CN101587182A - Locating method for RFID indoor locating system - Google Patents

Abstract

The present invention discloses a locating method for a RFID indoor locating system. The method includes that providing L of readers, placing M of actual reference label(s) and N of tracking label(s) according to a two-dimensional quadrate grid distribution, and calculating positions of virtual reference labels and corresponding RSSI of each reader through a virtual grid method; adjusting a suitable threshold based on the RSSI distribution status of the virtual labels adaptively, searching a neighboring virtual label of the tracking label by comparing the threshold with a RSSI difference value of the tracking label and the virtual label, implementing a cluster analysis to the selected neighboring virtual label, and determining the position of the tracking label accurately according to the position of the neighboring virtual label, a signal Euclidean distance weight value and a generic weight value thereof. By means of using a virtual label to replace an additional actual label, it is capable of reducing a system cost, eliminating a small probability position by regulating the threshold adaptively, enhancing the systemic flexibility, and improving a locating accuracy by the cluster analysis of a large probability position.

Description

A kind of localization method that is used for the RFID indoor locating system

Technical field

The present invention relates to a kind of localization method of the RFID of being used for indoor locating system, be specifically related to a kind of tracking tags localization method of the RFID of being used for indoor locating system.

Background technology

Radio-frequency (RF) identification (RFID) is a kind of automatic identification technology that begins to rise the nineties in 20th century, and its ultimate principle is to utilize the transport property of radiofrequency signal and space coupling (inductance or electromagnetic coupled), realizes by the automatic identification of certain objects.The RFID technical characterstic is as follows: noncontact, non line of sight, prolong in short-term, high precision, transmission range is big and cost is low.At present, adopt the positioning system of RFID technology mainly to comprise SpotON and LANDMARC.

SpotON: by using aggregation algorithms radio-frequency (RF) signal strength is analyzed, realized the Three-dimension Target location.But the data integration of system core and sensor fusion techniques are also among research, and its precision is not as good as the signal transmission delay method.

LANDMARC: adopt the reference label of extra fixed position to help position correction, these reference label are as the reference point in the system.The arrangement method of reader and reference label has very important influence to the overall accuracy of system.

LANDMARC system location model is easily realized, uses commonplace.But it has two shortcomings: the first, in sealing and there is in the environment of serious multipath effect locating effect unsatisfactory; The second, need intensive reference label to help estimate the position for improving precision, improve cost and may cause interference phenomenon.The present invention has proposed a kind of new localization method just on the basis of its deficiency.

Summary of the invention

The objective of the invention is to overcome the shortcoming of prior art, under the prerequisite that does not increase the additional reference label, adopt virtual label to replace physical tags, provide a kind of bearing accuracy height, the localization method that is used for the RFID indoor locating system that cost is low.

Purpose of the present invention is achieved through the following technical solutions:

Be used for the localization method of RFID indoor locating system, comprise the steps:

Step (1) is placed M actual reference label according to the square net distribution of two dimension, be divided into big virtual grid unit such as n * n by each physical grid unit that any 4 actual reference label covered, each virtual grid unit is covered by 4 virtual reference labels, and N tracking tags is placed on arbitrarily in the physical grid unit of M actual reference label covering; Wherein, M, N and n are integer, M 〉=4, n 〉=2, N 〉=1;

Step (2) is according to the positional information of M actual reference label and the regular distribution of virtual label, make all labels and reading device position all be in the coordinate axis first quartile and set initial point and coordinate axis, and and then the coordinate position of definite virtual label sum V and all virtual labels:

V＝[m ₁+(m ₁-1)×(n-1)]×[m ₂+(m ₂-1)×(n-1)]

$T_{i,b}=T_{a,b}+i\×\frac{T_{a+n,b}-T_{a,b}}{n}$

$T_{a,j}=T_{a,b}+j\×\frac{T_{a,b+n}-T_{a,b}}{n}$

Wherein actual reference label is put on two dimensional surface and is vertical m ₁Row, laterally m ₂OK; (a, b) the vertical a of expression is listed as the capable actual reference label of horizontal b, a ∈ (1, m ₁), b ∈ (1, m ₂); T _{A, b}, T _{A+n, b}And T _{A, b+n}The coordinate of representing actual reference label, T _{I, b}Denotation coordination is T _{A, b}And T _{A+n, b}Actual reference label between the coordinate of i virtual label, T _{A, j}Denotation coordination is T _{A, b}And T _{A, b+n}Actual reference label between the coordinate of j virtual label, parameter i ∈ (0, n-1), j ∈ (0, n-1);

Step (3) is gathered received signal intensity RSSI, L=2,3,4 or 5 to M actual reference label and N tracking tags respectively by L reader;

Step (4) adopts linear interpolation algorithm to determine the signal intensity of all virtual labels to the signal intensity RSSI of M actual reference label, and wherein the signal intensity of lateral rows virtual label is on two dimensional surface:

${\mathrm{\θ}}_k\left(T_{i,b}\right)={\mathrm{\θ}}_k\left(T_{a,b}\right)+i\×\frac{{\mathrm{\θ}}_k\left(T_{a+n,b}\right)-{\mathrm{\θ}}_k\left(T_{a,b}\right)}{n}$

The signal intensity of longitudinal row virtual label is:

${\mathrm{\θ}}_k\left(T_{a,j}\right)={\mathrm{\θ}}_k\left(T_{a,b}\right)+j\×\frac{{\mathrm{\θ}}_k\left(T_{a,b+n}\right)-{\mathrm{\θ}}_k\left(T_{a,b}\right)}{n}$

θ wherein _k(T _{A, b}), θ _k(T _{A+n, b}) and θ _k(T _{A, b+n}) respectively denotation coordination be T _{A, b}, T _{A+n, b}And T _{A, b+n}Actual reference label by k reader read the RSSI value, θ _k(T _{I, b}) and θ _k(T _{A, j}) respectively denotation coordination be T _{I, b}And T _{A, j}Virtual label by k reader read the RSSI value;

Step (5) determine tracking tags adjacent tags number range R (M, V)=(1, KV/M), wherein K is the parameters optimization in the LANDMARC method; Signal intensity self-adaptation according to virtual label is adjusted threshold value, and described self-adaptation is adjusted threshold value and is meant, is setting initial threshold value ds ₀On the basis, the relatively signal intensity of tracking tags and the signal intensity of virtual label respectively are when the signal strength difference of tracking tags and virtual label is not more than described initial threshold value ds ₀The time, select this virtual label as the contiguous virtual label of the plan of tracking tags; L reader selects to have the L group to intend contiguous virtual label, and the contiguous virtual label of the plan that chosen position is identical is as the adjacent tags of tracking tags; Judge this moment adjacent tags count t and R (M, V) relation of scope: if t＞(KV/M), adjustment threshold value ds ₀=ds ₀-tn * ds ₀/ V continues to compare the signal intensity of tracking tags and the signal intensity of virtual label, seeks the contiguous virtual label of tracking tags; If threshold value ds is adjusted in t＜1 ₀=ds ₀+ ds ₀/ V continues to compare the signal intensity of tracking tags and the signal intensity of virtual label, seeks the contiguous virtual label of tracking tags; Otherwise, stop ds ₀Successively decrease; Wherein tn=t/ (KV/M) rounds;

Step (6) is carried out cluster analysis to the contiguous virtual label of selecting, and described cluster analysis is that the contiguous virtual label that will abut against together is divided into a class, the weights p of each class _mFor:

$p_m=\frac{n_{\mathrm{cm}}}{{\mathrm{\Σ}}_{m=1}^{n_s}{n}_{\mathrm{cm}}}$

N wherein _CmRepresent adjacency contiguous virtual label number together in the m class, n _sRepresent the class sum, and m ∈ (1, n _s);

Step (7) is determined the weights of each class virtual label $w_q=\frac{1/E_q^2}{{\mathrm{\&Sigma;}}_{q=1}^{{\mathrm{<figure-callout\; id="1"\; label="n\; cm"\; filenames="A2009100405710016H1.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{cm}}}1/E_q^2},$ Q ∈ (1, n _Cm), E wherein _qIt is the signal intensity Euclidean distance of q virtual label and tracking tags; According to weights p _mAnd w _q, determine that the location estimation of tracking tags is:

$(x,y)={\mathrm{\&Sigma;}}_{m=1}^{n_s}{p}_m(x_m,y_m)={\mathrm{\&Sigma;}}_{m=1}^{n_s}{p}_m{\mathrm{\&Sigma;}}_{q=1}^{{\mathrm{<figure-callout\; id="1"\; label="n\; cm"\; filenames="A2009100405710016H1.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{cm}}}\frac{1/{E_q}^2}{{\mathrm{\&Sigma;}}_{q=1}^{{\mathrm{<figure-callout\; id="1"\; label="n\; cm"\; filenames="A2009100405710016H1.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{cm}}}1/{E_q}^2}(x_q,y_q)$

(x wherein _q, y _q) be the coordinate position of contiguous virtual label, (x _m, y _m) be the estimated position of tracking tags in the m class.

In the said method, the cluster analysis of step (6) specifically is that the contiguous virtual label with adjacency is divided into and is meant, in the contiguous virtual label of selecting, with the virtual label of grid adjacency as a class, wherein, square net of each virtual label covering.

In the said method, step (2) is set initial point and coordinate axis is meant, laterally the actual reference label of row is impartial to the distance of abscissa axis arbitrarily, and vertically the actual reference label of row is impartial to the distance of axis of ordinates arbitrarily.

Compared with prior art, the present invention has following advantage and beneficial effect:

(1) reduce system cost: the present invention adopts the virtual grid method to determine virtual label, replaces actual reference label, reduces the physical tags number greatly, has reduced system cost.

(2) enhanced system dirigibility: the present invention proposes a kind of new position and get rid of algorithm, each tracking tags is according to the RSSI of oneself and the RSSI of virtual label, self-adaptation is adjusted threshold value, get rid of the small probability position, avoid the big probability position that remains too much or very fewly cause bigger positioning error.

(3) improve bearing accuracy: the present invention adopts the big probability of clustering procedure analysis position, and the contiguous virtual label of adjacency is many more, and its weights are big more; The signal intensity Euclidean distance of virtual label and tracking tags is more little in similar, and its weights are also big more.Unite two class weights and can improve bearing accuracy.

Description of drawings

Fig. 1 is the arrangenent diagram that embodiment 1 is used in reader in the indoor environment, reference label.

Fig. 2 is to the cluster analysis synoptic diagram of the contiguous virtual label selected in the position fixing process of the present invention.

Fig. 3 is a localization process process flow diagram flow chart of the present invention.

Fig. 4 is a threshold value Adaptive adjusting algorithm process flow diagram in the position fixing process of the present invention.

Embodiment

The present invention is further illustrated below in conjunction with drawings and Examples, but the scope of protection of present invention is not limited to the scope of embodiment statement.

In the warehouse of 6m * 6m * 2.55m (length), the localization method that is used for the RFID indoor locating system can be estimated the position (tracking tags is embedded in article) of article, as shown in Figure 1, M=16 actual reference label placed in square net distribution according to two dimension, be spaced apart 1.2m between adjacent each label, N=6 tracking tags is placed on arbitrarily in the net region, and L=4 reader is placed on four corners, room.To further be divided into the big virtual grid unit of the individual grade of n * n (3 * 3) by 4 each physical grid unit that actual reference label covered, each virtual grid unit is covered by 4 virtual reference labels.

Selecting M=16, is that 16 actual reference label can be formed complete square net figure, also can select M=4,6,9,12 etc., form complete square net figure; 1.2m is selected at interval between the adjacent actual reference label, also can select 2m, determines according to actual needs at interval; Selecting n=3 is one of result of optimized choice, too small as n, can not embody the feature of the inventive method, and excessive as n, bearing accuracy there is no obvious raising, therefore can select n=4,5,6; Selecting L=4 is the parameters optimization of LANDMARC method.

By 4 readers reference label and tracking tags are gathered its received signal intensity RSSI, again test data is sent to server.Server is handled test data according to location algorithm, the position of estimation tracking tags.

According to the test data that system layout situation and the server of Fig. 1 receives, the process flow diagram of server process data estimation tracking tags position as shown in Figure 3, detailed process is as follows:

(1) according to the rule of putting of M=16 actual reference label: vertical m ₁=4 row, horizontal m ₂=4 row and virtual label correlation parameter n=3, obtain virtual label sum V=100 according to following formula:

V＝[m ₁+(m ₁-1)×(n-1)]×[m ₂+(m ₂-1)×(n-1)]

With coordinate axis and the initial point that Fig. 1 sets, bottom-up horizontal first row is 1.2m to the abscissa axis distance, and vertical from left to right first is listed as the axis of ordinates distance is 1.2.m.Then the coordinate of first actual reference label of the lower left corner is T _{A, b}=(1.2m, 1.2m), this actual reference label laterally the right first and vertically directly over the coordinate of first actual reference label be respectively T _{A+n, b}=(2.4m, 1.2m) and T _{A, b+n}=(1.2m, 2.4m), (a, b) the vertical a of expression is listed as the capable actual reference label of horizontal b, a, b=1,2,3,4; T _{I, b}Denotation coordination is T _{A, b}And T _{A+n, b}Actual reference label between the coordinate of i virtual label, T _{A, j}Denotation coordination is T _{A, b}And T _{A, b+n}Actual reference label between the coordinate of j virtual label, parameter i ∈ (0, n-1), j ∈ (0, n-1).Obtain coordinate according to following formula and be respectively T _{A, b}And T _{A+n, b}Actual reference label between laterally the coordinate position of all virtual labels be (1.2m, 1.2m), (1.6m, 1.2m), (2.0m, 1.2m):

$T_{i,b}=T_{a,b}+i\&times;\frac{T_{a+n,b}-T_{a,b}}{n}$

Obtain coordinate according to following formula and be respectively T _{A, b}And T _{A, b+n}Actual reference label between vertically the coordinate position of all virtual labels be (1.2m, 1.2m), (1.2m, 1.6m), (1.2m, 2.0m):

$T_{a,j}=T_{a,b}+j\&times;\frac{T_{a,b+n}-T_{a,b}}{n}$

The coordinate position of other virtual labels takes above-mentioned same method to calculate.

(2) obtain the RSSI of actual reference label and tracking tags, take linear interpolation algorithm to obtain the RSSI of virtual label.Learn by first reader and read: the RSSI of first actual reference label of the lower left corner is 8.86dBm, this actual reference label laterally the right first and vertically directly over the RSSI of first actual reference label be respectively 15.71dBm and 15.40dBm, the RSSI of the 5th tracking tags is 21.84dBm.Adopt following formula to obtain coordinate and be respectively T _{A, b}And T _{A+n, b}Actual reference label between laterally the RSSI of all virtual labels (corresponding first reader) be 8.86dBm, 11.14dBm, 13.43dBm:

${\mathrm{\&theta;}}_k\left(T_{i,b}\right)={\mathrm{\&theta;}}_k\left(T_{a,b}\right)+i\&times;\frac{{\mathrm{\&theta;}}_k\left(T_{a+n,b}\right)-{\mathrm{\&theta;}}_k\left(T_{a,b}\right)}{n}$

Obtain coordinate with following formula and be respectively T _{A, b}And T _{A+n, b}Actual reference label between vertically all are virtual

RSSI (corresponding first reader) is 8.86dBm, 11.04dBm, and 13.22dBm:

${\mathrm{\&theta;}}_k\left(T_{a,j}\right)={\mathrm{\&theta;}}_k\left(T_{a,b}\right)+j\&times;\frac{{\mathrm{\&theta;}}_k\left(T_{a,b+n}\right)-{\mathrm{\&theta;}}_k\left(T_{a,b}\right)}{n}$

Middle θ _k(T _{A, b}), θ _k(T _{A+n, b}) and θ _k(T _{A, b+n}) respectively denotation coordination be T _{A, b}, T _{A+n, b}And T _{A, b+n}Actual ginseng

By k reader read the RSSI value, θ _k(T _{I, b}) and θ _k(T _{A, j}) respectively denotation coordination be T _{I, b}And T _{A, j}

Label by k reader read the RSSI value.The above-mentioned same method of taking of the virtual label in other square nets is calculated.

(3) adjust threshold value according to the RSSI self-adaptation of virtual label, the vicinity of searching tracking tags is virtual learns that the adjacent tags number of the 5th tracking tags was 20 (detailed process as described in Figure 4).

(4) the contiguous virtual label of selecting is carried out cluster analysis, the contiguous virtual label of adjacency is divided into the weights p that asks for each class _mAccording to virtual label and tracking tags signal intensity in each class

From the weight w of asking this contiguous virtual label _qLearn that the adjacent tags of the 5th tracking tags can divide

Weights p ₁=0.2, p ₂=0.8.Position by contiguous virtual label, signal Euclidean distance weights and Weights can accurately try to achieve the position of tracking tags.

(5) judge whether to carry out the next round location, if change (2); Otherwise provide the end position fixing process of all tracking tags.

Flow process is shown in Figure 4, adjusts the concrete mistake of threshold value according to the RSSI distribution situation self-adaptation of virtual label

(1) counts M and virtual label sum V according to actual reference label, to the final neighbour who selects of tracking tags

Number determine scope R (M, V)=(1, KV/M)=(1,25), wherein K is at LANDMARC

The adjacent tags number of optimized choice, K=4;

(2) to i reader, ask for the RSSI difference of virtual label and tracking tags, form the RSSI difference value vector (d of i reader _I1, d _I2..., d _IV), the corresponding virtual label of RSSI difference.For each RSSI difference is provided with sign flag=1, select the initial threshold value ds of wherein maximum RSSI difference as this reader _I0Initial threshold value vector (ds at 4 readers ₁₀, ds ₂₀, ds ₃₀, ds ₄₀) in, select minimum value to adjust the initial threshold value ds of threshold value as self-adaptation ₀, the ds when learning the 5th tracking tags in location ₀=9.84;

(3) to i reader, comparison threshold value ds ₀With the RSSI difference of tape identification flag=1, get rid of greater than ds ₀The RSSI difference, promptly get rid of pairing virtual label.The virtual label that relatively remains in 4 groups of virtual labels again, the identical virtual label of chosen position is as the adjacent tags of tracking tags, and number is t, and keeps corresponding RSSI difference sign flag=1, puts other RSSI difference sign flag=0;

(4) relatively t and R (M, V), if t＞(KV/M), ds ₀=ds ₀-tn * ds ₀/ V, execution in step (3); If t＜1, ds ₀=ds ₀+ ds ₀/ V, execution in step (3); Otherwise, stop ds ₀Successively decrease.Wherein tn=t/ (KV/M) rounds.Learn the adjacent tags number t=20 of the 5th tracking tags after the threshold value adjustment.

As Fig. 2 the contiguous virtual label of final selection is carried out cluster and divide, detailed process is as follows:

(1) the contiguous virtual label with adjacency is divided into a class, the contiguous virtual label of adjacency is divided into a class is meant, in the contiguous virtual label of selecting, with the virtual label of grid adjacency as a class, wherein, square net of each virtual label covering; As class C among the figure ₁, C ₂, C ₃, wherein, C ₁The contiguous virtual label that 6 adjacency are arranged, C ₂The contiguous virtual label that 3 adjacency are arranged, C ₃The contiguous virtual label that has only 1 adjacency; The weights p of each class _mFor:

$p_m=\frac{n_{\mathrm{cm}}}{{\mathrm{\&Sigma;}}_{m=1}^{n_s}{n}_{\mathrm{cm}}}$

N wherein _CmRepresent adjacency contiguous virtual label number together in the m class, n _sRepresent the class sum, and m ∈ (1, n _s); According to weights p _mFormula obtain

The adjacent tags of learning the 5th tracking tags can be divided into 2 classes, weights p ₁=0.2, p ₂=0.8;

(2) the weights basis of virtual label in each class $w_q=\frac{1/E_q^2}{{\mathrm{\&Sigma;}}_{q=1}^{{\mathrm{<figure-callout\; id="1"\; label="n\; cm"\; filenames="A2009100405710016H1.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{cm}}}1/E_q^2}$ (q ∈ (1, n _Cm)) obtain E wherein _qIt is the signal intensity Euclidean distance of q virtual label and tracking tags; Learn the class weights p of the 5th tracking tags ₁=0.2 adjacent tags signal Euclidean distance weight w ₁～w ₄=0.2249,0.2310,0.2566,0.2875, class weights p ₂=0.8 adjacent tags signal Euclidean distance weight w ₁～w ₁₆=0.0343,0.0368,0.0554,0.0429,0.0366,0.0867,0.1096,0.0778,0.1283,0.0762,0.0518,0.0739,0.0523,0.0557,0.0364,0.0455;

(3) according to weights p _mAnd w _q, the location estimation formula of substitution tracking tags is determined the estimated position of tracking tags.The location estimation position of determining tracking tags is:

$(x,y)={\mathrm{\&Sigma;}}_{m=1}^{n_s}{p}_m(x_m,y_m)={\mathrm{\&Sigma;}}_{m=1}^{n_s}{p}_m{\mathrm{\&Sigma;}}_{q=1}^{n_{\mathrm{cm}}}{w}_q(x_q,y_q)$

The location estimation that obtains the 5th tracking tags is (2.8023,2.8259), and its physical location is (3.0000,3.6000).

According to error formula $e=\sqrt{{(x-x_0)}^2+{(y-y_0)}^2}$ Calculate the Estimated Position Error of each tracking tags, obtain new method positioning result, wherein (x as following table 1 ₀, y ₀) be the physical location of tracking tags.After testing, the method for the present invention's proposition has on average improved 26% than LANDMARC method bearing accuracy.

Table 1

Localization method	Average error	Least error	Maximum error
				The inventive method	0.53m	0.03m	1.16m
LANDMARC	0.72m	0.17m	1.33m

Above-described specific embodiment; purpose of the present invention, technical scheme and beneficial effect have been carried out further detailed description; institute is understood that; the above only is specific embodiments of the invention; be not restricted to the present invention; all in the spirit and principles in the present invention and so on, any modification of being made, be equal to replacement, improvement etc., all within protection scope of the present invention.

Claims (3)

1, a kind of localization method that is used for the RFID indoor locating system is characterized in that comprising the steps:

Step (1) is placed M actual reference label according to the square net distribution of two dimension, be divided into big virtual grid unit such as n * n by each physical grid unit that any 4 actual reference label covered, each virtual grid unit is covered by 4 virtual reference labels, and N tracking tags is placed on arbitrarily in the physical grid unit of M actual reference label covering; Wherein, M, N and n are integer, M 〉=4, n 〉=2, N 〉=1;

Step (2) is according to the positional information of M actual reference label and the regular distribution of virtual label, make all labels and reading device position all be in the coordinate axis first quartile and set initial point and coordinate axis, and and then the coordinate position of definite virtual label sum V and all virtual labels:

V＝[m ₁+(m ₁-1)×(n-1)]×[m ₂+(m ₂-1)×(n-1)]

$T_{i,b}=T_{a,b}+i\&times;\frac{T_{a+n,b}-T_{a,b}}{n}$

$T_{a,j}=T_{a,b}+j\&times;\frac{T_{a,b+n}-T_{a,b}}{n}$

Wherein actual reference label is put on two dimensional surface and is vertical m ₁Row, laterally m ₂OK; (a, b) the vertical a of expression is listed as the capable actual reference label of horizontal b, a ∈ (1, m ₁), b ∈ (1, m ₂); T _{A, b}, T _{A+n, b}And T _{A, b+n}The coordinate of representing actual reference label, T _{I, b}Denotation coordination is T _{A, b}And T _{A+n, b}Actual reference label between the coordinate of i virtual label, T _{A, j}Denotation coordination is T _{A, b}And T _{A, b+n}Actual reference label between the coordinate of j virtual label, parameter i ∈ (0, n-1), j ∈ (0, n-1);

Step (3) is gathered received signal intensity RSSI, L=2,3,4 or 5 to M actual reference label and N tracking tags respectively by L reader;

Step (4) adopts linear interpolation algorithm to determine the signal intensity of all virtual labels to the signal intensity RSSI of M actual reference label, and wherein the signal intensity of lateral rows virtual label is on two dimensional surface:

${\mathrm{\&theta;}}_k\left(T_{i,b}\right)={\mathrm{\&theta;}}_k\left(T_{a,b}\right)+i\&times;\frac{{\mathrm{\&theta;}}_k\left(T_{a+n,b}\right)-{\mathrm{\&theta;}}_k\left(T_{a,b}\right)}{n}$

The signal intensity of longitudinal row virtual label is:

${\mathrm{\&theta;}}_k\left(T_{a,j}\right)={\mathrm{\&theta;}}_k\left(T_{a,b}\right)+j\&times;\frac{{\mathrm{\&theta;}}_k\left(T_{a,b+n}\right)-{\mathrm{\&theta;}}_k\left(T_{a,b}\right)}{n}$

θ wherein _k(T _{A, b}), θ _k(T _{A+n, b}) and θ _k(T _{A, b+n}) respectively denotation coordination be T _{A, b}, T _{A+n, b}And T _{A, b+n}Actual reference label by k reader read the RSSI value, θ _k(T _{I, b}) and θ _k(T _{A, j}) respectively denotation coordination be T _{I, b}And T _{A, j}Virtual label by k reader read the RSSI value;

Step (5) determine tracking tags adjacent tags number range R (M, V)=(1, KV/M), wherein K is the parameters optimization in the LANDMARC method; Signal intensity self-adaptation according to virtual label is adjusted threshold value, and described self-adaptation is adjusted threshold value and is meant, is setting initial threshold value ds ₀On the basis, the relatively signal intensity of tracking tags and the signal intensity of virtual label respectively are when the signal strength difference of tracking tags and virtual label is not more than described initial threshold value ds ₀The time, select this virtual label as the contiguous virtual label of the plan of tracking tags; L reader selects to have the L group to intend contiguous virtual label, and the contiguous virtual label of the plan that chosen position is identical is as the adjacent tags of tracking tags; Judge this moment adjacent tags count t and R (M, V) relation of scope: if t＞(KV/M), adjustment threshold value ds ₀=ds ₀-tn * ds ₀/ V continues to compare the signal intensity of tracking tags and the signal intensity of virtual label, seeks the contiguous virtual label of tracking tags; If threshold value ds is adjusted in t＜1 ₀=ds ₀+ ds ₀/ V continues to compare the signal intensity of tracking tags and the signal intensity of virtual label, seeks the contiguous virtual label of tracking tags; Otherwise, stop ds ₀Successively decrease; Wherein tn=t/ (KV/M) rounds;

Step (6) is carried out cluster analysis to the contiguous virtual label of selecting, and described cluster analysis is that the contiguous virtual label that will abut against together is divided into a class, the weights p of each class _mFor:

$p_m=\frac{n_{\mathrm{cm}}}{{\mathrm{\&Sigma;}}_{m=1}^{n_s}{n}_{\mathrm{cm}}}$

N wherein _CmRepresent adjacency contiguous virtual label number together in the m class, n _sRepresent the class sum, and m ∈ (1, n _s);

Step (7) is determined the weights of each class virtual label $w_q=\frac{1/E_q^2}{{\mathrm{\&Sigma;}}_{q=1}^{n_{\mathrm{cm}}}1/E_q^2},$ Q ∈ (1, n _Cm), E wherein _qIt is the signal intensity Euclidean distance of q virtual label and tracking tags; According to weights p _mAnd w _q, determine that the location estimation of tracking tags is:

$(x,y)={\mathrm{\&Sigma;}}_{m=1}^{n_s}{p}_m(x_m,y_m)={\mathrm{\&Sigma;}}_{m=1}^{n_s}{p}_m{\mathrm{\&Sigma;}}_{q=1}^{n_{\mathrm{cm}}}\frac{1/{E_q}^2}{{\mathrm{\&Sigma;}}_{q=1}^{n_{\mathrm{cm}}}1/{E_q}^2}(x_q,y_q)$

(x wherein _q, y _q) be the coordinate position of contiguous virtual label, (x _m, y _m) be the estimated position of tracking tags in the m class.

2, according to right 1 described method, it is characterized in that described step (6) is divided into the contiguous virtual label of adjacency is meant, in the contiguous virtual label of selecting, with the virtual label of grid adjacency as a class, wherein, each virtual label covers a square net.

3, according to right 1 described method, it is characterized in that described step (2) is set initial point and coordinate axis is meant that laterally the actual reference label of row is impartial to the distance of abscissa axis arbitrarily, vertically the actual reference label of row is impartial to the distance of axis of ordinates arbitrarily.

Images