CN108462939B - Indoor positioning method for geomagnetic time sequence analysis - Google Patents
Indoor positioning method for geomagnetic time sequence analysis Download PDFInfo
- Publication number
- CN108462939B CN108462939B CN201810253960.7A CN201810253960A CN108462939B CN 108462939 B CN108462939 B CN 108462939B CN 201810253960 A CN201810253960 A CN 201810253960A CN 108462939 B CN108462939 B CN 108462939B
- Authority
- CN
- China
- Prior art keywords
- data
- geomagnetic
- time sequence
- points
- characteristic points
- 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
Images
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
Abstract
The invention relates to an indoor positioning method for geomagnetic time sequence analysis, which comprises the following steps: dividing an indoor environment into multiple paths, acquiring geomagnetic time sequence data of each path at a constant speed, and preprocessing the geomagnetic time sequence data to obtain a training set; finding out characteristic points in a training set, and carrying out data segmentation on geomagnetic time sequence data according to the characteristic points; taking geomagnetic time sequence data acquired by a mobile phone in a positioning process as a test set, and interpolating geomagnetic time sequence data samples in the test set by using an up-conversion strategy; taking data acquired by a mobile phone in the positioning process as a test set, identifying characteristic points in the test set, and matching the characteristic points with the characteristic points in the training set by using a classification algorithm; and calculating a distance accumulation matrix of the segmented data of the characteristic points and finding out the matched position of the test set to realize indoor positioning. The invention uses the characteristic points to segment the time sequence data, calculates the similarity after classification to obtain the user position, and greatly improves the time complexity of the algorithm.
Description
Technical Field
The invention relates to an indoor positioning technology, in particular to an indoor positioning method for geomagnetic time sequence analysis.
Background
An Indoor Positioning System (IPS) refers to a System that collects radio waves, magnetic fields, acoustic signals, or other sensory information using a mobile device to locate objects or people in a building. As the needs of users are deeply mined, applications based on indoor positioning technologies are also gradually moving closer to the field of view of people. Unmanned vehicles, whether in large stores or underground parking lots, and even in warehouses, operate automatically based on indoor location technology. Indoor positioning technology belongs to basic technology, and the accuracy and the timeliness of location can very big influence user experience. Although there are some commercial systems on the market today, the IPS still does not have a unified standard and needs to rely on external facilities, with poor positioning accuracy.
Disclosure of Invention
Aiming at the defects that an indoor positioning system in the prior art needs to depend on external facilities, has poor positioning accuracy and the like, the invention aims to provide an indoor positioning method based on geomagnetic time sequence analysis to achieve the aim of improving the positioning accuracy without depending on the external facilities.
In order to solve the technical problems, the invention adopts the technical scheme that:
the invention discloses an indoor positioning method for geomagnetic time sequence analysis, which comprises the following steps of:
1) dividing an indoor environment into multiple paths, acquiring geomagnetic time sequence data of each path at a constant speed, and preprocessing the geomagnetic time sequence data to obtain a training set;
2) finding out characteristic points in a training set, and carrying out data segmentation on geomagnetic time sequence data according to the characteristic points;
3) taking geomagnetic time sequence data acquired by a mobile phone in a positioning process as a test set, and interpolating geomagnetic time sequence data samples in the test set by using an up-conversion strategy;
4) taking data acquired by a mobile phone in the positioning process as a test set, identifying characteristic points in the test set, and matching the characteristic points with the characteristic points in the training set by using a classification algorithm;
5) and calculating a distance accumulation matrix of the segmented data of the characteristic points and finding out the matched position of the test set to realize indoor positioning.
The data segmentation in the step 2) according to the characteristic points is as follows:
using a bottom-up algorithm, firstly dividing a time sequence into a short sequence set of adjacent points;
connecting adjacent points, the data points in the sequence falling on the line segment;
connecting two adjacent line segments, wherein each line segment comprises 3 data points, and then calculating a fitting error of a middle point;
finding out a segment with the minimum error and the error smaller than a threshold value R according to the fitting error of the intermediate point, and taking the segment as a first line segment containing 3 points;
and sequentially calculating the 2 nd to N th segmented data according to the steps until all the data in the training set are segmented.
In the step 3), the interpolation of the geomagnetic time sequence data by using the frequency raising strategy is as follows:
when the acquisition frequency of the test mobile phone is lower than that of the training set, the number of samples of the test set is artificially increased by adopting an interpolation method, if the training set is represented by Q, the number of samples is m, the test set is represented by C, and the number of samples is n, the number of samples of the test set is expanded from n to the number of samples of which m is an order of magnitude.
In the step 4), the classification algorithm is used to match the feature points in the training set correspondingly as follows:
processing data from the perspective of geomagnetic time sequence data, converting feature point matching into a voice recognition problem according to a geomagnetic variation rule (hypothesis 1) contained in the data, and obtaining a matching point in training data by using a dynamic time warping algorithm.
The rules contained in the data are: dividing indoor environment into a plurality of paths, wherein the number of paths in a scene is N, and the path set is P ═ { P { (P) }1，p2…pNRepresents by "}; the physical distance of the two paths is represented by D (i, j), and the similarity degree of the data is represented by S (i, j); in the same scene, if D (i, j) is smaller, S (i, j) is larger, the path distance is longer, and the trend similarity of the data is different.
Calculating a distance accumulation matrix of the data of the characteristic point segments in the step 5) as follows: m is a matrix of M x n constructed from the time series of geomagnetism, and the cumulative distance matrix is McWherein M isc(0,0) ═ M (0,0), (0,0) is the first data of the matrix, i.e., the starting point of the initial positioning; the cumulative distance calculation formula for the 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 is the index number of the data in the matrix, 0< i < ═ m-1,0< j < ═ n-1. And taking the obtained cumulative distance as a similarity function of the two geomagnetic sequences.
The invention has the following beneficial effects and advantages:
1. the method carries out positioning in a time sequence data matching mode, and considers that the time complexity of direct matching is too high, the method defines the characteristic points, uses the characteristic points to segment time sequence data, calculates the similarity after classification to obtain the user position, and greatly improves the time complexity of the algorithm.
Drawings
FIG. 1 is a graphical representation of the results of using feature point matching in the present invention;
fig. 2 is a graph of average positioning accuracy for positioning in scene 1 using different mobile phones according to the present invention;
fig. 3 is a graph of average positioning accuracy for positioning in scene 2 using different mobile phones according to the present invention;
fig. 4 is a graph of average positioning accuracy of positioning in scene 1 by different mobile phones after using the frequency-up strategy in the present invention;
fig. 5 is a graph of average positioning accuracy of positioning in scene 2 by different mobile phones after using the frequency-up strategy in the present invention;
FIG. 6 is a graph of the average positioning time for positioning in scene 1 using different algorithms according to the present invention;
FIG. 7 is a graph of the average positioning time for positioning in scene 2 using different algorithms in the present invention;
FIG. 8 is a flow chart of a feature point segmentation algorithm in accordance with the present invention;
FIG. 9 is a flowchart of an up-conversion strategy algorithm of the present invention;
fig. 10 is a flow chart of a positioning algorithm of the present invention.
Detailed Description
The invention is further elucidated with reference to the accompanying drawings.
The invention discloses an indoor positioning method for geomagnetic time sequence analysis, which comprises the following steps of:
1) dividing an indoor environment into multiple paths, acquiring geomagnetic time sequence data of each path at a constant speed, and preprocessing the geomagnetic time sequence data to obtain a training set;
2) finding out characteristic points in a training set, and carrying out data segmentation on geomagnetic time sequence data according to the characteristic points;
3) taking geomagnetic time sequence data acquired by a mobile phone in a positioning process as a test set, and interpolating geomagnetic time sequence data samples in the test set by using an up-conversion strategy;
4) taking data acquired by a mobile phone in the positioning process as a test set, identifying characteristic points in the test set, and matching the characteristic points with the characteristic points in the training set by using a classification algorithm;
5) and calculating a distance accumulation matrix of the segmented data of the characteristic points and finding out the matched position of the test set to realize indoor positioning.
In the method, some initialized parameters need to be determined, so that geomagnetic time sequence data of a corridor needs to be acquired in a uniform speed mode, and then geomagnetic characteristic points are marked in an artificial mode. And then, the user can walk in the corridor at will, and the program of the client can detect whether the characteristic points appear in real time. And once the characteristic points appear, calling a characteristic point matching algorithm, and comparing the characteristic points with the characteristic points in the previous training data. Because the feature points are obvious, the accuracy of the classification stage is high. The essence of determining the feature points is to determine the general position of the user, and then the real-time data and the training data can be matched to obtain the matching points in the training data. Because the mode of uniform motion is used when training data is collected, the current position of the user can be calculated under the premise that the collection frequency of the sensor is always fixed.
In the method of the present invention, constructing the geomagnetic map is an indispensable process, and constructing the geomagnetic map means that a large amount of manpower is required to acquire indoor floor plans and sensing data. The higher the positioning requirement, the denser the acquired data and the more time it takes. In some specific scenes such as corridors, passageways and the like, the advantages of the geomagnetic map cannot be fully exerted, because the advancing directions of pedestrians in the scenes are basically unidirectional, and most geomagnetic map information is not used. And since the width of the corridor or the passageway is generally not too large, users are more concerned about the position of the users along the direction of the corridor or the passageway. That is, in this scenario, the map has a one-dimensional path from the original two-dimensional planar specification. Therefore, the invention firstly provides an indoor positioning method based on time sequence analysis to meet the positioning requirement of the user in the scene.
The method of the invention assumes and defines an indoor environment as follows:
assume that 1: dividing indoor environment into a plurality of paths, wherein the number of paths in a scene is N, and the path set is P ═ { P { (P) }1，p2…pNAnd (c) represents. The physical distance between the two paths is denoted by D (i, j), and the degree of similarity of the data is denoted by S (i, j) in the present invention.
Assume 2: in the same scene, if D (i, j) is smaller, S (i, j) is larger, and if the path distance is longer, the trend similarity of the data is different.
Assume that 3: in a small-width indoor scene, the user may take N paths, and the path in the indoor environment may be represented as { P }1，P2…PN}. Depending on the positioning accuracy requirements, the model may define N specific (any integer no less than 1) values depending on the specific width of the corridor. The walking track of the user can only be switched in a well-defined path and cannot exceed the range. That is, if the user's current path is PiAt a certain moment, the user may switch to path PjThe above.
Assume 4: each path can be divided into K segments according to the feature points, and then the t segment of the ith path is represented as Pi t。
Definition 1: geomagnetic time series. The two geomagnetic sequences are respectively Q (training set) and C (test set), wherein the length of Q is m, and the length of C is n. Then the sequence Q may be expressed as Q ═ Q1q2…qmAnd the same C may be represented by C ═ C1c2…cn。
Definition 2: and (5) geomagnetic sequence distance matrix. For two sequences in definition 1, a sequence distance matrix M is constructed, and the elements in the matrix are simple, that is, the euclidean distance from point to point in the corresponding sequence. As shown in equation 1.
Definition 3: the time sequence accumulates the distance matrix. M is a matrix of M x n, and the cumulative distance matrix M is calculatedc，Mc(0,0) is equal to M (0,0), and the cumulative distance calculation method for other positions is as shown in equation 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 degree of similarity of the two sequences is determined by Mc(m, n). The whole matching process is to find a path from the upper left corner to the lower right corner in the distance matrix, so that the sum of the passing elements is minimum. If the sum of the distances is smaller, it indicates that the two sequences are more similar. The cumulative distance is used in the present invention as a function of the similarity of the sequences.
Definition 4: time sequence similarity. The similarity function is an index for describing the degree of similarity of the time series. The smaller the distance D (i, j) between two paths, the larger S (i, j) is. Similarly, if the path distance is long, the tendency similarity of the data is considered to be different.
Definition 5: and (4) time sequence characteristic points. The characteristic points are points with obvious characteristics which are easy to identify, and the ith characteristic point F in the sequenceiCan be expressed as
Wherein m isi-1Showing the previous feature point Fi-1The value of the earth magnetism at (a),
number of samples, m, representing the difference between the current feature point and the previous feature pointiIs represented by Fi-1The earth magnetic value of (c).
As shown in fig. 8, the data segmentation in step 2) according to the feature points is as follows:
using a bottom-up algorithm, firstly dividing a time sequence into a short sequence set of adjacent points;
connecting adjacent points, the data points in the sequence falling on the line segment;
connecting two adjacent line segments, wherein each line segment comprises 3 data points, and then calculating a fitting error of a middle point;
finding out a segment with the minimum error and the error smaller than a threshold value R according to the fitting error of the intermediate point, and taking the segment as a first line segment containing 3 points;
and (3) calculating the line segments (namely segmented data) of the 2 nd to N th short data sequences in sequence according to the steps until all the data in the training set are segmented.
In this embodiment, in order to segment time series data, a bottom-up (a data pyramid structure is adopted, all bottom-layer data are paired and then grouped, the group is further classified, and the top layer is data that cannot be segmented) algorithm is used to segment the data. The time series is first divided into a short series set of adjacent points. There is no fitting error at this point because the data points in the sequence all fall on the line segment, connecting adjacent points. Next, two adjacent segments are connected, each segment containing three data points, at which time the median fitting error is calculated. The fitting error is generally the sum of the euclidean distances. And after the fitting errors of the intermediate points in the line segments of all the three points are calculated, finding out the segment with the minimum error and the error smaller than the threshold value R, using the segment as the first line segment containing the three points, and so on. The procedure performed is shown in algorithm 1 (fig. 8).
In the step 3), the interpolation of the geomagnetic time sequence data by using the frequency raising strategy is as follows:
when the acquisition frequency of the test mobile phone is lower than that of the training set, the number of samples of the test set is artificially increased by adopting an interpolation method, if the training set is represented by Q, the number of samples is m, the test set is represented by C, and the number of samples is n, the number of samples of the test set is expanded from n to the number of samples of which m is an order of magnitude.
In the matching process, the algorithm firstly identifies the feature points of the data in the current test set, then matches the feature points to the positions of the corresponding feature points in the training set, and needs to describe the feature points, and a three-dimensional vector is selected to describe the feature points, such as definition 3.
The matching process is essentially a classification process, and the feature points in the training set are essentially several classes in a classification algorithm, and the algorithm is to correctly classify the feature points in the test set into the corresponding classes.
Any classification algorithm cannot guarantee 100% accuracy, and the data matching of the algorithm depends on the result of feature point matching, and if the algorithm mismatches the currently detected feature point to the wrong path, the result of the algorithm is very bad, and the whole positioning process fails. Therefore, in the K-nearest neighbor classification algorithm model, a fault-tolerant mechanism is added into the algorithm. During the classification process, the algorithm retains the classification result, i.e. the set of possible classification results, the number of elements in the set does not exceed a set threshold k. In the process of matching, the algorithm matches all elements in the set to the training set. If k is set properly, this ensures that the correct time sequence slice can be found.
The data acquisition frequency of the sensors of different smart phones is different, so the lengths of data sequences acquired in the same path are different. The data matching algorithm can solve the problem of data misalignment, but it is also established when the lengths of the training data sequence and the test data sequence are not the same. If the magnitude difference of the data sequences is too large, the algorithm cannot ensure that correct matching is made, and the positioning precision is greatly reduced.
On the same indoor path, the data volume collected by the two mobile phones often differs by hundreds of data. The difference of the acquisition frequency can cause the loss of some trends and even influence the judgment of the characteristic points, and finally greatly influences the positioning precision. When training data is collected, the problem of positioning accuracy is considered, and mobile phone collection with high frequency is often used. Therefore, the algorithm of the present invention is to provide an up-conversion strategy to cope with this requirement.
When the acquisition frequency of the test mobile phone is lower than that of the training set, the invention adopts a method of artificially increasing the number of samples of the test set, and mainly adopts an interpolation method. If the training set is represented by Q, the number of samples is m, the test set is represented by C, the number of samples is n, and the algorithm is to expand the number of samples in the test set from n to a number of samples where the sum m is an order of magnitude. The algorithm is executed as shown in algorithm 2 (fig. 9).
In the step 4), the classification algorithm is used to match the feature points in the training set correspondingly as follows:
processing data from the perspective of geomagnetic time sequence data, converting feature point matching into a voice recognition problem according to a geomagnetic variation rule contained in the data, and obtaining matching points in training data by using a dynamic time warping algorithm.
The algorithm provided by the invention processes data from the perspective of time sequence data. By summarizing the laws contained in the data, the problem is transformed into a speech recognition problem, and then the dynamic time warping algorithm is used very naturally. The "trend-like" feature found by the present invention does not always exist through actual measurement of multiple paths. And the complexity of calculating the sequence accumulated distance is too high, and the method is not suitable for being used as an online positioning method, so that the method fully utilizes some characteristics of the geomagnetic peak value, gives the definition of the feature points, and uses a classification algorithm to match the feature points in the training data and the test data, thereby roughly determining the position of the user position in the whole scene. And then, acquiring the corresponding relation of the user position in the test data relative to the training data by a method of calculating the accumulated distance, and finally determining the user position by combining the label on the training data. The steps are shown in algorithm 3 (fig. 10):
in the invention, the algorithm needs to determine some initialized parameters, so that geomagnetic time sequence data of a corridor needs to be acquired in a uniform speed mode, and then geomagnetic characteristic points are marked in an artificial mode. And then, the user can walk in the corridor at will, and the program of the client can detect whether the characteristic points appear in real time. And once the characteristic points appear, calling a K-neighbor characteristic point matching algorithm, and comparing the K-neighbor characteristic point matching algorithm with the characteristic points in the previous training data. Because the feature points are obvious, the accuracy of the classification stage is high. The essence of determining the feature points is to determine the general position of the user, and then the real-time data and the training data can be matched to obtain the matching points in the training data. Because the mode of uniform motion is used when training data is collected, the current position of the user can be calculated under the premise that the collection frequency of the sensor is always fixed.
In the embodiment, a Samsung Note3 smart phone and a charm Pro6 mobile phone are selected, the two mobile phones are respectively based on an android4.4 system and an android6.0 system, and mysql is selected for data storage. The Python language has the advantages of rich tools and simple grammar in data processing, so the Python language is used for data processing. Also, to maintain compatibility as much as possible, a web service is built using the flash toolkit in the python language to receive parameters from the client. The experimental data graphs are all realized by using matplotlib library. The experimental data are shown in table 1. The data that mainly used are triaxial geomagnetic sensor reading and accelerometer's reading and have added the timestamp, and the precision of sensing data is 6 bits after keeping the decimal point, and the precision of timestamp is 0.1 s.
TABLE 1 Experimental data
In step 1), the invention has been experimented in two different indoor scenes, respectively. Scene 1 and scene 2 are divided into 5 paths, and geomagnetic data in the paths are collected at a constant speed by using the robot trolley. Specific scene parameters are shown in table 2:
TABLE 2 concrete scene parameters
In the step 2), the algorithm 1 is used to find out the feature points from bottom to top, 8 feature points are found in the scene 1, and 14 feature points are found in the scene 2. Scene 1 may be divided into 45 geomagnetic sequence segments and scene 2 may be divided into 75 sequence segments.
In the step 3), two mobile phone acquisition test sets are used, wherein the frequency of geomagnetic data acquisition of the samsung mobile phone is about one third of that of the charm mobile phone. Therefore, the data is interpolated by using the frequency-increasing strategy in algorithm 2, and the lengths of the collected sequences are guaranteed to be the same after interpolation, and the lengths of the sequences of scene 1 and scene 2 are 637 and 1538, respectively.
And 4) recording the feature points in the training set as categories, and classifying the data in the test set into the categories of the feature points in the training set. The invention uses a nearest neighbor algorithm, where k is selected to be 3. The output of the traditional k-nearest neighbor algorithm is directly the nearest neighbor category, and in order to ensure the fault tolerance of the algorithm, the invention also records the category of the next nearest neighbor. And in the operation of the next stage, a plurality of classified results are all brought into calculation to obtain a final result.
In step 5), because two feature points can determine one data segment, because the fault-tolerant mechanism is adopted in the previous step, each feature point has two possibilities, and therefore, 4 data segments are possible. And calculating the accumulated distance between the feature points, taking a segment with the minimum accumulated distance, and obtaining a matching path. And outputting the position of the user by utilizing the characteristic that the training set data is collected at a constant speed.
As can be seen from fig. 1, the success rate of matching data using feature points is about 91.5% on average, and the success rate can reach about 98% at inflection points (corners). This indicates that the identification accuracy of the inflection point is higher with respect to the feature point. The more complex the user path, the more advantageous the segment matching of the time series data.
As can be seen from fig. 2 and 3, the error of the device with a high sensor frequency will be relatively low. As the distance increases, the error of positioning also increases gradually, but its value is relatively stable. The error of the charm mobile phone in the algorithm of the invention is basically controlled within 2m, and the average positioning precision in the application scene reaches about 1.5 m. The error of the three-star mobile phone is basically controlled within 5m, and the average positioning precision in an application scene is 3.5 m.
After the frequency-up strategy is used, it can be seen that the error of the samsung handset is greatly reduced, as shown in fig. 4 and 5. The average accuracy of the charm mobile phone is still about 1.5m, and the average accuracy of the samsung mobile phone is reduced to about 2.5 m. Experimental results prove that the indoor positioning algorithm based on the time sequence can achieve the expected target in both positioning precision and positioning speed. And through the test to different mobile phones, the precision of positioning after using the frequency raising strategy is basically consistent, so the algorithm of the invention is stable. That is, the positioning method of the present invention is effective in a specific situation.
It can be easily found from fig. 6 and 7 that the computation time of the conventional DTW algorithm increases exponentially as the path distance increases. When the path length is greater than 20m, the positioning time exceeds 1s, and when the path distance exceeds 30m, the positioning time rises to about 3s, which is simply intolerable to the user. The algorithm proposed by the present invention maintains a constant level of runtime at all times. From the numerical view of time, the algorithm of the invention locates the time less than 0.1 s. Therefore, the method provided by the invention greatly improves the defect of long time consumption for time sequence data matching, and can be used as a real-time positioning method to provide real-time position service for users.
Claims (5)
1. An indoor positioning method for geomagnetic time sequence analysis is characterized by comprising the following steps:
1) dividing an indoor environment into multiple paths, acquiring geomagnetic time sequence data of each path at a constant speed, and preprocessing the geomagnetic time sequence data to obtain a training set;
2) finding out characteristic points in a training set, and carrying out data segmentation on geomagnetic time sequence data according to the characteristic points;
3) taking geomagnetic time sequence data acquired by a mobile phone in a positioning process as a test set, and interpolating geomagnetic time sequence data samples in the test set by using an up-conversion strategy;
4) taking data acquired by a mobile phone in the positioning process as a test set, identifying characteristic points in the test set, and matching the characteristic points with the characteristic points in the training set by using a classification algorithm;
5) calculating an accumulated distance matrix of the segmented data of the characteristic points and finding out the matched position of the test set to realize indoor positioning;
the data segmentation in the step 2) according to the characteristic points is as follows:
using a bottom-up algorithm, firstly dividing a time sequence into a short sequence set of adjacent points;
connecting adjacent points, the data points in the sequence falling on the line segment;
connecting two adjacent line segments, wherein each line segment comprises 3 data points, and then calculating a fitting error of a middle point;
finding out a segment with the minimum error and the error smaller than a threshold value R according to the fitting error of the intermediate point, and taking the segment as a first line segment containing 3 points;
and sequentially calculating the 2 nd to N th segmented data according to the steps until all the data in the training set are segmented.
2. The indoor geomagnetic timing analysis positioning method according to claim 1, wherein the step 3) uses an up-conversion strategy to interpolate the geomagnetic timing data by:
when the acquisition frequency of the test mobile phone is lower than that of the training set, the number of samples of the test set is artificially increased by adopting an interpolation method, if the training set is represented by Q, the number of samples is m, the test set is represented by C, and the number of samples is n, the number of samples of the test set is expanded from n to the number of samples of which m is an order of magnitude.
3. The indoor geomagnetic timing analysis positioning method as defined in claim 1, wherein the step 4) uses a classification algorithm to match the feature points in the training set to:
processing data from the perspective of geomagnetic time sequence data, converting feature point matching into a voice recognition problem according to a geomagnetic variation rule contained in the data, and obtaining matching points in training data by using a dynamic time warping algorithm.
4. The indoor geomagnetic timing analysis positioning method as defined in claim 1, wherein the data includes a rule: dividing indoor environment into a plurality of paths, wherein the number of paths in a scene is N, and the path set is P ═ { P { (P) }1，p2…pNRepresents by "}; the physical distance of the two paths is represented by D (i, j), and the similarity degree of the data is represented by S (i, j); in the same scenario, if D (i, j) is smaller, S (i, j) is larger.
5. The indoor geomagnetic timing analysis positioning method according to claim 1, wherein the cumulative distance matrix of the data of the feature point segments calculated in step 5) is: m is a matrix of M x n constructed from the time series of geomagnetism, and the cumulative distance matrix is McWherein M isc(0,0) ═ M (0,0), (0,0) is the first data of the matrix, i.e., the starting point of the initial positioning; the calculation formula of the accumulated distance matrix at 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 is the index number of the data in the matrix, 0< i < ═ m-1,0< j < ═ n-1; and taking the obtained accumulated distance matrix as a similarity function of the two geomagnetic sequences, wherein M is a row of the matrix M, and n is a column of the matrix M.
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 CN108462939A (en) | 2018-08-28 |
| CN108462939B true 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) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| 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 |
| 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 |
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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108462939A (en) | 2018-08-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109525935B (en) | Intelligent floor sensing method and control system for universal indoor environment | |
| CN111199564B (en) | Indoor positioning method and device of intelligent mobile terminal and electronic equipment | |
| CN110223324B (en) | Target tracking method of twin matching network based on robust feature representation | |
| CN105528359B (en) | For storing the method and system of travel track | |
| Cieslewski et al. | Point cloud descriptors for place recognition using sparse visual information | |
| CN108462939B (en) | Indoor positioning method for geomagnetic time sequence analysis | |
| WO2012035707A1 (en) | Map generation device, map generation method, method for moving mobile body, and robot device | |
| WO2019061949A1 (en) | Motion-behavior-assisted indoor fusion positioning method and apparatus and storage medium | |
| CN110400332A (en) | A kind of target detection tracking method, device and computer equipment | |
| CN104394588B (en) | Indoor orientation method based on Wi Fi fingerprints and Multidimensional Scaling | |
| AU2014306813A1 (en) | Visual-based inertial navigation | |
| CN111277946A (en) | Fingerprint database self-adaptive updating method in Bluetooth indoor positioning system | |
| CN111523447B (en) | Vehicle tracking method, device, electronic equipment and storage medium | |
| CN110426037A (en) | A kind of pedestrian movement track real time acquiring method under enclosed environment | |
| CN111595344B (en) | Multi-posture downlink pedestrian dead reckoning method based on map information assistance | |
| CN107622507B (en) | Air target tracking method based on deep learning | |
| CN112729301A (en) | Indoor positioning method based on multi-source data fusion | |
| CN111739053A (en) | Online multi-pedestrian detection tracking method under complex scene | |
| KR20220098312A (en) | Method, apparatus, device and recording medium for detecting related objects in an image | |
| CN113989333A (en) | Pedestrian tracking method based on face and head and shoulder information | |
| JP7339314B2 (en) | Deep Smartphone Sensor Fusion for Indoor Positioning and Tracking | |
| Khassanov et al. | Finer-level sequential wifi-based indoor localization | |
| Antsfeld et al. | Magnetic field sensing for pedestrian and robot indoor positioning | |
| Nowakowski et al. | Topological localization using Wi-Fi and vision merged into FABMAP framework | |
| CN116952277A (en) | Potential position estimation method and system based on artificial intelligence and behavior characteristics |
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 |