息用于表示上报所述MR数据的终端设备所属小区的属性信息；
    相应的，所述将所述第一特征信息集合作为所述MR数据的特征信息，包括：
    对所述第一特征信息集合和所述第二特征信息集合进行融合，得到所述MR数据的融合特征信息；
    将所述MR数据的融合特征信息作为所述MR数据的特征信息。
  7. 根据权利要求1-6任一项所述的方法，其特征在于，所述定位模型具体是基于机器学习方法以所述历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以所述历史位置MR数据的位置标签作为输出训练得到的模型；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
  8. 根据权利要求7所述的方法，其特征在于，所述历史位置MR数据的位置标签是对预设区域进行栅格化得到的，所述预设区域包括：上报所述历史位置MR数据的终端设备所属站点的位置确定的区域。
  9. 根据权利要求1-6任一项所述的方法，其特征在于，所述定位模型具体是基于所述历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的，所述时序关系为所述历史位置MR数据的时间戳信息对应的时间先后顺序。
  10. 一种定位方法，其特征在于，包括：
    获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
    对于所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；
    根据MR数据的特征信息，从指纹特征库中确定出与所述特征信息匹配的目标指纹文件，所述指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件；
    基于所述目标指纹文件和所述特征信息，确定出所述MR数据对应终端设备的位置信息。
  11. 根据权利要求10所述的方法，其特征在于，所述指纹特征库包括：历史位置MR数据集合中每条历史位置MR数据所在当前区域对应的指纹文件，所述指纹文件用于描述所述当前区域中每个栅格的特征信息与所述栅格的位置信息；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
  12. 一种定位装置，其特征在于，包括：获取模块、处理模块和定位模块；
    所述获取模块，用于获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
    所述处理模块，用于针对所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息；
    所述定位模块，用于将所述MR数据的特征信息输入到定位模型中，得到所述MR数据对应终端设备的位置信息，所述定位模型是以历史位置MR数据集合中每条历史位置MR数据的特征信息为输入，以所述历史位置MR数据对应终端设备的位置信息为输出训练得到的模型，所述历史位置MR数据是指携带位置信息的历史MR数据。
  13. 根据权利要求12所述的装置，其特征在于，所述处理模块，还用于在根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息之前，根据所述MR数据集合中每条MR数据携带的用户标识信息和时间戳信息，对所述MR数据集合中的所有MR数据进行排序处理，得到排序处理后的MR数据集合，所述排序处理后的MR数据集合中具有相同用户标识的MR数据按照时间戳信息排列在一起，以及将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合。
  14. 根据权利要求13所述的装置，其特征在于，所述处理模块，还用于在将所述排序处理后的MR数据集合中的具有相同用户标识信息且时间差值小于预设的第一时长的多条MR数据进行合并，得到合并处理后的MR数据集合之后，对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合；
    其中，每个MR数据子集合包括：按照时间戳信息排序且具有相同用户标识的多条MR数据，每个MR数据子集合中所有MR数据的总时长小于预设的第二时长，且每个MR数据子集合中相邻两条MR数据的时间戳信息对应的时间差值小于预设的第三时长，所述总时长为所述MR数据子集合中最后一条MR数据与第一条MR数据的时间戳信息对应的时间差值。
  15. 根据权利要求14所述的装置，其特征在于，所述处理模块，还用于在对所述合并处理后的MR数据集合进行拆分处理，得到多个MR数据子集合之后，对每个MR数据子集合中的所有MR数据进行滤波处理，得到处理后的MR数据集合。
  16. 根据权利要求12-15任一项所述的装置，其特征在于，所述处理模块，具有用于根据时间上位于所述MR数据之前的至少一条MR数据、所述MR数据以及时间上位于所述MR数据之后的至少一条MR数据，形成所述MR数据对应的MR通话序列，基于卷积神经网络和所述MR通话序列，对所述MR数据进行特征提取，得到所述MR数据的第一特征信息集合，以及将所述第一特征信息集合作为所述MR数据的特征信息。
  17. 根据权利要求16所述的装置，其特征在于，所述处理模块，还具有用于根据所述MR数据和所述MR数据对应的网络工参信息，提取所述MR数据的第二特征信息集合，以及对所述第一特征信息集合和所述第二特征信息集合进行融合，得到所述MR数据的融合特征信息，将所述MR数据的融合特征信息作为所述MR数据的特征信息；
    其中，所述MR数据对应的网络工参信息用于表示上报所述MR数据的终端设备所属小区的属性信息。
  18. 根据权利要求12-17任一项所述的装置，其特征在于，所述定位模型具体是基于机器学习装置以所述历史位置MR数据集合中每条历史位置MR数据的特征信息作为输入，以所述历史位置MR数据的位置标签作为输出训练得到的模型；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
  19. 根据权利要求18所述的装置，其特征在于，所述历史位置MR数据的位置标签是对预设区域进行栅格化得到的，所述预设区域包括：上报所述历史位置MR数据的终端设备所属站点的位置确定的区域。
  20. 根据权利要求12-17任一项所述的装置，其特征在于，所述定位模型具体是基于所述历史位置MR数据集合中所有历史位置MR数据之间的时序关系得到的特征信息和对应历史位置MR数据的位置信息的关联关系得到的，所述时序关系为所述历史位置MR数据的时间戳信息对应的时间先后顺序。
  21. 一种定位装置，其特征在于，包括：获取模块、处理模块和定位模块；
    所述获取模块，用于获取测量报告MR数据集合，所述MR数据集合包括：多条MR数据；
    所述处理模块，用于针对所述MR数据集合中的任意一条MR数据，根据所述MR数据和在时间上与所述MR数据相邻的至少一条MR数据，确定所述MR数据的特征信息，以及根据MR数据的特征信息，从指纹特征库中确定出与所述特征信息匹配的目标指纹文件，所述MR数据的特征信息是用于描述所述MR数据对应终端设备所在位置的信息，所述指纹特征库中存储有用于描述特征信息与终端设备所在位置的关联关系的指纹文件；
    所述定位模块，用于基于所述目标指纹文件和所述特征信息，确定出所述MR数据对应终端设备的位置信息。
  22. 根据权利要求21所述的装置，其特征在于，所述指纹特征库包括：历史位置MR数据集合中每条历史位置MR数据所在当前区域对应的指纹文件，所述指纹文件用于描述所述当前区域中每个栅格的特征信息与所述栅格的位置信息；所述历史位置MR数据的特征信息是通过对所述历史位置MR数据集合中的历史位置MR数据进行合并、拆分、滤波处理后提取得到的。
  23. 一种定位装置，其特征在于，包括：处理器、存储器及存储在所述存储器上并可在处理器上运行的计算机程序，所述处理器执行所述程序时实现如上述权利要求1-9任一项所述的方法。
  24. 一种定位装置，其特征在于，包括：处理器、存储器及存储在所述存储器上并可在处理器上运行的计算机程序，所述处理器执行所述程序时实现如上述权利要求10或11所述的定位方法。
  25. 一种存储介质，其特征在于，所述存储介质中存储有指令，当所述指令在计算机上运行时，使得计算机执行如上述权利要求1-9任一项所述的方法。
  26. 一种存储介质，其特征在于，所述存储介质中存储有指令，当所述指令在计算机上运行时，使得计算机执行如上述权利要求10或11所述的方法。
PCT/CN2019/129333 2019-04-25 2019-12-27 定位方法、装置及存储介质 WO2020215783A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910339092.9A CN111867049B (zh) 2019-04-25 2019-04-25 定位方法、装置及存储介质
CN201910339092.9 2019-04-25

Publications (1)

Publication Number Publication Date
WO2020215783A1 true WO2020215783A1 (zh) 2020-10-29

Family

ID=72941512

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/129333 WO2020215783A1 (zh) 2019-04-25 2019-12-27 定位方法、装置及存储介质

Country Status (2)

Country Link
CN (1) CN111867049B (zh)
WO (1) WO2020215783A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112218330A (zh) * 2020-11-19 2021-01-12 中国联合网络通信集团有限公司 定位方法及通信装置
CN113141616A (zh) * 2021-04-20 2021-07-20 博瑞得科技有限公司 O+b域数据+业务场景自适应识别的节能基站选择及节能模式选取的方法、装置和系统
CN113301646A (zh) * 2021-05-21 2021-08-24 恒安嘉新（北京）科技股份公司 一种定位方法、装置、电子设备及存储介质
CN113453334A (zh) * 2021-06-02 2021-09-28 中国联合网络通信集团有限公司 定位方法及定位装置
CN113810992A (zh) * 2021-11-15 2021-12-17 广州睿帆科技有限公司 数据处理方法及装置
CN115426713A (zh) * 2022-09-01 2022-12-02 内蒙古大学 一种基于图-时间卷积网络的室内定位方法及系统
CN116680468A (zh) * 2022-09-28 2023-09-01 荣耀终端有限公司 指纹库生成方法及电子设备
CN116761249A (zh) * 2022-09-23 2023-09-15 荣耀终端有限公司 室内定位方法、指纹库的构建方法、电子设备及存储介质

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114040494A (zh) * 2020-12-07 2022-02-11 北京市天元网络技术股份有限公司 一种mr数据用户的定位方法及系统
CN113207169B (zh) * 2021-03-26 2022-08-05 福建新大陆软件工程有限公司 一种基于多源异构数据的融合定位方法
CN115802282B (zh) * 2022-12-16 2024-06-07 兰笺(苏州)科技有限公司 无线信号场的协同定位方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20130065283A (ko) * 2011-12-09 2013-06-19 한국전자통신연구원 핸드오버 제어 방법, 무선 통신 단말의 동작 방법, 그리고 기지국
US8644818B1 (en) * 2010-10-14 2014-02-04 Cisco Technology, Inc. Radio access point device location verification based on wireless user device roaming history
CN108495364A (zh) * 2018-02-09 2018-09-04 北京天元创新科技有限公司 移动终端的定位方法及装置
CN108513251A (zh) * 2018-02-13 2018-09-07 北京天元创新科技有限公司 一种基于mr数据的定位方法及系统
CN108574984A (zh) * 2017-03-10 2018-09-25 中国移动通信集团上海有限公司 一种定位方法及装置

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101790189A (zh) * 2009-12-29 2010-07-28 广州杰赛科技股份有限公司 一种移动台定位的方法
CN106804046B (zh) * 2017-02-16 2019-12-03 广州杰赛科技股份有限公司 基于测量报告的移动台定位方法及装置
CN109151890A (zh) * 2017-06-19 2019-01-04 中国移动通信集团浙江有限公司 一种移动终端定位方法及装置
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
CN108901039B (zh) * 2018-07-24 2021-07-20 中国联合网络通信集团有限公司 交通线路区域中小区覆盖状况确定方法及装置
CN109600792B (zh) * 2019-02-01 2021-09-07 中南民族大学 一种lte mr数据的定位方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8644818B1 (en) * 2010-10-14 2014-02-04 Cisco Technology, Inc. Radio access point device location verification based on wireless user device roaming history
KR20130065283A (ko) * 2011-12-09 2013-06-19 한국전자통신연구원 핸드오버 제어 방법, 무선 통신 단말의 동작 방법, 그리고 기지국
CN108574984A (zh) * 2017-03-10 2018-09-25 中国移动通信集团上海有限公司 一种定位方法及装置
CN108495364A (zh) * 2018-02-09 2018-09-04 北京天元创新科技有限公司 移动终端的定位方法及装置
CN108513251A (zh) * 2018-02-13 2018-09-07 北京天元创新科技有限公司 一种基于mr数据的定位方法及系统

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112218330A (zh) * 2020-11-19 2021-01-12 中国联合网络通信集团有限公司 定位方法及通信装置
CN112218330B (zh) * 2020-11-19 2023-11-03 中国联合网络通信集团有限公司 定位方法及通信装置
CN113141616B (zh) * 2021-04-20 2022-07-29 博瑞得科技有限公司 O+b域数据+业务场景自适应识别的节能基站选择及节能模式选取的方法、装置和系统
CN113141616A (zh) * 2021-04-20 2021-07-20 博瑞得科技有限公司 O+b域数据+业务场景自适应识别的节能基站选择及节能模式选取的方法、装置和系统
CN113301646A (zh) * 2021-05-21 2021-08-24 恒安嘉新（北京）科技股份公司 一种定位方法、装置、电子设备及存储介质
CN113453334A (zh) * 2021-06-02 2021-09-28 中国联合网络通信集团有限公司 定位方法及定位装置
CN113453334B (zh) * 2021-06-02 2023-02-17 中国联合网络通信集团有限公司 定位方法及定位装置
CN113810992B (zh) * 2021-11-15 2022-02-11 广州睿帆科技有限公司 数据处理方法及装置
CN113810992A (zh) * 2021-11-15 2021-12-17 广州睿帆科技有限公司 数据处理方法及装置
CN115426713A (zh) * 2022-09-01 2022-12-02 内蒙古大学 一种基于图-时间卷积网络的室内定位方法及系统
CN116761249A (zh) * 2022-09-23 2023-09-15 荣耀终端有限公司 室内定位方法、指纹库的构建方法、电子设备及存储介质
CN116761249B (zh) * 2022-09-23 2024-04-12 荣耀终端有限公司 室内定位方法、指纹库的构建方法、电子设备及存储介质
CN116680468A (zh) * 2022-09-28 2023-09-01 荣耀终端有限公司 指纹库生成方法及电子设备
CN116680468B (zh) * 2022-09-28 2024-03-22 荣耀终端有限公司 指纹库生成方法及电子设备

Also Published As

Publication number Publication date
CN111867049B (zh) 2021-11-19
CN111867049A (zh) 2020-10-30

Similar Documents

Publication Publication Date Title
WO2020215783A1 (zh) 定位方法、装置及存储介质
Xia et al. Federated-learning-based client scheduling for low-latency wireless communications
US11762396B2 (en) Positioning system and positioning method based on WI-FI fingerprints
Cui et al. Received signal strength based indoor positioning using a random vector functional link network
US9092532B2 (en) Method and server for searching for nearby user in social networking services
Dai et al. Combination of DNN and improved KNN for indoor location fingerprinting
Luo et al. Enhancing Wi-Fi fingerprinting for indoor positioning using human-centric collaborative feedback
CN112218330B (zh) 定位方法及通信装置
CN103068035A (zh) 一种无线网络定位方法、装置及系统
CN108243451A (zh) 一种基于mr数据在td-lte栅格中定位的方法和系统
Aranda et al. Performance analysis of fingerprinting indoor positioning methods with BLE
CN112867147B (zh) 定位方法和定位装置
WO2019134549A1 (zh) 基于深度学习的定位方法、装置、计算机设备及存储介质
CN110298687B (zh) 一种区域吸引力评估方法及设备
CN112214677B (zh) 一种兴趣点推荐方法、装置、电子设备及存储介质
CN106465175A (zh) 用于收集和汇总网络质量数据的方法
Liu et al. Large-scale deep learning framework on FPGA for fingerprint-based indoor localization
CN114430530B (zh) 空间划分方法、装置、设备、介质和程序产品
Li et al. A novel method of Wi-Fi indoor localization based on channel state information
CN106470435B (zh) 识别WiFi群的方法及系统
Górak et al. Indoor localisation based on GSM signals: Multistorey building study
CN113055901B (zh) 一种终端定位的方法和装置
CN110619090B (zh) 一种区域吸引力评估方法及设备
CN112235723B (zh) 定位方法、装置、电子设备及计算机可读存储介质
Sukemi et al. Path Loss Prediction Accuracy Based on Random Forest Algorithm in Palembang City Area

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: 19926356

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19926356

Country of ref document: EP

Kind code of ref document: A1