CN110426031A - Indoor ground magnetic positioning method based on BP neural network and Pu Shi analysis - Google Patents
Indoor ground magnetic positioning method based on BP neural network and Pu Shi analysis Download PDFInfo
- Publication number
- CN110426031A CN110426031A CN201910534444.6A CN201910534444A CN110426031A CN 110426031 A CN110426031 A CN 110426031A CN 201910534444 A CN201910534444 A CN 201910534444A CN 110426031 A CN110426031 A CN 110426031A
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- training
- ground magnetic
- positioning method
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000004458 analytical method Methods 0.000 title claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 39
- 230000004807 localization Effects 0.000 claims abstract description 14
- 230000005389 magnetism Effects 0.000 claims abstract description 14
- 238000012545 processing Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 22
- 238000005259 measurement Methods 0.000 claims description 10
- 238000013519 translation Methods 0.000 claims description 10
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 5
- 238000013480 data collection Methods 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 238000013461 design Methods 0.000 claims description 2
- 238000001914 filtration Methods 0.000 claims description 2
- 238000012886 linear function Methods 0.000 claims description 2
- 241000228740 Procrustes Species 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 239000004035 construction material Substances 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/04—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by terrestrial means
- G01C21/08—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by terrestrial means involving use of the magnetic field of the earth
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
- G01C21/165—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation combined with non-inertial navigation instruments
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/029—Location-based management or tracking services
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- 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
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/38—Services specially adapted for particular environments, situations or purposes for collecting sensor information
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Automation & Control Theory (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geology (AREA)
- Navigation (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The invention discloses a kind of indoor ground magnetic positioning methods analyzed based on BP neural network and Pu Shi, belong to indoor positioning field.The present invention is aiming at the problem that ground magnetic positioning method is fluctuated etc. at any time by Smartphone device isomerism, data to be influenced, geomagnetic data is converted using general formula analysis (Procrustes Analysis, PA) to weaken or eliminate the geomagnetic data fluctuation problem as caused by equipment isomery, time change etc.；Come the corresponding relationship of training data and actual geographic position in such a way that segment processing initial data binding site coordinate inputs BP neural network parallel, the grid training pattern of localization region is established for indoor positioning, greatly reduces the workload of the fingerprint base building of indoor earth magnetism fingerprint location.
Description
Technical field
The present invention relates to indoor positioning fields, and in particular to a kind of indoor ground based on BP neural network and the analysis of general formula
Magnetic positioning method.
Background technique
By the development of many years, Global Satellite Navigation System (Global Navigation Satellite System) mesh
It is preceding that accurately outdoor positioning service can be provided for people.Satellite-signal is influenced to decline by wall, glass etc. under environment indoors
Subtract seriously, positioning accuracy is greatly reduced the demand for not being able to satisfy people's indoor positioning.To solve the problems, such as indoor positioning, with bluetooth,
WIFI, RFID, ultra wide band (UWB), LED visible light etc. have been widely studied for the indoor positioning technologies of signal source, but these
The realization of technology is dependent on signal transmitting base station, and widespread adoption will cause high cost input, therefore be not suitable for extensive
Using.Sensor-based pedestrian's dead reckoning (Pedestrian Dead Reckoning) technology can provide continuously for people
Relative position information, but application is there are biggish accumulated error for a long time, is not suitable for being used alone.
Earth's magnetic field is natural physical field existing for earth itself, indoors since construction material, electronic are set under environment
The standby influence waited makes indoor earth's magnetic field be distorted, and the characteristic value that can be used as indoor positioning uses.Due to not needing additional frame
If signal transmitting base station, it is low in cost the features such as, indoor ground magnetic orientation is increasingly becoming the hot research direction of indoor positioning.At present
Earth magnetism location technology mainly includes the earth magnetism fingerprint location and sequence earth magnetism fingerprint location two major classes of single-point.The ground magnetic orientation of single-point
It is affected by magnetic survey value, position error is larger；Sequences match positioning compared to single-point earth magnetism fingerprint location precision compared with
Height can overcome the problems, such as error hiding caused by similar fingerprints point to a certain extent, but position and be easy to be fluctuated shadow by geomagnetic data
It rings.The fluctuation of geomagnetic data is mainly as caused by the factors such as measuring device and time change, and major embodiment is on the whole
Magnetic data sequence variation trend is identical, and data waveform is similar, but the data of distinct device acquisition will appear difference on the whole again.
Geomagnetic data fluctuation can have a huge impact the precision of magnetic orientation over the ground with stability, solve the problems, such as that this is to improve earth magnetism to determine
The key of position stability.
Summary of the invention
It is influenced for prior art by equipment isomerism, data fluctuations etc., the present invention proposes a kind of based on BP nerve
The indoor ground magnetic positioning method of network and Pu Shi analysis, translates geomagnetic data using PA conversion, scaling processing, is set with eliminating
Data fluctuations caused by the factors such as standby isomery；Using BP neural network to the geomagnetic data sequence and position coordinates one after segmentation
It rises and is trained, obtain training pattern；The tuning on-line stage obtains the data sequence input training pattern acquired in real time in real time
Location information.
The present invention uses following technical scheme to solve above-mentioned technical problem:
Indoor ground magnetic positioning method based on BP neural network and Pu Shi analysis, specifically comprises the following steps:
Step 1, the neural network model of off-line training localization region；
Step 1.1, localization region geomagnetic data is acquired using smart phone, first calibration mobile phone sensor, using circle in figure 8
Mode dynamic calibration；The earth magnetism sequence of localization region is acquired after calibration, in collection process on the path of design at the uniform velocity
The original geomagnetic data of acquisition is done mean filter processing and weakens measurement influence of noise by walking；
Step 1.2, filtered geomagnetic data is segmented to obtain the set of several data segments according to sample frequency
(M1, M2..., Mn), PA is carried out to every one piece of data and is converted to new data sequence set (N1, N2..., Nn)；
Step 1.3, each section of data after PA is converted are inputted BP neural network with corresponding position coordinates, with ground
Magnetic data sequence is training sample data, is training label data with corresponding position coordinates, and input neural network is trained
Until reaching convergence, the neural network model of localization region geomagnetic data sequence and actual geographic coordinate is obtained；
Step 2, the tuning on-line stage；
Step 2.1, the window for setting real-time data collection is identical as sample frequency size, real-time acquisition geomagnetic data sequence
Column do mean filter processing, obtain the data segment T converted for PA；
Step 2.2, new data segment T ' is converted to as PA to filtered data, New Data Segment is inputted trained
Good neural network model, the output result of model is exactly current position coordinates；
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
The specific execution method of step 1.1 are as follows:
Step 1.11, the operation of correction mobile phone sensor is calibrated after log-on data acquisition software, and concrete operations are 8 words
Calibration, holding mobile phone, circle in figure 8 rotates back and forth, and prover time is one to three minutes；
Step 1.12, geomagnetic data is acquired after calibrating sensor, hand-held intelligent mobile phone is at the uniform velocity walked in localization region, along pre-
If acquiring path to walk from starting point to terminal, setting sample frequency is that 20HZ can be continuously to instruction for the effect for guaranteeing model training
Practice region acquisition data several times, mean filter processing is all made of to initial data, weakens the influence of measurement noise；
Step 1.13, data sectional processing is the data sectional according to sample frequency acquisition, i.e., the length of every data section
Degree is consistent with sample frequency, and entire data are divided into several segments (M1, M2..., Mn)；
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
The specific execution method of step 1.2 are as follows:
Translation data: the magnetic field data sequence of a certain section of measurement is set as Mi={ B1, B2..., Bn, it calculates according to the following formula equal
Value；
In formula,It is the mean value of magnetic field modulus value；According to the mean value of required every segment data, being subtracted with every one piece of data is worth
To new magnetic field data sequenceThis is carried out to all data segments to have operated
At translation data；
Scaled data: zoom factor s is calculated according to the following formula
After calculating zoom factor, with following formula to the data segment M ' after translationiIt zooms in and out to obtain data Ni, to all
Data segment obtains data acquisition system (N after having executed translation, scaling1, N2..., Nn):
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
The specific execution method of step 1.3 are as follows:
Step 1.31, corresponding position coordinates (x is added to each section of earth magnetism sequence datai, yi) constitute a training data
(N1, x, y)；Executing this operation to all data segments can be obtained training dataset:
Step 1.32, BP neural network is with (N1, N2..., Nn) it is input layer training data, with each section of location information
(xi, yi) it is training label data, all data segment input neural networks are trained, the model that training is obtained stores
Get up and is used for subsequent positioning.
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
In step 1.1, data acquire the magnetometer sensor for using smart phone, and sample frequency is set as 20~30HZ.
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
It is only that training data can not be influenced by mobile phone posture with earth magnetism modulus value in step 1.2.
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
In step 1.3, the specific parameter setting of neural network:
BP neural network is used here as three-layer network, i.e. input layer, middle layer, output layer；
Hidden layer transmission function is that S function tansig exports layer functions using linear function purelin, and learning algorithm uses
Gradient descent method, learning function use the learning function trainlm based on gradient descent method, and setting frequency of training is 500, convergence
Mean square deviation is 0.001, remaining parameter uses default setting.
As the interior analyzed the present invention is based on BP neural network and Pu Shi the further preferred scheme of magnetic positioning method,
The specific execution method of step 2.1 are as follows:
Step 2.11, it is identical with sample frequency that data window size is set, using the thought of linear list, In when real-time measurement
Gauge outfit deletes data, adds data in table tail, i.e., when size of data is greater than sample frequency, first deletes the data stored at first,
Then subsequent data are moved forward, the data newly come in is added in tail portion；
Step 2.12, for the data in window, the influence that mean filter processing weakens observation noise is done, after the completion of filtering
PA is to be converted to for matched vector.
The invention adopts the above technical scheme compared with prior art, has following technical effect that
1, geomagnetic data is converted using Pu Shi analysis method, the geomagnetic data sequence that different time, distinct device are acquired
It is transformed under similar template, can eliminate due to equipment isomery, geomagnetic data fluctuation caused by variation for a long time determines earth magnetism
The influence of position improves the stability of ground magnetic orientation；
2, using BP neural network training localization region geomagnetic data model, with mass data training geomagnetic data sequence with
The corresponding relationship of actual geographic coordinate guarantees the reliability of model.This method is with the corresponding network model output knot of measured data
Fruit eliminates the construction work of database in fingerprint location as positioning coordinate, reduces positioning cost input.
Detailed description of the invention
Fig. 1 is that the present invention is based on the indoor earth magnetism location technology flow charts of BP neural network and Pu Shi analysis；
Fig. 2 is PA flow path switch figure of the present invention.
Specific embodiment
In conjunction with the techniqueflow chart provided, illustrates that the present invention is specific in detail below and execute step, it should be noted that this
The protection scope of invention example without being limited thereto.
As shown in Figure 1, this method mainly includes two steps, off-line phase and on-line stage.Step 1 is acquisition positioning
Area data does PA conversion, and input BP neural network is trained to obtain grid model；By the data of real-time measurement in step 2
It inputs trained model and obtains positioning result.
The detailed implementation of the localization method is as follows:
With smart phone be acquisition data tool, open data acquisition software start acquire data, hand-held intelligent mobile phone around
" 8 " word rotation correction mobile phone sensor, correction course may be set in one to three minutes；Acquired number after correction
According to deletion, start to acquire geomagnetic data in localization region, multiple data files under the same paths of continuous acquisition, guarantee has big
Amount can be used for trained geomagnetic data.
After acquiring end of data, data are imported using Matlab software, mean filter processing is done to initial data and weakens survey
Data, are then divided into several data segment (M according to sample frequency by the influence for measuring noise1, M2..., Mn)。
PA conversion is made to data, the dimension of data used is first determined whether, as shown in Fig. 2, being then not required to if it is one-dimensional data
Rotation process is done again；It is then rotated if it is 2-D data.The present invention uses magnetic field modulus value for change data, is one
Dimension data needs to do Pan and Zoom operation, and specific implementation procedure is as follows:
Translation data: the magnetic field data sequence of a certain section of measurement is set as Mi={ B1, B2..., Bn, data are calculated as follows
The mean value of sequence:
According to the mean value of required every segment data, with sequence MiEach of data subtract mean valueObtain new magnetic field mould
Value sequence
Scaled data: zoom factor s is calculated according to the following formula
After calculating zoom factor, to the data sequence M ' after translationiScaling obtains data sequence Ni:
To all data segments after translation, scaling, final conversion data collection (N is obtained1, N2..., Nn)；It will
Corresponding geographical coordinate (the x of data segmenti, yi) data set for training neural network is formed with geomagnetic data collection, wherein (N1,
N2..., Nn) it is training sample data, (xi, yi) it is trained label data.
Training neural network is carried out in the case where Matlab develops environment, need to call Matlab grid creation function newff () wound
Build a grid, call format are as follows:
Net=newff (PR, [s1, s2..., sn], { TF1, TF2..., TFn, trianFun, BLF, PF)
The present invention uses three-layer neural network, and hidden layer transmission function is S function " tansig ", and output layer functions use line
Property function " purelin ", learning function uses the learning function " trainlm " based on gradient descent method, and frequency of training is appointed as
500, convergence mean square deviation is 0.001, remaining parameter uses default setting, call format are as follows:
Net=newff (PR, [s1, s2, sn], { ' tansig ', ' tansig ', ' purelin ' }, ' trainlm ')
In above formula, PR is the minimax value matrix of input vector, [s1, s2, sn] it is input layer, hidden layer, output layer
Neuron number；It has created grid and has called grid training function trian () training data in Matlab later.To turn by PA
Data segment N after changing is training sample data, is training label data with the corresponding position coordinates X of every segment data, N and X are done
Training function, call format are inputted after normalization are as follows:
[net, tr]=train (net, N, X)
Wherein, X=(xi, yi), i=1,2 ..., n, net are the grid of previous step creation；Training is called after reaching convergence
Function save () stores training grid model.
Positioning stage does mean filter to the data acquired in real time in setting window and PA is converted, obtains number to be matched
According to sequence N '；It calls function load () to load grid model, function sim () is called to input data sequence N ' and grid to be matched
Model net, it is current location data that grid, which exports result, and specific call format is as follows:
Result=sim (net, N ')
For researcher in this field, ground magnetic orientation can be done according to above-mentioned thinking by equipment isomerism, data fluctuations
Property research, various improvement and deformation can be done to the above method, BP neural network is not limited to the selection of neural network, but institute
Some improvement and deformation all should be within the scope of protection of the claims of the present invention.
Claims (8)
1. the indoor ground magnetic positioning method based on BP neural network and Pu Shi analysis, which is characterized in that specifically comprise the following steps:
Step 1, the neural network model of off-line training localization region；
Step 1.1, localization region geomagnetic data is acquired using smart phone, first calibration mobile phone sensor, using the side of circle in figure 8
Formula dynamic calibration；The earth magnetism sequence of localization region is acquired after calibration, is at the uniform velocity walked on the path of design in collection process,
The original geomagnetic data of acquisition is done mean filter processing and weakens measurement influence of noise；
Step 1.2, filtered geomagnetic data is segmented to obtain the set (M of several data segments according to sample frequency1,
M2..., Mn), PA is carried out to every one piece of data and is converted to new data sequence set (N1, N2..., Nn)；
Step 1.3, each section of data after PA is converted are inputted BP neural network with corresponding position coordinates, with ground magnetic number
According to sequence be training sample data, with corresponding position coordinates be training label data, input neural network be trained until
Reach convergence, obtains the neural network model of localization region geomagnetic data sequence and actual geographic coordinate；
Step 2, the tuning on-line stage；
Step 2.1, the window for setting real-time data collection is identical as sample frequency size, and real-time acquisition geomagnetic data sequence is done
Mean filter processing, obtains the data segment T converted for PA；
Step 2.2, new data segment T ' is converted to as PA to filtered data, New Data Segment is inputted trained
Neural network model, the output result of model are exactly current position coordinates.
2. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In the specific execution method of step 1.1 are as follows:
Step 1.11, the operation of correction mobile phone sensor is calibrated after log-on data acquisition software, and concrete operations are 8 word schools
Standard, holding mobile phone, circle in figure 8 rotates back and forth, and prover time is one to three minutes；
Step 1.12, geomagnetic data is acquired after calibrating sensor, hand-held intelligent mobile phone is at the uniform velocity walked in localization region, is adopted along default
Collect path to walk from starting point to terminal, setting sample frequency is that 20HZ can be continuously to training center for the effect for guaranteeing model training
Domain acquires data several times, is all made of mean filter processing to initial data, weakens the influence of measurement noise；
Step 1.13, data sectional processing is data sectional according to sample frequency acquisition, i.e., the length of every data section with
Sample frequency is consistent, and entire data are divided into several segments (M1, M2..., Mn)。
3. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In the specific execution method of step 1.2 are as follows:
Translation data: the magnetic field data sequence of a certain section of measurement is set as Mi={ B1, B2..., Bn, mean value is calculated according to the following formula；
In formula,It is the mean value of magnetic field modulus value；According to the mean value of required every segment data, mean value is subtracted with every one piece of data and is obtained newly
Magnetic field data sequenceThis operation is carried out to all data segments to complete to put down
Move data；
Scaled data: zoom factor s is calculated according to the following formula
After calculating zoom factor, with following formula to the data segment M ' after translationiIt zooms in and out to obtain data Ni, to all data
Data acquisition system (N is obtained after the complete translation of Duan Zhihang, scaling1, N2..., Nn):
4. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In the specific execution method of step 1.3 are as follows:
Step 1.31, corresponding position coordinates (x is added to each section of earth magnetism sequence datai, yi) constitute a training data (N1,
X, y)；Executing this operation to all data segments can be obtained training dataset:
Step 1.32, BP neural network is with (N1, N2..., Nn) it is input layer training data, with each section of location information (xi,
yi) it is training label data, all data segment input neural networks are trained, the model that training obtains is stored
It is used for subsequent positioning.
5. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In in step 1.1, data acquire the magnetometer sensor for using smart phone, and sample frequency is set as 20~30HZ.
6. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In only with earth magnetism modulus value being that training data can not be influenced by mobile phone posture in step 1.2.
7. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In, in step 1.3, the specific parameter setting of neural network:
BP neural network is used here as three-layer network, i.e. input layer, middle layer, output layer；
Hidden layer transmission function is that S function tansig exports layer functions using linear function purelin, and learning algorithm uses gradient
Descent method, learning function use the learning function trainlm based on gradient descent method, and setting frequency of training is 500, restrain square
Difference is 0.001, remaining parameter uses default setting.
8. the indoor ground magnetic positioning method according to claim 1 based on BP neural network and Pu Shi analysis, feature exist
In the specific execution method of step 2.1 are as follows:
Step 2.11, it is identical with sample frequency that data window size is set, using the thought of linear list, in gauge outfit when real-time measurement
Data are deleted, data is added in table tail, i.e., when size of data is greater than sample frequency, first deletes the data stored at first, then
Subsequent data are moved forward, the data newly come in are added in tail portion；
Step 2.12, for the data in window, the influence that mean filter processing weakens observation noise is done, PA is after the completion of filtering
It is converted to for matched vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910534444.6A CN110426031B (en) | 2019-06-19 | 2019-06-19 | Indoor geomagnetic positioning method based on BP neural network and Pu's analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910534444.6A CN110426031B (en) | 2019-06-19 | 2019-06-19 | Indoor geomagnetic positioning method based on BP neural network and Pu's analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110426031A true CN110426031A (en) | 2019-11-08 |
| CN110426031B CN110426031B (en) | 2024-01-16 |
Family
ID=68408759
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910534444.6A Active CN110426031B (en) | 2019-06-19 | 2019-06-19 | Indoor geomagnetic positioning method based on BP neural network and Pu's analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110426031B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112504265A (en) * | 2020-11-16 | 2021-03-16 | 中国科学院空天信息创新研究院 | Geomagnetic reference library construction method for indoor vehicle geomagnetic matching positioning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160210552A1 (en) * | 2013-08-26 | 2016-07-21 | Auckland University Of Technology | Improved Method And System For Predicting Outcomes Based On Spatio/Spectro-Temporal Data |
| CN109506647A (en) * | 2018-12-24 | 2019-03-22 | 哈尔滨工程大学 | A kind of INS neural network based and magnetometer combined positioning method |
| CN109579853A (en) * | 2019-01-24 | 2019-04-05 | 燕山大学 | Inertial navigation indoor orientation method based on BP neural network |
| CN109708628A (en) * | 2019-01-09 | 2019-05-03 | 浙江大学城市学院 | A kind of earth magnetism indoor orientation method based on integrated study and BP neural network |
| CN109855623A (en) * | 2019-01-09 | 2019-06-07 | 东南大学 | Geomagnetic model online approximating method based on Legendre multinomial and BP neural network |
-
2019
- 2019-06-19 CN CN201910534444.6A patent/CN110426031B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160210552A1 (en) * | 2013-08-26 | 2016-07-21 | Auckland University Of Technology | Improved Method And System For Predicting Outcomes Based On Spatio/Spectro-Temporal Data |
| CN109506647A (en) * | 2018-12-24 | 2019-03-22 | 哈尔滨工程大学 | A kind of INS neural network based and magnetometer combined positioning method |
| CN109708628A (en) * | 2019-01-09 | 2019-05-03 | 浙江大学城市学院 | A kind of earth magnetism indoor orientation method based on integrated study and BP neural network |
| CN109855623A (en) * | 2019-01-09 | 2019-06-07 | 东南大学 | Geomagnetic model online approximating method based on Legendre multinomial and BP neural network |
| CN109579853A (en) * | 2019-01-24 | 2019-04-05 | 燕山大学 | Inertial navigation indoor orientation method based on BP neural network |
Non-Patent Citations (2)
| Title |
|---|
| 乔玉坤等: "基于矩谐分析和BP神经网络的地磁基准图构建方法", 《兵工学报》 * |
| 徐龙阳等: "基于神经网络的多传感器融合PDR定位方法", 《传感技术学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112504265A (en) * | 2020-11-16 | 2021-03-16 | 中国科学院空天信息创新研究院 | Geomagnetic reference library construction method for indoor vehicle geomagnetic matching positioning |
| CN112504265B (en) * | 2020-11-16 | 2023-02-28 | 中国科学院空天信息创新研究院 | Geomagnetic reference library construction method for indoor vehicle geomagnetic matching positioning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110426031B (en) | 2024-01-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106525031B (en) | A kind of combined indoor orientation method | |
| CN104483658B (en) | Based on Wi-Fi and the indoor orientation method in earth's magnetic field | |
| CN106092093A (en) | A kind of indoor orientation method based on earth magnetism fingerprint matching algorithm | |
| CN110536256B (en) | Indoor positioning method based on double-layer grids | |
| CN104869541B (en) | A kind of indoor positioning method for tracing | |
| CN104507050B (en) | Probabilistic type finger print matching method in a kind of WiFi indoor positionings | |
| CN106054125B (en) | A kind of fusion indoor orientation method based on linear chain condition random field | |
| CN104320759B (en) | Based on fixedly target indoor locating system fingerprint base construction method | |
| CN105547301A (en) | Indoor map construction method and device based on geomagnetism | |
| CN107084737B (en) | Drainage pipe network inspection system and method based on AR real scene and voice navigation | |
| CN104215238A (en) | Indoor positioning method of intelligent mobile phone | |
| CN105898713A (en) | WiFi fingerprint indoor positioning method based on weighted cosine similarity | |
| CN103455811B (en) | Indoor wireless locating fingerprint collection method and system | |
| CN104936287A (en) | Sensor network indoor fingerprint positioning method based on matrix completion | |
| CN109323695A (en) | A kind of indoor orientation method based on adaptive Unscented kalman filtering | |
| CN101714211A (en) | Detection method of high-resolution remote sensing image street center line | |
| CN105208651A (en) | Wi-Fi position fingerprint non-monitoring training method based on map structure | |
| CN107179525A (en) | A kind of location fingerprint construction method of the Kriging regression based on Thiessen polygon | |
| CN104507097A (en) | Semi-supervised training method based on WiFi (wireless fidelity) position fingerprints | |
| CN106954190A (en) | A kind of WIFI indoor orientation methods based on index mapping domain | |
| CN103476113B (en) | System is set up based on MNL probability identification indoor locating system and method, location model | |
| CN110032939A (en) | A kind of remote sensing time series data approximating method based on gauss hybrid models | |
| CN110426031A (en) | Indoor ground magnetic positioning method based on BP neural network and Pu Shi analysis | |
| Wei et al. | MM-Loc: Cross-sensor indoor smartphone location tracking using multimodal deep neural networks | |
| CN106197418A (en) | The indoor orientation method that a kind of fingerprinting based on sliding window merges with sensor |
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 |