CN109348416B - Fingerprint indoor positioning method based on binary k-means - Google Patents
Fingerprint indoor positioning method based on binary k-means Download PDFInfo
- Publication number
- CN109348416B CN109348416B CN201811034691.1A CN201811034691A CN109348416B CN 109348416 B CN109348416 B CN 109348416B CN 201811034691 A CN201811034691 A CN 201811034691A CN 109348416 B CN109348416 B CN 109348416B
- Authority
- CN
- China
- Prior art keywords
- clustering
- cluster
- decision tree
- clusters
- stage
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
- G01S5/0252—Radio frequency fingerprinting
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Collating Specific Patterns (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The invention discloses an indoor positioning technology based on binary k-means clustering, which solves the problem of poor positioning performance caused by improper selection of an initial clustering center. The technical scheme is as follows: 1) in the off-line stage, fingerprint data of all reference points are collected, and the fingerprint data from the access point with stronger resolution capability is reserved; clustering and dividing the reference points by adopting a binary k-means algorithm; constructing a decision tree for each cluster; 2) in the online stage, fingerprint data is collected at an observation point; matching clusters to which the observation points belong according to the fingerprint data; and performing fine positioning by using a decision tree corresponding to the clustering to obtain the physical position of the observation point. Based on the fingerprint indoor positioning technology, the influence of the selection of the initial clustering center on the positioning effect is considered, and a binary k-means clustering algorithm is selected for clustering division, so that the positioning accuracy is improved. The method is used for indoor environment with wifi coverage or configurable access points, and is suitable for indoor rescue and personnel positioning and other scenes in large-scale places.
Description
Technical Field
The invention belongs to the technical field of communication, and particularly relates to a fingerprint indoor positioning method, in particular to a fingerprint indoor positioning method based on a binary k-means. The positioning system can be used for positioning various indoor environments covered with wifi, and provides accurate positioning and good user experience for users under the condition that additional equipment is not needed.
Background
With the development of communication technology, location-based services have become an emerging mobile internet industry and have good development prospects. Therefore, the need to quickly and accurately obtain location information of a mobile terminal is becoming increasingly urgent. The location information may also be used to support location-based services and to improve network management, improve the quality of location services and network performance. Therefore, a positioning technology and a related positioning system capable of rapidly, accurately and stably acquiring location information in a wireless network have become a current research hotspot. At present, the precision of outdoor positioning technology such as satellite positioning can reach centimeter level, but no economic and mature scheme is available in indoor environment, mainly because of the characteristics of complex indoor positioning environment, multiple interference sources, direct wave path loss, multipath propagation, variable environment and the like. Because of the shelter of the building, the loss of GPS signals penetrating the building is too large, and the positioning capability is greatly reduced, so that the GPS is not suitable for positioning in an indoor environment. In addition to the requirement for location-based technologies, some artificial constraints, such as security, privacy protection, etc., also pose new challenges for indoor location systems. The positioning precision and accuracy of the currently-used indoor positioning technology are not high enough, so that a new method needs to be found or the existing positioning method needs to be optimized to meet the positioning requirements of people.
The wireless indoor positioning technology can be divided into four types according to the signal measurement technology: time-of-arrival based measurements, angle-of-arrival based measurements, time-difference-of-arrival based measurements, and signal strength based measurements.
Positioning based on time of arrival measurements: the distance between the transmitting end and the receiving end can be calculated by measuring the transmission time t of the wireless signals at the transmitting end and the receiving end, and the positions can be determined by obtaining a plurality of distances.
Positioning based on angle of arrival measurements: and calculating an intersection point by using a geometric method according to the plurality of azimuth lines, wherein the intersection point is the position of the target to be measured.
Positioning based on time difference of arrival measurements: and simultaneously transmitting signals to a plurality of base stations with different distances from the position to be positioned, calculating the time difference of receiving the signals by the different base stations, and calculating to obtain the position to be positioned.
Signal strength based localization: the method is divided into a position fingerprint positioning method and a signal transmission loss method. The fingerprint positioning method uses the received signal strength indication RSSI data from all access points AP to describe physical positions, the received signal strength indication RSSIs of all the positions are collected into a fingerprint database, and in an online matching stage, the position matching is carried out according to a matching algorithm through the received signal strength indication RSSI value measured at a certain point, so that the specific positions of the test points are deduced. The signal transmission loss method comprises the steps of firstly establishing a received signal strength indicator RSSI-distance model, converting the RSSI into corresponding distance values according to the model, obtaining the distance from a position to be measured to each access point AP, and then obtaining the position of the position to be measured by using a trilateration method.
The existing fingerprint positioning method still has some problems to be solved: in order to have higher positioning accuracy, a large number of Access Points (AP) are often required to be laid in a positioning environment, which causes waste; the positioning result is closely related to the selection of the initial clustering center, and poor positioning result can be caused by improper selection of the initial clustering center; the positioning is unstable due to the change of the positioning environment and the movement of the person. These problems can make the positioning undesirable.
Disclosure of Invention
The invention aims to provide a fingerprint indoor positioning method based on a binary k-means, which can effectively improve poor positioning effect caused by unreasonable selection of an initial clustering center, aiming at the defects of the prior art.
The invention relates to a fingerprint indoor positioning method based on a binary k-means, which is characterized in that positioning is divided into an off-line stage and an on-line stage, wherein the off-line stage is divided into a data acquisition and processing sub-stage, a reference point clustering dividing sub-stage and a decision tree construction sub-stage; the on-line positioning stage is divided into a matching clustering sub-stage and a decision tree positioning observation point sub-stage. The method comprises the following steps:
the off-line stage comprises: a data acquisition and processing sub-stage, a reference point clustering and dividing sub-stage and a decision tree construction sub-stage. The specific process is as follows:
1) a data acquisition and processing sub-stage: acquiring data of all the reference points, processing the acquired data and generating a simplified fingerprint database;
1a) in the off-line stage, firstly, acquiring all reference point information of a positioning area; the positioning area is evenly divided into n small squares, i.e. n reference points. One fingerprint of the reference point is represented by a vector consisting of Received Signal Strength Indicator (RSSI) data collected by the reference point at the same time and coming from all Access Points (AP) in the environment, and all fingerprints of all the reference points form an original fingerprint database.
1b) According to the received signal strength indication RSSI data in the original fingerprint database; and selecting m access points which enable the information gain of the positioning area to be maximum, and deleting the fingerprint data from other access points in the original fingerprint database to generate a simplified fingerprint database.
2) Reference point clustering and dividing molecule stage: dividing n reference points into k clusters by adopting a binary k-means algorithm;
and clustering all the reference points by using a binary k-means clustering algorithm. In the binary k-means algorithm, k is the number of clusters to be generated. When the data in the simplified fingerprint database is subjected to cluster division, all reference points are regarded as belonging to the same cluster, and the cluster is divided into two sub-clusters through a k-means clustering algorithm. And calculating to obtain the cluster center of each sub-cluster and the square error and SSE of each sub-cluster, selecting the cluster with the largest square error and the largest square error from all generated clusters, and dividing the sub-clusters through k-means clustering until the number of the clusters reaches k, thereby obtaining k clusters. The principle of each partition is that the sum of the squared error of all clusters and the SSE is reduced and the sum of the squared error of the newly generated clusters is minimized. And updating the simplified fingerprint database into a clustered fingerprint database consisting of k clusters and centers thereof, and completing the division of the k clusters in the positioning area.
3) A decision tree construction sub-stage: constructing corresponding k decision trees for the k reference point clusters;
and after k clusters are generated, constructing a decision tree for each cluster by adopting a C4.5 algorithm. The training data set of the decision tree is fingerprint data of all reference points in the corresponding cluster, the attribute set of the decision tree is m access points, and the labels of leaf nodes of the decision tree are the reference points in the cluster. And taking the training data set and the attribute set as the input of the C4.5 algorithm to generate a decision tree. And after k decision trees are generated, adding the internal nodes, the test attributes, the segmentation points and the leaf nodes of each decision tree into a clustering fingerprint library to generate a clustering-decision tree fingerprint library.
In the online stage, a matching clustering sub-stage and a decision tree observation point positioning sub-stage are provided, and the specific process is as follows:
4) matching clustering sub-stage: realizing the matching of the observation points and the belonged clusters;
after entering an online stage, firstly, data are collected at an observation point, namely the observation point receives signal intensity indication data from m access points, then Euclidean distances between the data and each clustering center in k clusters in a clustering-decision tree fingerprint library are calculated, and the cluster with the minimum Euclidean distance is selected as the cluster to which the observation point belongs. Matching of the clusters to which the observation points belong is completed;
5) decision tree positioning observation point sub-stage: data matching between the observation point and the decision tree is realized, and the physical position of the observation point is obtained;
and matching the received signal strength indication data received by the observation point with the decision tree corresponding to the cluster. And testing the attributes corresponding to the internal nodes from the root node, distributing the observation points to the child nodes according to the test result, and recursively testing and distributing the observation points until the leaf nodes are reached, wherein the reference points corresponding to the leaf nodes finally distributed by the observation points are the positions obtained according to the positioning algorithm. The matching of the physical positions corresponding to the observation points is completed;
and after the clustering matching and the decision tree matching are finished, the physical position of the observation point is obtained, and the position positioning of the observation point is finished.
Compared with the prior art, the invention has the following advantages:
the positioning precision is improved: compared with the common clustering algorithm, the clustering number is increased from 1 to k by adopting a binary k-means algorithm, the sum of square errors of the whole is reduced by clustering division each time, and the cluster with the best effect is selected each time, so that the clustering algorithm is the global optimal clustering algorithm. Compared with a k-means clustering k-means algorithm, the method effectively solves the problem of poor clustering effect caused by improper initial clustering center selection, and greatly improves the indoor positioning precision.
The method has the following advantages: in the indoor positioning method, the received signal strength indication data contains noise due to the movement of people in the positioning area and the change of the environment. In the invention, a large amount of data is collected in an off-line sampling stage, and an access point capable of obviously improving the positioning precision is selected, so that the fingerprint data stored in the simplified fingerprint database are all data with small influence of noise, and the influence of the noise data on positioning is weakened, therefore, the method has robustness.
More economical and practical: the access points used in the invention are all the existing access points in the environment, and a new access point does not need to be installed in the environment, so that the method is more economical and practical.
Drawings
FIG. 1 is a model diagram of the fingerprint indoor positioning method based on binary k-means in the present invention;
FIG. 2 is a flow chart of the offline phase of the present invention;
FIG. 3 is a decision tree structure generated in the present invention;
FIG. 4 is a comparison chart of the positioning accuracy when different clustering numbers are taken under different positioning accuracies;
FIG. 5 is a comparison of the positioning accuracy of the present invention compared to positioning using the k-means method at a positioning accuracy of 2.0 m.
Detailed Description
The invention is described in detail below with reference to the figures and examples
Example 1
In the existing indoor positioning method, in order to have a good positioning effect, a large number of sensors specially used for positioning are often required to be laid in a positioning area, which causes waste; when the positioning environment is not changed, the positioning effect is better. However, when a person walks or the obstacle is increased or decreased in the positioning area, the positioning accuracy cannot be guaranteed, and the practicability is poor. In some positioning methods, a clustering algorithm, such as a k-means algorithm, is adopted, so that a positioning result strongly depends on initial clustering centers of k clusters, and when the initial clustering centers are not properly selected, the positioning effect is poor.
The invention provides an improvement aiming at the problems of the existing positioning method and provides a fingerprint indoor positioning method based on a binary k-means. Referring to fig. 1, the positioning of the present invention is divided into an off-line stage and an on-line stage according to the sequence, wherein the off-line stage is divided into a data acquisition and processing sub-stage, a reference point clustering division sub-stage and a decision tree construction sub-stage; the on-line positioning stage is divided into a matching clustering sub-stage and a decision tree positioning observation point sub-stage. The indoor positioning process comprises the following steps:
the off-line stage comprises: a data acquisition and processing sub-stage, a position clustering and molecular dividing stage and a decision tree construction sub-stage.
1) A data acquisition and processing sub-stage: acquiring data of all the reference points, processing the acquired data and generating a simplified fingerprint database;
1a) in the off-line stage, firstly, acquiring all reference point information of a positioning area; the positioning area is evenly divided into n small squares, i.e. n reference points. Each reference point data is represented by a vector formed by Received Signal Strength Indication (RSSI) data of Access Points (AP) which can be detected by the reference point, the RSSI vectors of the received signal strength indications of all the Access Points (AP) form fingerprints of the reference point, and the fingerprints of all the reference points form an original fingerprint library.
1b) According to received signal strength indication RSSI data in an original fingerprint library, m access points AP with the largest influence on the positioning effect are selected by combining an information theory, and a simplified fingerprint library is generated. According to the information theory, the access point with the largest information gain of the positioning area is the access point AP with the largest influence on the positioning effect. And calculating the information gains of all the access points AP, and selecting m access points AP with the maximum information gain. In the invention, the positioning effect is measured by the positioning accuracy under a certain positioning precision. m may be chosen as the minimum number of access points that will result in the best positioning. And deleting the fingerprint data from other access points in the original fingerprint database, and generating a simplified fingerprint database from the original fingerprint database. In other words, all the access points in the positioning environment are screened to obtain m access points having a large influence on the positioning effect.
2) Position clustering and dividing molecule stage: dividing n reference points into k clusters by adopting a binary k-means algorithm;
the binary k-means algorithm is a clustering method in a specific form, and compared with other clustering algorithms (such as a k-means algorithm), the effect is greatly improved. In the binary k-means algorithm, k is the number of clusters to be generated, and is typically selected from a range of square values less than the number n of reference points using an exhaustive method. Compared with a k-means clustering algorithm, the binary k-means algorithm increases the number of clusters from 1 to k step by step, the reference points which belong to the same cluster and are generated by each division have the largest similarity, and the reference points which belong to different clusters have the smallest similarity. Therefore, the result of each clustering division is globally optimal, and the finally generated k clusters are also globally optimal.
When clustering division is carried out on the reference points, all the reference points are regarded as a cluster, the cluster is a father cluster, and the cluster is divided into two sub-clusters through a k-means clustering algorithm. And averaging the fingerprint data of all the reference points in each sub-cluster to obtain the cluster center of each sub-cluster, and deleting the parent cluster. And calculating the square error sum SSE of each cluster, selecting the cluster with the larger square error sum as a parent cluster, performing cluster division on the parent cluster by using a k-means algorithm, and deleting the parent cluster. And selecting the cluster with the largest sum of square errors from all clusters as a parent cluster each time, dividing the parent cluster into sub-clusters and deleting the parent cluster until the number of the clusters reaches k, so that k clusters are obtained, and constructing a clustering fingerprint database by the k clusters and the cluster centers. The principle of each division is that the sum of the square error sum SSE of the sub-cluster reaches the minimum and is smaller than the square error sum of the parent cluster. And updating the simplified fingerprint database into a clustering fingerprint database consisting of k clusters and centers thereof, and completing the division of the k clusters in the positioning area.
3) A decision tree construction sub-stage: constructing corresponding k decision trees for the k reference point clusters;
the clustering algorithm divides all reference points into k clusters, if the clusters constructed in 2) are used for positioning, only the possible range of the observation points can be obtained, and the positioning precision is very low. In order to reduce the positioning range of the observation point and obtain the specific physical position of the observation point, only one reference point can be arranged in each cluster, so that the positioning precision is low, the positioning time is greatly increased, and the real-time performance is reduced. Therefore, the invention further reduces the positioning range by adopting a decision tree algorithm and carries out accurate positioning.
In the decision tree, each internal node represents a test for an attribute, each branch represents the result output of the test for the attribute, and each leaf node represents a classification result, i.e., a reference point location. When a decision tree is constructed, the principle of selecting attributes is as follows: the method comprises the steps of firstly selecting an access point which enables the information gain rate to be maximum as a test attribute of a root node, constructing a preselected segmentation point for data of the attribute, selecting the preselected segmentation point with the maximum information gain as the segmentation point of the attribute, and generating left and right child nodes according to the segmentation point. And then the child nodes are tested and divided to generate new left and right child nodes. When a classification result can be determined by using all test rules on a branch from a root node to a certain node, the child node is a leaf node, and the leaf node is not tested and divided. And when the decision tree does not have nodes which can be continuously tested and segmented, finishing the construction of the decision tree. And after the k decision trees are constructed, adding the internal nodes, the test attributes, the segmentation points and the leaf nodes of each decision tree into a clustering fingerprint library to generate a clustering-decision tree fingerprint library.
In the online phase: and matching the clustering sub-stage and the decision tree positioning observation point sub-stage.
4) Matching clustering sub-stage: realizing the matching of the observation points and the belonged clusters;
after entering an online stage, firstly, data of an observation point is collected, the observation point is a point to be located in the invention, namely, the observation point receives signal intensity indication data from m access points, Euclidean distances between the data and k clustering centers in a clustering-decision tree fingerprint library are calculated, and a cluster with the minimum Euclidean distance is selected as a cluster to which the observation point belongs. The possible range of the observation points is narrowed from all the reference point sets to the reference point subset in the cluster, and the matching of the cluster to which the observation points belong is completed.
5) Decision tree positioning observation point sub-stage: data matching between the observation point and the decision tree is realized, and the physical position of the observation point is obtained;
and matching the received signal strength indication data received by the observation point with the decision tree corresponding to the matched cluster. And testing attributes corresponding to the root node and the internal node from the root node, distributing the observation point to the left child node or the right child node according to the test result, testing and distributing the observation point recursively in such a way until the leaf node is reached, wherein the label of the leaf node finally distributed by the observation point is the position obtained according to the positioning algorithm. And completing data matching of the observation points and the decision tree, and obtaining the physical positions of the observation points.
In the on-line positioning stage, after the data collected by the observation points are subjected to cluster matching and decision tree matching, the physical positions of the observation points are obtained, and the position positioning of the observation points is completed.
The technical idea of the invention is as follows: because the data from the same reference point have the maximum similarity, the Euclidean distance is adopted to calculate the similarity between the data, and the similarity between the data is utilized to judge whether different data come from the same reference point. In the invention, clustering is adopted to divide the reference points into k clusters, and the smaller the sum of squared errors of the clusters is, the greater the similarity of the reference point data in the clusters is. After data are collected in a positioning area in an off-line stage, m access point APs with the largest information gain are selected by processing received signal strength indication data from access point APs in the environment collected at all reference points in the off-line stage and combining an information theory. And then, clustering and dividing all the reference points, and dividing the reference points with high similarity into the same cluster. And then constructing a decision tree for each cluster, wherein the labels of leaf nodes of the decision tree are classified, and the position of a reference point is a category, so as to generate a cluster-fingerprint database. In the on-line positioning stage, fingerprint data collected at an observation point and from m access points AP are matched with the fingerprints of each clustering center, Euclidean distances between the fingerprints and each clustering center are calculated, the clustering with the minimum Euclidean distance is used as the clustering to which the observation point belongs, and finally, the observation point is matched to a reference point with the maximum similarity by a decision tree algorithm through a series of tests and classification to obtain the physical position of the observation point.
The method and the device utilize the existing access points in the positioning environment to position, if the access points do not exist in the environment originally, the access points need to be laid, and the newly laid access points can be used for positioning, communication or other purposes, so that waste is avoided; in order to reduce the influence of the positioning environment change on the positioning precision, the access points in the environment are screened, and only the access points capable of improving the positioning precision are used, so that the influence of the acquired fingerprint data on the environmental noise is small; the invention adopts a binary k-means algorithm, and compared with a k-means clustering algorithm, a clustering result is not influenced by an initial clustering center. The number of clusters is increased from 1 to k, so that the cluster division result obtained every time is globally optimal, and the positioning effect is obviously improved.
Example 2
The fingerprint indoor positioning method based on the dichotomy k-means is the same as that in the embodiment 1, m APs with the largest influence on the positioning result are selected in the step 1b), and a simplified fingerprint library is generated, and the method specifically comprises the following steps;
1b1) the n reference points in the positioning area are denoted as (G)1,G2,…Gn) And m access points with the largest information gain in the positioning area are selected by using the information theory. When access point information is not used, i.e. received signal strength indication data is not collected, the uncertainty of all reference points is:
in the above formula, GjDenotes the jth reference point, P (G)j) Is the proportion of the jth reference area in the positioning environment in the positioning area.
1b2) When measured from the ith access point APiAfter receiving the signal strength indication data, the ith access point AP is adoptediWhen positioning is performed, the uncertainty of all reference points becomes:
in the above formula, APiI.e. from the i-th access point APiHas a received signal strength indication data value of v, P (G)j,APiV) is at the jth reference point GjCollected from the ith access point APiThe probability that the received signal strength indication data of (b) is v, P (G)j/APiV) is the AP from the i-th access point known to be acquired at some reference pointiWhen the received signal strength indication data is v, the reference point is the jth reference point GjThe probability of (c).
1b3) Is measuring from the ith access point APiAfter receiving the signal strength indication data, the access point AP of the ith is utilizediThe uncertainty reduction amount of all reference points is:
InfoGain(G,APi)＝H(G)-H(G/APi)；
1b4) the m access points with larger information gain InfoGain values are screened out by the above formula, and the received signal strength indication data from the m access points are less influenced by environmental changes.
And reserving the received signal strength indication data collected by the selected m access points in the original fingerprint database, and deleting the fingerprint data irrelevant to the m access points to form the simplified fingerprint database.
Example 3
The fingerprint indoor positioning method based on the dichotomous k-means is the same as the embodiment 1-2, and the cluster center calculation formula in the step 2) is as follows:
all fingerprint data used by the above calculation is from a reduced fingerprint library. Wherein N isiAs a cluster CiNumber of fingerprint data in, NciAs a cluster CiNumber of reference points in, ciAs a cluster CiCenter of (A), RjlAs a cluster CiAt the jth reference pointFirst fingerprint data of a set, njAs a cluster CiThe number of fingerprint data of the jth reference point in (1). In the invention, the ith cluster C can be calculated by using the formulaiC cluster center ofi，ciCan be used to calculate to obtain cluster CiThe sum of the squares of the errors of (1).
The sum of squared errors is calculated as:
the ith cluster C can be calculated by using the formulaiSum of squared errors SSE ofi，SSEiCan be used for judging the clustering division result and selecting the cluster to be divided from the constructed clusters. The two calculation formulas of this example will be used in the reference point cluster partitioning stage of the present invention to construct k clusters.
Example 4
The fingerprint indoor positioning method based on the dichotomous k-means is the same as that in the embodiment 1-3, the dichotomous k-means algorithm is adopted in the step 2), n reference point data are divided into k clusters according to data in the simplified fingerprint database, and the process of generating the clustered fingerprint database is as follows:
2.1) regarding all reference points as a cluster, wherein the cluster is a father cluster, randomly initializing a cluster center of the father cluster, generating a cluster fingerprint database by utilizing the cluster and the cluster center, and executing 2.3);
2.2) calculating the square error and SSE of each generated cluster, selecting the cluster with the maximum square error and SSE as a parent cluster, and storing the rest clusters in a cluster fingerprint library to keep unchanged;
and 2.3) carrying out binary classification on the parent cluster through a k-means algorithm to generate two sub-clusters. The classification principle is as follows: a clustering method is selected such that the sum of the square error sums SSE of the two child clusters is minimized and smaller than the sum of the square errors of the parent cluster. After clustering division is finished, adding 1 to the number of clusters, and deleting a father cluster and a clustering center thereof from a clustering fingerprint database;
2.4) calculating the centers of the two sub-clusters, and adding the sub-clusters and the corresponding cluster centers to a cluster fingerprint database;
2.5) judging the clustering number, if the clustering number reaches k, finishing the clustering construction, storing a clustering fingerprint library, deleting the simplified fingerprint library, and ending the position clustering and dividing stage; otherwise, the number of clusters is less than k, and the execution is returned to 2.2), and the iteration is repeated until the number of clusters reaches k.
Through the steps, k clusters can be constructed, and a cluster fingerprint database is generated.
Example 5
The fingerprint indoor positioning method based on the dichotomous k-means is the same as the embodiment 1-4, and the calculation formula of the information gain rate in the step 3) is as follows:
in the above formula, DjI.e. the set of reference points in the data set for node j in the decision tree, InfoGain (D)j,APi) As an access point APiInformation gain at node j, H (D)j) Gain _ ratio (D), the uncertainty of node jj,APi) Is an access point APiThe information gain rate at node j. The method can be used for constructing a decision tree stage by calculating the information gain rate of each access point at the node j through the formula, selecting the access point with the maximum information gain rate as the attribute of the node, and generating the decision tree through a series of processing.
Example 6
The fingerprint indoor positioning method based on the dichotomous k-means is the same as that in the embodiment 1-5, the process of clustering and constructing the corresponding k decision trees for the k reference points in the step 3) and generating the clustering-decision tree fingerprint database is as follows:
3.1) generating a clustering-decision tree fingerprint database:
constructing an empty fingerprint database which is called a clustering-decision tree fingerprint database;
3.2) selecting the attribute of the root node of the decision tree:
selecting one cluster from the clusters of the non-generated decision tree, wherein the data set is the fingerprint data of the cluster in the cluster fingerprint database, and keeping the existing cluster-decision tree fingerprint database unchanged. Calculating the information gain rate of each access point in the data set, and selecting the access point with the largest information gain rate as the attribute of the root node;
3.3) selecting the partition points of each node in the decision tree:
and sorting the data in the data set of the node from small to large according to the selected attribute values, wherein if N different values are shared, N-1 preselected segmentation points exist. The value of each pre-selected segmentation point is the average value of front and back continuous different elements in the sorted values. Calculating the information gain of each pre-selection partition point, selecting the pre-selection partition point with the largest information gain as the partition point of the node, deriving left and right nodes, wherein the data set of the left node is the data set of which the attribute value in the data set of the father node is smaller than that of the partition point, the data set of the right node is the data set of which the attribute value in the data set of the father node is larger than that of the partition point, and the left and right nodes are set as leaf nodes;
3.4) processing leaf nodes:
selecting a leaf node, if the data in the data set of the node comes from the same reference point, setting the reference point as the label of the node, and executing 3.5); otherwise, this node becomes an internal node and is no longer a leaf node. Calculating the information gain rate of each access point by using the data set of the node, selecting the access point which enables the information gain rate of the node to be maximum as the attribute of the access point, and executing 3.3);
3.5) judging whether a decision tree is generated:
if all leaf nodes in the decision tree have labels, the clustering decision tree structure is finished, and the internal nodes of the decision tree, the test attributes and the segmentation points thereof, the leaf node labels and the corresponding clustering fingerprints of the clustering fingerprint library are combined and added to the clustering-decision tree fingerprint library, and the step 3.5) is executed. Otherwise, selecting a leaf node without a label, updating the node as an internal node, selecting the access point with the maximum information gain rate of the node as the attribute of the node, and executing 3.3);
3.6) judging whether k decision trees are generated:
if k decision trees are generated, the decision tree construction stage is ended, the clustering-decision tree fingerprint library construction is completed, and the clustering fingerprint library is deleted; otherwise, the number of the decision trees is less than k, and 3.2) is returned.
A more detailed example is given below to further illustrate the invention:
example 7
The fingerprint indoor positioning method based on the dichotomous k-means is the same as the embodiment 1-6, and the attached figure 1 is a model diagram of the invention, and the specific process is as follows:
the off-line stage is divided into a data acquisition and processing sub-stage, a position clustering and molecular dividing stage and a decision tree construction sub-stage, each sub-process is shown as the attached figure 2, and the specific process is as follows:
1. a data acquisition and processing sub-stage:
selecting an area of 64m in an indoor environment with wifi coverage2The area of (1) is a positioning area, the positioning area is divided into 100 small squares, each square is a reference point, and 100 reference points are marked as G1,G2,…G100. For the jth reference point G thereinjThe probability of its occurrence is P (G)j) Value of
The total number of detectable access points in the positioning area is 17, and the received signal strength indication data from the 17 access points is collected at each reference point. If at a certain reference point GjCan not detect the ith access point APiThe corresponding data is recorded as the minimum value of all the received signal strength indication data. These data are processed according to information theory to select m access points that minimize the uncertainty of the location area information the most. m may be chosen from all integer values smaller than 17 to be the one that gives the best positioning effect.
The uncertainty of all locations when no access point is used, i.e. no received signal strength indication data from an access point is collected, is first calculated:
after collecting the received signal strength indication data in the positioning environment, if from the i-th APiIf the received signal strength indication data is v, it is recorded as APiAt reference point G ═ vjCollected from APiHas a probability of P (G) of v being the received signal strength indication data ofj,APiV) known to be acquired from the AP at some reference pointiIs v, and the reference point is GjHas a probability of P (G)j/APiV), then the uncertainty of all positions at this time becomes:
therefore, using the received signal strength indication data of the access point, the uncertainty of the location is reduced by:
InfoGain(APi)＝H(G)-H(G/APi)
therefore, m access points with larger information gain InfoGain values are screened out, and the access points can obviously reduce the uncertainty of the positioning area. The received signal strength indication data from the m access points collected from the 100 reference points is saved as a reduced fingerprint library. At reference point GjThe vector formed by the signal strength indicating data from the m access points collected at the same time is GjA fingerprint of, reference point GjThere are several fingerprints, and all the fingerprints of all the reference points constitute the simplified fingerprint library.
2. Position clustering and dividing molecule stage:
the optimal number of clusters is related to the data set, and the optimal number of clusters for different data sets is not necessarily the same. In order to select the optimal clustering number of the simplified fingerprint database, the example sequentially selects 6 k values: 3, 4, 5, 6, 7 and 8 are used as the number of the target clusters to compare the positioning effect. And selecting a k value each time, and dividing the fingerprints in the simplified fingerprint database into k clusters by using a binary k-means clustering algorithm. The dividing process is as follows: and when the number of the clusters is 1, selecting the cluster as a parent cluster, directly using a k-means algorithm to perform cluster division on the parent cluster, generating a child cluster with the minimum sum of square errors, and deleting the parent cluster. And selecting the cluster with the largest sum of squared errors from all clusters as a parent cluster, performing cluster division by using a k-means algorithm to generate a child cluster with the smallest sum of the two sum of squared errors and smaller than the sum of the squared errors of the original cluster, and deleting the parent cluster. And repeating the partition clustering until the number of clusters reaches k.
When the clustering division is carried out, the data in the simplified fingerprint library is used for calculating the sum of the center error and the square error of each cluster. The specific calculation formula is as follows:
ith cluster CiAfter generation, the cluster center c of the cluster is calculatediThe formula of (1) is:
in the above formula, NiAs a cluster CiNumber of fingerprint data in, NciAs a cluster CiNumber of reference points in, ciAs a cluster CiCenter of (1), njAs a cluster CiNumber of fingerprint data of jth reference point in, RjlAs a cluster CiThe ith fingerprint data acquired at the jth reference point. Obtaining the clustering center c according to the formulai。
Cluster CiThe equation for the sum of squared errors of (1) is:
according to the above formula, the cluster C can be obtainediSum of squared errors SSE ofi。
And after the k clustering partitions are finished, generating a clustering fingerprint database by using the k clusters and the clustering centers, and finishing the position clustering partition stage.
3. A decision tree construction sub-stage:
and constructing a decision tree for the k clusters, wherein the adopted algorithm is C4.5 algorithm. C4.5 Algorithm machineThe goal of the algorithm for generating decision trees in a learner learning algorithm is to find a mapping from attribute values to classes for a given data set of multiple entities, and this mapping can be used to classify new unknown classes of entities. In this example, for one cluster C of k clustersjWhen constructing the decision tree, the access point is attribute, and the data set is CjFrom the ith access point APiThe fingerprint data of (2) is the attribute value of the attribute, and the reference point position is the category. When a C4.5 algorithm is adopted to construct a decision tree, an access point with the largest information gain rate is selected for each internal node as the attribute of the node, then a partition point is selected for the attribute, and left and right subtrees are constructed until all leaf nodes have category labels. In the constructed decision tree, each leaf node represents a category, and the label of the category is the position of the reference point. The generated decision tree is shown in fig. 3, wherein circles represent internal nodes, access points in the circles are attributes of the nodes, and squares represent leaf nodes. For convenience of representation, the segmentation limit is labeled as the absolute value of the actual segmentation limit in fig. 3. As a cluster CjThe constructed decision tree is shown in FIG. 3, from the root node to the label G8The branch construction process of the leaf node is as follows: the data set of the root node is CjThe information gain rate of each access point at the root node is calculated first according to all the fingerprint data in the database, and the AP2Has the largest information gain rate, so the attribute of the root node is AP2Is AP2Selects a division point, thereby generating two branches for attribute values ranging from 71-82 and 83-88, generates two child nodes, and divides the data set of the root node into left and right child nodes according to the branch range. Setting the left child node as a father node, and selecting the attribute which enables the data set information gain rate to be maximum for the father node, wherein the attribute is AP5For which a split point is selected, resulting in two branches and child nodes. The data set of the left child node is the attribute AP in the data set of the father node5Has an attribute value of 64-70, since there is only a reference point G in the data set8So that the node is a leaf node and the label is G8. The rest child nodes of the root node are subjected to branch construction untilAnd all the generated branches can correctly classify the data in the data set, and the decision tree construction is completed.
And combining the decision tree internal nodes, the test attributes and the segmentation points thereof, the leaf node labels and the corresponding clustering fingerprints of the clustering fingerprint library to generate a clustering-decision tree fingerprint library.
The position of a position to be determined in the positioning area is set as an observation point in the online stage, and the position of the observation point is one of the positions of 100 reference points because the position of the offline stage is the same as the positioning area of the online stage. The online stage aims to find the physical position of the observation point according to the fingerprint data of the observation point, namely, the reference point with the maximum similarity to the online fingerprint data is selected. The current stage is divided into a clustering matching sub-stage and a decision tree matching sub-stage, and the main process is as follows:
4. clustering matching sub-stage:
and calculating Euclidean distances between the fingerprint data of the observation point and each cluster center, wherein the cluster with the minimum Euclidean distance is the cluster with the maximum similarity to the fingerprint data of the observation point, and selecting the cluster as the cluster to which the observation point belongs.
5. Decision tree matching sub-stage:
and performing a series of tests and distribution on the fingerprint data of the observation points through the decision tree of the cluster to which the observation points belong, and finally distributing the fingerprint data to leaf nodes, wherein the labels of the leaf nodes are the specific positions of the observation points obtained according to the invention. The location of the observation point in the online phase is completed.
The technical advantages of the present invention can be further illustrated by the following examples:
example 8
The fingerprint indoor positioning method based on the dichotomous k-means is the same as the embodiment 1-7, and the attached figure 4 is a comparison graph of the positioning accuracy when the invention takes different target clustering numbers under different positioning accuracy:
Example 9
The fingerprint indoor positioning method based on binary k-means is the same as the embodiment 1-7, and the attached figure 5 is a comparison graph of the positioning accuracy of the fingerprint indoor positioning method based on binary k-means positioning method and the fingerprint indoor positioning method based on binary k-means positioning method, wherein the comparison graph is shown in the figure, under the positioning accuracy of 2.0 m:
and 2, simulating, namely comparing the fingerprint positioning method based on the binary k-means with the fingerprint positioning method based on the k-means clustering, wherein the result is shown in the figure 5. The positioning accuracy of the two methods is compared when the positioning accuracy is 2.0 m. The solid line represents the positioning accuracy when the fingerprint positioning method based on the binary k-means is adopted, and the dotted line represents the positioning accuracy when the fingerprint positioning method based on the k-means clustering is adopted. As can be seen from FIG. 5, when the number of clusters is the same, the positioning accuracy of the method is higher than that of the fingerprint positioning method based on k-means clustering. Compared with a fingerprint positioning method adopting a k-means algorithm, the method has the advantage that the positioning effect is improved.
Claims (2)
1. A fingerprint indoor positioning method based on a binary k-means is characterized in that positioning is divided into an off-line stage and an on-line stage, wherein the off-line stage is divided into a data acquisition and processing sub-stage, a reference point clustering dividing sub-stage and a decision tree construction sub-stage; the on-line positioning stage is divided into a matching clustering sub-stage and a decision tree positioning observation point sub-stage; the method comprises the following steps:
the off-line stage comprises: a data acquisition and processing sub-stage, a reference point clustering division sub-stage and a decision tree construction sub-stage; the specific process is as follows:
1) a data acquisition and processing sub-stage: acquiring data of all the reference points, processing the acquired data and generating a fingerprint database;
1a) in the off-line stage, firstly, acquiring all reference point information of a positioning area; the positioning area is evenly divided into n small squares, namely n reference points, one fingerprint of each reference point is represented by a vector formed by received signal strength indication data which are collected by the reference point at the same moment and come from all access points in the environment, and all fingerprints of all the reference points form an original fingerprint database;
1b) indicating data according to the received signal strength in the original fingerprint database; selecting m access points which enable the information gain of the positioning area to be maximum, deleting fingerprint data from other access points in the original fingerprint database, and generating a simplified fingerprint database;
2) reference point clustering and dividing molecule stage: dividing n reference points into k clusters by adopting a binary k-means algorithm;
clustering and dividing all the reference points by using a binary k-means clustering algorithm; in the binary k-means algorithm, k is the number of clusters to be generated; when the data in the simplified fingerprint database is subjected to cluster division, all reference points are regarded as belonging to the same cluster, and are divided into two sub-clusters through a k-means clustering algorithm; calculating to obtain the cluster center of each sub-cluster and the square error sum SSE of each sub-cluster; selecting a mean error and a maximum cluster from all generated clusters, dividing sub-clusters through k-means clustering until the number of clusters reaches k, and obtaining k clusters; the principle of each division is that the sum of square errors of all clusters is reduced, and the sum of square errors of newly generated clusters reaches the minimum; updating the simplified fingerprint database into a clustering fingerprint database consisting of k clusters and centers thereof, and completing the division of the k clusters in the positioning area; the binary k-means algorithm divides n reference point data in the simplified fingerprint database into k clusters, and the process of generating the clustered fingerprint database is as follows:
2.1) regarding all reference points as a cluster, wherein the cluster is a father cluster, randomly initializing a cluster center, generating a cluster fingerprint database by utilizing the cluster and the cluster center, and executing 2.3);
2.2) calculating the sum of square errors of all generated clusters, selecting the cluster with the largest sum of square errors as a parent cluster, and storing the rest clusters in a cluster fingerprint database to keep unchanged;
2.3) performing binary classification on the parent cluster through a k-means algorithm to generate two sub-clusters; the classification principle is as follows: selecting a clustering method which enables the sum of the square errors of the two sub-clusters to be smaller than the sum of the square errors of the parent cluster and enables the sum of the square errors of the two sub-clusters to be minimum; after clustering division is finished, adding 1 to the number of clusters, and deleting a father cluster and a clustering center thereof from an original clustering fingerprint library;
2.4) calculating the centers of the two sub-clusters, and adding the sub-clusters and the corresponding cluster centers to a cluster fingerprint database;
2.5) judging the clustering number, if the clustering number reaches k, finishing the clustering construction, storing a clustering fingerprint library, and ending the position clustering and molecular dividing stage; otherwise, the clustering number is less than k, and the execution is returned to 2.2), and iteration is repeated until the clustering number reaches k;
3) a decision tree construction sub-stage: constructing corresponding k decision trees for the k reference point clusters;
after k clusters are generated, a decision tree is constructed for each cluster by adopting a C4.5 algorithm; the training data set of the decision tree is fingerprint data of all reference points in the corresponding cluster, the attribute set of the decision tree is m access points, and the labels of leaf nodes of the decision tree are the reference points in the cluster; taking the training data set and the attribute set as the input of a C4.5 algorithm to generate a decision tree; after k decision trees are generated, adding internal nodes, test attributes, segmentation points and leaf nodes of each decision tree into a clustering fingerprint library to generate a clustering-decision tree fingerprint library;
in the online stage, a matching clustering sub-stage and a decision tree observation point positioning sub-stage are provided, and the specific process is as follows:
4) matching clustering sub-stage: realizing the matching of the observation points and the belonged clusters;
after entering an online stage, firstly collecting fingerprint data at an observation point, namely receiving signal intensity indication data from m access points received by the observation point, then calculating Euclidean distances between the data and each clustering center in k clusters in a clustering-decision tree fingerprint library, and selecting the cluster with the minimum Euclidean distance as the cluster to which the observation point belongs; matching of the clusters to which the observation points belong is completed;
5) decision tree positioning observation point sub-stage: data matching between the observation point and the decision tree is realized, and the physical position of the observation point is obtained;
matching the fingerprint data received by the observation point with a decision tree corresponding to the cluster; testing attributes corresponding to the internal nodes from the root nodes, distributing the observation points to the child nodes according to the test results, and testing and distributing the observation points recursively in such a way until the leaf nodes are reached, wherein the reference points corresponding to the leaf nodes finally distributed by the observation points are positions obtained according to the positioning algorithm; the matching of the physical positions corresponding to the observation points is completed;
and after the clustering matching and the decision tree matching are finished, the physical position of the observation point is obtained, and the position positioning of the observation point is finished.
2. The method for indoor fingerprint positioning based on binary k-means as claimed in claim 1, wherein the process of constructing the corresponding k decision trees for the k reference point clusters in step 3) and generating the cluster-decision tree fingerprint database is as follows:
3.1) selecting the attribute of the root node of the decision tree:
selecting one cluster without generating a decision tree from the k reference point clusters, wherein the data set is the fingerprint data of the reference point in the corresponding cluster, if a cluster-decision tree fingerprint library is not generated, generating an empty cluster-decision tree fingerprint library, otherwise, keeping the existing cluster-decision tree fingerprint library unchanged; calculating the information gain rate of each access point in the data set, and selecting the access point with the largest information gain rate as the attribute of the root node;
3.2) selecting the partition points of each node in the decision tree:
sorting the data in the data set of the node from small to large according to the selected attribute values, wherein if N different values are shared, N-1 preselected segmentation points exist; the value of each pre-selected segmentation point is the average value of front and back continuous different elements in the sorted values; calculating the information gain of each pre-selection partition point, selecting the pre-selection partition point with the largest information gain as the partition point of the node, deriving left and right nodes, wherein the data set of the left node is the data set of which the attribute value in the data set of the father node is smaller than that of the partition point, the data set of the right node is the data set of which the attribute value in the data set of the father node is larger than that of the partition point, and the left and right nodes are set as leaf nodes;
3.3) selecting a leaf node, if the data in the data set of the node comes from the same reference point, setting the reference point as the label of the node, and executing 3.4); otherwise, selecting the access point with the largest information gain rate as the attribute of the node, and executing 3.2);
3.4) if all leaf nodes in the decision tree have labels, the construction of the clustering decision tree is finished, the internal nodes of the decision tree, the test attributes and the partition points thereof, the leaf nodes and the corresponding clustering fingerprints of the clustering fingerprint library are combined and added into the clustering-decision tree fingerprint library, the number of the decision trees is increased by one, and 3.5) is executed, otherwise, one leaf node without labels is selected, the access point which enables the information gain rate of the node to be maximum is selected as the attribute of the node, and 3.2) is executed;
3.5) if the number of the decision trees is equal to k, finishing the construction stage of the decision trees and finishing the construction of the clustering-decision tree fingerprint database; otherwise, the number of the decision trees is less than k, and 3.1) is returned.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811034691.1A CN109348416B (en) | 2018-09-06 | 2018-09-06 | Fingerprint indoor positioning method based on binary k-means |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811034691.1A CN109348416B (en) | 2018-09-06 | 2018-09-06 | Fingerprint indoor positioning method based on binary k-means |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109348416A CN109348416A (en) | 2019-02-15 |
| CN109348416B true CN109348416B (en) | 2020-08-25 |
Family
ID=65292174
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811034691.1A Active CN109348416B (en) | 2018-09-06 | 2018-09-06 | Fingerprint indoor positioning method based on binary k-means |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109348416B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110213710A (en) * | 2019-04-19 | 2019-09-06 | 西安电子科技大学 | A kind of high-performance indoor orientation method, indoor locating system based on random forest |
| CN110728309B (en) * | 2019-09-27 | 2023-05-02 | 中国铁道科学研究院集团有限公司通信信号研究所 | Track clustering method based on railway signals and Beidou positioning |
| CN110719566A (en) * | 2019-10-18 | 2020-01-21 | 安徽农业大学 | WiFi clustering algorithm-based agricultural machinery signal-free positioning method |
| CN111417067B (en) * | 2020-03-13 | 2021-05-07 | 智慧足迹数据科技有限公司 | Method and device for positioning visited position of user |
| EP4264309A4 (en) | 2020-12-15 | 2024-05-01 | Nokia Technologies Oy | Enhanced fingerprint positioning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103476115A (en) * | 2013-09-22 | 2013-12-25 | 中国地质大学（武汉） | Method for WiFi fingerprint positioning based on AP set similarity |
| CN103678500A (en) * | 2013-11-18 | 2014-03-26 | 南京邮电大学 | Data mining improved type K mean value clustering method based on linear discriminant analysis |
| CN103745200A (en) * | 2014-01-02 | 2014-04-23 | 哈尔滨工程大学 | Facial image identification method based on word bag model |
| CN106658570A (en) * | 2016-12-23 | 2017-05-10 | 西安电子科技大学 | Mobile sink information collection path constructing method based on secondary grid partitioning |
| CN107727095A (en) * | 2017-09-08 | 2018-02-23 | 西安电子科技大学 | 3D indoor orientation methods based on spectral clustering and weighting reverse transmittance nerve network |
| CN107948296A (en) * | 2017-11-30 | 2018-04-20 | 中电科卫星导航运营服务有限公司 | A kind of indoor locating system and method based on bluetooth |
-
2018
- 2018-09-06 CN CN201811034691.1A patent/CN109348416B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103476115A (en) * | 2013-09-22 | 2013-12-25 | 中国地质大学（武汉） | Method for WiFi fingerprint positioning based on AP set similarity |
| CN103678500A (en) * | 2013-11-18 | 2014-03-26 | 南京邮电大学 | Data mining improved type K mean value clustering method based on linear discriminant analysis |
| CN103745200A (en) * | 2014-01-02 | 2014-04-23 | 哈尔滨工程大学 | Facial image identification method based on word bag model |
| CN106658570A (en) * | 2016-12-23 | 2017-05-10 | 西安电子科技大学 | Mobile sink information collection path constructing method based on secondary grid partitioning |
| CN107727095A (en) * | 2017-09-08 | 2018-02-23 | 西安电子科技大学 | 3D indoor orientation methods based on spectral clustering and weighting reverse transmittance nerve network |
| CN107948296A (en) * | 2017-11-30 | 2018-04-20 | 中电科卫星导航运营服务有限公司 | A kind of indoor locating system and method based on bluetooth |
Non-Patent Citations (1)
| Title |
|---|
| Power-Efficient Access-Point Selection for Indoor Location Estimation;Yiqiang Chen, Qiang Yang, Jie Yin, Xiaoyong Chai;《IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING》;20060731;第4节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109348416A (en) | 2019-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109348416B (en) | Fingerprint indoor positioning method based on binary k-means | |
| CN110012428B (en) | Indoor positioning method based on WiFi | |
| CN106646338B (en) | A kind of quickly accurate indoor orientation method | |
| CN106851571B (en) | Decision tree-based rapid KNN indoor WiFi positioning method | |
| KR102116824B1 (en) | Positioning system based on deep learnin and construction method thereof | |
| CN105813194B (en) | Indoor orientation method based on fingerprint database secondary correction | |
| CN108919177B (en) | Positioning map construction method based on virtual information source estimation and track correction | |
| Zhou et al. | Indoor fingerprint localization based on fuzzy c-means clustering | |
| CN110049549B (en) | WiFi fingerprint-based multi-fusion indoor positioning method and system | |
| CN112135248B (en) | WIFI fingerprint positioning method based on K-means optimal estimation | |
| CN112748397A (en) | UWB positioning method based on self-adaptive BP neural network under non-line-of-sight condition | |
| CN109121083B (en) | Indoor positioning method based on fingerprint similarity of AP (Access Point) sequence | |
| CN110213710A (en) | A kind of high-performance indoor orientation method, indoor locating system based on random forest | |
| CN113596989A (en) | Indoor positioning method and system for intelligent workshop | |
| CN109143161A (en) | High-precision indoor orientation method based on mixed-fingerprint Environmental Evaluation Model | |
| Anuwatkun et al. | A diff-based indoor positioning system using fingerprinting technique and k-means clustering algorithm | |
| CN112533144B (en) | Indoor positioning method, device, computing equipment and computer storage medium | |
| Torres-Sospedra et al. | Scalable and efficient clustering for fingerprint-based positioning | |
| CN107526058A (en) | Indoor orientation method, apparatus and system | |
| CN108680897B (en) | Indoor positioning method and device, electronic equipment and storage medium | |
| CN108521631B (en) | Mobile AP (access point) identification method for indoor positioning | |
| Chanama et al. | A comparison of decision tree based techniques for indoor positioning system | |
| CN110636437A (en) | Fingerprint pair-based auxiliary positioning method | |
| Soro et al. | Performance comparison of indoor fingerprinting techniques based on artificial neural network | |
| CN109121081A (en) | A kind of indoor orientation method based on position Candidate Set Yu EM algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |