CN104185275B

CN104185275B - A kind of indoor orientation method based on WLAN

Info

Publication number: CN104185275B
Application number: CN201410458932.0A
Authority: CN
Inventors: 诸彤宇; 刘帅; 宋志新
Original assignee: Beihang University
Current assignee: Beihang University
Priority date: 2014-09-10
Filing date: 2014-09-10
Publication date: 2017-11-17
Anticipated expiration: 2034-09-10
Also published as: CN104185275A

Abstract

The invention discloses a kind of indoor orientation method based on WLAN, belong to indoor wireless communication and network technique field.Method includes：The RSSI data predictions for each AP that sampled point is collected, a peacekeeping bivector is therefrom extracted respectively as characteristic vector；Feature vector clusters are analyzed area to be targeted is divided into multiple positioning subregions；For every group of characteristic vector, respective disaggregated model is respectively trained out；Poll highest subregion set is chosen from all subregions with reference to " ballot " mechanism based on disaggregated model；Positioned using two-wheeled and reduce subregion range of convergence, improve positioning precision.Fully digging utilization of the invention RSSI spatial distribution characteristic, solves the problems such as large-scale indoor positioning search package space is excessive, and computation complexity is high；New location model is established, is solved in existing WLAN indoor orientation methods, the problems such as can not effectively learning and adapt to RSSI signals non-linear caused by the reason such as non line of sight transmission effects, RSSI attenuation laws be abnormal, non-gaussian statistical property.

Description

Indoor positioning method based on WLAN

Technical Field

The invention relates to a positioning method in the field of WLAN indoor positioning, and belongs to the technical field of indoor wireless communication and networks.

Background

In recent years, with the continuous improvement of the physical living standard of people, the demand of people on location services is increasing day by day, such as the wide demand on location in many aspects of personnel scheduling, asset management, emergency rescue, safety monitoring, safety scheduling, intelligent transportation, map navigation, travel guidance and the like; particularly, in emergency situations, such as emergency rescue, disaster relief emergency command and dispatch and other special application scenarios, the positioning information is more important.

With the intensive research of general computers and distributed communication technologies, indoor Wireless communication and network technologies develop rapidly, and indoor positioning modes based on Wireless Local Area Networks (WLAN), Bluetooth, Wireless Sensor Networks (WSN) and the like and indoor positioning methods based on fingerprints and probability methods are derived.

Based on the positioning technology of WLAN, Bluetooth, WSN, etc., the terminal detects the RSSI (Received Signal Strength Indication) of multiple APs Received in each mesh by performing mesh division indoors and deploying a large number of APs (Access points) indoors, and the RSSI of each node Received in each mesh is used as the characteristic quantity of the network to complete positioning because the Signal Strength sent by each Signal node Received at different positions is different.

Indoor positioning based on fingerprints is realized by collecting RSSI of different APs in an indoor area, storing addresses and coordinates of corresponding wireless access points in a database, measuring the intensity of surrounding wireless signals by a terminal user, and carrying out matching positioning on the intensity and the RSSI which is stored in the database in advance in a proper amount, so that coordinate information of a positioned terminal user is obtained.

The probability method utilizes the existing training samples on the reference points to obtain the RSSI signal probability distribution on each reference point. Generally, a gaussian function is used to perform probability distribution fitting, and the mean value and bandwidth of gaussian probability distribution of each reference point are obtained. The probability method fully utilizes the statistical characteristics of signal distribution, and the positioning precision is generally higher than that of the weighted nearest neighbor method.

However, they also have respective problems. In practical application, for large-scale indoor positioning, the indoor positioning method based on the fingerprint has the defects of large space matching search range, high calculation complexity and large requirement on storage space, and in practical application, the indoor positioning method based on the probability method has the characteristics of non-Gaussian, non-linear and multi-modal probability distribution of RSSI signals on a certain fixed reference point, so that the difference between a fitted probability distribution function and the actual probability distribution is large, and a large matching error is caused during positioning.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: the defects in the prior art are overcome, and the indoor positioning method based on the WLAN is provided, so that the matching search range can be reduced, the prediction model according with the actual situation can be obtained, and the calculation complexity and the time complexity are reduced to a certain extent.

The technical problem to be solved by the invention is as follows: the matching search range is reduced, a prediction model conforming to the actual situation is established, and an indoor positioning method based on the WLAN is provided, and the method comprises the following steps:

the method comprises the following steps: preprocessing the RSSI data of each AP collected by the sampling points, and extracting one-dimensional and two-dimensional vectors from the preprocessed RSSI data as characteristic vectors respectively.

The necessary preprocessing of the scanned RSSI data includes: and deleting the data with the RSSI less than-100 dB and deleting the data of the non-positioning AP. The deleting of the data of the non-positioning AP means deleting the RSSI of the AP which is not suitable for positioning. The use of APs that are not suitable for positioning are characterized by too low a strength (RSSI less than-95 dB) or poor stability (variance greater than 20), which increases computational complexity, reduces positioning accuracy and is therefore excluded.

Various feature vectors capable of accurately quantifying the RSSI distribution rule are extracted from the original data by adopting different extraction methods. The method comprises the following steps:

(1) sequencing all scanned APs in an ascending order according to MAC addresses, and numbering all original data scanned during offline acquisition according to corresponding sampling points on the acquisition position marks;

(2) the respective feature vectors can be extracted in two ways:

a. combining the sequenced APs pairwise, i.e. dividing the APs into groups according to MAC addressesGroups, each group of APs denoted As (AP)_i,AP_j) (wherein, 0<i<j is less than or equal to m, m represents the number of all APs), and the RSSI vector of the corresponding AP combination and the corresponding sampling points are extracted from the original data marked with the sampling points;

b. each AP is independently used as one group, namely all off-line collected data are divided into m groups according to the MAC address of the AP, and each group of APs is expressed as an AP_i(wherein, 0<i is less than or equal to m, m represents the number of all APs), and the RSSI one-dimensional vector of the corresponding AP and the corresponding sampling point are extracted from the original data marked with the sampling points.

Step two: and performing cluster analysis on the feature vectors, and dividing the region to be positioned into a plurality of positioning sub-regions, wherein each sub-region reflects an RSSI distribution feature.

And taking the feature vectors constructed in the step one as input, and taking the distance between the feature vectors as a similarity measurement function to perform cluster analysis. Optionally, the cluster analysis uses an X-means algorithm that automatically finds the number of clusters. The X-means clustering algorithm improves the K-means algorithm, the clustering number K does not need to be specified in advance during initial operation of the algorithm, only a value range [ K1, K2] (K1< K2) of K needs to be specified, and the algorithm finds an optimal clustering number K in the specified range to realize clustering division. The X-means algorithm takes Bayesian information criterion as guidance, and continuously traverses the clustering centers of different clusters to represent different signal characteristics, wherein the signal characteristics reflect the aggregation phenomenon of signal distribution in a certain area.

Step three: respectively training corresponding classification models for each group of feature vectors in combination with clustering results; and selecting a subregion set with the highest vote number from all subregions based on a classification model and combining a 'voting' mechanism. Which comprises the following steps:

in the off-line stage, a Support Vector Machine (SVM) classification model corresponding to each feature Vector of each construction method is trained for the feature vectors constructed by the two construction methods provided in the step two. The SVM is based on the VC dimension (VC dimension) theory of statistical learning and the principle of structural risk minimization. SVMs attempt to maximize the generalization of classifiers by trading off classification accuracy (correctness of classification for a particular sample) and classification ability (error-free classification for any sample). The characteristic value is used as the input of the SVM classifier and is abstract description of data, so that the selection of the characteristic value is very important, and the final classification effect can be directly influenced by accurately reflecting the characteristics of the data to be classified.

And in the online stage, extracting classification characteristic vectors from real-time data, reading corresponding SVM classification models trained in the offline stage, calculating the probability of the vectors to be classified corresponding to different regions according to the support vector polynomial expansion term values, and selecting a region set R with the highest vote number from all the regions by combining a 'voting' mechanism.

The specific operation of the online positioning stage comprises the following steps:

(1) reading a trained SVM classification model, and calculating a support vector polynomial expansion term value;

(2) reading the currently acquired RSSI and extracting a classification feature vector;

(3) mapping the classified feature vectors to a high-dimensional space through a polynomial kernel function, and calculating the probability of the vectors to be classified corresponding to different regions according to the support vector polynomial expansion term values;

(4) for each AP group (AP)_i,AP_j) Judging whether each divided sub-region meets the condition, if a plurality of sub-regions meet the condition, the SVM model considers that the current equipment is possibly in a union set of the sub-regions;

the eligible area is defined as an AP group (AP)_i,AP_j) The prediction probability in a certain sub-region is larger than a certain threshold (0)<<1) Then, the region is considered eligible;

(5) combining a 'voting' mechanism to select an area set R with the highest number of votes from all the areas, and the method specifically comprises the following steps:

if AP group (AP)_i,AP_j) If the sample data of (1) is determined to be in a certain area through SVM prediction, then the ticket number of the area is increased by 1. Geometrically, selecting the area with the most covering times as the positioning area with coarse granularity, wherein the ticket number of each area should be 0 to 0In the meantime.

Step four: and the area set range is narrowed by two-wheel positioning, and the positioning precision is improved. The method specifically comprises the following steps:

(2) reading the currently acquired RSSI, extracting a classification feature vector, and standardizing classification features;

(3) mapping the classified characteristic vectors to a high-dimensional space through a polynomial kernel function, calculating the probability of the vectors to be classified corresponding to different regions according to the support vector polynomial expansion term values, and selecting the probability of each region in the coarse-grained positioning region R obtained in the third step;

(4) for each AP_iJudging whether each divided sub-region meets the condition, wherein the sub-region is a subset of the coarse-grained positioning region R obtained in the third step, and if a plurality of sub-regions meet the condition, the SVM model considers that the current equipment is possibly in a union of the sub-regions;

the qualified area is when the AP is_iThe prediction probability in a certain region is greater than a certain threshold (0)<<1) Then, the region is considered eligible;

(5) combining a ' voting ' mechanism to select the region set R ' with the highest vote number from the R, the method specifically comprises the following steps: if AP_iIf the sample data of (1) is determined to be in a certain area through SVM prediction, then the ticket number of the area is increased by 1. Geometrically, the area with the most covering times is selected as a fine-grained positioning area, and the ticket number of each area should be between 0 and m.

The technical scheme provided by the invention has the beneficial effects that: the invention fully utilizes the spatial distribution characteristics of the RSSI, and reduces the deviation of the positioning area caused by improper area division; a novel positioning model is established, and the problems that in the existing WLAN indoor positioning method, the nonlinearity and non-Gaussian statistical characteristics of RSSI signals caused by non-line-of-sight transmission effect, multipath propagation effect, abnormal RSSI attenuation rule and the like cannot be effectively learned and adapted, large-scale indoor positioning, overlarge search matching space, high calculation complexity and the like are solved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a flow chart of a method implementation of the present invention;

FIG. 2 is a flow chart of the clustering process of the method of the present invention;

FIG. 3 is another clustering flow chart of the method of the present invention;

FIG. 4 is a flow chart of the training of the method of the present invention;

FIG. 5 is a coarse grain location flow chart of the method of the present invention;

fig. 6 is a fine-grained positioning flowchart of the method of the present invention.

Detailed Description

The following describes the embodiments of the present invention with reference to the flow chart and the specific examples.

FIG. 2 is a flow chart of the clustering process of the method of the present invention, which is part of the off-line phase. The method specifically comprises the following steps:

201. peripheral AP signals are scanned at each calibration point by using a smart phone at high frequency, and the scanned data format is shown in Table 1. It should be noted that the number of data collected at each calibration point is not fixed, and varies with the length of the collection time. And if the RSSI of the corresponding AP cannot be acquired at the current position, filling by-100 dB.

TABLE 1 Scan data Format

Number of calibration points	AP1	AP2	AP3	AP4	AP5	AP6	AP7
								1	-85	-97	-63	-100	-100	-90	-72
1	-83	-92	-65	-100	-98	-85	-69
								2	-70	-73	-95	-82	-63	-100	-100
……	……	……	……	……	……	……	……

202. And extracting all APs from the collected data, and sequencing the APs according to the ascending order of the MAC addresses. The aim is that the SVM algorithm used in the positioning phase is related to the vector order, so that a vector ordering must be determined artificially. In the embodiment of the present invention, the MAC ascending order of the APs is used as the sorting method.

203. Combining the sequenced APs into a whole (AP)_i,AP_j) (wherein, 0<i<j is less than or equal to m, m represents the number of all APs), i.e. the APs are divided into MAC addressesAnd (4) grouping. And extracting a two-dimensional RSSI vector corresponding to the AP combination from each RSSI data marked with the sampling points to serve as classified original data. As shown in tables 2 and 3.

Table 2 extracted data format

Number of calibration points	AP1	AP2
			1	-85	-97
1	-83	-92
			2	-70	-73
……	……	……

Table 3 extract data format

Number of calibration points	AP2	AP3
			1	-97	-63
1	-92	-65
			2	-73	-95
……	……	……

204. And taking the vectors constructed in the step 203 as input, taking the distance between the vectors as a similarity measurement function, and performing clustering analysis by adopting an X-means algorithm capable of automatically finding the clustering number. And respectively recording the division condition of each two-dimensional AP combination on the whole positioning area.

The specific implementation process of the X-means algorithm clustering analysis is as follows:

step1. specify the clustering number k Range [ k ]_min,k_max]And initializing k ═ k_min. The range of K is selected according to the size of the actual region to be measured, and the range of each sub-region is 200m²To 700m²In this way, [ k ] is calculated_min,k_max]；

Step2, randomly selecting k AP combination data points u from the feature vector set EV extracted in the step 202₁,u₂,u₃...u_kAs an initial clustering center; the feature vector set EV is as shown in table 2 and table 3, and k feature vectors are selected from the feature vector set EV as initial centers;

step3. combine data points x for each AP in the feature vector set EVⁱJudging the cluster to which the similarity belongs according to the similarity,wherein s (arg)₁,arg₂) Calculating a function for the similarity;

step4, repeating the above process, assigning all data points to the most similar class clusters, thereby preliminarily dividing all AP group data points into corresponding class clusters;

step5, for each cluster, recalculating the cluster center,wherein, c⁽ⁱ⁾Represents the data point xⁱPreliminarily determining the type of the user; c. C⁽ⁱ⁾J means: if the data point xⁱBelongs to class j, then (c)⁽ⁱ⁾J) 1, otherwise (c)⁽ⁱ⁾J) 0; the center represents a weighted average center point position of each cluster class;

step6. calculating a criterion function,wherein xⁱIs a data point in the data set, u_jIs the cluster center of class j; k refers to the number of cluster centers;

step7, if the criterion function does not change any more, turning to Step8, indicating that the clustering result is stable; otherwise, jumping to Step3, and clustering again;

step8, further dividing each clustered cluster and calculating Bayesian information criterion BIC before and after division_pre,BIC_post(ii) a The Bayesian Information Criterion (BIC) is an important component of Bayesian theory, can evaluate different models on the same data set based on posterior probability, and is suitable for being used as a reference basis for selecting models with low complexity and better description on the data set.

For the clustering model corresponding to the clustering number k, the calculation formula of the Bayesian information criterion is as follows:wherein EV is the set of feature vectors extracted in step 202; r is the number of the eigenvectors contained in the EV, wherein the number of the eigenvectors is equal to the number of RSSI combinations at all positions acquired by the AP group; p represents the number of parameters, called Schwarz criterion, and the calculation formula is p ═ k + k · d in the invention, wherein d is the dimension of the feature vector in EV, i.e. d ═ 2;can be regarded as a penalty on the complexity of the clustering model;is a clustering model M_kThe maximum a posteriori log likelihood estimation on the feature vector set EV is calculated as the following formula

Wherein,u_(i)the cluster center is a cluster center of a cluster i;

step9. if BIC_pre＞BIC_postChecking whether the result model is higher than the original score, if the score is high, accepting splitting, turning to Step10, otherwise, making k equal to k +1 and jumping to Step 8;

step10. if k > k_maxThen clustering needs to be carried out again, and the process goes to Step 7; otherwise, let k be k +1 and jump to Step2, and calculate the clustering condition of adding a class;

step11, selecting the dividing mode with the largest BIC as a clustering result;

assuming that M is a model set corresponding to different clustering numbers k, then Namely the best clustering model. Each type is represented as a signal feature that reflects the clustering of signal distributions over a certain area.

FIG. 3 is another flow chart of the clustering process of the present invention, which is part of the off-line sampling stage. The method specifically comprises the following steps:

301. peripheral AP signals are scanned at each calibration point by using a smart phone at high frequency, and the scanned data format is shown in Table 1. It should be noted that the number of data collected at each calibration point is not fixed, and varies with the length of the collection time. And if the RSSI of the corresponding AP cannot be acquired at the current position, filling by-100 dB.

302. The APs are divided into m groups according to the MAC address (m represents the number of all APs). And extracting a one-dimensional RSSI vector of the corresponding AP from each RSSI data marked with the sampling points to serve as classified original data. As shown in tables 4 and 5.

Table 4 extracted data format

Number of calibration points	AP1
		1	-85
1	-83
		2	-70
……	……

Table 5 extract data format

Number of calibration points	AP2
		1	-97
1	-92

2	-73
		……	……

303. And (2) taking the vectors constructed in the step 302 as input, taking the distance between the vectors as a similarity measurement function to perform cluster analysis, wherein the cluster analysis adopts an X-means algorithm capable of automatically finding the number of clusters, and the clustering method is similar to the step 203, and is characterized in that the feature vector in the step 203 is changed from a two-dimensional vector formed by combining APs into a one-dimensional vector of a single AP. The signal mode reflects the aggregation phenomenon of signal distribution in a certain area, and the division condition of each AP on the whole positioning area is recorded respectively. It should be noted that, for the same positioning area, the division of the area by different APs may be different because the deployment positions of the APs are far apart in space, and are subject to non-line-of-sight transmission effects, multipath propagation effects and RSSI fading rules which are different from one another, so that the division results may be different.

Fig. 4 is a training flow chart of the method of the present invention, which belongs to a part of the off-line sampling stage. The method specifically comprises the following steps:

401. two-dimensional RSSI vectors of AP pairwise combination are extracted from each index point, as shown in tables 2 and 3, and the numbers of the index points are replaced by the category numbers after corresponding clustering.

402. One-dimensional RSSI vectors of a single AP are extracted from each index point, as shown in tables 4 and 5, and the numbers of the index points are replaced by the numbers of the corresponding clustered classes.

403. And (4) respectively carrying out SVM training on the vectors obtained in the steps 401 and 402, and calculating a classification characteristic value of the support vector machine. And the calculation of the classification characteristic value provides data support for the subsequent judgment of the effectiveness of the initialization range and the positioning range reduction. The classification feature selected in this embodiment is the RSSI vector obtained in steps 401 and 402.

Fig. 5 is a coarse-grained location flowchart of the method of the present invention, which is part of the on-line phase. The method specifically comprises the following steps:

501. loading each trained two-dimensional AP group (AP)_i,AP_j) The SVM classification model reads the currently acquired RSSI, sorts the RSSI in an ascending order according to the MAC addresses of the APs, and extracts each group of AP combination as a classification vector. It should be noted that the SVM algorithm is related to the vector order, and therefore a vector arrangement order must be artificially determined. In this example, the MAC ascending order of the APs is used as the sorting method.

502. And (3) predicting the classification vector formed by the AP combination extracted in the step 501 by using a corresponding SVM model, and respectively calculating the probability of each group of APs in each region under the corresponding region division mode. As the current position may be located at the edges of multiple regions, or as RSSI fluctuation may occur due to the fact that one or some APs in the two-dimensional AP combination are subjected to non-line-of-sight transmission effect, multipath propagation effect, RSSI attenuation rule abnormality, and the like, the multiple regions may all meet the requirements, and the selection may be performed according to the following method:

the feature vector extracted by sampling the AP each time may have a plurality of prediction results meeting requirements after being predicted by the corresponding SVM model, and each prediction result corresponds to a sub-region divided by the region to be positioned by the AP group. In the above formula, s represents the number of prediction results meeting the conditions, namely represents that the SVM model determines that the current equipment is possibly in a plurality of sub-regions; area_kThe k-th eligible region is represented, i.e. the representative SVM model identifies in which sub-regions the current device is likely to be. Area (AP)_i,AP_j) Representative AP group (AP)_i,AP_j) The determined area where the current position is located, namely the representative SVM model, considers that the current device is possibly in the union of the several sub-areas. The selection method of the sub-region which meets the requirement is that if the probability of predicting that the current feature vector is in a certain sub-region is not less than a certain threshold (0)<<1) The sub-region is considered satisfactory. In this example, the value of n is 1/n, and n represents the number of sub-regions divided by the AP combination.

503. For all AP groups (APs) obtained in step 502_i,AP_j) Area (AP) of_i,AP_j) And calculating a positioning result in a voting mode. If sample data for a certain combination of APs is predicted to be deemed likely to be within a certain area, via step 502, then the area ticket number is incremented by 1. Area (AP) traversing all AP combinations_i,AP_j) Voting, selecting the region with the most votes as the coarse-grained location region, the votes in each region being 0 to 0In the meantime. Geometrically expressed as selection by Area (AP)_i,AP_j) The region with the most number of times of coverage is regarded as the coarse-grained positioning region. If all the regionsIf the number of votes in a plurality of regions is the maximum and is larger than ξ, the union of the regions is obtained and used as a coarse-grained location region for location, and since normal location requires at least 3 APs to participate in calculation, in this example, the value of ξ is 4.

Fig. 6 is a fine-grained location flowchart of the method of the present invention, which is part of the online phase. The method specifically comprises the following steps:

601. loading the SVM classification model of each AP trained before, reading the currently acquired RSSI, forming a one-dimensional classification vector by the RSSI, predicting the classification vector by using the SVM classification model of each AP trained before, respectively calculating the probability of each AP in each area under the corresponding area division mode, and selecting the probability of each area in the coarse-grained positioning area R calculated in the last step. As the current position may be located at the edges of multiple regions, or as the AP is subjected to RSSI fluctuation due to non-line-of-sight transmission effect, multipath propagation effect, abnormal RSSI attenuation rule, and the like, a situation that multiple regions all meet requirements may occur, and the selection may be performed according to the following method:

the feature vector extracted by sampling the AP each time may have a plurality of prediction results meeting requirements after being predicted by the corresponding SVM model, and each prediction result corresponds to a sub-region divided by the region to be positioned by the AP. In the above formula, s represents the number of prediction results meeting the conditions, namely represents that the SVM model determines that the current equipment is possibly in a plurality of sub-regions; area_kThe k-th eligible region, i.e., the sub-region in which the SVM model identifies the current device as being likely, must be a subset of the coarse-grained location region R found in step four. Area (AP)_i) Representing AP_iThe determined area of the current position, namely representing the current setting considered by the SVM modelIt is possible to locate the union of these several sub-regions. The selection method of the sub-region which meets the requirement is that if the probability of predicting that the current feature vector is in a certain sub-region is not less than a certain threshold (0)<<1) The sub-region is considered satisfactory. In this example, the value of n is 1/n, and n represents the number of sub-regions divided by the AP.

602. For all APs in step 601_iArea (AP) of_i) It is calculated in a "voting" manner, i.e. if the sample data of an AP is deemed to be within a certain area as predicted by step 601, the area vote count is incremented by 1. Area (AP) traversing all APs_i) And voting, and selecting the area with the most votes as a fine-grained positioning area for positioning, wherein the number of votes in each area is between 0 and m. Geometrically expressed as selection by Area (AP)_i) If the number of votes of all the areas is less than a certain threshold ξ, the positioning is considered to be failed, and the positioning is ended, if the number of votes of the areas is the most and is greater than ξ, the union of the areas is obtained, the coordinates and the radius of the center point of the area are obtained to be used as the final positioning area, and the ξ value is 4 in the present example because the normal positioning requires at least 3 APs to participate in the calculation.

The invention has not been described in detail and is within the skill of the art.

The above description is only a part of the embodiments of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention.

Claims

1. An indoor positioning method based on WLAN is characterized by comprising the following steps:

the method comprises the following steps: preprocessing RSSI data of each AP acquired by a sampling point, and extracting a one-dimensional vector and a two-dimensional vector from the preprocessed RSSI data as characteristic vectors respectively;

step two: performing clustering analysis on the characteristic vectors, and dividing a region to be positioned into a plurality of positioning sub-regions;

step three: respectively training corresponding classification models for each group of feature vectors in combination with clustering results; selecting a subregion set with the highest vote number from all subregions based on a classification model and a 'voting' mechanism;

step four: the sub-area set range is reduced by two-wheel positioning, and the positioning precision is improved;

the first step of extracting one-dimensional and two-dimensional vectors from the preprocessed data as feature vectors respectively comprises the following steps:

(1) sequencing all the scanned APs in an ascending order according to the MAC addresses;

(2) one-dimensional and two-dimensional vectors are extracted as feature vectors according to the following two methods:

a. combining the sequenced APs pairwise, and dividing the APs into the groups according to the MAC addressesGroups, each group of APs denoted As (AP)_i,AP_j) Wherein, 0<i<j is less than or equal to m, m represents the number of all APs, and a vector formed by combining the APs is extracted from the preprocessed data to serve as a feature vector;

b. each AP is independently used as one group, namely all off-line collected data are divided into m groups according to the MAC address of the AP, and each group of APs is expressed as an AP_iWherein, 0<i is less than or equal to m, m represents the number of all APs, and a vector formed by the APs is extracted from the preprocessed data to be used as a feature vector;

the third step, the concrete implementation process includes an off-line stage and an on-line stage;

in the off-line stage, a Support Vector Machine (SVM) classification model corresponding to each feature Vector of each construction method is trained respectively aiming at the feature vectors constructed by the two construction methods provided in the step one;

in the online stage, a classification feature vector is extracted from real-time data, an SVM classification model trained in the offline stage is read, the probability that the vector to be classified corresponds to different regions is calculated according to support vector polynomial expansion term values, and a region set R with the highest vote number is selected from all the regions by combining a voting mechanism;

the voting mechanism refers to if a group of APs (APs) is present_i,AP_j) Is determined to be in a certain area through SVM prediction, thenAdding 1 to the ticket number of the region; EV (AP) traversing all AP groups_i,AP_j) Voting, selecting the region with the most votes as the coarse-grained location region, wherein the votes of each region should be 0 to 0EV is a feature vector set;

and step four, adopting two rounds of positioning to reduce the range of the region set, and specifically realizing the following steps:

(5) combining a ' voting ' mechanism to select the region set R ' with the highest vote number from the R, the method specifically comprises the following steps: if AP_iThe sample data of the AP is predicted by the SVM and is determined to be in a certain area, the number of votes in the area is added with 1, the area with the largest number of votes is selected as a positioning fine-grained positioning area according to the positioning area votes of each AP, and the number of votes in each area is between 0 and m.

2. The WLAN based indoor positioning method of claim 1, wherein: the first step of preprocessing the RSSI data of each AP collected by the sampling points comprises the following steps: deleting data with too low RSSI, deleting data of non-positioning AP, and filling up RSSI data which is not scanned;

the step of deleting the data with the low RSSI refers to deleting the data with the RSSI intensity lower than a certain threshold value; the data of the non-positioning AP is deleted, namely the RSSI of the AP which is not suitable for positioning is deleted, and the characteristic that the RSSI is not suitable for positioning is too low, namely the RSSI is less than-95 dB or the stability is poor, namely the variance is more than 20.

3. The WLAN based indoor positioning method of claim 1, wherein: in the second step, the feature vector is subjected to cluster analysis, and the region to be positioned is divided into a plurality of positioning sub-regions, and the specific steps are as follows: and (4) taking the characteristic vectors constructed in the step one as input, taking the distance between the characteristic vectors as a similarity measurement function to perform cluster analysis, wherein the cluster analysis adopts an X-means algorithm capable of automatically finding the cluster number.

4. The WLAN based indoor positioning method of claim 1, wherein: the specific operation of the online positioning stage comprises:

(4) for each AP group (AP)_i,AP_j) And judging whether each divided sub-region meets the condition, and if a plurality of sub-regions meet the condition, the SVM model considers that the current equipment is possibly in the union of the sub-regions.