RFID label positioning method based on generalized multidimensional scale
阅读说明:本技术 一种基于泛化多维标度的rfid标签定位方法 (RFID label positioning method based on generalized multidimensional scale ) 是由 马永涛 田成龙 于 2019-11-22 设计创作,主要内容包括:本发明涉及一种基于泛化多维标度的大规模RFID标签定位方法包括数据预处理阶段、地图碎片生成阶段和地图碎片装配阶段。其中,数据预处理阶段包括以下步骤:根据获得的距离信息构建距离矩阵;根据距离矩阵构建指示矩阵。地图碎片生成阶段包括以下步骤:Bron-Kerbosch算法处理指示矩阵I,得到一系列标签均可通信的子网络,称其为极大团;将阅读的编号集合并入极大团中,得增广极大团;利用增广极大团对距离矩阵切片,得由增广极大团确定的标签簇的距离矩阵;使用处理所得距离矩阵,得相应的地图碎片。(The invention relates to a large-scale RFID label positioning method based on generalized multidimensional scale, which comprises a data preprocessing stage, a map fragment generating stage and a map fragment assembling stage. Wherein, the data preprocessing stage comprises the following steps: constructing a distance matrix according to the obtained distance information; and constructing an indication matrix according to the distance matrix. The map fragment generation phase comprises the following steps: processing the indication matrix I by a Bron-Kerbosch algorithm to obtain a series of subnetworks with labels capable of communicating, and calling the subnetworks as a maximum clique; merging the read number set into the maximum clique to obtain an augmented maximum clique; slicing the distance matrix by using the augmented maximum cluster to obtain a distance matrix of the label cluster determined by the augmented maximum cluster; and obtaining the corresponding map fragments by using the distance matrix obtained by processing.)
1. A large-scale RFID label positioning method based on generalized multidimensional scale comprises a data preprocessing stage, a map fragment generating stage and a map fragment assembling stage. Wherein the content of the first and second substances,
the data preprocessing stage comprises the following steps:
1) constructing a distance matrix D according to the obtained distance information:
in the formula:the distance between the labels is the distance between the labels,
2) Constructing an indication matrix I according to the distance matrix:
in the formula: p, q ═ 1,2, …, M; [*]pqRepresenting the element at the position of the p row and the q column of the matrix;
the map fragment generation phase comprises the following steps:
1) processing the indication matrix I by a Bron-Kerbosch algorithm to obtain a series of subnetworks with labels capable of communicating, and calling the subnetworks as a maximum clique;
2) merging the read number set into the maximum clique to obtain an augmented maximum clique;
3) slicing the distance matrix by using the augmented maximum cluster to obtain a distance matrix of the label cluster determined by the augmented maximum cluster;
4) and processing the obtained distance matrix by using a multidimensional scaling algorithm to obtain a corresponding map fragment.
2. The method of claim 1, wherein the map fragment assembling stage comprises the following steps:
1) selecting a suitable map fragment
2) if the candidate set C is empty, turning to step 5); otherwise, selecting the map fragment with the most common nodes with the intermediate map
3) assemblyAnd
4) Updating candidate set C ═ C \ ikK is k + 1; turning to step 2);
5) absolute intermediate map: determining a rigid transformation by Procrustes analysis, so that the distribution of the reader nodes in the transformed intermediate map is matched with the distribution of a real reader as much as possible; this transformation is then applied to the entire intermediate map, resulting in the estimated position coordinates of the tag.
Technical Field
The invention belongs to the technical field of RFID positioning, and aims to solve the problem of positioning tags with low complexity, high precision and high concurrency by obtaining the distance between tags by utilizing a backscatter-based inter-tag communication network.
Background
With the explosive development of integrated circuits and Internet of Things (IoT), more and more researchers have been invested in the field of radio-frequency identification tag (RFID) based indoor positioning, especially Ultra High Frequency (UHF) based RFID, in the last decades. Sensing location information of an object to which an RFID tag is attached is particularly important. In many internet of things systems, the existence of position information directly determines whether the system can exert all functions and provide satisfactory services for users. The development of the future 5G network puts higher requirements on the accuracy, concurrency and real-time performance of indoor positioning.
Existing indoor positioning technologies can be classified into ranging-based and non-ranging-based positioning methods according to positioning principles. Typical ranging techniques include Received Signal Strength (RSS), time of arrival (ToA), time difference of arrival (TDoA), angle of arrival (AoA), phase of arrival (PoA), phase difference of arrival (PDoA), and organic combinations thereof. Positioning methods based on ranging often only can achieve single-target positioning, however, many RFID applications require multi-target concurrent positioning technology.
Non-ranging-based localization technologies mainly include fingerprinting (fingerprint), radio frequency holographic imaging (radio frequency tomography) and radio frequency tomography (RTI). A representative work of fingerprinting is LANDMARC, a location-aware prototype system that uses reference tags to locate objects in a room. A.Buffi et al put forward a Synthetic Aperture Radar (SAR) -based positioning method in an unmanned plane scene equipped with UHFRFID reader. In addition, l.yang et al propose an RFID-based location system, Tagoram. The system uses a differential enhanced hologram (DAH) to realize real-time tracking and high-precision positioning of the mobile RFID tag. The other RFID positioning method of the brand-new outcrop corner is RTI, can realize the accurate positioning of multiple targets under the background that a fingerprint library is not established, and has the application precondition that the number of labels to be positioned must be known in advance. Y.ma et al propose a novel RTI positioning technique, which can accurately position multiple targets in advance under the background of unknown target number, and make up for the technical gap in RTI. These non-ranging methods either can achieve single-target continuous positioning, small-scale high-concurrency positioning, or large-scale low-concurrency positioning, and cannot meet the large-scale high-concurrency positioning requirements. A novel positioning technology is urgently needed in the industry, and the multi-label concurrent positioning under a large-scale deployment scene is supported.
Disclosure of Invention
The invention relates to a RFID label positioning method based on generalized multidimensional scaling, which utilizes a backscatter-based communication network between labels to obtain distance estimation between the labels and executes a multidimensional scaling algorithm from a distributed angle. Aiming at the situation that the labels cannot be communicated in a large-scale label deployment scene, the invention firstly excavates the sub-networks which can be communicated among the labels in the network; then, using MDS (multidimensional scaling) algorithm for each sub-network to obtain a pool of map fragments; finally, the fragmented maps are assembled to form a complete map. The positioning effect of low complexity, high precision and high concurrency in a large-scale label deployment scene is achieved. The technical scheme of the invention is as follows:
a large-scale RFID label positioning method based on generalized multidimensional scale comprises a data preprocessing stage, a map fragment generating stage and a map fragment assembling stage. Wherein the content of the first and second substances,
the data preprocessing stage comprises the following steps:
1) constructing a distance matrix D according to the obtained distance information:
in the formula:
i, j is 1,2, …, M, is the distance between tags,i 1,2, …, M, j 1,2, … N, tag-reader distance,i, j is 1,2, … N, which is the inter-reader distance; m is the number of tags, and N is the number of readers; if communication between tags is not possible, provision is made for2) Constructing an indication matrix I according to the distance matrix:
in the formula: p, q ═ 1,2, …, M; [*]pqRepresenting the element at the position of the p row and the q column of the matrix;
the map fragment generation phase comprises the following steps:
1) processing the indication matrix I by a Bron-Kerbosch algorithm to obtain a series of subnetworks with labels capable of communicating, and calling the subnetworks as a maximum clique;
2) merging the read number set into the maximum clique to obtain an augmented maximum clique;
3) slicing the distance matrix by using the augmented maximum cluster to obtain a distance matrix of the label cluster determined by the augmented maximum cluster;
4) and processing the obtained distance matrix by using a multidimensional scaling algorithm to obtain a corresponding map fragment. In the map fragment assembling stage, the method comprises the following steps:
1) selecting a suitable map fragment
i0Numbering the map fragments, initializing an intermediate mapLet k be 1 and the candidate set C be {1,2, …, n } \ i0N is the number of map tiles, "\" is the difference operator of the set;2) if the candidate set C is empty, turning to step 5); otherwise, selecting the map fragment with the most common nodes with the intermediate mapikNumbering the map fragments;
3) assembly
Andfirst, a rigid transformation is determined by Procrustes analysis, so that the distribution of common nodes within the transformed map fragments matches as closely as possible the intermediate mapDistribution of internal common nodes; then, the intermediate map and the transformed map fragments are stitched, the average value of the coordinates of the common nodes is taken, the coordinates of the respective unique nodes are kept at the same time, and the updated intermediate map is obtained4) Updating candidate set C ═ C \ ikK is k + 1; turning to step 2);
5) absolute intermediate map: determining a rigid transformation by Procrustes analysis, so that the distribution of the reader nodes in the transformed intermediate map is matched with the distribution of a real reader as much as possible; this transformation is then applied to the entire intermediate map, resulting in the estimated position coordinates of the tag.
The invention relates to a large-scale RFID label positioning based on generalized multidimensional scale, which utilizes obtained distance estimation information to construct a distance matrix and an indication matrix of a network, introduces a Bron-Kerbosch algorithm in a graph theory to obtain a potential completely-communicable sub-network in the network, expands the sub-network to ensure the feasibility of assembly, utilizes an MDS algorithm to obtain a relative map of each expanded network, and obtains the position information of a large-scale label through hierarchical assembly. Compared with the traditional RFID label positioning method, the method is based on the cooperation, dimension reduction and statistical thought, executes the multi-dimensional scale algorithm from a distributed angle, greatly expands the application scene of the multi-dimensional scale algorithm, and realizes the positioning effect with low complexity, high precision and high concurrency. Meanwhile, the operation mechanism of the hierarchical assembly enables the assembly error of the front stage to rise in a controllable and stable mode, and the rear stage assembly has strong anti-interference capability. In the face of map fragments with severe quality fluctuation, the input mismatch rate can be greatly reduced by multi-stage assembly, and a middle map with stable quality is output.
Drawings
FIG. 1 is a generalized multi-dimensional scaling technique based label localization scenario.
Fig. 2 is an algorithm flow chart.
Detailed Description
The following describes a large-scale RFID tag positioning method based on generalized multi-dimensional scale according to the present invention with reference to the accompanying drawings.
A label localization scenario based on generalized multi-dimensional scaling techniques is shown in fig. 1. Several tags are scattered randomly in a large-scale (10m × 10m) positioning scene, and isolated tags in the scene are removed.
The positioning method estimates the position of the positioning device according to the estimated distance information, the algorithm flow is shown in figure 2, and the steps are as follows:
1) constructing a distance matrix according to the obtained distance information:
in the formula:i, j is 1,2, …, M, is the distance between tags,
i 1,2, …, M, j 1,2, … N, tag-reader distance,i, j is 1,2, … N, which is the inter-reader distance; m is the number of tags, and N is the number of readers. If communication between tags is not possible, provision is made for2) Constructing an indication matrix according to the distance matrix:
in the formula: p, q ═ 1,2, …, M; [*]pqRepresenting the element at the position of the p-th row and q-th column of the matrix.
In the map fragment generation phase, the method comprises the following steps:
3) the Bron-Kerbosch algorithm processes the indication matrix I to obtain a series of subnetworks where the tags can communicate. Set label number set as Lt1,2, …, M, the number set corresponding to the label in the subnet is
α is 1,2, …, Ω where Ω is the number of subnetsαIs a very large group of networks.4) Let the number set of the reader be Lr(M, M +1, …, M + N), LrIncorporated into the maximal pellet to obtain the augmented maximal pellet
β=1,2,…,Ω。5) And slicing the distance matrix by using the augmented maximum cluster to obtain the distance matrix of the label cluster determined by the augmented maximum cluster.
6) And processing the obtained distance matrix by using a multidimensionalscaling algorithm to obtain a corresponding map fragment.
In the map fragment assembling stage, the method comprises the following steps:
7) selecting a suitable map fragment
i0Numbering the map fragments, initializing an intermediate mapLet k be 1 and the candidate set C be {1,2, …, n } \ i0And n is the number of map tiles.8) If candidate set C is empty, go to step 11). Otherwise, selecting the map fragment with the most common nodes with the intermediate map
ikThe map tiles are numbered.9) Assembly
Andfirst, a rigid transformation is determined by Procrustes analysis, so that the distribution of common nodes within the transformed map fragments matches as closely as possible the intermediate mapDistribution of internal common nodes; then, the intermediate map and the transformed map fragments are stitched, the average value of the coordinates of the common nodes is taken, the coordinates of the respective unique nodes are kept at the same time, and the updated intermediate map is obtained10) Updating candidate set C ═ C \ ikAnd k is k + 1. And 8) turning.
11) The intermediate map is absolute. Determining a rigid transformation by Procrustes analysis, so that the distribution of the reader nodes in the transformed intermediate map is matched with the distribution of a real reader as much as possible; this transformation is then applied to the entire intermediate map, resulting in the estimated position coordinates of the tag.