CN105960011B - Indoor objects localization method based on Sensor Network and bayes method - Google Patents
Indoor objects localization method based on Sensor Network and bayes method Download PDFInfo
- Publication number
- CN105960011B CN105960011B CN201610244430.7A CN201610244430A CN105960011B CN 105960011 B CN105960011 B CN 105960011B CN 201610244430 A CN201610244430 A CN 201610244430A CN 105960011 B CN105960011 B CN 105960011B
- Authority
- CN
- China
- Prior art keywords
- value
- bayes
- received signal
- signal strength
- target
- 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.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
- H04W64/006—Locating users or terminals or network equipment for network management purposes, e.g. mobility management with additional information processing, e.g. for direction or speed determination
Abstract
The invention discloses a kind of indoor objects localization method based on Sensor Network and bayes method classifies to the received signal strength indication for monitoring target using bayes classification method, realizes positioning function.Use the low data rate of 2.4GHz (250kbps) wireless sense network as locating infrastructure, the positioning of Target is directly completed using the changing pattern of RSSI (received signal strength), without considering the error problem of ranging and fingerprint matching, therefore there is system to lay the advantages such as simple, locating speed is fast.
Description
Technical field
The invention belongs to indoor objects to be set to technical field, be related to a kind of indoor mesh based on Sensor Network and bayes method
Mark localization method.
Background technique
It has a wide range of applications taking human as the indoor positioning of target in business and public safeties field.Since GPS satellite is fixed
Environment is difficult to obtain effective positioning accuracy position signal indoors, or even is not available, at present the indoor locating system of mainstream according to
Rely and lay indoors wireless anchor node (Anchor node), monitors the signal that target (Target node) issues, and then pass through three
Angle positioning or fingerprint location strategy carry out position solution.For the triangulation in plane domain, need to know Target
To the physical distance (as shown in Figure 1) of at least three Anchor, according to three hyp friendships centered on three Anchor
Put the position to determine Target.The stringent clock that triangulation closely relies between the precision and Anchor of ranging is same
Step carries out Target confirmation using voice signal, the synchronous error of 1ms also results in the error of 0.34m in ranging, and equation
Solution can further accumulate range error, so that last position error is difficult to control, may reach rice, even more than ten meters
Rank, it is therefore desirable to the controlling mechanism of special hardware or complexity.Fingerprint positioning method needs to generate one before positioning
Finger print data collection can be matched to complete ranging in positioning according to collected Target fingerprint with finger print data collection
Or it is done directly positioning.
In existing indoor orientation method, commonly assume that there are multiple Wi-Fi hotspots as Anchor.However, from implementation
Angle sees that this mode faces both sides problem.First, it is not total although Wi-Fi hotspot has been laid in advance in some places
Energy gain access, therefore RSSI (received signal strength) value can not be obtained；In addition, for some places (be related to safety and
Privacy), do not allow additionally to lay Wi-Fi hotspot；Therefore, WiFi has certain limitation as the infrastructure of indoor positioning.The
Two, wireless signal is easy the interference by indoor complex environment, and is in nonlinear attenuation feature, (it is strong to receive signal using RSSI
Degree) variation carry out ranging often bring it is difficult to predict error.
In recent years, it in fingerprint location research, is proposed in succession based on data mining, based on the fingerprint location of probability statistics
Strategy.Above-mentioned work has used WiFi signal, needs to generate fingerprint base, and the precision of One-Point Location by largely training
It is limited, need to design complicated algorithm to complete the matching of actual measurement sample to fingerprint base；This is mainly by WiFi signal itself
Caused by feature.Comparatively, although the radio signal attenuation of the low-power consumption of Sensor Network, low data rate is obvious, it is easy area
Point, hardware cost is relatively low, and system is laid more flexible.
Summary of the invention
The object of the present invention is to provide a kind of indoor objects localization method based on Sensor Network and bayes method, solves
The problems such as implementation existing in the prior art is complicated, clock stepped cost is high, positioning accuracy is low, stability is poor.
The technical scheme adopted by the invention is that a kind of indoor objects positioning side based on Sensor Network and bayes method
Method classifies to the received signal strength indication for monitoring target using bayes classification method, realizes positioning function.
It is of the invention to be further characterized in that, further, specifically follow the steps below:
Step 1,
Using IEEE802.15.4 standard, it is assumed that containing n wireless anchor nodes in Sensor Network, monitor target and have sent detection
Wrap Pi, will form record: R in a networki={ ri1,ri2,…,rij,rin, wherein RiExpression is acquired by wireless anchor node
The detection packet P arrivediThe sequence of values that received signal strength indication is constituted, corresponds to the Yi Tiaobao in computing terminal database
Record containing n attribute value, rijIt is the detection packet P for numbering the wireless anchor node for being j and calculatingiReceived signal strength indication,
Received signal strength indication is sent to acquisition terminal by multi-hop wireless link by each wireless anchor node in real time；
In the training stage, monitors target and be respectively in application scenarios in prior ready-portioned trained region, carry and monitor
The subject of target in each trained region by a small margin, slow random movement 30 seconds to 1 minute, monitor target interval 64 to
256 milliseconds send primary detection packet (frequency of giving out a contract for a project that current mainstream radio node can support at least 1KHz)；In general,
To given trained region, in order to guarantee training precision, Target node should at least send 100 detection packets (can be according to positioning
The actual size in region come adjust subject traveling time and detection packet time interval of giving out a contract for a project)；Complete all trained regions
It is every record one new attribute of increase, i.e. position attribution after interior received signal strength indication is collected, as class label,
For next stage training Bayes classifier；
Step 2,
Use the exceptional value of k- mean value strategy identification received signal strength indication:
In formula (a), p is the point in received signal strength indication data set, ciIt is cluster CiCentroid, d (p, ci) indicate p and
ciBetween Euclidean distance；A monitoring target position is given, after obtaining all received signal strength indications, is believed from receiving
K point, and the centroid as k cluster are randomly choosed in number intensity value data set；In next each iteration, each cluster makes
The new mean value of the calculation and object of the cluster assigned by last iteration, then uses updated mean value as new centroid, repeatedly
Until in generation, can be continued until that distribution is stablized；
Step 3,
Assuming that R is for training Bayes's location classifier, the training set with class label, its record is indicated
For R={ r1,r2,...rn, riIndicate ith attribute value；There is m trained region in entire located space, is denoted as L respectively1,
L2...Lm, the core missions of Bayes's location classifier are exactly to calculate Pr (Li| R) maximum posteriori probability value and corresponding position
Class label；According to Bayes' theorem, formula (b) gives calculation:
It is a constant for given R ∈ R, Pr (R) in training set R；Assume that monitoring target is in positioning sky simultaneously
Between in the prior probability in each trained region be equal, i.e. Pr (Li)=Pr (Lj), wherein 1≤i < j≤m, then Bayes position
Set classifier only and need to consider maximize Pr (R | Li)；When wireless anchor node quantity increases, the dimension of R also can be mentioned accordingly
It is high；Therefore, in order to reduce Pr (R | Li) computing cost, improve system locating speed, it is assumed that it is any record R attribute between
And any dependence, i.e. r is not presentiIndependently of rj, wherein 1≤i < j≤n；Therefore searched for using formula (c) Pr (R | Li)
Maximum value:
Pr(R|Li)=Pr (r1|Li)×Pr(r2|Li)×…Pr(rn|Li) (c)
In positioning application, the distribution in training region is discrete, but RSSI value is bounded and continuous data, according to public affairs
Formula (d) finds out the probability distribution of each category value needed for formula (c) by gauss of distribution function:
Wherein, uLiWith σLiL is used respectivelyiIn all records j-th of attribute (rj) value mean value and standard deviation replace.
Beneficial effects of the present invention: the present invention uses the low data rate of 2.4GHz (250kbps) wireless sense network as fixed
Position infrastructure, directly completes the positioning of Target, without considering ranging using the changing pattern of RSSI (received signal strength)
And the error problem of fingerprint matching, therefore there is system to lay the advantages such as simple, locating speed is fast.It is answered in many indoor positionings
In, for example, the positioning of exhibition center's tourist's guide to visitors, office building employee or the positioning of cleaner position, often do not need to carry out single
Point location, zone location can meet demand, i.e., only which region identification Target is located at, rather than identifies which it is located at
Specific coordinate points.Positioning accuracy is high, and stability is good.
Detailed description of the invention
Fig. 1 is RSSI distribution density schematic diagram.
Fig. 2 is corridor space positioning scene figure.
Fig. 3 is lobby space positioning scene figure.
Fig. 4 is each localization region recall rate of corridor space.
Fig. 5 is each localization region recall rate of lobby space.
Fig. 6 is each localization region accuracy rate of corridor space.
Fig. 7 is each localization region accuracy rate of lobby space.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
1. sensing network architecture and working method
The present invention use IEEE802.15.4 standard, using multiple Anchor joint structures at a wireless sense network as base
Infrastructure；Target sends detection packet periodically, after Anchor node receives, calculates its corresponding RSSI value.In general,
RSSI value is reduced with the increase of distance, and therefore, the RSSI value of distance Target different Anchor nodes, acquisition is also
Different.Classified herein using bayes classification method to the RSSI of Target, and then completes positioning function.
It is assumed that containing n Anchor node in Sensor Network, Target has sent detection packet Pi, will form one in a network
" record ": Ri={ ri1,ri2,…,rij,rin, wherein rijIt is the P for numbering the Anchor for being j and calculatingiRSSI value, often
RSSI value is sent to acquisition terminal by multi-hop wireless link by a Anchor in real time.In the training stage, Target locates respectively
In application scenarios in prior ready-portioned trained region, carry the subject of Target in each trained region by a small margin,
Slow random movement 40 seconds, the interval Target 256ms sent primary detection packet, to given trained region (in Fig. 6 and Fig. 7
Li), Target node sends altogether 160 detection packets, if do not considered missing and invalid value, can generate 160 records in terminal.
Complete after RSSI in all trained regions collects, be every record one new attribute of increase, i.e., " position " attribute (as
Class label), for next stage training Bayes classifier.For example, when Target is in training region Li, then corresponding record
" position " attribute value both be set to Li。
2. outlier identification mechanism
Find that RSSI value will appear a small amount of exception during actual measurement, especially there are more personnel to walk in space indoors
In the case where dynamic.Record comprising exceptional value is unfavorable for training efficient Bayes classifier, quasi- in order to improve position identification
True rate, it is necessary to reject RSSI exceptional value.RSSI exceptional value is identified using k- mean value strategy herein.
In formula (a), p is the point in RSSI data set, ciIt is cluster CiCentroid, d (p, ci) indicate p and ciBetween Europe
Family name's distance.A position Target is given, after obtaining all RSSI values, k point is randomly choosed from RSSI data set,
And the centroid as k cluster.In next each iteration, each cluster uses the object meter of the cluster assigned by last iteration
Then new mean value uses updated mean value as new centroid.Until iteration can be continued until that distribution is stablized.
3. Bayes's classification strategy
Assuming that R is for training Bayes's location classifier, the training set with class label, its record is indicated
For R={ r1,r2,...rn, riIndicate ith attribute value；There is m trained region (i.e. m kind position category in entire located space
Number), it is denoted as L respectively1, L2...Lm.The core missions of Bayes's location classifier are exactly to calculate Pr (Li| R) maximum a posteriori it is general
Rate value and corresponding position class label.According to Bayes' theorem, formula (b) gives calculation.
It is a constant for given R ∈ R, Pr (R) in training set R；Assume that Target is in located space simultaneously
In the prior probability in each trained region be equal, i.e. Pr (Li)=Pr (Lj) (1≤i < j≤m), then, the Bayes of this paper
Location classifier need to only consider maximize Pr (R | Li).When Anchor quantity increases, the dimension of R can also be correspondinglyd increase,
Therefore, in order to reduce Pr (R | Li) computing cost, improve system locating speed, it is assumed that it is any record R attribute between simultaneously
There is no any dependences, i.e. riIndependently of rj(1≤i < j≤n), thus can be used formula (c) come search for Pr (R | Li)
Maximum value.
(c) Pr(R|Li)=Pr (r1|Li)×Pr(r2|Li)×…Pr(rn|Li)
In positioning application, the distribution in training region is discrete, but RSSI value is bounded and continuous data, Fig. 1 exhibition
RSSI distribution situation after having shown excluding outlier, in a trained region, it can be seen that the distribution of RSSI value is overall
The feature (the case where being located at other training regions for Target, as a result similar) for showing approximate Gaussian distribution, shows shellfish
The feasibility in theory of this classifier of leaf.Each category value needed for formula (c) can be found out by gauss of distribution function according to formula (d)
Probability distribution.
Wherein, uLiWith σLiL can be used respectivelyiIn all records j-th of attribute (rj) value mean value and standard deviation replace.
The present invention considers the scene of the indoor positioning of two quasi-representatives: the hall of the corridor sum in building, horizontal layout point
Not not as shown in Figures 2 and 3.In two kinds of scenes, the metope that it is 2 meters apart from ground level that Anchor, which is fixed on, positioning area
Domain carries out natural division according to floor tile on ground, and corridor space Tile dimensions are 1.8 × 0.9m2, lobby space Tile dimensions are
1.2×1.2m2If localization region levels off to infinitesimal, One-Point Location is evolved into, meanwhile, the scale of training set is also corresponding
Tend to be infinitely great；But for taking human as target positioning application in, according to human motion speed, 1~2m2Localization region be close
Reason.In addition, during the experiment, other than carrying the subject personnel of Target, having about 15 during data set is collected
The test zone that people (in 11 minutes) walks through corridor space has 62 people (in 29 minutes) to walk through the test of lobby space
Region.
Fig. 4 and Fig. 5 are described in corridor space and lobby space respectively, the recall rate of this paper localization method.Recall rate body
Show Target and is in region LiIt is correctly identified as L simultaneouslyiFrequency.As can be seen that recall rate is basic in corridor space
Greater than 80%, the recall rate in more than half region can be close to 100%；In lobby space, most of recall rate is close
100%, minimum recall rate is 83.3%.
Fig. 6 and Fig. 7 describes the accuracy rate that this paper positioning system reaches under two kinds of scenes.Accuracy rate describes Target
In region LiIt is correctly identified as L simultaneouslyi, Target be in region Lj(j ≠ i) is not erroneously identified as L simultaneouslyiThis feelings
The frequency of condition.Each region in corridor space, locating accuracy is nearly all more than 80%, Average Accuracy 87%；?
In lobby space, minimum the 84.5% of locating accuracy, up to 100%, average case 92%.
Claims (1)
1. a kind of indoor objects localization method based on Sensor Network and bayes method, which is characterized in that use Bayes's classification
Method classifies to the received signal strength indication for monitoring target, realizes positioning function；
Specifically follow the steps below:
Step 1,
Using IEEE802.15.4 standard, it is assumed that containing n wireless anchor nodes in Sensor Network, monitor target and have sent detection packet Pi,
It will form record: R in a networki={ ri1,ri2,…,rij,rin, wherein RiIt indicates collected by wireless anchor node
Detection packet PiThe sequence of values that received signal strength indication is constituted, one corresponded in computing terminal database include n
The record of a attribute value, rijIt is the detection packet P for numbering the wireless anchor node for being j and calculatingiReceived signal strength indication, each
Received signal strength indication is sent to acquisition terminal by multi-hop wireless link by wireless anchor node in real time；
In the training stage, monitors target and be respectively in application scenarios in prior ready-portioned trained region, carry and monitor target
Subject in each trained region by a small margin, slow random movement 30 seconds to 1 minute, monitor 64 to 256 milli of target interval
Second sends primary detection packet；To given trained region, in order to guarantee training precision, Target node at least sends 100 detections
Packet；After completing the received signal strength indication collection in all trained regions, it is that every record increases a new attribute, ascends the throne
Attribute is set, as class label, for next stage training Bayes classifier；
Step 2,
Use the exceptional value of k- mean value strategy identification received signal strength indication:
In formula (a), p is the point in received signal strength indication data set, ciIt is cluster CiCentroid, d (p, ci) indicate p and ciBetween
Euclidean distance；A monitoring target position is given, after obtaining all received signal strength indications, from received signal strength
Value Data concentrates k point of random selection, and the centroid as k cluster；In next each iteration, each cluster uses last time
Then the new mean value of the calculation and object of the cluster assigned by iteration uses updated mean value as new centroid, iteration meeting one
Until directly continueing to that distribution is stablized；
Step 3,
Assuming that T is for training Bayes's location classifier, the training set with class label, its record is expressed as R
={ r1,r2,...rn, riIndicate ith attribute value；There is m trained region in entire located space, is denoted as L respectively1,
L2...Lm, the core missions of Bayes's location classifier are exactly to calculate Pr (Li| R) maximum posteriori probability value and corresponding position
Class label；According to Bayes' theorem, formula (b) gives calculation:
It is a constant for given R ∈ T, Pr (R) in training set T；Assume that monitoring target is in located space simultaneously
The prior probability in each trained region is equal, i.e. Pr (La)=Pr (Lb), wherein 1≤a≤b≤m, then Bayes position
Classifier need to only consider maximize Pr (R | Li)；When wireless anchor node quantity increases, the dimension of T can also be correspondinglyd increase；
Therefore, in order to reduce Pr (R | Li) computing cost, improve system locating speed, it is assumed that it is any record R any two attributes rp
With rqBetween and be not present any dependence, i.e. rpIndependently of rq, wherein 1≤p < q≤n；Therefore it is searched for using formula (c)
Pr(R|Li) maximum value:
Pr(R|Li)=Pr (r1|Li)×Pr(r2|Li)×ΛPr(rn|Li) (c)
In positioning application, the distribution in training region is discrete, but RSSI value is bounded and continuous data, according to formula
(d), the probability distribution of each category value needed for formula (c) is found out by gauss of distribution function:
Wherein, uLiWith σLiL is used respectivelyiIn all records j-th of attribute rjThe mean value and standard deviation of value replaces.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610244430.7A CN105960011B (en) | 2016-04-18 | 2016-04-18 | Indoor objects localization method based on Sensor Network and bayes method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610244430.7A CN105960011B (en) | 2016-04-18 | 2016-04-18 | Indoor objects localization method based on Sensor Network and bayes method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105960011A CN105960011A (en) | 2016-09-21 |
| CN105960011B true CN105960011B (en) | 2019-06-18 |
Family
ID=56917656
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610244430.7A Expired - Fee Related CN105960011B (en) | 2016-04-18 | 2016-04-18 | Indoor objects localization method based on Sensor Network and bayes method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105960011B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108169734A (en) * | 2017-12-07 | 2018-06-15 | 国网山东省电力公司烟台供电公司 | A kind of method of locating terminal and system based under fiber mode |
| CN110940951A (en) * | 2018-09-25 | 2020-03-31 | 北京四维图新科技股份有限公司 | Positioning method and device |
| CN112073902A (en) * | 2020-08-25 | 2020-12-11 | 中国电子科技集团公司第五十四研究所 | Multi-mode indoor positioning method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102638889A (en) * | 2012-03-21 | 2012-08-15 | 浙江大学 | Indoor wireless terminal positioning method based on Bayes compression sensing |
| CN102821465A (en) * | 2012-09-07 | 2012-12-12 | 哈尔滨工业大学 | WLAN (Wireless Local Area Network) indoor positioning method based on subregion information entropy gain |
| CN104185270A (en) * | 2013-05-28 | 2014-12-03 | 中国电信股份有限公司 | Indoor positioning method, system and positioning platform |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9955300B2 (en) * | 2012-12-31 | 2018-04-24 | Texas Instruments Incorporated | Method for incorporating invisible access points for RSSI-based indoor positioning applications |
-
2016
- 2016-04-18 CN CN201610244430.7A patent/CN105960011B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102638889A (en) * | 2012-03-21 | 2012-08-15 | 浙江大学 | Indoor wireless terminal positioning method based on Bayes compression sensing |
| CN102821465A (en) * | 2012-09-07 | 2012-12-12 | 哈尔滨工业大学 | WLAN (Wireless Local Area Network) indoor positioning method based on subregion information entropy gain |
| CN104185270A (en) * | 2013-05-28 | 2014-12-03 | 中国电信股份有限公司 | Indoor positioning method, system and positioning platform |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105960011A (en) | 2016-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Toward location-enabled IoT (LE-IoT): IoT positioning techniques, error sources, and error mitigation | |
| CN103905992B (en) | Indoor positioning method based on wireless sensor networks of fingerprint data | |
| CN106248107B (en) | A kind of track deduction calibration method and device based on indoor earth magnetism path matching | |
| CN103561462B (en) | Indoor positioning system and method totally based on smart mobile terminal platform | |
| CN110913344B (en) | Collaborative target tracking system and method | |
| CN105704652B (en) | Fingerprint base acquisition and optimization method in a kind of positioning of WLAN/ bluetooth | |
| CN108709557B (en) | Indoor map generation method based on multi-user track fitting | |
| CN104076327B (en) | Continuous positioning method based on search space reduction | |
| CN109874112A (en) | A kind of method and terminal of positioning | |
| Cengiz | Comprehensive analysis on least-squares lateration for indoor positioning systems | |
| CN110933599B (en) | Self-adaptive positioning method fusing UWB and WIFI fingerprints | |
| CN105635956B (en) | One kind is based on indoor virtually target localization method and device | |
| CN110213710A (en) | A kind of high-performance indoor orientation method, indoor locating system based on random forest | |
| Sun et al. | Cortina: Collaborative context-aware indoor positioning employing RSS and RToF techniques | |
| CN105960011B (en) | Indoor objects localization method based on Sensor Network and bayes method | |
| CN106793085A (en) | Fingerprint positioning method based on normality assumption inspection | |
| Sou et al. | Indoor location learning over wireless fingerprinting system with particle Markov chain model | |
| CN110345935A (en) | A kind of method of indoor map matching positioning | |
| Yu et al. | Floor positioning method indoors with smartphone’s barometer | |
| CN106686720A (en) | Wireless fingerprint positioning method and system based on time dimension | |
| Kolakowski | Automatic radio map creation in a fingerprinting‐based BLE/UWB localisation system | |
| Wang et al. | The technology of crowd-sourcing landmarks-assisted smartphone in indoor localization | |
| Nguyen et al. | Smartphone-based indoor tracking in multiple-floor scenarios | |
| Wang et al. | Simultaneous navigation and pathway mapping with participating sensing | |
| Shin et al. | Received signal strength-based robust positioning system in corridor environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190618 Termination date: 20200418 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |