CN103763769B - Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods - Google Patents

Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods Download PDF

Info

Publication number
CN103763769B
CN103763769B CN201310733950.0A CN201310733950A CN103763769B CN 103763769 B CN103763769 B CN 103763769B CN 201310733950 A CN201310733950 A CN 201310733950A CN 103763769 B CN103763769 B CN 103763769B
Authority
CN
China
Prior art keywords
cluster
ap
fingerprint
clusters
sub
Prior art date
Application number
CN201310733950.0A
Other languages
Chinese (zh)
Other versions
CN103763769A (en
Inventor
梁栋
毕真
周盈君
曾书磊
刘敬智
Original Assignee
北京邮电大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 北京邮电大学 filed Critical 北京邮电大学
Priority to CN201310733950.0A priority Critical patent/CN103763769B/en
Publication of CN103763769A publication Critical patent/CN103763769A/en
Application granted granted Critical
Publication of CN103763769B publication Critical patent/CN103763769B/en

Links

Abstract

本发明公开了一种基于AP重选择和自适应簇分裂的室内指纹定位方法,包括离线数据采集阶段和在线匹配定位阶段:离线数据采集阶段运用自适应簇分裂聚簇方法对栅格数量进行压缩,运用AP重选择策略使每个簇选出定位能力最强的AP集合,最终得到新定义的V‑RSS指纹库;在线匹配定位阶段采用基于V‑RSS指纹库的自适应簇分裂定位方法;本发明的优点是计算量和计算复杂度大大低于传统方法,能够节省定位时间和待定位的移动终端在定位过程中消耗的电能;本发明有效地解决了传统指纹室内定位过程中存在的存储量大、计算量大、定位精度不可控的问题。 The present invention discloses a method of indoor positioning based on fingerprint AP reselection and adaptive split clusters, including online and offline data collection phase matching positioning phase: the offline data collection phase using adaptive clustering method for clustering the number of division grids compression use AP reselection policy so that each cluster selected the strongest set of positioning capability AP, end up V-RSS fingerprint database newly defined; positioning stage match online using split positioning method based on adaptive cluster V-RSS fingerprint database; advantage of the invention is that the amount of calculation and the computation complexity is much lower than the conventional method, the positioning time can be saved and the mobile terminal to be located in the power consumption of the positioning process; the present invention effectively solves the conventional fingerprint storage chamber during positioning exist large, large amount of calculation is not controllable positioning accuracy problem.

Description

基于接入点重选择和自适应簇分裂的室内指纹定位方法 Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods

技术领域 FIELD

[0001] 本发明涉及基于接入点(Access Point,简称AP)重选择和自适应簇分裂的室内指纹定位方法,属于无线通信系统中的无线定位技术领域。 [0001] The present invention relates to an indoor location based fingerprint access point (Access Point, abbreviated AP) and adaptive reselection split clusters, belonging to the wireless location technology in the field of wireless communication systems.

背景技术 Background technique

[0002] 目前,随着无线网络的广泛普及和移动智能终端的迅猛发展,基于位置的服务受到越来越多的关注,在紧急救援、医疗保健、社交网络、导航和监控等领域已获得广泛的应用并展示出巨大的市场前景。 [0002] Currently, with the rapid development of widespread popularity of wireless networks and mobile intelligent terminal, location-based services are more and more attention, emergency rescue, medical care, social networking, navigation, and surveillance and other fields has been widely applications and demonstrate the huge market prospects.

[0003] 在开阔的室外环境中,全球定位系统(GPS),网络辅助全球卫星定位系统(A-GPS) 和蜂窝网定位系统可以提供满足各种精度需求的室外环境定位信息。 [0003] in an open outdoor environment, a global positioning system (GPS), a network Assisted GPS (A-GPS) systems and cellular location accuracy may be provided to meet the needs of a variety of outdoor environment positioning information. 在室内环境中,特别是热点区域,如超市、展厅、医院、会馆、剧院、图书馆、监狱等地,人们对室内位置信息的需求也越来越强烈,以实现导航定位、上下文感知、人员物资监控等全方位的智能服务。 In the indoor environment, especially in hot areas, such as supermarkets, exhibition halls, hospitals, Hall, theaters, libraries, prisons and other places, the demand for indoor location information have become stronger, to achieve navigation and positioning, context-aware, staff supplies a full range of intelligent monitoring service. 但由于GPS信号在很多室内区域无法有效接收,如停车场、医院等,且其定位精度较低,因此室外定位技术无法满足室内定位的需求。 But because GPS signals can not be effectively received a lot of indoor areas, such as parking lots, hospitals, and low positioning accuracy, so outdoor indoor location positioning technology can not meet the demand.

[0004] 已有的传统的室内定位技术大多数需要额外的专用硬件设施,使用成本高,覆盖范围小,限制了基于位置的服务在室内环境的普及。 [0004] existing traditional indoor positioning technology most need additional dedicated hardware facilities, high cost, low coverage, limited the popularity of the indoor environment in the location-based services. 基于无线局域网(WLAN)和接收信号强度(RSS)的室内定位技术,完全基于现有的WLAN基础设施和移动终端就能独立实现定位,无需任何额外的专用设备,成本低,分布广泛,且能满足大多数室内定位应用的精度需求,成为室内定位技术的研究热点。 Based on a wireless LAN (WLAN) and a received signal strength (RSS) of indoor positioning technologies can be implemented independently positioned entirely on existing WLAN infrastructure and the mobile terminal, without any additional special equipment, low cost, widely distributed, and can indoor positioning accuracy to meet the needs of most applications, become a hot topic of indoor positioning technology.

[0005] WLAN室内定位技术可以分成两种方法:几何测量法和指纹定位法。 [0005] WLAN indoor positioning technologies can be divided into two methods: the geometric measurement and fingerprinting positioning method. 几何测量法是指利用几何学的原理和路损模型计算用户的位置。 Geometry measuring method, using the principle of geometry and path loss model calculates the user's location. 但是由于室内无线电传播环境的极度复杂性,几何测量法的RSS传播模型预测精度往往较低,定位精度不高,且需预先知道接入点(Access Point,简称AP)的具体位置,不能满足实际需求。 However, due to the extreme complexity of indoor radio propagation environment, geometry measuring method RSS propagation model prediction accuracy tends to be lower, the positioning accuracy is not high, and the need to know in advance an access point (Access Point, abbreviated AP) specific location, can not meet the actual demand. 指纹定位法采集参考点位置的RSS样本,构建指纹数据库,并通过RSS样本的指纹匹配得出定位结果,指纹定位法定位精度明显高于几何法,且不需要知道AP的位置。 Localization fingerprint collection of RSS samples of the reference point position, build a fingerprint database, and a positioning result obtained by the fingerprint matching samples RSS, positioning accuracy of the fingerprint was higher than those geometric positioning method, and need not know the location of the AP.

[0006] 但是,鉴于越来越多的用户主要依靠电池供电的移动设备,而现有的传统WLAN指纹定位方法都存在存储量大、计算量大的问题,终端的负荷和能耗高成为了亟待解决的问题。 [0006] However, in view of the more and more users rely mainly on battery-powered mobile device, and the existing conventional WLAN positioning fingerprinting methods have storage capacity, computational load, load and energy consumption of the terminal becomes high Problems to be solved.

[0007] 此外,传统的WLAN指纹定位方法定位精度不可控,无法实现对精度有限定的用户的需求,如何实现精度可控以满足不同用户对所需精度的不同需求、如何在WLAN室内定位方法中节约存储量、减小计算复杂度等成为当前一个重要的急待解决的技术问题。 [0007] Further, the conventional WLAN fingerprinting positioning method for positioning a controllable precision, accuracy can not be achieved with a defined user's needs, how to achieve an accuracy controlled to meet different customer needs of different precision required, how WLAN indoor positioning method in the storage saving, and so reduce the computational complexity of the current has become an important technical problem need to be solved.

发明内容 SUMMARY

[0008] 本发明的目的在于提供一种能够克服上述技术问题的基于接入点重选择和自适应簇分裂的室内指纹定位方法,本发明是在传统指纹定位方法的基础上,结合新定义的自动簇分类方法和接入点(Access Point,简称AP)重选择策略,对室内待定位的移动终端进行定位,用以解决传统指纹室内定位过程中存在的存储量大、计算量大、定位精度不可控的问题。 [0008] The object of the present invention to provide a method of indoor positioning fingerprint cluster splitting and adaptive reselection based on the access point capable of overcoming the above-mentioned problems, the present invention is based on the conventional positioning method fingerprints, newly defined binding automatic classification and clustering access point (access point, abbreviated AP) reselection policy to the mobile terminal to be positioned is positioned interior to solve the conventional fingerprint storage capacity indoor positioning process, and calculates a large amount, the positioning accuracy uncontrollable problem.

[0009] 本发明的核心技术特征和思路是基于地理位置和辨识度的最佳AP集合重选择、基于多叉树的自适应簇分裂定位方法。 [0009] The technical features and the core idea of ​​the invention is the location and identification of the best AP reselection set, splitting positioning method based on an adaptive multi-tree-based clustering. 此处多叉树泛指四叉树或八叉树,对于平面定位问题将采用四叉树,对于空间定位问题将采用八叉树。 Here refers to multi-branch tree quadtree or octree, the plane positioning of the quadtree, the issue will be used for spatial orientation octree.

[0010] 本发明包括以下步骤: [0010] The present invention comprises the steps of:

[0011] 步骤一:按照传统方法将待定位区域均匀划分为若干栅格(grid),选取每个栅格的几何中心作为参考点位置,在每个参考点位置测量能接收到的所有接入点(Access Point)的信号强度,生成接收信号强度指纹向量(RSS向量),建立基础RSS指纹库; [0011] Step a: according to conventional methods to be positioned evenly divided into several raster areas (Grid), select all access the geometric center of each grid point as a reference position, the measurement can be received in each reference position point (Access point) signal intensity, the fingerprint vector to generate a received signal strength (RSS vectors foundation), RSS fingerprint database;

[0012] 步骤二:结合基于地理位置和辨识度的AP重选择方法、自适应簇分裂聚簇方法和自学习机制,生成栅格大小可变的RSS指纹库(V-RSS指纹库); [0012] Step Two: based on a combination of location and identification of the AP reselection method, adaptive clustering method for clustering split and self-learning mechanisms, variable size grid generated RSS fingerprint database (V-RSS fingerprint database);

[0013] 步骤三:将待定位的移动终端移动到某个位置,测量实时接收的RSS向量,多次测量取平均值并进行去噪处理; [0013] Step 3: the mobile terminal moves to be positioned to a location, real-time measurement vector the received RSS, multiple measurements and averaged denoising;

[0014] 步骤四:将步骤三中测得的RSS向量与V-RSS指纹库进行比对,运用基于多叉树的自适应簇分裂定位方法估计待定位的移动终端的位置; [0014] Step four: three step RSS measured vector V-RSS fingerprint database for comparison, using an adaptive multi-cluster-tree division based positioning method to be positioned estimated position of the mobile terminal;

[0015] 步骤五:输出定位结果,所述定位结果包括最佳匹配的簇编号、待定位的移动终端的位置估计和不确定度。 [0015] Step 5: To a positioning result, the positioning result including the best matching cluster number, the location of the mobile terminal to be positioned and estimate uncertainty.

[0016] 本发明具有以下创新特点和有益效果: [0016] The present invention has the following innovative features and advantages:

[0017] (1)基于地理位置和辨识度的最佳AP集合重选择; [0017] (1) based on the optimum location and identification of the AP reselection set;

[0018] 在传统指纹定位方法中AP选择是经过一次选择后即应用于整个待定位区域;或者经过聚簇算法,一个聚簇(cluster)应用一组最佳AP集合。 [0018] In the conventional fingerprinting positioning method is selected through the AP after a selection to be applied to the entire location area; or via clustering algorithm, a clustering (Cluster) applying a set of optimum collection AP.

[0019] 在本发明中,AP选择虽然也是基于簇为单位,对每个簇都要进行AP选择,但不同的是:首先,本发明针对不同的地理位置(地理簇)计算不同的最佳AP集合;第二,最佳AP集合选择时采用辨识度作为衡量准则;第三,在本发明中,子簇虽然是由母簇分裂而来的,但是子簇与母簇采用的最佳AP集合却可能是不同的,这是因为在计算最佳AP集合时,只基于该簇内的所有参考点进行计算,也就是说,母簇与子簇包含不同的参考点,经过AP选择方法, 得出的最佳AP集合也很可能是不同的。 First, the present invention is different optimum calculated for different geographical location (cluster): [0019] In the present invention, although the AP selection is based on units of clusters, each cluster should be carried out for AP selection, except that AP set; second, using the best AP recognizable as measured when set selection criteria; third, in the present invention, although the sub-cluster split from the cluster by a parent, but parent sub-clusters and clusters using the best AP set it may be different, this is because when calculating the optimal AP set, all calculated based on only the reference point within the cluster, i.e., the sub-cluster comprises a cluster master different reference points, after the AP selection method, AP obtained a collection of the best is also likely to be different.

[0020] AP重选择机制为不同地理位置的簇寻找到空间区分能力最强的AP,其好处是去除了对定位贡献较小甚至起副作用的AP,既减少了指纹的存储维度,又提高了定位精度。 [0020] AP re-selection mechanism to find the strongest ability to distinguish between the space for the AP cluster geographically, its benefits are removed to locate even smaller contribution from the AP side effects, not only reduces the dimensions of storage of fingerprints, but also improves positioning accuracy.

[0021] (2) V-RSS 指纹库; [0021] (2) V-RSS fingerprint database;

[0022] 传统指纹定位方法采用的是栅格大小固定、参与定位的AP集合不变的位置指纹库,其存储方式以二维(平面定位)或三维(空间定位)矩阵为主;与传统指纹定位方法相比本发明的V-RSS指纹库具有3个创新点:栅格(簇)大小不固定、每个栅格(簇)的AP集合可变、 采用四叉树(平面定位)或八叉树存储(空间定位)。 [0022] The conventional fingerprint positioning method uses a fixed grid size, the AP set position engaged in the positioning invariant fingerprint database that stores a two-dimensional manner (plane location) or three dimensional (spatial orientation) based matrix; and conventional fingerprint positioning method having three V-RSS innovation compared to the fingerprint database according to the invention: a grid (cluster) size is not fixed, the AP variable set in each grid (clusters), quadtree (plane location) or eight tree storage (spatial location).

[0023] 本发明的以上创新点是基于AP重选择、自适应簇分裂聚簇方法和自学习机制实现的:首先,对不同的簇,选取在该簇中对空间区分能力最强的AP作为最佳AP集合,在V-RSS指纹库只需要存储最佳AP集合的指纹即可,其它AP的指纹被抛弃,并且不同簇所存指纹对应的AP集合是不完全相同的,因此需要在指纹库中记录AP的编号。 [0023] The innovation of the present invention, the above is based on the AP reselection cluster splitting adaptive clustering method and mechanism to achieve self-learning: First, different clusters, the cluster in selecting the strongest spatial separation capabilities of the AP as the best AP set, the fingerprint can be V-RSS fingerprint database stores only a set of the best AP, AP other fingerprint is discarded, and the AP collection of different clusters corresponding to the fingerprint deposit is not identical, it is necessary in the fingerprint database recording the number of AP.

[0024] 其次,采用自适应簇分裂方法,将整个待定位区域分割为大小不一的簇,簇的大小取决于所有AP在该簇的空间区分能力,如果某个子区域存在强空间区分能力的AP集合,则该子区域能分割为很多较小的簇,反之则该区域只能分割为少数几个较大的簇甚至只能归为1个簇,其好处是减少了指纹的存储数量。 [0024] Next, adaptive cluster splitting method, the entire region is divided into different sizes to be positioned clusters, cluster size depends on the ability to distinguish all of the space in the AP cluster, if a sub-region there is a strong spatial separation ability AP collection, the sub-region can be divided into many smaller clusters, and vice versa in the region can only be divided into a few larger clusters can even be classified as a cluster, its benefits are reducing the number of stored fingerprints. 在实际定位过程中,在不具备强空间区分能力的AP的子区域,即使分割得再细,存储再多的指纹,对于提高定位精度也没有增益。 In the actual positioning process, the AP does not have the sub-regions strong spatial separation ability, even though the fine division have another, more fingerprints stored, for improving the positioning accuracy is also no gain.

[0025] 另外,辨识度阈值是决定当前簇是否能继续分裂的关键,该阈值既可以由人工根据实际经验设置,也可以由自学习机制自行获得,其方法是将辨识度阈值定为预期定位误差、无线信号波动特征和簇尺寸的函数,通过自学习的方式确定该函数的表达式或者数值对应关系。 [0025] In addition, the recognition threshold is the key to determine the current cluster whether it can continue to divide, and the threshold either manually based on actual experience setting can also be made self-learning mechanism on their own to obtain, the method is to identify the threshold set for the expected positioning error, the radio signal characteristics and fluctuations in function of cluster size determining the expression or function of the correspondence between the value by self-learning manner.

[0026] (3)本发明的基于多叉树的自适应簇分裂的在线定位方法; [0026] (3) On-line adaptive location based multi-cluster-tree split according to the invention;

[0027] 传统指纹定位方法多采用将在线实时指纹与离线指纹进行穷举对比,寻找欧式距离(或者加权欧式距离)最小的指纹对应的位置作为定位估计;与之相比,本发明采用的是基于四叉树(平面定位)或八叉树(空间定位)的搜索式定位方法。 [0027] The conventional fingerprint positioning method to use the real-time online and offline fingerprint fingerprint exhaustive comparison, to find the Euclidean distance (or a weighted Euclidean distance) corresponding to the minimum position of the fingerprint as the location estimate; In contrast, the present invention uses a search expression quadtree-based positioning method (plane location) or octree (spatial positioning).

[0028] 本发明的优点是不需要计算在线实时指纹与所有离线指纹的欧式距离,且在计算欧式距离时只需要考虑最佳AP集合组成的线性空间,本发明的计算量和计算复杂度大大低于传统方法,与传统方法相比,本发明在能够保证定位精度的前提下,既减少了需要存储的指纹数量又降低了指纹向量的维度;还能够节省定位时间和待定位的移动终端在定位过程中消耗的电能;由于传统指纹定位方法只能给出定位估计但无法对定位估计进行评价,而本发明不但可以给出定位估计还可以给出定位估计的不确定度;本发明有效地解决了传统指纹室内定位过程中存在的存储量大、计算量大、定位精度不可控的问题,具有较强的实用价值和现实意义。 [0028] The advantage of the present invention is not necessary to calculate the Euclidean distance with all the off-line real-time fingerprint fingerprints, in calculating the Euclidean distance and need only consider the linear space of a collection of the best AP, and calculates the amount of the present invention greatly computational complexity lower than the conventional method, compared to conventional methods, the present invention can ensure the positioning accuracy of the premise, reducing the number of fingerprints to be stored in turn reduces the dimension of the fingerprint vector; positioning time can also be saved and the mobile terminal to be positioned positioning the electric power consumed; targeting methods because conventional fingerprint analysis can not be evaluated, but the location estimate of the location estimate, but the present invention can not only be given location estimate may also be given the uncertainty of the location estimate; the present invention is effectively solve the storage capacity of traditional fingerprint indoor positioning process, and calculates volume, positioning accuracy uncontrollable problems, with a strong practical value and practical significance.

附图说明 BRIEF DESCRIPTION

[0029] 图1是本发明的总体流程图; [0029] FIG. 1 is a general flow chart of the present invention;

[0030] 图2是本发明的离线数据采集过程中生成V-RSS指纹库示例图; [0030] FIG. 2 is a V-RSS fingerprint database off-line data collection process of FIG example of the present invention generates;

[0031] 图3是本发明的离线阶段自适应簇分裂聚簇流程图; [0031] FIG. 3 is the offline phase the invention flowchart adaptive clustering split clusters;

[0032] 图4是本发明的用四叉树存储的V-RSS指纹库示例图; [0032] FIG. 4 is a quadtree V-RSS stored fingerprint database exemplary view of the invention;

[0033] 图5是本发明的在线阶段自适应簇分裂定位流程图; [0033] FIG. 5 is an online phase of the present invention positioned flowchart adaptive split clusters;

[0034] 图6是本发明的在线定位结果示例图; [0034] FIG. 6 is an example of a positioning result online view of the invention;

[0035] 图7是本发明的在线定位过程示例图。 [0035] FIG. 7 is an exemplary line positioning process of the present invention of FIG.

具体实施方式 Detailed ways

[0036] 下面结合附图和实施例对本发明进行详细描述。 Drawings and embodiments of the present invention will be described in detail [0036] below in conjunction. 如图1所示,本发明分为离线指纹采集阶段和在线匹配定位阶段两个阶段。 1, the present invention is classified into offline and online fingerprint matching stage positioning stage two stages.

[0037] 本发明的离线指纹采集阶段,是在传统RSS指纹库基础上,运用自适应簇分裂聚簇方法对栅格数量进行压缩,运用AP重选择策略使每个簇选出定位能力最强的AP集合,最终得到新定义的V-RSS指纹库,在保证定位精度的前提下,既减少了需要存储的指纹数量,又降低了指纹向量的维度,运用簇分裂自学习机制后,以上过程均自动完成,无需人工干预。 [0037] Offline fingerprint stage of the present invention, in a conventional RSS fingerprint database is based on the use of an adaptive clustering method for clustering the number of division grids compression, using AP reselection policy is selected so that each cluster positioned strongest the AP collection, end up V-RSS fingerprint database new definition, under the premise of ensuring the positioning accuracy, reducing the number of fingerprints to be stored, but also reduces the dimensions fingerprint vector, the use of the cluster split self-learning mechanism, the above process all done automatically, without human intervention.

[0038] 本发明的离线指纹采集阶段主要完成指纹采集和指纹库的建立,包括建立基础RSS指纹库和在此基础上建立V-RSS指纹库; [0038] Offline fingerprint collection stage to complete the establishment of the present invention is mainly fingerprint and the fingerprint library, including the establishment of RSS based fingerprint database and V-RSS established on the basis of the fingerprint database;

[0039] 首先将待定位区域均匀划分为若干栅格,选择各栅格的几何中心作为参考点;然后在每个参考点进行多次测量,存储所有能测量到的AP的信号强度,取平均值后生成RSS向量,存入基础指纹数据库; [0039] First to be positioned evenly divided into several raster areas, the geometric center of each raster is selected as a reference point; and a plurality of measurements at each reference point, can be stored all the measured signal strength of the AP, averaging after the value generated RSS vectors stored in the base fingerprint database;

[0040] V-RSS指纹数据库的建立,首先对待定位区域进行簇预分裂,然后选出该簇中定位能力最强的AP组成该簇的最佳AP集合,再依据一定原则(例如,最佳AP集合对该区域的辨识能力是否超过阈值)判定该簇是否有必要继续簇分裂,并将该簇的相关数据存入V-RSS指纹库中。 [0040] established V-RSS fingerprint database, the first cluster area to treat pre-positioning division, and then select the best collection of most of the AP cluster positioning capability AP composition of the cluster, and then based on certain principles (for example, the best AP discerning collection of the region exceeds the threshold value) determine whether it is necessary to continue the cluster cluster split, and the cluster of related data stored in the V-RSS fingerprint database. 此后,对于有必要继续簇分裂的簇再次进行簇预分裂、AP重选择、判断能否继续分裂的流程,直到所有簇都不需要再分裂为止。 Since then, the need to continue to cluster cluster cluster split again pre-split, AP reselection to determine whether to continue splitting process until all the clusters do not need to split up. 自适应簇分裂完成后,整个区域被分割为大小不一致的许多簇,这些簇都不需要继续分裂(因为该区域中的所有AP都不具备将该簇细分的能力),簇的较长边长可以视为该簇的最小定位精度;本发明的V-RSS指纹数据库以四叉树(平面定位)或八叉树(空间定位)的形式存储。 After the completion of adaptive split clusters, the entire region is divided into a number of inconsistencies cluster size, these clusters do not need to continue to divide (as all the area AP do not have the ability to subdivide the cluster), cluster longer sides It can be regarded as the minimum length of the positioning accuracy of the cluster; V-RSS fingerprint database according to the present invention quadtree (plane location) or octree (spatial positioning) stored.

[0041] 本发明的在线匹配定位阶段打破了基于传统RSS指纹库穷举匹配的方法,本发明采用了基于V-RSS指纹库的自适应簇分裂定位方法,从而大大降低了计算复杂度。 Online [0041] The present invention breaks the positioning phase matching method based on a conventional RSS exhaustive matching fingerprint database, the present invention uses a split V-RSS based on adaptive fingerprint database cluster positioning method, thereby greatly reducing the computational complexity. 本发明的在线匹配定位阶段的主要步骤是从V-RSS指纹库中选取与实时指纹最匹配的簇,完成空间定位。 The main steps of online matching positioning phase of the present invention is to select clusters with real-time fingerprint best matches from the V-RSS fingerprint database, perform spatial orientation. 首先对待定位的移动终端接收到的所有AP的信号强度进行实时采样和去噪,记录RSS 向量;然后运用基于四叉树(平面定位)或八叉树(空间定位)搜索的自适应簇定位算法,从根簇开始,将当前簇分为四个或八个子簇,计算实时指纹与四个或八个子簇的重心指纹的欧式距离,选择欧式距离最近的子簇作为粗定位结果,然后再对该子簇继续簇分裂、欧式距离计算与选择等流程,直到当前子簇不能再分裂为止,将该簇作为定位结果输出;以该簇的几何中心作为待定位的移动终端的位置估计,以该簇的较长边长作为定位结果的不确定度。 AP signal strength of the first treatment all positioned mobile terminal receives real-time sampling and noise removal, recording RSS vector; and then use an adaptive clustering algorithm based on quadtree Location (plane location) or octree (spatial localization) search , starting from the root cluster, this cluster will be divided into four or eight sub-clusters, and calculating a real fingerprint four or eight sub-cluster Euclidean distance of the center of gravity of fingerprints, selecting the nearest Euclidean distance coarse positioning sub-clusters as a result, and then to the split sub-cluster clustering continues, Euclidean distance calculation and selection process, until the current sub-clusters can not be split up, the cluster as the positioning result output; position of the geometric center of the cluster to be positioned as a mobile terminal is estimated, in this the longer side length cluster uncertainty as a result of positioning. 以下将以平面定位问题为例,说明本发明的具体实施方式。 The following problems will be positioned plane to illustrate specific embodiments of the present invention. 下面是本发明的附图中涉及的各变量所表示的含义的详细列表: The following is a detailed list of the meaning of each variable in the drawings according to the present invention is represented by:

[0042] [0042]

Figure CN103763769BD00081

[0043] [0043]

Figure CN103763769BD00091

[0044] 为了方便表述,本发明首先定义簇的编号方式:定义待定位区域的整体为根簇,簇编号为〇。 [0044] For convenience of description, the present invention first define the cluster numbering: location area is defined to be the root of the entire cluster, the cluster number is square. 如果根簇可以分裂,则将四个子簇依次编号为1,2,3,4,这四个簇为第1层簇。 If the root cluster can be split, then the four sub-clusters are numbered 1, 2, four clusters Tier 1 cluster. 再对第1层的所有簇进行分裂判决,以簇2为例(其他同理),如果簇2可以被继续分裂,则将四个子簇依次编号为21,22,23,24,这些簇为第2层簇;如果簇2不可以继续分裂,则不存在簇21, 22,23,24。 Then all the clusters of the first layer is split decision, for example in cluster 2 (other empathy), if the cluster 2 may continue to be split, then the four sub-clusters are numbered 21, 22, these clusters layer 2 clusters; cluster 2 if not continue to divide, no clusters 21, 22, 23 exists. 依次类推可以得到所有簇的编号规则,例如在图7中,红色标记位置的簇号(1= 423,它表示簇4的第2个子簇的第3个孙子簇,该簇为第3层簇。 And so on can be obtained all clusters numbering plan, for example, in FIG. 7, the cluster number of red marking locations (1 = 423, which represents the second sub-clusters of 3 grandchild clusters with the cluster 4, the cluster is a Layer 3 clusters .

[0045] 本发明包括以下步骤: [0045] The present invention comprises the steps of:

[0046] 步骤一:按照传统方法将待定位区域均匀划分为若干栅格(grid),选取每个栅格的几何中心作为参考点位置,在每个参考点位置测量能接收到的所有接入点(Access Point,AP)的信号强度,生成接收信号强度指纹向量(RSS向量),建立基础RSS指纹库;将第i 个参考点的指纹标记为 [0046] Step a: according to conventional methods to be positioned evenly divided into several raster areas (Grid), select all access the geometric center of each grid point as a reference position, the measurement can be received in each reference position point (Access point, AP) signal strength, received signal strength to generate the fingerprint vector (vector RSS foundation), RSS fingerprint database; fingerprint mark i-th reference points is

Figure CN103763769BD00101

则基础RSS指纹库为 The basis for the RSS fingerprint database

[0047] [0047]

Figure CN103763769BD00102

[0048] 步骤二:结合基于地理位置和辨识度的AP重选择方法、自适应簇分裂的聚簇方法和自学习机制,生成栅格大小可变的RSS指纹库(V-RSS指纹库); [0048] Step Two: clustering method based AP binding location and recognizable reselection method, adaptive and self-learning cluster splitting mechanism, generating a grid of variable size RSS fingerprint database (V-RSS fingerprint database);

[0049] 图2展示了待定位区域如何生成V-RSS指纹库的具体步骤。 [0049] Figure 2 shows how to generate specific steps V-RSS fingerprint database region to be positioned. 图3则具体展示了每一层栅格是如何分裂和确定生成的流程,依照图3对每一步做具体说明; Figure 3 shows a specific grid for each layer to determine how to divide and generate process, in accordance with specific instructions to do every step of FIG. 3;

[0050] (0)初始时,设置一个存储待分裂簇的空栈,并将当前簇设置为根簇; [0050] (0) initially, setting a storage cluster empty stack to be split, and current cluster to the root cluster;

[0051] (1)设当前簇为q,对q做预分裂,如图2所示,对簇q作虚拟的十字均分,预分裂后得至1J四个子簇qi,q2,q3,q4; [0051] (1) is provided as the current cluster q, q-made pre-split, as shown, to cross the virtual cluster for sharing q 2, after the pre-split to obtain four sub-clusters 1J qi, q2, q3, q4 ;

[0052] (2)计算各子簇的重心指纹向量(简称重心指纹);以子簇qi为例,其重心指纹rqi= (rqi,i,rqi,2. . .rqi,M),其中rqi,j表不第j个AP在qi子簇内的平均接收信号强度; [0052] (2) Calculate the centroid vector of each sub-cluster fingerprint (fingerprint center of gravity referred to); qi an example of sub-clusters, which the center of gravity fingerprint rqi = (rqi, i, rqi, 2 .rqi, M..), Where RQI , j table is not the j-th average received signal strength of the AP qi within the sub-cluster;

[0053] (3) AP重选择;AP重选择指当前簇改变后需要重新进行AP选择,最终为不同地理位置的簇定义自己的最佳AP集合;最佳AP集合的选择方法是从AP集合中选取定位能力最强的若干AP,其目的是滤除定位能力差、稳定度差的AP以提高定位精度,并减少在线定位的计算量;AP选择已有很多算法,本发明采用辨识度作为AP选择的依据,辨识度定义为簇间距离与簇内方差的比值,选取辨识度最大的若干AP组成最佳AP集合;在该算法中,辨识度的定义引入了Fi sher准则函数;Fi sher准则是判别个体所属类别的一种多元统计分析方法,它能够对两个类别间的分离度进行定量描述;其定义式为: [0053] (3) AP reselection; AP reselection refers to the current need to re-select the AP after changing the cluster, ultimately defines its own set of best AP cluster geographically; method of selecting the best AP from the AP collection set select a plurality of strongest AP positioning capability, its purpose is to filter positioned poor, poor stability of the AP to improve the positioning accuracy, and reduce the amount of the calculated positioning line; AP selection algorithm has a lot of, the present invention is employed as a recognizable based on the AP selection, and the ratio of the distance between the clusters in the cluster variance is defined as the recognition, select the maximum number of recognizable best AP AP composition set; in this algorithm, the definition introduces a recognizable function Fi sher norms; Fi sher guidelines are a multivariate statistical analysis of discriminating individual category, it is possible to quantitatively describe the degree of separation between the two classes; formula defined as:

Figure CN103763769BD00103

[0054] 其中,J (Y)越大则说明两类之间分离度越大;且该准则可以推广到对多个类间的分离度进行描述;应用到本发明的AP算法中,tq, j表示第j个AP在簇q内的辨识度,tq, j越大则表示第j个AP在簇q的四个子簇91冲,阳冲间的区分度越大(对簇9的区分能力越强);以上AP选择算法仅为举例,本发明方法亦可采用其他AP选择算法; [0054] wherein, J (Y) then the greater the greater the degree of separation between the two types; and the criteria can be extended to the degree of separation between the plurality of classes is described; AP algorithm applied to the present invention, tq, j represents the j-th degree of recognition within the AP cluster of q, TQ, the greater j indicates the j-th cluster in the punch AP q four sub-clusters 91, the greater the discrimination between a male punch (ability to distinguish clusters 9 stronger); AP selection algorithm above is only an example, the method of the present invention may also employ other AP selection algorithm;

[0055] 选择最佳AP集合的具体步骤为: [0055] In particular the step of selecting the best set of AP:

[0056] (31)先计算子簇的重心指纹和簇内方差; [0056] (31) to the center of gravity is calculated fingerprint and variance within the sub-cluster of the cluster;

[0057] 以簇qi中第j个AP为例,其重心指纹 [0057] In cluster qi j-th AP as an example, the fingerprint center of gravity

Figure CN103763769BD00104

[0058] 其簇内方差 [0058] within its cluster variance

Figure CN103763769BD00105

[0059] 同理可得 [0059] Similarly available

Figure CN103763769BD00106

[0060] (32)对簇q,第j个AP的辨识度定义为: [0060] (32) of the cluster q, identification of the j-th AP is defined as:

[0061] [0061]

Figure CN103763769BD00111

[0062] 辨识度向量tq= (tq,l,tq,2···,tq,M); [0062] Identification of vector tq = (tq, l, tq, 2 ···, tq, M);

[0063] (33)对tq, j按由大到小的顺序排序,取出前N个组成最佳AP集合,把对应的AP编号存在向量Vq中; [0063] (33) for tq, j sorted in descending order, the best AP composition taken before a set of N, the number corresponding to the AP in the presence of the vector Vq;

[0064] (34)分别得到四个子簇的簇指纹。 [0064] (34) respectively fingerprint clusters of four sub-clusters. 例如,子簇qi簇指纹定义为是指从rqi中选出Vq 对应的指纹,得到经过AP选择后子簇qi的重心指纹uql,uql可以表示为: For example, a fingerprint sub-clusters qi defined to mean cluster selected from rqi Vq corresponding to the fingerprint, the obtained AP after selecting the center of gravity of the sub-clusters qi fingerprint uql, uql may be expressed as:

[0065] [0065]

Figure CN103763769BD00112

[0066] 同理可得 [0066] Similarly available

Figure CN103763769BD00113

with

Figure CN103763769BD00114

的区别是后者包括了所有AP在簇qi的指纹,其维度为M,而前者只包括了最佳AP集合的指纹,其维度为N,N〈M; The difference is that the latter includes all fingerprints AP cluster qi, which dimension is M, while the former includes only the best fingerprint collection AP, which dimension is N, N <M;

[0067] (4)定义函数 [0067] (4) defined function

Figure CN103763769BD00115

用于判断当前簇可否分裂: Used to determine whether the current cluster division:

[0068] [0068]

Figure CN103763769BD00116

[0069] 其中ε为可分裂门限(辨识度阈值),既可根据经验由人工设置,也可由自学习机制自动获得,其方法是将辨识度阈值定为预期定位误差、无线信号波动特征和簇尺寸的函数, 通过自学习的方式确定该函数的具体表达式或者数值对应关系。 [0069] where ε is a splittable threshold (recognition threshold), either empirically by the manually set, can also be automatically self-learning mechanism, which is the recognition threshold set to the expected position error, radio signal fluctuation characteristics and clusters function of the size, to determine the specific value of the expression or function of a correspondence relationship by self-learning manner.

[0070] (5)存储数据:将簇编号q、判别函数Γ (tq)的值、当前簇最佳AP集合Vq存储到V-RSS 指纹库中; [0070] (5) for storing data: the cluster number q, the value of discriminant function Γ (tq), the collection AP Vq best current cluster is stored in the V-RSS fingerprint database;

[0071] 如果Γ (tq)=l,跳至丨」步骤(6); [0071] If Γ (tq) = l, skip Shu "in step (6);

[0072] 如果Γ (tq) =0,跳到步骤(7); [0072] If Γ (tq) = 0, skip to step (7);

[0073] (6)对簇q进行分裂;实际上就是对簇预分裂的确认过程,对簇q进行十字均分,得至1J四个子簇qi,q2,q3,q4,存储Vq在四个子簇产生的重心指纹Uql,Uq2,Uq3,Uq4,将簇q2,q3,q4压入栈中,当前簇设置为如,跳到步骤(I); [0073] (6) to the cluster to divide q; the validation process is actually a cluster of pre-splitting, sharing of the cluster q be a cross, four sub-clusters have to 1J qi, q2, q3, q4, stored in the four sub Vq Uql cluster centroid fingerprint generated, Uq2, Uq3, Uq4, a cluster q2, q3, q4 pushed onto the stack, is set as the current cluster, skip to step (the I);

[0074] (7)当前簇q不能再分裂,返回栈中取出一个新簇作为当前簇,跳到步骤(1);如果栈空则整个簇分裂过程结束,V-RSS指纹库生成完毕。 [0074] (7) can no longer divide q current cluster, a new cluster taken as the current cluster return stack, skip to step (1); if the stack is empty, the entire cluster splitting process ends, V-RSS fingerprint database generation is complete.

[0075] V-RSS指纹库将以四叉树方式存储,图4给出了一个示例。 [0075] V-RSS fingerprint database will be stored quadtree, FIG. 4 shows an example.

[0076] 下面举例说明生成V-RSS指纹库的过程。 [0076] The following example illustrates the process of generating a V-RSS fingerprint database. 如图2所示,以第1层的簇4的分裂过程为例: 2, the division process of clusters the first layer of Example 4:

[0077] (1)簇4经过预分裂,生成四个子簇41,42,43,44; [0077] (1) pre-4 cluster splitting, generates four sub-clusters 41, 42;

[0078] (2)第j个AP在四个子簇的位置指纹和方差分别是Γ41, j,Γ42, j,Γ43, j,Γ44, j和 [0078] (2) the position of j-th AP at the fingerprint and variance four sub-clusters respectively Γ41, j, Γ42, j, Γ43, j, Γ44, j and

Figure CN103763769BD00117

由此可以算出它在簇4的辨识度t4,j; Thus it can be calculated recognizable cluster 4 t4, j;

[0079] (3)假设空间中共出现了10个AP的信号,考虑到计算复杂度需要从中选出4个AP参加定位(即M=IO,N=4),对t4, j (j=l,2··· 10)排序,取出最大的4个作为最佳AP集合;假设辨识度排名前四的AP分别是第7,9,1,5个AP,则V4= (7,9,1,5),U4= (r4,i,r4,5,r4j,r4,9),这两个数据将存入V-RSS指纹库的节点4中。 [0079] (3) appeared in the hypothesis space CCP AP 10 signals, taking into account the computational complexity required to choose the positioning participate AP 4 (i.e., M = IO, N = 4), to t4, j (j = l , 2 · · · 10) sorted, the maximum taken as an optimum set of four AP; AP front four rankings are assumed recognition of a 7,9,1,5 AP, then V4 = (7,9,1 , 5), U4 = (r4, i, r4,5, r4j, r4,9), two data will be stored in the fingerprint database nodes V-RSS 4.

[0080] (4)计算最佳AP集合的平均辨识度,若它大于阈值ε则Γ (t4)=l,否则Γ (t4)=0;假设Γ (t4)=l,则簇4的预分裂被确认有效,形成四个新簇41,42,43和44。 [0080] (4) calculating the average recognizable best AP set, if it is larger than the threshold ε is Γ (t4) = l, otherwise Γ (t4) = 0; Suppose Γ (t4) = l, then the cluster pre-4 the division was validated, the formation of four new clusters 42, 43 and 44.

[0081] 对簇41,42,43和44依次进行类似簇4的流程,可以对整个待定位区域进行更细化的分割。 [0081] The clusters 42, 43 and 44 sequentially 4 clusters similar process can be performed more refined division of the entire area to be located. 值得注意的是,判定簇4是否能分裂的依据是Γ (t4),它用到的AP是所有AP中区域分割能力最强的前四个,即(7,9,1,5);而判定簇41是否能分裂的依据是Γ (t41),它用到的AP与(7,9,1,5)不一定相同,此即AP重选择。 It is noteworthy that, according to determine whether to split a cluster 4 is Γ (t4), which uses the AP is the strongest of all in the AP region segmentation capability of the first four, namely (7,9,1,5); and determine whether the cluster 41 division is based Γ (t41), which uses the AP and (7,9,1,5) are not necessarily the same, namely AP reselection.

[0082] 计算离线阶段指纹数据的存储量: [0082] The calculated amount of stored fingerprint data offline phase:

[0083] 以图2中的V-RSS指纹库为例,图中C=16*16=256,M=10,N=4,V-RSS指纹库共有第1 层簇1个,第2层簇10个,第3层簇4个,第4层簇16个,其中每个簇需要存储1个簇编号、1分裂判决变量和4个指纹,总存储量为:(1 + 10+4+16) * (1+1+4) =186;而传统RSS指纹库的存储量为:256* (1+1+10) =3072,可以算出本发明方法的指纹存储量与传统方法的比值: [0083] In FIG 2 V-RSS fingerprint database, for example, in FIG C = 16 * 16 = 256, M = 10, N = 4, V-RSS consensus fingerprint database cluster a first layer, the second layer cluster 10, cluster 4 layer 3, fourth layer 16 clusters, wherein each cluster need to store a cluster number, a four split decision variables and fingerprints, the total storage capacity of: (1 + 4 + 10 + 16) * (1 + 1 + 4) = 186; stored amount of conventional RSS fingerprint database is: 256 * (1 + 1 + 10) = 3072, can be calculated by the method of the present invention, the fingerprint storage with traditional methods the ratio of:

[0084] [0084]

Figure CN103763769BD00121

[0085] 步骤三:将待定位的移动终端移动到某个位置,测量实时接收的RSS向量: [0085] Step 3: the mobile terminal moves to be positioned to a position, real-time measurement vector the received RSS:

[0086] W= (wi, · · · ,Wj, · · -wm) [0086] W = (wi, · · ·, Wj, · · -wm)

[0087] 步骤四:将步骤三中测得的RSS向量与V-RSS指纹库进行比对,运用基于多叉树的自适应簇分裂定位方法估计待定位的移动终端的位置。 [0087] Step four: three step RSS measured vector V-RSS fingerprint database for comparison, using an adaptive multi-cluster-tree based splitting method for estimating position of the mobile terminal is positioned to be located. 现结合图5对步骤四做具体说明。 In conjunction with FIG. 5 is now specifically described by four steps. 为了方便表示,以四叉树中的节点表示簇,即当前当前节点Ak等价于当前簇q。 For ease of presentation, the quadtree nodes represent clusters, i.e., the current equivalent to the current node in the current cluster Ak q.

[0088] (1)从根节点Ak,k=0开始; [0088] (1) from the root node Ak, k = 0 starts;

[0089] (2)读取V-RSS指纹库存储的数据,判断当前节点11{是否可以继续分裂:如果可以继续分裂,继续执行步骤(3),否则跳至步骤(7); [0089] (2) V-RSS reading data stored in the fingerprint database, it is determined whether the current node can continue to divide {11: If you can continue to divide, proceed to step (3), otherwise skip to step (7);

[0090] (3)当前节点下一层的四个子节点记为Θ=1,2,3,4;从V-RSS指纹库中读取每个子节点对应的重心指纹ue,0e [1,4]; [0090] (3) at the current node level four sub-node denoted Θ = 1,2,3,4; reading each child node corresponding to the center of gravity from the V-RSS ue fingerprint fingerprint database, 0e [1,4 ];

[0091] (4)待定位的移动终端的实时指纹经当前节点的最佳AP集合选择后得到 [0091] (4) a mobile terminal in real-time fingerprint to be positioned to give the best AP after selecting the set of the current node

[0092] [0092]

Figure CN103763769BD00122

[0093] (5)从四个子节点中找出与 [0093] (5) from four child nodes to identify with

Figure CN103763769BD00123

欧式距离最近的,其编号记为Ak+1; Nearest Euclidean distance, denoted by its number Ak + 1;

[0094] [0094]

Figure CN103763769BD00124

[0095] 欧氏距离公式: [0095] Euclidean distance formula:

Figure CN103763769BD00125

[0096] (6)Ak+1成为当前节点,跳至步骤(2); [0096] (6) Ak + 1 becomes the current node, go to step (2);

[0097] (7)当前节点对应的簇即为最佳匹配簇,输出簇信息作为定位结果,包括簇编号(节点编号)和簇的尺寸。 [0097] (7) a cluster corresponding to the current node is the best match cluster, the cluster information is output as a result of positioning, including a cluster number (node ​​number) and size of the cluster.

[0098] 步骤五:输出定位结果,所述定位结果包括最佳匹配的簇编号、待定位的移动终端的估计位置和不确定度,具体实施方案为:以最佳匹配簇的几何中心作为待定位的移动终端的位置估计,以最佳匹配簇的较长边长作为定位结果的不确定度。 [0098] Step 5: To a positioning result, the positioning result including the best matching cluster number, and the estimated uncertainty of the position of the mobile terminal to be positioned, as specific embodiments: cluster best match as determined geometric center bit position estimate for the terminal, a longer side length of best match cluster uncertainty as positioning result.

[0099] 下面举例说明步骤四和步骤五。 [0099] The following illustrates steps four and five steps.

[0100] 如图7所示,假设待定位的移动终端在图中标记的位置,且待定位区域是一个边长为16m的正方形区域,下面结合动态图解释定位算法的流程。 [0100] As shown in FIG. 7, the position of the mobile terminal is assumed to be located in the figures mark, Judai positioning region is a square region of side length of 16m, the flow explained below in connection with dynamic positioning algorithm of FIG.

[0101] (1)当前簇为根簇,假设根簇可以分裂为四个子簇1、2、3、4,从V-RSS指纹库中可以读出根簇的最佳AP集合VQ以及四个子簇的重心指纹Ul、U2、U3和U4 ; [0101] (1) the current clusters clump, assuming the root cluster may split into four sub-clusters 1,2,3,4, V-RSS from the fingerprint database can be read out of the best AP clump and four sub-set of VQ the center of gravity of the cluster fingerprint Ul, U2, U3 and U4;

[01 02] (2)设待定位的移动终端实时测量得到的指纹向量是w,从w中选出VQ中AP对应的指纹,组成AP重选择后的实时指纹向量 [0102] (2) real-time measurements fingerprint vector is provided a mobile terminal to be located is obtained w, select the VQ corresponding to the fingerprint from the AP w in real time fingerprint vector composition after AP reselection

Figure CN103763769BD00131

[0103] (3)计算 [0103] (3) Calculate

Figure CN103763769BD00132

与Ul、U2、U3和U4的欧式距离,设 And Ul, U2, U3 and U4, Euclidean distance, provided

Figure CN103763769BD00133

与U4的欧式距离最小,贝Ij判定待定位的移动终端在簇4中; With a minimum Euclidean distance U4, shellfish Ij determined to be located in the cluster of the mobile terminal 4;

[0104] (4)假设簇4可以继续分裂,得到四个子簇41、42、43、44,从V-RSS指纹库中可以读出簇4的最佳AP集合V4以及四个子簇的重心指纹U41、U42、U43和U44 ;从W中选出V4中AP对应的指纹,组成AP重选择后的实时指纹向量 [0104] (4) assuming clusters 4 may continue to divide, to obtain four sub-clusters 41, 42, V-RSS from the fingerprint database can be read out cluster centroid fingerprint best set the AP 4 and four sub-clusters V4 U41, U42, U43 and U44; V4 are selected from W AP corresponding fingerprint, the fingerprint in real time the composition of the vector AP reselection

Figure CN103763769BD00134

计算 Compute

Figure CN103763769BD00135

与U41、U42、U43和U44的欧式距离,设 And U41, U42, U43 and U44 Euclidean distance, provided

Figure CN103763769BD00136

与U42的欧式距离最小,则判定待定位的移动终端在簇42中; U42 with minimum Euclidean distance, is determined to be located in the mobile terminal 42 in the cluster;

[0105] (5)重复上一步骤,继续判定待定位的移动终端在簇423中;接下来发现簇423已经不可分裂,因此把待定位的移动终端定位到簇423的几何中心,其坐标为(9,5);考虑到簇423的边长为2米,因此定位不确定度取为2米,最终的定位结果为: [0105] (5) Repeat the previous step, the mobile terminal continues to be determined in the tufts 423 positioned; next cluster 423 has been found to not split, so the mobile terminal to be positioned is positioned to the geometric center of the cluster 423, which coordinates (9,5); take into account the cluster 423 side length of 2 meters, so the positioning uncertainty is taken to be 2 m, for the final positioning result:

Figure CN103763769BD00137

米。 Meter.

[0106] 图6是为待定位移动终端寻找最佳匹配簇的示意图;四叉树是指V-RSS的存储形式,利用四叉树计算最佳匹配点;该过程既可由具有相应处理能力的待定位智能移动终端自行完成,也可由待定位终端上报AP后由AP完成。 [0106] FIG. 6 is a schematic view of a cluster to find the best match for the mobile terminal to be located; quadtree storage means in the form of V-RSS is calculated using the best matching point quadtree; this process can having a respective processing capacity intelligent mobile terminal to be located complete its own to be located after the terminal may also be completed by the AP AP reports.

[0107] 在线定位的计算复杂度: [0107] positioned online computational complexity:

[0108] 在线定位的计算量主要消耗在位置指纹与实时指纹的欧式距离计算上,下面分别从理论和实例两个角度计算本发明的方法和传统指纹定位方法在计算复杂度上的对比。 [0108] positioned online calculation mainly consumed in the Euclidean distance calculation in real time the position of the fingerprint with the fingerprint, the following method of the present invention were calculated and the comparison conventional fingerprinting positioning method in computational complexity theory and examples from two angles.

[0109] 传统指纹定位方法如果采用穷举法寻找欧式距离最近的指纹位置,则计算复杂度应为 [0109] If the conventional fingerprinting positioning method using brute-force method to find the nearest Euclidean distance fingerprint position, the computational complexity should be

Figure CN103763769BD00138

,本发明的方法只需要与部分AP在部分位置的指纹进行比对,最大计算复杂度应为 The method of the present invention requires only a portion are aligned in the AP position of the fingerprint portion, the maximum computational complexity should be

Figure CN103763769BD00139

,可以算出计算复杂度的比率: It can be calculated from the ratio of the computational complexity:

[0110] [0110]

Figure CN103763769BD001310

[0111] 以上述例子为例, [0111] In the above example, for example,

Figure CN103763769BD001311

传统方法的欧式距离计算次数为256* 10=2560次,如果传统方法也只采用4个AP完成定位过程,则计算次数下降为256*4=1024次; 本发明的方法的欧式距离计算次数为(4+4+4) *4=48次,可以算出: Euclidean distance of the conventional method of calculating the number of 256 * 10 = 2560 times, if the conventional method of only using the 4 AP complete the positioning process, the calculated decrease in the number of 256 * 4 = 1024; Euclidean distance method of the present invention calculates the number of (4 + 4 + 4) * 4 = 48 times, can be calculated:

[0112] [0112]

Figure CN103763769BD001312

[0113] 即本发明的方法在在线定位阶段的计算量是传统方法计算量的5%,即本发明的方法大大节省了定位时间和定位过程中待定位的移动终端消耗的电能。 [0113] The method of the present invention, i.e., the conventional method is to calculate an amount of 5% of the calculated amount of online positioning stage, i.e., the method of the present invention saves power consumption of the mobile terminal during positioning time and location to be positioned.

[0114] 为描述方便,本发明的说明书中均以平面定位问题为例子展开叙述,但本发明也可用于空间定位问题,只需将本说明书中的定位区域改为定位空间,栅格改为立体栅格,四叉树改为八叉树,步骤一的栅格改为立体栅格,将步骤二和步骤四中的四叉树改为八叉树, 其他步骤、流程和算法均相同。 [0114] For ease of description, the description of the present invention are planar positioning problems described as an example deployment, the present invention can also be used for positioning of the space, only the location area in this specification is positioned to space the grid to dimensional grid quadtree to octree grid step to a three-dimensional grid, and the step of the step two to four quadtree octree, other steps, processes and algorithms are the same.

[0115] 以上所述,仅为本发明的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明公开的范围内,能够轻易想到的变化或替换,都应涵盖在本发明权利要求的保护范围内。 [0115] The above are only specific embodiments of the present invention, but the scope of the present invention is not limited thereto, any skilled in the art within the scope of the art disclosed in the present invention, can easily think of variations or Alternatively, it shall fall within the scope of the claims of the invention.

Claims (6)

1.基于接入点重选择和自适应簇分裂的室内指纹定位方法,其特征在于,包括以下步骤: 步骤一:按照传统方法将待定位区域均匀划分为若干栅格,选取每个栅格的几何中心作为参考点位置,在每个参考点位置测量能接收到的所有接入点的信号强度,生成接收信号强度指纹向量,即RSS向量,建立基础RSS指纹库; 步骤二:结合基于地理位置和辨识度的AP重选择方法、自适应簇分裂聚簇方法和自学习机制,生成栅格大小可变的RSS指纹库,S卩V-RSS指纹库; (0) 初始时,设置一个存储待分裂簇的空栈,并将当前簇设置为根簇; (1) 设当前簇为q,对q做预分裂,对簇q作虚拟的十字均分,预分裂后得到四个子簇ql·, Q2,q3,q4; ⑵计算各子簇的重心指纹向量;以子簇qi为例,其重心指纹rqi= (rqi,i,rqi,2"_rqi,M), 其中rql>j表示第j个六?在(11子簇内的平均接收信号强度; (3 1. Based on the access point and the adaptive reselection cluster splitting chamber fingerprinting positioning method, characterized by comprising the following steps: Step 1: according to the conventional method the location area to be evenly divided into several grids, each grid select signal strength geometric center position as a reference point, the measurement can be received at each reference point positions of all the access points, received signal strength to generate a fingerprint vector, i.e. a vector RSS, RSS established based fingerprint database; step two: location-based binding AP reselection method and recognizable, adaptive clustering method cluster splitting and self-learning mechanism, generating a grid of variable size RSS fingerprint database, S V-RSS Jie fingerprint database; (0) initially, setting a memory to be empty stack split clusters, and the current cluster to the root cluster; (1) set the current get four sub-clusters after cluster ql · q, q to do pre-split, to make virtual cluster q cross-sharing, pre-split, Q2, q3, q4; ⑵ calculated for each sub-cluster centroid fingerprint vector; qi an example of sub-clusters, which the center of gravity fingerprint rqi = (rqi, i, rqi, 2 "_rqi, M), wherein rql> j represents the j th ? six average received signal strength (within the sub-cluster 11; (3 ) AP重选择,当前簇改变后需要重新进行AP选择,最终为不同地理位置的簇定义自己的最佳AP集合;选择最佳AP集合的具体步骤为: (31) 先计算子簇的重心指纹和簇内方差; 以簇qi中第j个AP为例,其重心指纹 ) AP reselection, need to be changed after the current cluster the AP selection, ultimately defines its own set of best AP cluster geographically; AP selecting the best set of specific steps as follows: (31) first calculates the center of gravity of the fingerprint sub-clusters and cluster variances; qi in cluster j-th AP as an example, the fingerprint center of gravity
Figure CN103763769BC00021
表示第j个AP在第i个参考点的位置指纹;Cqi表示簇qi中参考点数目; 其族内方差 AP denotes the j-th position in the i-th fingerprint reference points; CQI qi represents the number of reference points in the cluster; its variance Group
Figure CN103763769BC00022
同理可得rq2,j Similarly available rq2, j
Figure CN103763769BC00023
(32) 对簇q,第j个AP的辨识度定义为: (32) for clusters of q, the j-th identification of AP is defined as:
Figure CN103763769BC00024
辨识度向量tq= (tq,l,tq,2"_,tq,M); (33) 对tq,j按由大到小的顺序排序,取出前N个组成最佳AP集合,把对应的AP编号存在向量Vq中; (34) 分别得到四个子簇的簇指纹;例如,子簇qi簇指纹定义为是指从中选出Vq对应的指纹,得到经过AP选择后子簇qi的重心指纹Uq1,Uqi可以表示为: Identification of vector tq = (tq, l, tq, 2 "_, tq, M); (33) for tq, j sorted in descending order, the best composition before removing the N AP set, the corresponding AP vector Vq number exists; and (34) respectively four clusters fingerprint sub-cluster; for example, a fingerprint sub-clusters qi is defined to mean a cluster selected from the corresponding fingerprint Vq to obtain the center of gravity after the fingerprint Uq1 AP selecting sub-clusters of qi , Uqi can be expressed as:
Figure CN103763769BC00025
同理可得1VlVuq4 和1;的区别是后者包括了所有AP在簇qi的指纹,其维度为M, 而前者只包括了最佳AP集合的指纹,其维度为N,N〈M; ⑷定义函数F (tq),用于判断当前簇可否分裂: Similarly available 1VlVuq4 and 1; the difference is that the latter includes all fingerprints AP cluster qi, which dimension is M, while the former includes only the best fingerprint collection AP, which dimension is N, N <M; ⑷ defined function F (tq), used to determine whether the current cluster division:
Figure CN103763769BC00031
其中ε为可分裂门限,既可根据经验由人工设置,也可由自学习机制自动获得,其方法是将辨识度阈值定为预期定位误差、无线信号波动特征和簇尺寸的函数,通过自学习的方式确定该函数的具体表达式或者数值对应关系; ⑶存储数据:将簇编号q、判别函数Γ (tq)的值、当前簇最佳AP集合vq存储至IjV-RSS指纹库中; 如果Γ (tq)=l,跳到步骤(6); 如果Γ (tq)=0,跳到步骤⑵; (6) 对簇q进行分裂;实际上就是对簇预分裂的确认过程,对簇q进行十字均分,得到四个子簇qi,q2,q3,q4,存储Vq在四个子簇产生的重心指纹Uql,Uq2,Uq3,Uq4,将簇q2,q3,q4压入栈中,当前簇设置为如,跳到步骤(I); (7) 当前簇q不能再分裂,返回栈中取出一个新簇作为当前簇,跳到步骤(1);如果栈空则整个簇分裂过程结束,V-RSS指纹库生成完毕; 步骤三:将待定位的移动终端移动到某个位置,测 Wherein ε is a splittable threshold, can also be obtained automatically by the manual setting self-learning mechanism based on experience, which is the identification threshold as a function of expected position error, radio signal fluctuation characteristics and the cluster size, the self-learned the way to determine the specific expression or function value corresponding relation; ⑶ storing data: the cluster number q, discriminant function Γ (tq) value, set the current cluster best AP IjV-RSS vq stored in the fingerprint database; Gamma] if ( tq) = l, skip to step (6); if Γ (tq) = 0, skip to step ⑵; (6) to the cluster to divide q; the validation process is actually a cluster of pre-splitting, to be cross cluster q average, to obtain four sub-clusters Qi, the center of gravity fingerprint Uql q2, q3, q4, Vq stored in the four sub-clusters generated, Uq2, Uq3, Uq4, a cluster q2, q3, q4 pushed onto the stack, is set as the current cluster , skip to step (I); (7) no longer divide the current cluster q, return stack out a new cluster as the current cluster, skip to step (1); if the stack is empty, the entire cluster splitting process ends, V-RSS fingerprint library generation is completed; step 3: the mobile terminal moves to be positioned to a location, measured 实时接收的RSS向量,多次测量取平均值并进行去噪处理; 步骤四:将步骤三中测得的RSS向量与V-RSS指纹库进行比对,运用基于多叉树的自适应簇分裂定位方法估计待定位的移动终端的位置;以四叉树中的节点表示簇,即当前节点Xk等价于当前簇q; (1-1)从根节点Xk,k = 0开始; (1-2)读取V-RSS指纹库存储的数据,判断当前节AAk是否可以继续分裂:如果可以继续分裂,继续执行步骤(1-3),否则跳至步骤(1-7); (1-3)当前节点下一层的四个子节点记为Θ = 1,2,3,4;从V-RSS指纹库中读取每个子节点对应的重心指纹ue,0e [1,4]; (1-4)待定位的移动终端的实时指纹经当前节点的最佳AP集合选择后得到 RSS vector received in real time, multiple measurements averaged and denoising; Step four: measured in step 3 and V-RSS RSS vector fingerprint database for comparison, the use of multiple adaptive tree-based clustering split estimated position of the mobile terminal positioning method to be positioned; quadtree representation of nodes in the cluster, i.e., the current equivalent to the current cluster node Xk q; (1-1) starting from the root node Xk, k = 0; (1- 2) V-RSS reads data stored fingerprint database to determine whether the current section AAk can continue to divide: If you can continue to divide, proceed to step (1-3), otherwise skip to step (1-7); (1-3 ) four sub-node under the current layer node denoted Θ = 1,2,3,4; reading each child node corresponding to the center of gravity from the V-RSS ue fingerprint fingerprint database, 0e [1,4]; (1- 4) real-time fingerprint mobile terminal to be positioned to give the best AP after selecting the set of the current node
Figure CN103763769BC00032
(1-5)从四个子节点中找出与%1(欧式距离最近的,其编号记为Ak+1; (1-5) 1% identify with the (nearest Euclidean distance, which is referred to as number Ak + 1 from four child nodes;
Figure CN103763769BC00033
欧氏距离公式: Euclidean distance formula:
Figure CN103763769BC00034
(1-6) Ak+1成为当前节点,跳至步骤(1-2); (1-7)当前节点对应的簇即为最佳匹配簇,输出簇信息作为定位结果,包括簇编号和簇的尺寸; 步骤五:输出定位结果,所述定位结果包括最佳匹配的簇编号、待定位的移动终端的位置估计和不确定度。 (1-6) Ak + 1 becomes the current node, go to step (1-2); (1-7) a cluster corresponding to the current node is the best match cluster, the cluster information is output as the positioning result, cluster numbers including cluster and size; step 5: to a positioning result, the positioning result including the best matching cluster number, the location of the mobile terminal to be positioned and estimate uncertainty.
2.根据权利要求1所述的基于接入点重选择和自适应簇分裂的室内指纹定位方法,其特征在于,接入点选择虽然也是基于簇为单位,对每个簇都要进行AP选择,但不同的是:第一,针对不同的地理位置计算不同的最佳AP集合;第二,最佳AP集合选择时采用辨识度作为衡量准则;第三,子簇虽然是由母簇分裂而来的,但是子簇与母簇采用的最佳AP集合却可能是不同的,这是因为在计算最佳AP集合时,只基于该簇内的所有参考点进行计算,也就是说,母簇与子簇包含不同的参考点,经过AP选择方法,得出的最佳AP集合也很可能是不同的。 The indoor positioning method based on fingerprints access point cluster splitting and adaptive reselection according to claim 1, wherein, while the access point selection is based on units of clusters, each cluster must be selected AP , but the difference is: first, different optimal AP calculated for a set of different geographic locations; second, using the best AP recognizable as a measure of the time set selection criteria; third, although the sub-clusters and cluster split from a master to, but sub-optimal set of clusters and the parent AP cluster uses it may be different, this is because when calculating the optimal AP set, all based on only the reference point within the cluster is calculated, that is, the cluster master sub-cluster contains different reference points, after the AP selection method, the best set of AP derived likely be different.
3. 根据权利要求1所述的基于接入点重选择和自适应簇分裂的室内指纹定位方法,其特征在于,接入点AP重选择机制为不同地理位置的簇寻找到空间区分能力最强的AP,其好处是去除了对定位贡献较小甚至起副作用的接入点AP。 The indoor positioning method based on fingerprints access point cluster splitting and adaptive reselection according to claim 1, wherein the access point AP reselection mechanism for a cluster of different geographic locations to find the strongest spatial separation the AP, its benefits are removed to locate even smaller contribution from the access point AP side effects.
4. 根据权利要求1所述的基于接入点重选择和自适应簇分裂的室内指纹定位方法,其特征在于,首先,对不同的簇,选取在该簇中对空间区分能力最强的AP作为最佳AP集合,在V-RSS指纹库只需要存储最佳AP集合的指纹即可,其它AP的指纹被抛弃,并且不同簇所存指纹对应的AP集合是不完全相同的,因此需要在指纹库中记录AP的编号。 The indoor positioning method based on fingerprints access point cluster splitting and adaptive reselection according to claim 1, characterized in that, first, the different clusters, the cluster in selecting the strongest spatial ability to distinguish AP AP as the optimal set, V-RSS fingerprint to the fingerprint database stores only a set of the best AP, AP other fingerprint is discarded, and different clusters corresponding to the stored fingerprint collection AP is not identical, it is necessary in the fingerprint AP numbering database records.
5. 根据权利要求1所述的基于接入点重选择和自适应簇分裂的室内指纹定位方法,其特征在于,采用自适应簇分裂方法,将整个待定位区域分割为大小不一的簇,簇的大小取决于所有AP在该簇的空间区分能力,如果某个子区域存在强空间区分能力的AP集合,则该子区域能分割为很多较小的簇,反之则该区域只能分割为少数几个较大的簇甚至只能归为1 个簇,以减少了指纹的存储数量。 The access point-based reselection according to claim 1 and a cluster division adaptive fingerprint indoor positioning method, wherein the method of adaptive cluster splitting, dividing the entire area to be positioned cluster sizes, cluster size depends on the ability to distinguish all of the space in the AP cluster, if a sub-set of strong regional presence AP ability to distinguish between space, then the sub-region can be divided into many smaller clusters, and vice versa in the region can only be divided into a small number of a few even larger clusters can only be classified as a cluster, in order to reduce the number of stored fingerprints.
6. 根据权利要求1所述的基于接入点重选择和自适应簇分裂的室内指纹定位方法,其特征在于,辨识度阈值是决定当前簇是否能继续分裂的关键,该阈值既能够由人工根据实际经验设置,也能够由自学习机制自行获得,其方法是将辨识度阈值定为预期定位误差、无线信号波动特征和簇尺寸的函数,通过自学习的方式确定该函数的表达式或者数值对应关系。 The indoor positioning method based on fingerprints access point cluster splitting and adaptive reselection according to claim 1, wherein the identification key threshold is determined whether the current cluster continue to divide, and the threshold value can be either manually the practical experience provided, or by itself obtained by the self-learning mechanism, which is the identification threshold as a function of expected position error, radio signal fluctuation characteristics and the cluster size is determined expression or the value of the function by means of self-learning correspondence.
CN201310733950.0A 2013-12-26 2013-12-26 Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods CN103763769B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310733950.0A CN103763769B (en) 2013-12-26 2013-12-26 Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310733950.0A CN103763769B (en) 2013-12-26 2013-12-26 Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods

Publications (2)

Publication Number Publication Date
CN103763769A CN103763769A (en) 2014-04-30
CN103763769B true CN103763769B (en) 2017-07-04

Family

ID=50530910

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310733950.0A CN103763769B (en) 2013-12-26 2013-12-26 Selection and adaptive re-division of the cluster-based fingerprint access point indoor positioning methods

Country Status (1)

Country Link
CN (1) CN103763769B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103997783B (en) * 2014-05-14 2017-03-29 无锡北邮感知技术产业研究院有限公司 Targeting one kind outdoor clustering method and apparatus
KR101846831B1 (en) * 2014-06-30 2018-04-09 후아웨이 테크놀러지 컴퍼니 리미티드 Method and apparatus for indoor positioning
CN105120476B (en) * 2015-07-20 2018-10-16 北京邮电大学 Based on cochannel interference compensation and the modified indoor orientation method of RSS fingerprints
CN105657736A (en) * 2016-03-22 2016-06-08 清华大学 Method and device for starting access points of wireless local area network
CN105916113B (en) * 2016-06-01 2019-04-12 北京邮电大学 A kind of localization method and device
CN108107407A (en) * 2017-12-18 2018-06-01 黑龙江大学 Three-dimensional sound source positioning method based on spatial distributed microphones

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101657014A (en) * 2009-09-11 2010-02-24 哈尔滨工业大学 WLAN indoor positioning method based on nearest neighbor probability method
CN102170697A (en) * 2011-04-06 2011-08-31 北京邮电大学 Indoor positioning method and device

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1820120B1 (en) * 2004-10-29 2012-07-25 Skyhook Wireless, Inc. Location beacon database and server, method of building location beacon database, and location based service using same
US8660577B2 (en) * 2009-12-04 2014-02-25 Nokia Corporation Method and apparatus for on-device positioning using compressed fingerprint archives

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101657014A (en) * 2009-09-11 2010-02-24 哈尔滨工业大学 WLAN indoor positioning method based on nearest neighbor probability method
CN102170697A (en) * 2011-04-06 2011-08-31 北京邮电大学 Indoor positioning method and device

Also Published As

Publication number Publication date
CN103763769A (en) 2014-04-30

Similar Documents

Publication Publication Date Title
Alzantot et al. Crowdinside: automatic construction of indoor floorplans
JP2015531053A (en) System, method, and computer program for dynamically creating a radio map
Yim Introducing a decision tree-based indoor positioning technique
Li et al. Experiencing and handling the diversity in data density and environmental locality in an indoor positioning service
CN102932911B (en) Positioning method and positioning system of location fingerprints
CN103402258B (en) One kind of Wi-Fi-based indoor positioning system and method
CN103476118B (en) wlan indoor location fingerprinting positioning method for real-time monitoring
Dawes et al. A comparison of deterministic and probabilistic methods for indoor localization
CN102170697B (en) Indoor positioning method and device
CN101778399A (en) Method for optimizing WLAN (Wireless Local Area Network) indoor ANN (Artificial Neural Network) positioning based on FCM (fuzzy C-mean) and least-squares curve surface fitting methods
CN103068035B (en) A wireless network positioning method, apparatus and system for
CN102573049B (en) An in-home positioning method and system
CN101639527A (en) K nearest fuzzy clustering WLAN indoor locating method based on REE-P
CN102821465B (en) WLAN (Wireless Local Area Network) indoor positioning method based on subregion information entropy gain
Jiang et al. Indoor mobile localization based on Wi-Fi fingerprint's important access point
CN105026888B (en) Moving means for positioning
CN102638889B (en) Indoor wireless terminal positioning method based on Bayesian Compressive Sensing
Panagiotakis et al. Interactive image segmentation based on synthetic graph coordinates
CN102802260A (en) WLAN indoor positioning method based on matrix correlation
CN103747524A (en) Android terminal indoor positioning method based on cloud platform
CN102932738A (en) Improved positioning method of indoor fingerprint based on clustering neural network
CN103120000B (en) Coverage areas of the model and used to produce
CN101839720A (en) Safety-based road map navigation
CN102186239A (en) Radio frequency (RF) fingerprint positing method, device and network equipment
CN102711239B (en) RSS (received signal strength) fingerprint database based secondary fuzzy clustering indoor-positioning method

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
GR01