CN113038370B - Offline fingerprint library construction method, position fingerprint positioning method and system - Google Patents
Offline fingerprint library construction method, position fingerprint positioning method and system Download PDFInfo
- Publication number
- CN113038370B CN113038370B CN202110246161.9A CN202110246161A CN113038370B CN 113038370 B CN113038370 B CN 113038370B CN 202110246161 A CN202110246161 A CN 202110246161A CN 113038370 B CN113038370 B CN 113038370B
- Authority
- CN
- China
- Prior art keywords
- fingerprint
- edge computing
- mobile edge
- clustering
- target point
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000010276 construction Methods 0.000 title claims abstract description 17
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 239000013598 vector Substances 0.000 claims description 48
- 238000005070 sampling Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000003860 storage Methods 0.000 abstract description 11
- 230000005540 biological transmission Effects 0.000 abstract description 3
- 238000000926 separation method Methods 0.000 abstract description 3
- 238000003064 k means clustering Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/023—Services making use of location information using mutual or relative location information between multiple location based services [LBS] targets or of distance thresholds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- 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
- H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention provides an offline fingerprint library construction method, a position fingerprint positioning method and a system, wherein the offline fingerprint library construction method comprises the steps of respectively deploying a mobile edge computing server at each wireless access point device, and the method comprises the following steps: obtaining a set number of clusters and cluster centers by adopting a clustering algorithm for the obtained fingerprint data; and storing the data in each class into a corresponding mobile edge computing server according to the position of the initial clustering center to construct and complete the database separation of each offline fingerprint database. According to the invention, the MEC server is deployed at the AP equipment, and the preprocessing, storage and matching tasks of the fingerprint data are 'sunk' to the MEC server to finish, so that the MEC server is very close to a user, and the transmission delay of the fingerprint data can be greatly reduced. By using the MEC server, storage and computation pressures of the cloud server and the mobile terminal can be relieved.
Description
Technical Field
The invention belongs to the technical field of positioning, relates to a position fingerprint positioning method, and in particular relates to an offline fingerprint library construction method, a position fingerprint positioning method and a position fingerprint positioning system.
Background
With the increasing maturity of wireless network technology, location awareness is one of the more important services, and indoor positioning technology plays an increasing role in our daily lives. Technology for realizing indoor positioning by utilizing wireless signals is also mature, and existing algorithms are roughly divided into two types: one is to estimate the distance by parameters, common types of parameters are time of arrival (TOA), time difference of arrival (TDOA), angle of arrival (AOA), etc.; the other is to use a position fingerprint to locate, wherein the position fingerprint refers to a wireless signal power value which can reflect the current propagation environment and has multipath characteristics. Conventional fingerprint positioning algorithms generally involve two phases: an off-line stage of generating a fingerprint database and an on-line stage of implementing positioning. The information of the fingerprint library is filled by the position fingerprint information of a plurality of sampling points, and the position fingerprint information of the sampling points comprises two pieces of information: the MAC address and its corresponding received signal strength (RECEIVED SIGNAL STRENGTH, RSS). In the online stage, the fingerprint information of the current position of the user is matched with the fingerprint information of the fingerprint library through certain matching algorithms, and the current position is calculated, so that positioning is realized.
Along with the maturation of WIFI technology, most of large-scale indoor environments have now all arranged the WIFI focus, can cover almost all intelligent terminal products, such as cell-phone, dull and stereotyped, intelligent wrist-watch etc. people can easily access the WIFI focus through these equipment and acquire the service, and more researchers have thrown the research of WIFI fingerprint indoor location with the eye. In the research of fingerprint positioning, fingerprint library establishment modes and position estimation algorithms are more involved, and the position estimation algorithms comprise deterministic algorithms, non-deterministic algorithms, machine learning algorithms and the like. Common positioning algorithms are K-Nearest Neighbor (KNN), support vector machine (Support Vector Machine, SVM), random Forest (RF), etc.
In a large indoor environment, to ensure the accuracy of fingerprint positioning, more WIFI signal transmitting devices and more reference points are generally required, which directly results in a substantial increase in the data size of an offline fingerprint database, while fingerprint positioning generally requires a mobile device to transmit fingerprint data to a remote cloud server, and the cloud server stores the fingerprint data. Besides bringing huge pressure to the storage of fingerprint data of a cloud server, calculation time delay of fingerprint matching is increased, so that instantaneity is reduced, and positioning effect is affected.
Disclosure of Invention
Aiming at the technical problems that the existing fingerprint positioning method brings great pressure to the storage of fingerprint data by a cloud server and increases the calculation time delay of fingerprint matching, the invention provides an offline fingerprint library construction method, a position fingerprint positioning method and a position fingerprint positioning system.
In one aspect, the present invention provides a method for constructing an offline fingerprint library, in which mobile edge computing servers are deployed at each wireless access point device, the method comprising the steps of: obtaining a set number of clusters and cluster centers by adopting a clustering algorithm for the obtained fingerprint data; and storing the data in each class into a corresponding mobile edge computing server according to the position of the initial clustering center to construct and complete the database separation of each offline fingerprint database.
Further, a K-clustering algorithm is adopted for the obtained fingerprint data to obtain a set number of clusters and cluster centers, and the method specifically comprises the following steps:
Step 101: taking k pieces of fingerprint data randomly from m pieces of fingerprint data as initial clustering centers, namely { C 1,C2,…,Ck };
step 102: traversing all fingerprint data, respectively calculating Euclidean distances between the corresponding reference point positions and the clustering centers of k clusters, and classifying the fingerprint data into the class of the clustering center with the minimum Euclidean distance;
Step 103: recalculating the cluster center of each cluster, namely replacing the initial cluster center by the average value of all data in each class to serve as a new cluster center of the cluster, thereby completing an iterative operation;
Step 104: repeating the steps 102 and 103 until the iteration times reach the set maximum iteration times, or the cluster center positions of all the classes are converged, namely the deviation between the current calculated cluster center position of each cluster and the cluster center position calculated last time is smaller than a set threshold value, so that the whole clustering operation is completed;
Step 105: the new cluster and its cluster center { C 1 *,C2 *,…,Ck * } are output.
Further, the data in each class is stored in a corresponding mobile edge computing server according to the position of the initial clustering center, and the database of each offline fingerprint database is built, specifically comprising the following steps:
Step 201: taking the position information of n wireless access point devices as initial cluster centers of n clusters, wherein C= { C 1,C2,…,Cn}＝{L1,L2,…,Ln }, C 1,C2,…,Cn is a fixed language of the cluster centers, L 1,L2,…,Ln is each reference point, and each cluster center corresponds to one mobile edge computing server;
Step 202: traversing m reference positions by a mobile equipment end, respectively calculating Euclidean distances between the current reference point position and n clustering centers, finding out a clustering center with the minimum Euclidean distance, transmitting a sampling value of the current reference point to a mobile edge computing server corresponding to the clustering center, classifying all the reference points into n different classes, and transmitting RSS sampling information of the reference points of the same class to the same mobile edge computing server;
Step 203: the mobile edge computing server preprocesses the reference point sampling data in the corresponding class, and after preprocessing the reference point sampling data by all the mobile edge computing servers is completed, one-time clustering operation is completed;
step 204: each mobile edge computing server recalculates the clustering center of the corresponding subclass, takes the average value of the positions of all the reference points contained in the subclass as a new clustering center, and sets the RSS vector of the new clustering center as the average value of the RSS vectors of all the reference points in the subclass;
Step 205: each mobile edge computing server updates and stores the n new clustering centers, performs clustering operation again on the reference points in the classes according to Euclidean distance rules, and if the reference points are clustered and then are classified into other classes, the current mobile edge computing server transmits data to the corresponding mobile edge computing server;
step 206: repeating the steps 204 and 205 until the cluster center change value of each class is smaller than a set threshold value or the cluster iteration number reaches a set maximum iteration number;
step 207: and outputting the RSS vector of the clustering center of each class, and storing the RSS vector at the mobile device.
In a second aspect, the present invention provides a location fingerprint positioning method, comprising the steps of:
the method for constructing the off-line fingerprint library provided by any one of the possible embodiments of the technical scheme is adopted to construct sub-libraries of the off-line fingerprint library;
each mobile edge computing server is matched with a database constructing an offline fingerprint database based on the RSS vector of the target point, and the matching result of each mobile edge computing server is weighted to obtain the final estimated position of the target point.
Further, the method for obtaining the final estimated position of the target point comprises the following steps:
Acquiring an RSS vector of a target point, and determining similarity weights of the target point and each cluster center based on the RSS vector of the target point and the RSS vector of each cluster center;
Sorting the similarity weight values, comparing the maximum similarity weight with a set maximum weight threshold, and if the maximum similarity weight is greater than the set maximum weight threshold, setting the maximum similarity weight as a first weight (alternatively set to 1 in the embodiment), and setting the other similarity weights as a second weight (alternatively set to 0 in the embodiment); otherwise, comparing the similarity weights with the set minimum weight threshold values sequentially from large to small, setting all the similarity weights smaller than the set minimum weight threshold value as second weights if the similarity weights are smaller than the set minimum weight threshold value, and recalculating the similarity weights larger than or equal to the set minimum weight threshold value;
if the similarity weight of the target point and each cluster center is the second weight, the mobile edge computing server storing the cluster center does not perform position matching operation on the target point;
If the similarity weight of the target point and each cluster center is not the second weight, the mobile edge computing server storing the corresponding cluster center performs position matching operation on the target point to obtain a matching result; and carrying out weighted calculation on the matching results of the mobile edge calculation servers to obtain the final estimated position.
Still further, the method for determining the similarity weight of the target point and each cluster center comprises the following steps: the similarity between the RSS vector of the target point and the ith clustering center is calculated by using the Euclidean distance D i of the RSS vector, and the expression is as follows:
wherein rsti is the RSS vector of the target point, n is the number of elements of the RSS vector of the target point and is also the number of categories, RSS vector, i=1, 2, …, n,
Converting the Euclidean distance D i into similarity weight of the class and the target point, wherein the expression is as follows:
further, the expression of the similarity weight equal to or larger than the set minimum weight threshold is recalculated as follows:
Wherein W t represents similarity weights each smaller than a set minimum weight threshold.
Further, the expression of the final estimated position is as follows:
The matching result of the corresponding mobile edge computing server is:
L＝{L1,L2,…,Lr}＝{(x1,y1),(x2,y2),…,(xr,yr)},
where L 1,L2,…,Ln is each reference point, (x 1,y1),(x2,y2),…,(xr,yr) is the positional information of each reference point.
In a third aspect, the present invention provides a location fingerprint positioning system, each of which deploys a mobile edge computing server at each wireless access point device, the system comprising: the system comprises an offline fingerprint database construction module, a position fingerprint positioning module and an estimated position determining module, wherein the offline fingerprint database construction module is used for obtaining a set number of clusters and cluster centers for obtained fingerprint data by adopting a clustering algorithm; storing the data in each class into a corresponding mobile edge computing server according to the position of the initial clustering center to construct and complete the sub-library of each off-line fingerprint library;
The position fingerprint positioning module is deployed in each mobile edge computing server and is used for matching with a database which is stored by the RSS vector of the target point and is used for constructing an offline fingerprint database;
the estimated position determining module is used for weighting the matching result of each mobile edge computing server to obtain the final estimated position of the target point.
The beneficial technical effects obtained by the invention are as follows:
According to the invention, the MEC server is deployed at the AP equipment, and the preprocessing, storage and matching tasks of the fingerprint data are 'sunk' to the MEC server to finish, so that the MEC server is very close to a user, and the transmission delay of the fingerprint data can be greatly reduced. Through using MEC server, can alleviate cloud server and mobile terminal's storage and calculation pressure to MEC server security is higher, and the fingerprint data of only part of storage, and leakage risk is little, has still avoided the risk of data in the way of transmitting to the high in the clouds.
According to the method, when an offline fingerprint library is constructed, the fingerprint library is divided according to the thought of a K-means algorithm, and initial clustering is set as the AP position so as to improve the clustering effect. Finally, in an online positioning stage, an adaptive weight matching algorithm based on similarity is provided, a set threshold value is improved to increase a weight value for clustering, and adaptive change is carried out according to the position condition of a target point, so that the positioning accuracy of the clustering matching algorithm is improved.
Drawings
FIG. 1 is a flow chart of a conventional fingerprint positioning process;
fig. 2 is a schematic view of the basic architecture of the MEC system;
FIG. 3 is a schematic diagram of an offline fingerprint library construction method according to an embodiment;
FIG. 4 is a diagram of a fingerprint library splitting principle based on a K-means algorithm in an embodiment;
FIG. 5 is a schematic diagram of a target point at a class junction;
FIG. 6 is a flowchart of a weight process in a position fingerprint positioning method according to an embodiment;
fig. 7 is a schematic diagram of a MEC-based fingerprint positioning process in a position fingerprint positioning method according to an embodiment.
Detailed Description
The invention is further described below with reference to the drawings and specific examples.
For a better understanding of the present invention, a conventional fingerprint locating manner is described below.
It is assumed that, in the target positioning area, the number of wireless AP (Access Point) devices available for indoor positioning is n, and the number of sampling points (position reference points) included in the fingerprint database is m, and a specific positioning model is shown in fig. 1. In the off-line stage, the RSS values received by different sampling points from different AP devices are stored to construct a fingerprint database, RSS i,j is the RSS value received by the jth sampling point from the ith AP, and the position of each sampling point is represented by coordinates and corresponds to a unique RSS vector. In the online stage, after the RSS vector line at the current position is properly processed, a specific position matching algorithm is adopted to match with m fingerprint information in a fingerprint library, and the current position of the user is calculated.
The traditional fingerprint positioning mode is dependent on the computing power of the mobile device and the remote server, and more AP devices and reference positions are introduced when a fingerprint library is generated in an offline state in order to achieve higher precision. The large amount of fingerprint data brings larger pressure to the mobile equipment and the cloud application server, and the operation time delay of fingerprint matching in an online stage is increased.
The invention combines the traditional centralized data processing flow with the mobile network by adopting the mobile edge calculation, and processes the data by the mobile edge calculation (Mobile Edge Computing, MEC) server before the data is transmitted to the core network, namely, the storage and calculation capacity of the far end of the original center is reduced to the MEC server (the basic structure diagram of the MEC system is shown in figure 2), so that the MEC server takes on most or all of the roles of data calculation, storage and communication.
Example 1: the method for constructing the offline fingerprint library comprises the following steps: in the off-line stage, before the mobile equipment end transmits a plurality of sampling values at the current reference point to the MEC server, the fingerprint data is split by a clustering algorithm, namely, the reference point sampling data is transmitted to different MEC servers according to a clustering rule. The construction principle of the off-line fingerprint database is shown in figure 3, the number of fingerprint database sub-databases is c, the number of reference points of each sub-database is { n 1,n2,…,nc }, and
In the positioning model after the clustering operation is added, a clustering operation is newly added in a task list of the MEC server, the classified storage of fingerprint data is completed, and the specific flow of the offline fingerprint library construction is as follows:
Input: n pieces of AP device location information, m pieces of reference point location information, and the number of cluster centers c=n.
And (3) outputting: and finally, n sub-fingerprint libraries and corresponding n sub-fingerprint library clustering centers.
Step1: the position information of n AP devices is used as an initial cluster center of n clusters, C= { C 1,C2,…,Cn}＝{L1,L2,…,Ln }, each cluster center corresponds to one MEC server, and the nearest reference point to the AP devices is used as the initial cluster center.
Step2: the mobile equipment end traverses m reference positions, respectively calculates Euclidean distances between the current reference point position and n clustering centers, finds out a clustering center with the minimum Euclidean distance, transmits a sampling value of the current reference point to a MEC server corresponding to the clustering center, divides all the reference points into n different classes, and sends RSS sampling information of the reference points of the same class to the same MEC server.
Step3: the MEC server preprocesses the reference point sampling data in the corresponding class, and after preprocessing work of all MEC servers on the reference point sampling data is completed, one-time clustering operation is completed.
Step4: each MEC server recalculates the clustering center of the corresponding subclass, takes the average value of the positions of all the reference points contained in the subclass as a new clustering center, and sets the RSS vector of the new clustering center as the average value of the RSS vectors of all the reference points in the subclass, wherein the calculation formula is as follows:
Wherein C i (i=1, …, n) is the cluster center of the ith class, and n i is the number of reference points contained in the ith class.
Wherein, RSS c i is the RSS vector value of the cluster center of the ith class, ni is the number of reference points contained in the ith class, RSS j i is the RSS vector of the jth reference point in the ith class, and n is the AP number.
Step5: and updating and storing the n new clustering centers by each MEC server, carrying out clustering operation again on the reference points in the classes according to Euclidean distance rules, and transmitting data to the corresponding MEC servers by the current MEC server if the reference points are clustered and then are classified into other classes.
Step6: and (3) repeating the step (4) and the step (5) until the cluster center change value of each class is smaller than a set threshold value or the cluster iteration number reaches a set maximum iteration number, and then the clusters tend to be stable.
Step7: and outputting the RSS vector of the clustering center of each class, and storing the RSS vector at the mobile device.
At this time, the database separation operation of the offline fingerprint database is finished, and the similarity of the RSS vectors of the reference points stored by each MEC server is higher, and the difference of the RSS vectors of the reference points between different MEC servers is larger by adding a clustering algorithm.
Implementation 2: based on the embodiment 1, the embodiment adopts a K-means clustering algorithm to obtain a sub fingerprint library clustering center. K-means clustering, namely K-means clustering, wherein in the position fingerprint, the Euclidean distance is used as a similarity standard by the K-means algorithm, the reference points are classified into K classes, and the similarity of the reference points of different classes is lower. Assume that m reference points { L 1,L2,…,Lm } are present in the current whole area, and the number of APs used for creating the fingerprint offline fingerprint library is n and { AP 1,AP2,…,APn }, respectively. The coordinates of the reference point L i are (xi, y i), and the corresponding fingerprint is RSS i＝{rss1,i,rss2,i,…,rssn,i. The specific flow of K-means clustering is as follows:
input: the number k (1 < k < m+1) of the classes and the offline fingerprint data m.
And (3) outputting: k clusters and cluster centers satisfying the condition.
Step1: k are randomly taken from m pieces of fingerprint data as initial cluster centers, namely { C 1,C2,…,Ck }.
Step2: traversing all fingerprint data, respectively calculating Euclidean distances between the corresponding reference point positions and the clustering centers of k clusters, and dividing the fingerprint data into the class of the clustering center with the smallest Euclidean distance.
Step3: the cluster center of each cluster is recalculated, that is, in each class, the average value of all the data is used to replace the initial cluster center as a new cluster center of the cluster, and an iterative operation is completed.
Step4: repeating Step2 and Step3 until the iteration times reach the set maximum iteration times, or the cluster center positions of all the classes are converged, namely, the deviation between the current calculated cluster center position of each cluster and the last calculated cluster center position is smaller than a set threshold value, and completing the whole clustering operation.
Step5: the new cluster and its cluster center { C 1 *,C2 *,…,Ck * } are output.
The K-means clustering algorithm has simple principle, can reduce the calculated amount in the online matching stage, and is more suitable for the traversal query problem of a large amount of data. However, the initial clustering center is selected randomly, and the situation that the initial clustering similarity is high can occur, so that the clustering result is not ideal enough, and the final positioning effect is affected.
Because the selection of the initial clustering center can influence the clustering effect, the deployment positions of the AP equipment are far away from each other in order to improve the difference of fingerprint data. Assuming that the number of the AP devices participating in the offline fingerprint library is n, the deployed positions of the AP devices are L＝{L1,L2,…,Ln}＝{(x1,y1),(x2,y2),…,(xn,yn)},, the k value is set to be the number n of MEC servers, the initial clustering center is set at the deployed positions of the n AP devices, and the difference of the initial clustering centers is improved. After the clustering operation is completed on the data of the offline fingerprint library, the data in each class is stored in the corresponding MEC server according to the position of the initial clustering center, and the fingerprint library splitting principle after the clustering operation is added is shown in fig. 4.
When the reference points in the target positioning area are denser, the application range of the matching algorithm can be reduced through the K-means clustering algorithm, and the calculation amount of the server is reduced. However, when the points at the class boundary are processed, if the similarity between the target point and the cluster centers of the classes is similar, only one class can be selected to execute the matching algorithm, so that the error is larger.
As shown in fig. 5, the target point 1 is located at the intersection of the class a and the class C, and the cluster centers of the target point 1 and the class a are close according to the euclidean distance, so the belonging class is class a, and the belonging class of the target point B is class D. In practice, however, the distance between the target point 1 and the target point 2 is relatively close, the similarity is relatively high, but the attribution classes are different, and the similarity between the class a reference point and the class D reference point is relatively low, so that the position estimated values of the final target point 1 and the target point 2 are greatly different.
Considering existence of these junction points, the location fingerprint positioning method provided in the following embodiment 3 selects an MEC server performing a matching operation by adopting a self-adaptive weight calculation mode based on fingerprint similarity, performs a secondary calculation on the weight by setting a threshold value, does not perform a matching operation on the MEC server corresponding to the class with the weight of 0, and finally performs a weighting process on the matching result.
Implementation 3: the position fingerprint positioning method comprises the following steps: the method for constructing the off-line fingerprint library is adopted to construct sub-libraries of the off-line fingerprint library;
each mobile edge computing server is matched with a database constructing an offline fingerprint database based on the RSS vector of the target point, and the matching result of each mobile edge computing server is weighted to obtain the final estimated position of the target point. The specific processing logic is as follows:
Input: maximum weight threshold T high, minimum weight threshold T low.
Step1: after preprocessing is completed on the target point sampling data, the RSS vector of the target point is r= { RSS 1,rss2,…,rssn }, the RSS vector of the cluster center of the ith (i=1, 2, …, n) class is RSS c i＝{rssc i,1,rssc i,2,…,rssc i,n }, and the similarity between the target point and the ith cluster center is calculated by using the euclidean distance D i of the RSS vector:
step2: converting the Euclidean distance between the clustering center and the target point into similarity weight between the class and the target point, wherein the similarity weight conversion formula is shown in formula (4):
Step3: the weight values are ordered from big to small, w= { W 1,W2,…,Wn }, W 1 is compared with a maximum weight threshold T high, then the smaller weight value is compared with a minimum weight threshold T low, and the weight processing logic flow is as shown in fig. 6:
If the maximum weight is greater than T high, it indicates that the target point is not at the boundary of the class and has high similarity with the reference point in the class corresponding to the maximum weight, and at this time, the weight is set to 1, and other weight values are set to 0. If the weight value is less than T low, the similarity between the current position and the current position is extremely low, and the weight value less than T low is set to 0. When the weight is recalculated, the calculation rule is as follows:
Step4: the server corresponding to the class with the similarity weight value not being 0 performs the position matching operation on the target point, and the server with the weight value being 0 does not perform the matching operation.
Step5: and carrying out weighted calculation on the matching results of the MEC servers to obtain a final estimated position. Assuming that r w= { W 1,W2,…,Wr }, the weight value of which is not 0 after the weight processing, the matching result of the corresponding MEC server is L＝{L1,L2,…,Lr}＝{(x1,y1),(x2,y2),…,(xr,yr),},, and the final position estimation value is:
According to the position fingerprint positioning method provided by the embodiment, the MEC server is deployed at the AP equipment, and the fingerprint data preprocessing task and the cloud server position matching task of the mobile equipment are submitted to the MEC server to finish the task. If the data of the offline fingerprint library is transmitted to the cloud server for storage when the MEC server finishes data processing, the data of the fingerprint library is further acquired from the cloud when the MEC server finishes fingerprint matching later. The embodiment can store the data of the offline fingerprint database at the MEC server, so that fingerprint data is not required to be transmitted to the cloud server in the fingerprint positioning process, a positioning task is localized, the transmission time delay of the offline fingerprint database from the mobile device to the cloud server is greatly shortened, and the fingerprint positioning flow of the new model is shown in fig. 7.
Example 4: a location fingerprinting positioning system, each deploying a mobile edge computing server at each wireless access point device, the system comprising: the system comprises an offline fingerprint database construction module, a position fingerprint positioning module and an estimated position determining module, wherein the offline fingerprint database construction module is used for obtaining a set number of clusters and cluster centers for obtained fingerprint data by adopting a clustering algorithm; storing the data in each class into a corresponding mobile edge computing server according to the position of the initial clustering center to construct and complete the sub-library of each off-line fingerprint library;
The position fingerprint positioning module is deployed in each mobile edge computing server and is used for matching with a database which is stored by the RSS vector of the target point and is used for constructing an offline fingerprint database;
And the estimated position determining module is used for weighting the matching result of each mobile edge computing server to obtain the final estimated position of the target point.
It should be noted that, for convenience and brevity of description, specific working processes of the system, apparatus, unit or module described in the present application may refer to corresponding processes in the foregoing method, and thus are not described in detail in the present document.
The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the claims, which are all within the protection of the present invention.
Claims (7)
1. The method for constructing the off-line fingerprint library is characterized in that mobile edge computing servers are deployed at each wireless access point device respectively, and comprises the following steps:
step 1: obtaining a set number of clusters and cluster centers by adopting a clustering algorithm for the obtained fingerprint data;
step 2: storing the data in each class into a corresponding mobile edge computing server according to the position of the initial clustering center to construct and complete the sub-library of each off-line fingerprint library;
The method comprises the steps of obtaining a set number of clusters and cluster centers from the obtained fingerprint data by adopting a K-clustering algorithm, and specifically comprises the following steps:
Step 101: taking k pieces of fingerprint data randomly from m pieces of fingerprint data as initial clustering centers, namely { C 1, C2, …, Ck };
Step 102: traversing all fingerprint data, respectively calculating Euclidean distances between the corresponding reference point positions and the clustering centers of k clusters, and classifying the fingerprint data into the class of the clustering center with the minimum Euclidean distance;
Step 103: re-calculating the cluster center of each cluster, namely replacing the initial cluster center with the average value of all data in each class to serve as a new cluster center of the cluster, thereby completing an iterative operation;
Step 104: repeating the steps 102 and 103 until the iteration times reach the set maximum iteration times, or the cluster center positions of all the classes are converged, namely the deviation between the current calculated cluster center position of each cluster and the cluster center position calculated last time is smaller than a set threshold value, so that the whole clustering operation is completed;
step 105: outputting a new cluster and a cluster center { C 1 *, C2 *, …, Ck * };
the step2 specifically comprises the following steps:
Step 201: taking the position information of n wireless access point devices as initial cluster centers of n clusters, wherein C= { C 1,C2,…,Cn}={L1,L2,…,Ln }, each cluster center corresponds to a mobile edge computing server, C 1,C2,…,Cn is the cluster center of n clusters, and L 1,L2,…,Ln is each reference point;
Step 202: traversing m reference positions by a mobile equipment end, respectively calculating Euclidean distances between the current reference point position and n clustering centers, finding out a clustering center with the minimum Euclidean distance, transmitting a sampling value of the current reference point to a mobile edge computing server corresponding to the clustering center, classifying all the reference points into n different classes, and transmitting RSS sampling information of the same class of reference points to the same mobile edge computing server;
Step 203: the mobile edge computing server preprocesses the reference point sampling data in the corresponding class, and after preprocessing the reference point sampling data by all the mobile edge computing servers is completed, one-time clustering operation is completed;
step 204: each mobile edge computing server recalculates the clustering center of the corresponding subclass, takes the average value of the positions of all the reference points contained in the subclass as a new clustering center, and sets the RSS vector of the new clustering center as the average value of the RSS vectors of all the reference points in the subclass;
Step 205: each mobile edge computing server updates and stores the n new clustering centers, performs clustering operation again on the reference points in the classes according to Euclidean distance rules, and if the reference points are clustered and then are classified into other classes, the current mobile edge computing server transmits data to the corresponding mobile edge computing server;
step 206: repeating the steps 204 and 205 until the cluster center change value of each class is smaller than a set threshold value or the cluster iteration number reaches a set maximum iteration number;
step 207: and outputting the RSS vector of the clustering center of each class, and storing the RSS vector at the mobile device.
2. The position fingerprint positioning method is characterized by comprising the following steps of:
Constructing a sub-library of an offline fingerprint library by adopting the offline fingerprint library construction method as claimed in claim 1;
Each mobile edge computing server is matched with a database constructing an offline fingerprint database based on the RSS vector of the target point, and the matching result of each mobile edge computing server is weighted to obtain the final estimated position of the target point.
3. The position fingerprint positioning method according to claim 2, wherein the method of obtaining the final estimated position of the target point comprises the steps of:
Acquiring an RSS vector of a target point, and determining similarity weights of the target point and each cluster center based on the RSS vector of the target point and the RSS vector of each cluster center;
Sorting the similarity weight values, comparing the maximum similarity weight with a set maximum weight threshold, and setting the maximum similarity weight as a first weight and the other similarity weights as second weights if the maximum similarity weight is larger than the set maximum weight threshold; otherwise, comparing the similarity weights with the set minimum weight threshold values sequentially from large to small, setting all the similarity weights smaller than the set minimum weight threshold value as second weights if the similarity weights are smaller than the set minimum weight threshold value, and recalculating the similarity weights larger than or equal to the set minimum weight threshold value;
if the similarity weight of the target point and each cluster center is the second weight, the mobile edge computing server storing the cluster center does not perform position matching operation on the target point;
if the similarity weight of the target point and each cluster center is not the second weight, the mobile edge computing server storing the cluster center carries out position matching operation on the target point; and carrying out weighted calculation on the matching results of the mobile edge calculation servers to obtain the final estimated position.
4. A method of locating a position fingerprint according to claim 3,
The method for determining the similarity weight of the target point and each cluster center comprises the following steps:
the similarity between the RSS vector of the target point and the ith cluster center is calculated by using the Euclidean distance D i of the RSS vector, and the expression is as follows:
，
wherein rsti is the RSS vector of the target point, n is the number of elements of the RSS vector of the target point and is also the number of categories, RSS vector, i=1, 2, …, n, which is the cluster center of the i-th class;
Converting the Euclidean distance D i into similarity weight of the class and the target point, wherein the expression is as follows:
。
5. A position fingerprint positioning method according to claim 3, wherein the expression of recalculating the similarity weight equal to or greater than the set minimum weight threshold is as follows:
，
Wherein D i represents the Euclidean distance of the RSS vector for the similarity of the RSS vector of the target point and the ith clustering center, D j represents the Euclidean distance of the RSS vector for the similarity of the RSS vector of the target point and the jth clustering center, i=1, 2, …, n, n is the number of elements of the RSS vector of the target point, Representing similarity weights each less than a set minimum weight threshold.
6. A position fingerprinting positioning method according to claim 3, wherein the expression of the final estimated position is as follows:
，
wherein, it is assumed that r w= { W 1, W2, …,Wr } with weight value not being the second weight after the weight processing, the matching result of the corresponding mobile edge computing server is L={L1, L2, …,Lr}={(x1, y1), (x2, y2),…, (xr, yr)},
Where L 1,L2,…,Ln is each reference point, (x 1, y1), (x2, y2),…, (xr, yr) is the positional information of each reference point.
7. A location fingerprint positioning system, employing the offline fingerprint library construction method of claim 1 and the location fingerprint positioning method of claim 2, wherein mobile edge computing servers are deployed at each wireless access point device, respectively, comprising: the system comprises an offline fingerprint database construction module, a position fingerprint positioning module and an estimated position determining module, wherein the offline fingerprint database construction module is used for obtaining a set number of clusters and cluster centers for obtained fingerprint data by adopting a clustering algorithm; storing the data in each class into a corresponding mobile edge computing server according to the position of the initial clustering center to construct and complete the sub-library of each off-line fingerprint library;
The position fingerprint positioning module is deployed in each mobile edge computing server and is used for matching with a database which is stored by the RSS vector of the target point and is used for constructing an offline fingerprint database;
And the estimated position determining module is used for weighting the matching result of each mobile edge computing server to obtain the final estimated position of the target point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110246161.9A CN113038370B (en) | 2021-03-05 | 2021-03-05 | Offline fingerprint library construction method, position fingerprint positioning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110246161.9A CN113038370B (en) | 2021-03-05 | 2021-03-05 | Offline fingerprint library construction method, position fingerprint positioning method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113038370A CN113038370A (en) | 2021-06-25 |
| CN113038370B true CN113038370B (en) | 2024-05-14 |
Family
ID=76468136
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110246161.9A Active CN113038370B (en) | 2021-03-05 | 2021-03-05 | Offline fingerprint library construction method, position fingerprint positioning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113038370B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105044662A (en) * | 2015-05-27 | 2015-11-11 | 南京邮电大学 | Fingerprint clustering multi-point joint indoor positioning method based on WIFI signal intensity |
| CN105372628A (en) * | 2015-11-19 | 2016-03-02 | 上海雅丰信息科技有限公司 | Wi-Fi-based indoor positioning navigation method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9237418B2 (en) * | 2013-07-26 | 2016-01-12 | Marvell World Trade Ltd. | Positioning database management applied to server-based positioning system |
| KR101606382B1 (en) * | 2014-12-04 | 2016-03-25 | 현대모비스 주식회사 | Method for building database of fingerprinting positioning and fingerprinting position method using the built database |
| US10670728B2 (en) * | 2018-11-05 | 2020-06-02 | Mapsted Corp. | Method and system for crowd-sourced trusted-GPS region for mobile device localization |
-
2021
- 2021-03-05 CN CN202110246161.9A patent/CN113038370B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105044662A (en) * | 2015-05-27 | 2015-11-11 | 南京邮电大学 | Fingerprint clustering multi-point joint indoor positioning method based on WIFI signal intensity |
| CN105372628A (en) * | 2015-11-19 | 2016-03-02 | 上海雅丰信息科技有限公司 | Wi-Fi-based indoor positioning navigation method |
Non-Patent Citations (3)
| Title |
|---|
| Dynamic Scheduling Strategy of Smart Manufacture System;Kai Zhao;《2019 4th International Conference on Communication and Information Systems (ICCIS)》;20200729;全文 * |
| K-means++和WKNN自适应指纹算法在井下定位***中的应用研究;刘夏;莫树培;罗浩;陈明;;计算机应用与软件;20200612(06);全文 * |
| 复杂室内环境下的WiFi定位技术研究;陆音;《计算机科学》;20161115;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113038370A (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bae et al. | Large-scale indoor positioning using geomagnetic field with deep neural networks | |
| CN108540929B (en) | Indoor fingerprint positioning method based on RSSI signal strength sequencing | |
| CN110113709B (en) | Ultra-wideband indoor positioning error elimination algorithm based on support vector machine | |
| CN105120479B (en) | The signal intensity difference modification method of terminal room Wi-Fi signal | |
| Zhang et al. | An efficient machine learning approach for indoor localization | |
| CN110602651A (en) | Positioning method based on WIFI position fingerprint and positioning system of robot | |
| CN107703480A (en) | Mixed kernel function indoor orientation method based on machine learning | |
| CN109991591B (en) | Positioning method and device based on deep learning, computer equipment and storage medium | |
| CN102480678A (en) | Fingerprint positioning method and system | |
| CN111079780A (en) | Training method of space map convolution network, electronic device and storage medium | |
| Elesawi et al. | Hierarchical multi-building and multi-floor indoor localization based on recurrent neural networks | |
| Zhang et al. | TIPS: transformer based indoor positioning system using both CSI and DoA of WiFi signal | |
| Hsu et al. | An adaptive Wi-Fi indoor localisation scheme using deep learning | |
| Chen et al. | A wifi indoor localization method based on dilated cnn and support vector regression | |
| CN113038370B (en) | Offline fingerprint library construction method, position fingerprint positioning method and system | |
| Wei et al. | IoT-aided fingerprint indoor positioning using support vector classification | |
| Yang et al. | Research on high precision indoor positioning method based on low power Bluetooth technology | |
| Tang et al. | Wi-Fi fingerprint localization using RSSI-probability radio map and AP weight clustering | |
| Zhou et al. | Integrated location fingerprinting and physical neighborhood for WLAN probabilistic localization | |
| Yan et al. | Indoor position method of industrial robot based on wifi fingerprint position technology | |
| Guo et al. | A hybrid indoor positioning algorithm for cellular and Wi-Fi networks | |
| Wu et al. | Research on RSS based indoor location method | |
| Quezada-Gaibor et al. | Lightweight Hybrid CNN-ELM Model for Multi-building and Multi-floor Classification | |
| CN115226027A (en) | WiFi indoor fingerprint positioning method and device | |
| Nie et al. | Joint access point fuzzy rough set reduction and multisource information fusion for indoor Wi-Fi positioning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |