CN108462939A - A kind of indoor orientation method of earth magnetism Time-Series analysis - Google Patents
A kind of indoor orientation method of earth magnetism Time-Series analysis Download PDFInfo
- Publication number
- CN108462939A CN108462939A CN201810253960.7A CN201810253960A CN108462939A CN 108462939 A CN108462939 A CN 108462939A CN 201810253960 A CN201810253960 A CN 201810253960A CN 108462939 A CN108462939 A CN 108462939A
- Authority
- CN
- China
- Prior art keywords
- data
- earth magnetism
- characteristic point
- time series
- test set
- 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.)
- Granted
Links
Classifications
-
- 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
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- 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
- H04W4/025—Services making use of location information using location based information parameters
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The present invention relates to a kind of indoor orientation method of earth magnetism Time-Series analysis, step is:Indoor environment is subjected to multipath division, and at the uniform velocity acquires the earth magnetism time series data of each path, earth magnetism time series data is pre-processed, obtains training set;The characteristic point in training set is found out, and data sectional is carried out to earth magnetism time series data according to characteristic point;Using the collected earth magnetism time series data of mobile phone in position fixing process as test set, using raising frequency strategy to the earth magnetism time series sample data in test set into row interpolation;Using the collected data of mobile phone in position fixing process as test set, the characteristic point in test set is identified, and match it with the characteristic point in training set using sorting algorithm;It calculates the distance accumulation matrix of the data of characteristic point segmentation and finds the matched position of test set, realize indoor positioning.The present invention is segmented time series data using characteristic point, and similarity is calculated after classification and obtains user location, greatly improves the time complexity of algorithm.
Description
Technical field
The present invention relates to a kind of indoor positioning technologies, specially a kind of indoor orientation method of earth magnetism Time-Series analysis.
Background technology
Indoor locating system (Indoor Positioning System, IPS) refers to using mobile device between floors
The system for collecting radio wave, magnetic field, acoustic signals, or other heat transfer agents to position object or person.With the need of user
It asks by profound excavation, the visual field of people is also gradually walked close in the application based on indoor positioning technologies.Either in megastore
Or underground parking, or even the unmanned van operated automatically in warehouse are all built upon on indoor positioning technologies basis
's.Indoor positioning technologies belong to basic technology, and the precision and timeliness of positioning can strong influence user experiences.Although mesh
Preceding have some commercialized systems in the market, but the standard that IPS still neither ones are unified, and needs to rely on external set
It applies, positioning accuracy is poor.
Invention content
For the deficiencies of indoor locating system in the prior art needs dependence outside plant, positioning accuracy is poor, the present invention wants
The technical issues of solution, is to provide a kind of indoor orientation method based on earth magnetism Time-Series analysis, and outside plant is not depended on to reach,
Improve the target of positioning accuracy.
In order to solve the above technical problems, the technical solution adopted by the present invention is:
A kind of indoor orientation method of earth magnetism Time-Series analysis of the present invention, includes the following steps:
1) indoor environment is subjected to multipath division, and at the uniform velocity acquires the earth magnetism time series data of each path, when to earth magnetism
Ordinal number obtains training set according to being pre-processed;
2) characteristic point in training set is found out, and data sectional is carried out to earth magnetism time series data according to characteristic point;
3) using the collected earth magnetism time series data of mobile phone in position fixing process as test set, using raising frequency strategy to test set
In earth magnetism time series sample data into row interpolation;
4) using the collected data of mobile phone in position fixing process as test set, the characteristic point in test set is identified, and make
It is matched with the characteristic point in training set with sorting algorithm;
5) it calculates the distance accumulation matrix of the data of characteristic point segmentation and finds the matched position of test set, it is indoor fixed to realize
Position.
Carrying out data sectional according to characteristic point in step 2) is:
Using bottom-up algorithm, time series is divided into the short arrangement set of consecutive points first;
Adjacent point is connected, the data point in sequence is fallen on line segment;
Two neighboring line segment is connected, every line segment includes 3 data points, then calculates intermediate point error of fitting;
According to intermediate point error of fitting error identifying, minimum and error is less than the segmentation of threshold values R, includes 3 as first
The line segment of point;
Calculate the data after the 2~N articles segmentation successively by above-mentioned steps, all data sectionals are complete in training is gathered
At.
It is into row interpolation to earth magnetism time series data using raising frequency strategy in step 3):
When the frequency acquisition of testing mobile phone is less than the frequency acquisition of training set, test set is artificially increased using interpolation method
Sample size, if training set is indicated with Q, sample size m, test set is indicated with C, sample size n, by test set sample
This quantity extended to from n with m be an order of magnitude sample size.
It corresponding with the characteristic point in training set done matching and be using sorting algorithm in step 4):
Data are handled from the angle of earth magnetism time series data, according to contain in data GEOMAGNETIC CHANGE rule (assuming that
1) it converts Feature Points Matching to speech recognition problem, using dynamic time warping algorithm, obtains the match point in training data.
The rule contained in data is:Indoor environment is divided into mulitpath, the number of paths under a scene is N,
Set of paths P={ p1, p2…pNIndicate;The physical distance of two paths this indicated with D (i, j), the similar journey of data
Degree is indicated with S (i, j);Under Same Scene, if D (i, j) is smaller, S (i, j) is bigger, path distance farther out, then data
Tendency similarity it is different.
The distance accumulation matrix of the data of calculating characteristic point segmentation is in step 5):M is one according to earth magnetism time series
The matrix of the m*n of structure, Cumulative Distance matrix are Mc, wherein Mc(0,0)=M (0,0), (0,0) are first data of matrix,
The namely starting point of initial alignment;The Cumulative Distance calculation formula of other positions is as follows:
Mc(i, j)=Min (MC(i-1,j-1),Mc(i-1,j),Mc(i,j-1),Mc(i,j))+M(i,j)
Wherein i, j be matrix in data call number, 0<i<=m-1,0<j<=n-1.Using obtained Cumulative Distance as
The similarity function of two earth magnetism sequences.
The invention has the advantages that and advantage:
1. the method for the present invention is positioned by the matched mode of time series data, it is contemplated that the direct matched time is multiple
Miscellaneous Du Taigao, invention defines characteristic points, are segmented to time series data using characteristic point, and similarity is calculated after classification and is obtained
User location is obtained, the time complexity of algorithm is greatly improved.
Description of the drawings
Fig. 1 is that the result in the present invention using Feature Points Matching illustrates;
Fig. 2 is the average positioning accuracy curve graph positioned in scene 1 using different mobile phones in the present invention;
Fig. 3 is the average positioning accuracy curve graph positioned in scene 2 using different mobile phones in the present invention;
Fig. 4 is to be write music using the average positioning accurate that different mobile phones are positioned in scene 1 after raising frequency strategy in the present invention
Line chart;
Fig. 5 is to be write music using the average positioning accurate that different mobile phones are positioned in scene 2 after raising frequency strategy in the present invention
Line chart;
Fig. 6 is the average positioning time curve graph positioned in scene 1 using different algorithms in the present invention;
Fig. 7 is the average positioning time curve graph positioned in scene 2 using different algorithms in the present invention;
Fig. 8 is characteristic point segmentation algorithm flow chart in the present invention;
Fig. 9 is raising frequency policing algorithm flow chart in the present invention;
Figure 10 is location algorithm flow chart in the present invention.
Specific implementation mode
The present invention is further elaborated with reference to the accompanying drawings of the specification.
A kind of indoor orientation method of earth magnetism Time-Series analysis of the present invention, includes the following steps:
1) indoor environment is subjected to multipath division, and at the uniform velocity acquires the earth magnetism time series data of each path, when to earth magnetism
Ordinal number obtains training set according to being pre-processed;
2) characteristic point in training set is found out, and data sectional is carried out to earth magnetism time series data according to characteristic point;
3) using the collected earth magnetism time series data of mobile phone in position fixing process as test set, using raising frequency strategy to test set
In earth magnetism time series sample data into row interpolation;
4) using the collected data of mobile phone in position fixing process as test set, the characteristic point in test set is identified, and make
It is matched with the characteristic point in training set with sorting algorithm;
5) it calculates the distance accumulation matrix of the data of characteristic point segmentation and finds the matched position of test set, it is indoor fixed to realize
Position.
In the method for the present invention, it is thus necessary to determine that the parameter of some initialization, it is therefore desirable to acquire out corridor in a manner of at the uniform velocity
Earth magnetism time series data, earth magnetism characteristic point is then marked out by artificial mode.User can be random later in corridor row
It walks, the program of client can also detect whether characteristic point appearance in real time.Once characteristic point occurs, then Feature Points Matching is called
Algorithm is compared with the characteristic point in training data before.Since characteristic point is all obvious, the standard of sorting phase
True rate is relatively high.It is exactly that the position of user substantially is determined to determine characteristic point substantially, next, can be by real time data and instruction
Practice data to be matched, obtains the match point in training data.Due to being the side using uniform motion when acquiring training data
Where formula can extrapolate current user location in the case where sensor frequency acquisition is always fixed precondition.
In the methods of the invention, structure earth magnetism map is an essential process, and builds earth magnetism map and then mean
It and a large amount of manpower is needed to go to acquire indoor plan view and sensing data.Positioning requirements are higher, and the data of acquisition will also be got over
It is dense, need the time expended also more.And in some specific scenes such as corridor, passageway etc., the advantage of earth magnetism map is simultaneously
It cannot bring into play completely, because the direction of advance of pedestrian in such a scenario is substantially unidirectional, most earth magnetism
There is no used figure information.And since the width in corridor, passageway is generally not too large, user more concerned be oneself along
Corridor, passageway direction position.That is, in such a scenario, map is by original two dimensional surface stipulations to one-dimensional
Path.Therefore, present invention firstly provides a kind of based on the indoor orientation method of Time-Series analysis to cope with user in this scene
Location requirement.
The method of the present invention is assumed and is defined as follows to indoor environment:
Assuming that 1:Indoor environment is divided into mulitpath, the number of paths under a scene is N, set of paths P=
{p1, p2…pNIndicate.The physical distance of two paths indicates with D (i, j), the similarity degree present invention of data with S (i,
J) it indicates.
Assuming that 2:Under Same Scene, if D (i, j) is smaller, S (i, j) is bigger, if path distance farther out, counts
According to tendency similarity it is different.
Assuming that 3:Under the smaller indoor scene of width, there are N items in the path that user may walk, then the road in indoor environment
Diameter can be expressed as { P1, P2…PN}.According to the requirement of positioning accuracy, it is specific that model can define N according to the specific width in corridor
The numerical value of (arbitrary integer for being not less than 1).The run trace of user can only switch in the path defined, without departing from this
Range.That is, if the current path of user is Pi, at a time, user may be switched to path PjOn.
Assuming that 4:Each paths can be according to feature points segmentation at K section, then t-th of tunnel of the i-th paths
Segment table is shown as Pi t。
Define 1:Earth magnetism time series.Two earth magnetism sequences are respectively Q (training set) and C (test set), the wherein length of Q
Length for m, C is n.So sequence Q can be expressed as Q=q1q2…qm, same C can be expressed as C=c1c2…cn。
Define 2:Earth magnetism sequence distance matrix.For two sequences defined in 1, a sequence distance matrix M, square are built
Element in battle array is very simple, is exactly the Euclidean distance of point-to-point in corresponding sequence.It is specific as shown in formula 1.
Define 3:Sequential Cumulative Distance matrix.M is the matrix of a m*n, next will calculate Cumulative Distance matrix Mc,
Mc(0,0) is equal to M (0,0), and the Cumulative Distance computational methods of other positions are as shown in formula 2.
Mc(i, j)=Min (MC(i-1,j-1),Mc(i-1,j),Mc(i,j-1),Mc(i,j))+M(i,j)(2)
The similarity degree M of two sequencesc(m, n) is indicated.Entire matched process is exactly in distance matrix, from a left side
A path for leading to the lower right corner is looked at upper angle so that the sum of element of process minimum.If sum of the distance is smaller, illustrate two
Sequence is more similar.Use Cumulative Distance as the similarity function of sequence in the present invention.
Define 4:Sequential similarity.The index of the similarity degree of ordinal series when similarity function is for describing.If two
The distance between paths D (i, j) is smaller, then S (i, j) is bigger.Likewise, if path distance farther out, then it is assumed that data
Tendency similarity is different.
Define 5:Time sequence characteristic point.Characteristic point is some readily identified points for having obvious characteristic, the ith feature in sequence
Point FiIt can be expressed asWherein, mi-1That indicate is previous characteristic point Fi-1The ground magnetic value at place,It indicates
The sample size of current signature point and upper characteristic point difference, miIndicate Fi-1The ground magnetic value at place.
As shown in figure 8, being according to characteristic point progress data sectional in step 2):
Using bottom-up algorithm, time series is divided into the short arrangement set of consecutive points first;
Adjacent point is connected, the data point in sequence is fallen on line segment;
Two neighboring line segment is connected, every line segment includes 3 data points, then calculates intermediate point error of fitting;
According to intermediate point error of fitting error identifying, minimum and error is less than the segmentation of threshold values R, includes 3 as first
The line segment of point;
The line segment (data after being segmented) for calculating the 2~N articles short data sequence successively by above-mentioned steps, until training set
All data sectionals are completed in conjunction.
In the present embodiment, in order to be split to time series data, used it is bottom-up (use data pyramid structure,
Will all bottom datas pairing after be grouped again, to a group re-grading, top layer be the data that cannot divide again) algorithm by data divide
Section.Time series is divided into the short arrangement set of consecutive points first.At this time without error of fitting, because adjacent point is connected
Come, the data point in sequence is all fallen on line segment.Next two neighboring line segment is connected, every line segment includes three at this time
A data point at this time calculates intermediate error of fitting again.Error of fitting generally selects the sum of Euclidean distance.The line of all three points
After the error of fitting of intermediate point in section is calculated, the segmentation of error identifying minimum and error less than threshold values R, as first
Include the line segment of three points, and so on.Shown in the process of execution such as algorithm 1 (Fig. 8).
It is into row interpolation to earth magnetism time series data using raising frequency strategy in step 3):
When the frequency acquisition of testing mobile phone is less than the frequency acquisition of training set, test set is artificially increased using interpolation method
Sample size, if training set is indicated with Q, sample size m, test set is indicated with C, sample size n, by test set sample
This quantity extended to from n with m be an order of magnitude sample size.
During matched, algorithm first has to identify the characteristic point of current test intensive data, then feature
Point is matched to the position of corresponding characteristic point in training set, needs that characteristic point is described, and one three has been selected in the present invention
Dimensional vector carrys out Expressive Features point as defined 3.
Matching process is substantially exactly the process classified, and the characteristic point in training set is substantially exactly several in sorting algorithm
A class, algorithm are correctly assigned to the characteristic point in test set in corresponding classification.
Any sorting algorithm all cannot be guaranteed 100% accuracy rate, and the Data Matching of algorithm depends on characteristic point
It is matching as a result, if algorithm accidentally in the Feature Points Matching that goes out current detection to the path of mistake, the result right and wrong of algorithm
Often bad, the also therefore failure of entire position fixing process.So in k nearest neighbor sorting algorithm model, algorithm adds fault tolerant mechanism.
During classification, algorithm retains classification results, that is, possible classification results set, and the quantity of element does not surpass in set
Cross the threshold value k of setting.During matched, algorithm all matches the element in set with training set.If k settings
When, in this way since ensure that and can find correct temporal sequence.
The data acquiring frequency of the included sensor of different smart mobile phones is different, therefore under same path
The length of the data sequence of acquisition is also to make a world of difference.Data Matching algorithm can solve the problems, such as that data are misaligned, but this
Also when being built upon the length of training data sequence and sequence of test data and be not much different.If data sequence magnitude differs
Excessive, then algorithm cannot ensure to make correct matching, the precision of positioning can also substantially reduce.
On the indoor path of same, the data volume of two mobile phones acquisition often differs hundreds of datas.Frequency acquisition
Difference can lead to the missings of some trend, or even can effect characteristics point judgement, the final strong influence precision of positioning.
And when acquiring training data, it is contemplated that often frequency of use higher mobile phone acquisition the problem of positioning accuracy.Therefore,
The algorithm of the present invention will provide a kind of raising frequency strategy to cope with this demand.
When the frequency acquisition of testing mobile phone is less than the frequency acquisition of training set, what the present invention took is that artificial increase is tested
The sample size of collection, mainly using a kind of method of interpolation.If training set is indicated with Q, sample size m, test set
It is indicated with C, sample size n, algorithm is an order of magnitude to do is to extend to test set sample size from n and with m
Sample size.Shown in the implementation procedure of algorithm such as algorithm 2 (Fig. 9).
It corresponding with the characteristic point in training set done matching and be using sorting algorithm in step 4):
Data are handled from the angle of earth magnetism time series data, according to the GEOMAGNETIC CHANGE rule contained in data by feature
Point matching is converted into speech recognition problem, using dynamic time warping algorithm, obtains the match point in training data.
Algorithm proposed by the present invention is handled data from the angle of time series data.Pass through the rule to containing in data
The summary of rule has used dynamic time warping calculation with being then very natural as soon as converting problem to the problem of speech recognition
Method.And by the practical measurement of multipath, present invention discover that the characteristic of " trend is similar " is not to be permanently present.And calculate sequence
The complexity of row Cumulative Distance is excessively high, is not appropriate for the localization method online as one, therefore, the present invention takes full advantage of ground
Some characteristics of magnetic hill value give the definition of characteristic point, and use sorting algorithm by the spy in training data and test data
Sign point matches, and substantially can determine that out position of the user location in entire scene.Then pass through the side of calculating Cumulative Distance
Method, correspondence of the user location relative to training data in test data in acquisition, in conjunction with the mark on training data, most
User location is determined eventually.Shown in step such as algorithm 3 (Figure 10):
In the present invention, algorithm it needs to be determined that some initialization parameter, it is therefore desirable to corridor is acquired out in a manner of at the uniform velocity
Earth magnetism time series data, earth magnetism characteristic point is then marked out by artificial mode.User can be random later in corridor row
It walks, the program of client can also detect whether characteristic point appearance in real time.Once characteristic point occurs, then k nearest neighbor feature is called
Point matching algorithm is compared with the characteristic point in training data before.Since characteristic point is all obvious, rank of classifying
The accuracy rate of section is relatively high.It is exactly that the position of user substantially is determined to determine characteristic point substantially, next, will can in real time count
It is matched according to training data, obtains the match point in training data.Due to being to use at the uniform velocity to transport when acquiring training data
Dynamic mode can extrapolate current user location at which in the case where sensor frequency acquisition is always fixed precondition
In.
The present embodiment selects Samsung Note3 smart mobile phones and Meizu Pro6 mobile phones, this two mobile phones to be based on respectively
Android4.4 systems and Android6.0 systems, that data storage is selected is mysql.Python has work in data processing
Tool enriches and the simple advantage of grammer, therefore the python language that data processing uses.Equally, in order to as keep compatible as possible
Property, build the parameter that web services are transmitted for receiving client using the flask kits in python language.Test number
All it is to be realized using the libraries matplotlib according to figure.Experimental data is as shown in table 1.The data mainly used are that the earth magnetism of three axis passes
Sensor is read and the reading of accelerometer simultaneously adds timestamp, and the precision of sensing data is all to retain after decimal point 6, when
Between the precision stabbed be 0.1s.
1 experimental data of table
In step 1), the present invention is tested in two different indoor scenes respectively.Scene 1 and scene 2 are all divided into 5
Paths, and use the geomagnetic data in the acquisition path of robot car at the uniform velocity.Concrete scene parameter is as shown in table 2:
2 concrete scene parameter of table
It is bottom-up to find out characteristic point using algorithm 1 in step 2), 8 characteristic points, scene 2 are had found in scene 1
Have found 14 characteristic points.So scene 1 can be divided into 45 earth magnetism tracts, and scene 2 can be divided into 75 sequences
Section.
In step 3), two kinds of mobile phone collecting test collection, the wherein frequency of Samsung mobile phone acquisition geomagnetic data have been used about
It is the one third of Meizu mobile phone.Therefore, interpolation is carried out to data using the raising frequency strategy in algorithm 2, ensures to adopt after interpolation
The sequence length of collection is the same, and the sequence length of scene 1 and scene 2 is 637 and 1538 respectively.
It in step 4), is recorded using the characteristic point in training set as classification, the data in test set is assigned into training
In the classification for collecting characteristic point.The nearest neighbor algorithm that the present invention uses, the k selections 3 in algorithm.Traditional k nearest neighbor algorithms export straight
Connect is that the classification of arest neighbors is also recorded for the classification of time neighbour in order to ensure the fault-tolerance of algorithm in the present invention.In next stage
Operation in multiple results of classification are all brought into calculating, obtain final result.
In step 5) because two characteristic points can determine a data sectional, due to previous step take it is fault-tolerant
Mechanism, each characteristic point there are two types of may, accordingly, it is possible to data sectional just have 4.Calculate characteristic point between accumulation away from
From take total Cumulative Distance minimum one is segmented, and obtains coupling path.It is the characteristics of at the uniform velocity acquisition using training set data,
Export the position of user.
It will be seen from figure 1 that carrying out matched success rate averagely 91.5% or so to data using characteristic point, and turning
Success rate can reach 98% or so at point (corner).This just illustrates that the recognition accuracy of inflection point wants high by one relative to characteristic point
A bit.If complex situations are more in user path, be more conducive to the two stage cultivation of time series data.
From Fig. 2,3 as can be seen that the error of the high equipment of sensor frequency can be relatively lower.With the growth of distance,
The error of positioning is also gradually increasing, but its numerical value is metastable.Error of the Meizu mobile phone in the algorithm of the present invention
Within 2m, the average positioning accuracy under application scenarios has reached 1.5m or so for basic control.The error of Samsung mobile phone is basic
Within 5m, the average positioning accuracy under application scenarios is 3.5m for control.
After having used raising frequency strategy, it can be seen that the error of Samsung mobile phone is greatly diminished, as shown in Figure 4,5.Meizu
The mean accuracy of mobile phone is still 1.5m or so, and the mean accuracy of Samsung mobile phone has been lowered to 2.5m or so.Experimental result is demonstrate,proved
Bright, the indoor positioning algorithms proposed by the present invention based on sequential are either attained by expection in positioning accuracy and locating speed
Target.And by the test to different mobile phones, the precision using the positioning after raising frequency strategy is almost the same, therefore this hair
Bright algorithm is stable.That is under special scenes, localization method of the invention is effective.
The calculating time of traditional DTW algorithms can be easily found from Fig. 6,7 with the lengthening of path distance and
It is constantly increased with the magnitude of index.When path length is more than 20m, positioning time has been more than just 1s, and works as path distance
When more than 30m, positioning time has just increased to about 3s, this duration is just insufferable for users.And it is of the invention
The algorithm of proposition is but always maintained at the run time of constant rank.From the time numerically from the point of view of, algorithm of the invention positioning
Time is less than 0.1s.Therefore, method proposed by the present invention greatly improves time series data matching the shortcomings that time-consuming, and can be with
As real-time localization method, real-time location-based service is provided to the user.
Claims (6)
1. a kind of indoor orientation method of earth magnetism Time-Series analysis, it is characterised in that include the following steps:
1) indoor environment is subjected to multipath division, and at the uniform velocity acquires the earth magnetism time series data of each path, to earth magnetism sequential number
According to being pre-processed, obtain training set;
2) characteristic point in training set is found out, and data sectional is carried out to earth magnetism time series data according to characteristic point;
3) using the collected earth magnetism time series data of mobile phone in position fixing process as test set, using raising frequency strategy in test set
Earth magnetism time series sample data is into row interpolation;
4) using the collected data of mobile phone in position fixing process as test set, the characteristic point in test set is identified, and use and divide
Class algorithm matches it with the characteristic point in training set;
5) it calculates the distance accumulation matrix of the data of characteristic point segmentation and finds the matched position of test set, realize indoor positioning.
2. the indoor orientation method of earth magnetism Time-Series analysis as described in claim 1, it is characterised in that according to feature in step 2)
Row data are clicked through to be segmented into:
Using bottom-up algorithm, time series is divided into the short arrangement set of consecutive points first;
Adjacent point is connected, the data point in sequence is fallen on line segment;
Two neighboring line segment is connected, every line segment includes 3 data points, then calculates intermediate point error of fitting;
According to intermediate point error of fitting error identifying, minimum and error is less than the segmentation of threshold values R, includes 3 points as first
Line segment;
Calculate the data after the 2~N articles segmentation successively by above-mentioned steps, all data sectionals are completed in training is gathered.
3. the indoor orientation method of earth magnetism Time-Series analysis as described in claim 1, it is characterised in that use raising frequency in step 3)
Strategy is into row interpolation to earth magnetism time series data:
When the frequency acquisition of testing mobile phone is less than the frequency acquisition of training set, artificially increase the sample of test set using interpolation method
This quantity, if training set is indicated with Q, sample size m, test set is indicated with C, sample size n, by test set sample number
Amount extended to from n with m be an order of magnitude sample size.
4. the indoor orientation method of earth magnetism Time-Series analysis as described in claim 1, it is characterised in that use classification in step 4)
It corresponding with the characteristic point in training set done matching and is by algorithm:
Data are handled from the angle of earth magnetism time series data, it will according to the GEOMAGNETIC CHANGE rule (assuming that 1) contained in data
Feature Points Matching is converted into speech recognition problem, using dynamic time warping algorithm, obtains the match point in training data.
5. the indoor orientation method of earth magnetism Time-Series analysis as described in claim 1, it is characterised in that the rule contained in data
For:Indoor environment is divided into mulitpath, the number of paths under a scene is N, set of paths P={ p1, p2…pNCome
It indicates;The physical distance of two paths this indicate that the similarity degree of data is indicated with S (i, j) with D (i, j);In same field
Under scape, if D (i, j) is smaller, S (i, j) is bigger, and farther out, then the tendency similarity of data is different for path distance.
6. the indoor orientation method of earth magnetism Time-Series analysis as described in claim 1, it is characterised in that calculate feature in step 5)
The distance accumulation matrix of data of point segmentation is:M is the matrix of a m*n built according to earth magnetism time series, Cumulative Distance
Matrix is Mc, wherein Mc(0,0)=M (0,0), (0,0) are first data of matrix, that is, the starting point of initial alignment;Other
The Cumulative Distance calculation formula of position is as follows:
Mc(i, j)=Min (MC(i-1,j-1),Mc(i-1,j),Mc(i,j-1),Mc(i,j))+M(i,j)
Wherein i, j be matrix in data call number, 0<i<=m-1,0<j<=n-1.Using obtained Cumulative Distance as two
The similarity function of earth magnetism sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810253960.7A CN108462939B (en) | 2018-03-26 | 2018-03-26 | Indoor positioning method for geomagnetic time sequence analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810253960.7A CN108462939B (en) | 2018-03-26 | 2018-03-26 | Indoor positioning method for geomagnetic time sequence analysis |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108462939A true CN108462939A (en) | 2018-08-28 |
CN108462939B CN108462939B (en) | 2020-10-02 |
Family
ID=63236490
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810253960.7A Active CN108462939B (en) | 2018-03-26 | 2018-03-26 | Indoor positioning method for geomagnetic time sequence analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108462939B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113657184A (en) * | 2021-07-26 | 2021-11-16 | 广东科学技术职业学院 | Evaluation method and device for piano playing fingering |
CN117271987A (en) * | 2023-11-23 | 2023-12-22 | 国网吉林省电力有限公司长春供电公司 | Intelligent acquisition and processing method for environmental state data of power distribution equipment |
CN117289184A (en) * | 2023-11-07 | 2023-12-26 | 中国科学院空天信息创新研究院 | Method for assisting in identifying interference magnetic field by using double magnetic sensors |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104105067A (en) * | 2013-04-09 | 2014-10-15 | 广东美晨通讯有限公司 | Real-time positioning system and method based on geomagnetic information |
CN104897154A (en) * | 2015-04-29 | 2015-09-09 | 中测高科(北京)测绘工程技术有限责任公司 | Mobile phone-based indoor geomagnetic navigation method and apparatus thereof |
WO2016088346A1 (en) * | 2014-12-05 | 2016-06-09 | 株式会社デンソー | Wireless positioning system, wireless positioning terminal, and location information transmission device |
CN106092093A (en) * | 2016-05-26 | 2016-11-09 | 浙江工业大学 | A kind of indoor orientation method based on earth magnetism fingerprint matching algorithm |
-
2018
- 2018-03-26 CN CN201810253960.7A patent/CN108462939B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104105067A (en) * | 2013-04-09 | 2014-10-15 | 广东美晨通讯有限公司 | Real-time positioning system and method based on geomagnetic information |
WO2016088346A1 (en) * | 2014-12-05 | 2016-06-09 | 株式会社デンソー | Wireless positioning system, wireless positioning terminal, and location information transmission device |
CN104897154A (en) * | 2015-04-29 | 2015-09-09 | 中测高科(北京)测绘工程技术有限责任公司 | Mobile phone-based indoor geomagnetic navigation method and apparatus thereof |
CN106092093A (en) * | 2016-05-26 | 2016-11-09 | 浙江工业大学 | A kind of indoor orientation method based on earth magnetism fingerprint matching algorithm |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113657184A (en) * | 2021-07-26 | 2021-11-16 | 广东科学技术职业学院 | Evaluation method and device for piano playing fingering |
CN113657184B (en) * | 2021-07-26 | 2023-11-07 | 广东科学技术职业学院 | Piano playing fingering evaluation method and device |
CN117289184A (en) * | 2023-11-07 | 2023-12-26 | 中国科学院空天信息创新研究院 | Method for assisting in identifying interference magnetic field by using double magnetic sensors |
CN117289184B (en) * | 2023-11-07 | 2024-02-02 | 中国科学院空天信息创新研究院 | Method for assisting in identifying interference magnetic field by using double magnetic sensors |
CN117271987A (en) * | 2023-11-23 | 2023-12-22 | 国网吉林省电力有限公司长春供电公司 | Intelligent acquisition and processing method for environmental state data of power distribution equipment |
CN117271987B (en) * | 2023-11-23 | 2024-02-06 | 国网吉林省电力有限公司长春供电公司 | Intelligent acquisition and processing method for environmental state data of power distribution equipment |
Also Published As
Publication number | Publication date |
---|---|
CN108462939B (en) | 2020-10-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107038717B (en) | A method of 3D point cloud registration error is automatically analyzed based on three-dimensional grid | |
Cieslewski et al. | Point cloud descriptors for place recognition using sparse visual information | |
CN105745687B (en) | Context aware Moving target detection | |
CN104602342B (en) | A kind of high efficiency indoor localization method based on iBeacon equipment | |
CN105716604A (en) | Mobile robot indoor positioning method and system based on geomagnetic sequences | |
CN110536245B (en) | Deep learning-based indoor wireless positioning method and system | |
CN109115205A (en) | A kind of indoor fingerprint positioning method and system based on geomagnetic sensor array | |
CN107688665A (en) | A kind of indoor map method for auto constructing, device and storage medium | |
CN106780552B (en) | Anti-shelter target tracking based on regional area joint tracing detection study | |
CN104394588B (en) | Indoor orientation method based on Wi Fi fingerprints and Multidimensional Scaling | |
CN112258580B (en) | Visual SLAM loop detection method based on deep learning | |
CN104599286B (en) | A kind of characteristic tracking method and device based on light stream | |
Yue et al. | Robust loop closure detection based on bag of superpoints and graph verification | |
CN108462939A (en) | A kind of indoor orientation method of earth magnetism Time-Series analysis | |
CN108627798B (en) | WLAN indoor positioning algorithm based on linear discriminant analysis and gradient lifting tree | |
CN108984785A (en) | A kind of update method and device of the fingerprint base based on historical data and increment | |
CN110536257B (en) | Indoor positioning method based on depth adaptive network | |
CN110175615A (en) | The adaptive visual position recognition methods in model training method, domain and device | |
CN104504709A (en) | Feature ball based classifying method of three-dimensional point-cloud data of outdoor scene | |
CN110388926A (en) | A kind of indoor orientation method based on mobile phone earth magnetism and scene image | |
CN102567702A (en) | Method for automatically identifying valleys and ridge lines based on ChangE DEM (Dynamic Effect Model) data | |
CN110503643A (en) | A kind of object detection method and device based on the retrieval of multiple dimensioned rapid scene | |
CN111739053A (en) | Online multi-pedestrian detection tracking method under complex scene | |
CN107274679A (en) | Vehicle identification method, device, equipment and computer-readable recording medium | |
CN109297489A (en) | A kind of indoor navigation method based on user characteristics, electronic equipment and storage medium |
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 |