CN113055814A

CN113055814A - Indoor positioning method for improving WKNN

Info

Publication number: CN113055814A
Application number: CN202110160749.2A
Authority: CN
Inventors: 王昌达; 赵显�
Original assignee: Jiangsu University
Current assignee: Jiangsu University
Priority date: 2021-02-05
Filing date: 2021-02-05
Publication date: 2021-06-29
Anticipated expiration: 2041-02-05
Also published as: CN113055814B

Abstract

The invention relates to an improved WKNN indoor positioning method, which comprises the steps of performing fencing in a positioning area, establishing a reference point, selecting an AP set suitable for subsequent positioning through RSS of APs collected from the reference point, and storing related data into a fingerprint database. Recording RSS of an AP (access point) of a positioning point by using a portable device, judging a positioning area according to the same Mac address, selecting a reference point similar to the change trend of the AP value of the positioning point according to the quantity difference of peak-to-valley values between the positioning point and the reference point, selecting K reference points with the minimum Euclidean distance by a KNN (K-nearest neighbor) algorithm, giving weights to the reference points through normalization and wasserstein distance, judging whether the reference points are similar or not, giving weights to the reference points, and taking weighted average as a final result.

Description

Indoor positioning method for improving WKNN

Technical Field

The invention belongs to the field of indoor positioning, and particularly relates to an indoor positioning method for improving WKNN.

Background

At present, the location-based service is divided into indoor location and outdoor location, in the field of outdoor location, the GPS occupies a dominant position, and with the advent of Russian GLONNASS satellite navigation system and Chinese Beidou satellite navigation system, the related technology of outdoor location is mature.

In the field of indoor positioning, the signal sources can be classified into positioning technologies such as infrared rays, visible light, ultrasonic waves, ultra wide bands, bluetooth, Zigbee and WIFI. Wherein the hardware cost of infrared, ultrasonic and ultra-wideband technologies is high; the visible light and Bluetooth technologies are greatly influenced by environmental factors and are not easy to expand; the Zigbee technology requires a large amount of batteries for power supply, and the battery handling is not easy for environmental protection.

At present, two main types of methods for solving the problem of device heterogeneity in indoor positioning based on WIFI fingerprints are as follows: a calibration method and a calibration-free method. The calibration method needs to learn a relation model of reference device RSS value and user mobile terminal RSS value during calibration. The method has high time delay and cannot meet the real-time application of large-scale access of equipment. The calibration-free method is used for uniformly processing the original fingerprints to obtain new fingerprints, and although the defects of low automation degree and large time delay of the early calibration method are overcome, the corresponding fingerprint characteristics are reduced, and the precision is reduced.

Disclosure of Invention

Aiming at the problems in the prior art, the invention provides an indoor positioning method for improving the influence of the isomerism of WKNN reduction equipment on positioning accuracy, and aims to solve the problem that the influence of equipment for acquiring data in an off-line stage and different equipment used in the positioning stage on the positioning accuracy is reduced. The method comprises the steps of screening out unstable AP sets providing similar information through a new AP selection method, selecting an AP set suitable for positioning, reducing the burden of a database for subsequent positioning, selecting reference points to be positioned with the same change trend as the AP values of the positioning points, calculating Wasserstein distances between the positioning points and the reference points through K reference points selected in a KNN algorithm, and generating weights for the WKNN positioning algorithm.

In order to achieve the above object, the present invention provides an indoor positioning method for improving WKNN, including the following steps:

1) dividing an indoor scene into k positioning areas and giving labels 1, 2 and 3 … … k;

2) performing palisade in each positioning area, establishing reference point, and recording position information of the reference point, such as ith reference point coordinate p_i＝(x_i,y_i) Acquiring the RSS of the current reference point every delta T by the mobile equipment, and recording the RSS of the ith reference point as: s_i＝{S_i,1,S_i,2,...,S_i,j,...,S_i,zZ is the number of wireless access points AP, s_i,jDenotes the ith referenceThe RSS of the jth wireless access point AP received by the test point; s_i,j＝{S_i,j,1,S_i,j,2,...,S_i,j,x,...,S_i,j,tT represents the acquisition times;

3) selecting a wireless access point AP subset for positioning through a wireless access point AP selection strategy;

4) storing each reference point information as a fingerprint in a fingerprint database;

5) a user collects RSS of a positioning point through a mobile device carried by the user;

6) calculating the score of each positioning area by using the AP subset and the RSS value, judging whether the highest score is a single area, finding out the RSS value of the AP subset, and determining the positioning area to which the positioning point belongs;

7) finding out a reference point similar to the positioning point according to the quantity difference of the peak-to-valley values between the positioning point and the reference point;

8) and (5) finding k fingerprints closest to the Euclidean distance of the RSS value acquired in the step (5) in a fingerprint database, judging whether the fingerprints are similar reference points or not by utilizing the normalization and wasserstein distances, and weighting the positions corresponding to the k fingerprints and the distance weight to obtain the current position of the user.

Further, in step 3), the AP selection policy includes the following steps:

3.1) carrying out denoising treatment on the collected RSS data: for each s_i,jThe gaussian filtering is carried out and the number of the gaussian filters,

t is the number of acquisitions, and the measurement f (S)_i,j) Has a density function of

F (S) is more than or equal to 0.6_i,j)≤1,0.15σ₁+μ≤f(S_i,j)≤0.39σ₁+ mu, S in this range_i,j,gTaking out the value and averaging to obtain the current RSS value, i.e. the jth AP information is S_i＝{RSS_j,1,RSS_j,2,...,RSS_j,nN is the number of reference points;

3.2) complementing missing data: the data missing from AP at a certain reference point is RSS_j,x＝RSS_j,min-(RSS_j,min2-RSS_j,min)，RSS_j,minRepresents the minimum value of the AP among all reference points, RSS_j,min2A second smallest value among all reference points representing the AP;

3.3) set a set C_qAnd make an order

Calculating the frequency of occurrence of each AP signal and setting a threshold σ₂If, if

Discarding the current AP; otherwise enter set C_qWherein g is the number of reference points, N_jIndicating that when the reference point has RSS data, the data is larger than>-80db, then 1, otherwise 0;

3.4) set a set C_rAnd make an order

3.5) finding AP respectively_i,AP_jClosest point F, AP to source_i,AP_jPoints F of (a) are denoted as iF and jF, respectively;

3.6) setting the threshold σ₄，σ₅，

E.g. d₁＜σ₄And d is₂＜σ₅Then AP will be started_i,AP_jIs put into C_rPerforming the following steps;

3.7) set C_rIf an AP has multiple tags, resetting the AP and the AP having the same tag as the AP to have the same tag;

3.8) for C_rThe same tags in (a) are screened according to the reference point scoring capability T,for the same tag, the AP with the largest score is kept and put into the set C_qWhile the others are discarded; note AP_iScore capability at the ith reference point of

RSS_iIs the RSS value at the ith reference point_maxIs the maximum value of the RSS among all reference points of the AP_minIs the minimum value of RSS among all reference points of the AP, d_i,iFIs the distance from the ith reference point to point F; AP (Access Point)_iTotal scoring ability

3.9) set of pairs C_qAll APs in the system calculate the total score capability T of the reference points_allAnd arranged in descending order;

3.10) determining the number of the APs participating in positioning according to the size of the positioning area, and sequentially selecting a wireless access point AP subset for positioning from the AP set.

Further, the step 3.5) includes the following steps:

3.5.1) separately determining AP_i,AP_jWhether the reference points q, w, e for which the first three maximum RSS values are taken are less than the threshold value sigma₃，

d_q,wIs the distance between the positions of reference point q and reference point w, d_q,eIs the position distance of the reference point q and the reference point e, d_w,eIs the position distance of reference point w and reference point e;

3.5.2) if d_q,w<σ₃And d is_q,e<σ₃And d is_w,e<σ₃If there are only two satiations and one is not, then the source is close to a reference point common to both satiations, such as: d_q,w<σ₃And d is_q,e<σ₃But d_w,e>＝σ₃Coordinate of q (x)_q,y_q) Is F;

3.5.3) if d_q,w<σ₃And d is_q,e<σ₃And d is_w,e<σ₃Only 1 is satisfied, two are not, e.g. d_q,w<σ₃But d_q,e>σ₃,d_w,e>＝σ₃If the source is between the coordinates q and w, let F be

3.5.4) if d_q,w<σ₃And d is_q,e<σ₃And d is_w,e<σ₃If none of the three are satisfied, finding another point which can meet the requirement of 3.5.2) or 3.5.3) according to the sequence of RSS size values, and recording the left side of the point as F;

3.5.5) if d_q,w<σ₃And d is_q,e<σ₃And d is_w,e<σ₃And if all three are satisfied, the reference point coordinate with the maximum RSS value is F.

Further, the step 6) includes the steps of:

6.1) marking the maximum value and the minimum value of the AP selected by the positioning area in each positioning area respectively, and marking the AP_j,maxFor the maximum value of the jth AP, note AP_j,minIs the minimum value of the jth AP;

6.2) the RSS information of the collected locating points is selected according to the MAC address, the RSS of the AP with the same MAC address is marked as x_tp＝(RSS_tp,1,RSS_tp,2,...,RSS_tp,j,RSS_tp,m) M is the number of the APs selected in the step 3) for positioning;

6.3) set score 0 if AP_j,min≤RSS_tp,j≤AP_j,maxIf so, the score is score +1 until all the m APs are aligned, and a total score is calculated;

6.4) circularly comparing all the positioning areas according to the method, and calculating the score of each positioning area;

6.5) selecting the location area with the highest score and recording x in 6.2)_tpA value;

6.6) if the same division situation exists, observing whether the same division positioning areas are adjacent, if so, recording the same division positioning areas and the RSS values selected in the 6.2), and if not, jumping back to the step 6 for re-acquisition.

Further, the step 7) includes the steps of:

7.1) set a set C_zAnd make an order

7.2) make possible reference point decisions, let j equal 1, 2, …, m, note that all meet RSS_tp,j-1≤RSS_tp,j≤RSS_tp,j+1Or RSS_tp,j-1≤RSS_tp,j≤RSS_tp,j+1j and the number of satisfied j is recorded as g₁，

7.3) comparing all satisfied j values in step 7.2) with the ith reference point, RSS_i,j-1≤RSS_i,j≤RSS_i,j+1Or RSS_i,j-1≥RSS_i,j≥RSS_i,j+1Let g₂0, for each value of j satisfied, g₂＝g₂+1, if not, but now | RSSS_i,j-1-RSS_i,j< 5 or | RSS_i,j+1-RSS_i,jI < 5, at which time g₂＝g₂+1；

7.4) let g1-g2 ═ σ₆If, if

Put the ith reference point label into set C_zM is the number of APs selected for positioning.

Further, the step 8) includes the steps of:

when the highest score in step 6) is a single region:

8.1) setting RSS of AP with the same MAC address as RSS of anchor point to x_tpGenerating matrix x from reference point data in fingerprint database_rp＝[x₁,x₂,...,x_i,...,x_g]^TIn which x is_iIs the firsti reference points, x_i＝(RSS_i,1,RSS_i,2,...,RSS_i,j,...,RSS_i,m)，RSS_i,jRepresenting the RSS value of the jth AP at the ith reference point, m being the number of selected APs used for positioning, and the positioning point data generating a row vector x_tp＝(RSS_tp,1,RSS_tp,2,...,RSS_tp,j,RSS_tp,m)；

8.2) set a set C_pAnd make an order

8.3) calculating x separately_tpAnd Euclidean distance, x, of each reference point in the fingerprint database_tpAnd the Euclidean distance of the ith reference point in the fingerprint database

Selecting the minimum K reference points, recording the reference point labels, and storing the data of the K reference points in C_pIn, with C_zComparing the number of the same reference points with the reference point labels, and marking as U;

8.4) for C_pNormalizing the data in each reference point and positioning point to obtain a new value RSSNEW_i,j，RSSNEW_tp,j；

Wherein the RSS_i,jIs the RSS value of the jth AP at the ith reference point_i,minFor the minimum RSS value at the ith reference point_i,maxThe maximum RSS value at the ith reference point; RSS (really simple syndication) data base)_tp,jFor the RSS value of the jth AP at the anchor point, RSS_tp,minFor the minimum RSS value at the anchor point, RSS_tp,maxThe maximum RSS value at the anchor point;

8.5) calculating the location points and C_pThe Wasserstein distance of the ith reference point in is

8.6) setting the cycle times K times, and calculating each Wasserstein distance between the positioning point and each reference point;

8.7) generating weights based on Wasserstein distance

8.8) locating position

Wherein the reference number when the reference point is being calculated is C_zIn the middle, V is taken

Otherwise V gets

Z_iWeights generated for the ith Wasserstein distance, (x)_i,y_i) Is the ith reference point coordinate;

when the highest score in step 6) is not a single region

8.1) record x for each region_tpThen x_tp＝(x_tp1,x_tp2,...,x_tpc,...,x_tpL) L is the number of the positioning areas with the highest score being the same;

8.2) set a set C_p,C_p＝(C_p1,C_p2,...,C_pc,...,C_pL) And order

Selecting the minimum K reference points, recording the reference point labels, and storing the data of the K reference points in C_pcIn, with C_zComparing the number of the same reference points with the reference point labels, and marking as U;

8.4) for C_pcNormalizing the data in each reference point and positioning point to obtain a new value RSSNEW_i,j，RSSNEW_tp,j；

8.5) calculating the location points and C_pcThe Wasserstein distance of the ith reference point in is

8.7) generating weights based on Wasserstein distance

8.8) locating position

Otherwise V gets

8.9) locating position

Compared with the prior art, the method has the advantages that the change trend of the AP signal selected from the positioning point and the reference point is compared, the reference point which is possibly matched is selected, and then the weight is generated through the Wasserstein distance formed by the absolute value of the reference point, so that the problem of reduction of useful information in the relative value is solved. Secondly, aiming at the existing strategy of AP selection, the redundant information removal is too complicated, and the invention provides a simpler and more convenient method. The method has the advantages that the possible reference points to be matched are determined by comparing the AP value in the positioning points with the trend of the change of the AP value in the reference points, and the weight generated according to the Wasserstein distance is used in the WKNN positioning algorithm, so that the problem of fingerprint characteristic reduction when the original fingerprint is uniformly processed to obtain a new fingerprint is solved to a certain extent, and meanwhile, the complexity of calculation of all data points and the influence of irrelevant reference points on positioning are reduced.

Drawings

Fig. 1 is a flow chart of an indoor positioning method for improving WKNN according to the present invention.

Detailed Description

The invention will be further described with reference to the accompanying drawings and specific embodiments, as shown in fig. 1, comprising the steps of:

2) performing palisade in each positioning area, establishing reference point, and recording position information of the reference point, such as ith reference point coordinate p_i＝(x_i,y_i) Acquiring the RSS of the current reference point every delta T by the mobile equipment, and recording the RSS of the ith reference point as: s_i＝{S_i,1,S_i,2,...,S_i,j,...,S_i,zZ is the number of wireless access points AP, s_i,jIndicating the ith reference pointThe received RSS of the jth wireless access point AP; s_i,j＝{S_i,j,1,S_i,j,2,...,S_i,j,x,...,S_i,j,tT represents the acquisition times;

As a preferred embodiment of the present invention, in step 3), the AP selection policy includes the following steps:

F (S) is more than or equal to 0.6_i,j)≤1,0.15σ₁+μ≤f(S_i,j)≤0.39σ₁+ mu, S in this range_i,j,gTaking out the value and averaging it to obtain the current RSS value, i.e. the jth AP informationIs S_i＝{RSS_j,1,RSS_j,2,...,RSS_j,nN is the number of reference points;

3.3) set a set C_qAnd make an order

3.4) set a set C_rAnd make an order

3.6) setting the threshold σ₄，σ₅As a preferred embodiment of the present invention, the number of reference points in a region is generally 70-120, σ₄The value range is between 20 and 30, sigma₅The value range is between 4.5 and 7,

e.g. d₁＜σ₄And d is₂＜σ₅Then AP will be started_i,AP_jIs put intoC_rPerforming the following steps;

3.8) for C_rThe same label is screened according to the reference point scoring capability T, and for the same label, the AP with the maximum score is kept and put into a set C_qWhile the others are discarded; note AP_iScore capability at the ith reference point of

As a preferred embodiment of the present invention, in step 3.5), it is established that the F point of the signal is the closest position to the AP signal emitting source in the area, and the following two functions are performed: 1. in step 3.5), it is determined whether the two AP signals provide redundant information, and if the two F points are too close, the possibility of providing redundant information is high. 2. For 3.8) the importance of the reference points in their scoring ability.

The step 3.5) comprises the following steps:

3.5.1) separately determining AP_i,AP_jWhether the reference points q, w, e for which the first three maximum RSS values are taken are less than the threshold value sigma₃As a preferred embodiment of the invention, σ₃The value range is 4.5Between the number of the particles of the group-7,

As a preferred embodiment of the present invention, the step 6) includes the steps of:

6.1) marking the location area in each location area separatelyThe maximum value and the minimum value of the AP in the positioning area are recorded as the AP_j,maxFor the maximum value of the jth AP, note AP_j,minIs the minimum value of the jth AP;

As a preferred embodiment of the present invention, step 7) comprises the steps of:

7.1) set a set C_zAnd make an order

7.4) let g1-g2 ═ σ₆If, if

As a preferred embodiment of the present invention, step 8) comprises the steps of:

when the highest score in step 6) is a single region:

8.1) setting RSS of AP with the same MAC address as RSS of anchor point to x_tpGenerating matrix x from reference point data in fingerprint database_rp＝[x₁,x₂,...,x_i,...,x_g]^TIn which x is_iIs the ith reference point, x_i＝(RSS_i,1,RSS_i,2,...,RSS_i,j,...,RSS_i,m)，RSS_i,jRepresenting the RSS value of the jth AP at the ith reference point, m being the number of selected APs used for positioning, and the positioning point data generating a row vector x_tp＝(RSS_tp,1,RSS_tp,2,...,RSS_tp,j,RSS_tp,m)；

8.2) set a set C_pAnd make an order

8.7) generating weights based on Wasserstein distance

8.8) locating position

Otherwise V gets

when the highest score in step 6) is not a single region

8.2) set a set C_p,C_p＝(C_p1,C_p2,...,C_pc,...,C_pL) And order

8.7) generating weights based on Wasserstein distance

8.8) locating position

Otherwise V gets

8.9) locating position

Claims

1. An improved WKNN indoor positioning method is characterized by comprising the following steps:

2) performing palisade in each positioning area, establishing reference point, and recording position information of the reference point, such as ith reference point coordinate p_i＝(x_i,y_i) Acquiring the RSS of the current reference point every delta T by the mobile equipment, and recording the RSS of the ith reference point as: s_i＝{S_i,1,S_i,2,...,S_i,j,...,S_i,zZ is the number of wireless access points AP, s_i,jRepresenting the RSS of the jth wireless access point AP received by the ith reference point; s_i,j＝{S_i,j,1,S_i,j，2,...，S_i，j，x，...，S_i，j，tT represents the acquisition times;

2. The method as claimed in claim 1, wherein the AP selection strategy in step 3) comprises the following steps:

F (S) is more than or equal to 0.6_i,j)≤1,0.15σ₁+μ≤f(S_i,j)≤0.39σ₁+ mu, S in this range_i,j,gThe value extraction is averaged intoThe current RSS value, i.e. the jth AP information is S_j＝{RSS_j,1,RSS_j,2,...,RSS_j,nN is the number of reference points;

3.2) complementing missing data: the data missing at the xth reference point by the AP is RSS_j,x＝RSS_j,min-(RSS_j,min2-RSS_j，min)，RSS_j，minRepresents the minimum value of the AP among all reference points, RSS_j，min2A second smallest value among all reference points representing the AP;

3.3) set a set C_qAnd make an order

3.4) set a set C_rAnd make an order

3.6) setting the threshold σ₄，σ₅，

3. The method of claim 2, wherein said step 3.5) includes the steps of:

3.5.2) if d_q,w<σ₃And d is_q,e<σ₃And d is_w,e<σ₃If there are only two satiations and one is not, then the source approaches a parameter common to both satiationsExamination points, such as: d_q,w<σ₃And d is_q,e<σ₃But d_w,e>＝σ₃Coordinate of q (x)_q,y_q) Is F;