CN110996280B - RFID indoor positioning fingerprint database updating system and method - Google Patents
RFID indoor positioning fingerprint database updating system and method Download PDFInfo
- Publication number
- CN110996280B CN110996280B CN201911075626.8A CN201911075626A CN110996280B CN 110996280 B CN110996280 B CN 110996280B CN 201911075626 A CN201911075626 A CN 201911075626A CN 110996280 B CN110996280 B CN 110996280B
- Authority
- CN
- China
- Prior art keywords
- anchor
- label
- signal intensity
- target
- value
- 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
- H04W4/30—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
-
- 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
- G01S5/14—Determining absolute distances from a plurality of spaced points of known location
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/23—Updating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
Abstract
The utility model provides a RFID indoor location fingerprint storehouse update system and method, through using each anchor label and each target label's actual coordinate, Euclidean distance, X axle and Y axle mapping coordinate difference data as input data, target label signal strength carries out GBDT model training for the target value, has established anchor label and target label's real-time change relation, solve among the prior art when carrying out fingerprint storehouse renewal inefficiency, time human cost is high, the fingerprint renewal is inaccurate scheduling problem, realize RFID indoor location fingerprint storehouse's accurate quick renewal, under the accurate online renewal's of fingerprint storehouse circumstances of guaranteeing, very big reduction signal sampling working strength.
Description
Technical Field
The disclosure relates to the technical field of indoor positioning, in particular to a system and a method for updating an RFID indoor positioning fingerprint database.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the development of the internet of things, the indoor positioning system based on the internet of things has wide market prospect and wide product application. Radio Frequency Identification (RFID) technology is an advanced technology for information identification by radio frequency signals. Because of the advantages of low price, small size and the like of the RFID tag, indoor positioning by using the radio frequency identification technology becomes a mainstream indoor positioning scheme.
In the RFID indoor positioning system, a common method is a distance measurement method based on an angular distance, and a fingerprint positioning method based on RSSI (received signal strength). The fingerprint positioning method mainly comprises the steps of collecting fingerprints in an off-line stage and constructing a positioning model, acquiring a signal intensity value of an RFID label in real time in the on-line stage, inputting the signal intensity value into the model constructed in the off-line stage, and outputting the position of the label. One of the key difficulties affecting the accuracy of RFID fingerprint location methods is the accurate acquisition of the RFID tag signal strength values in the fingerprint library. In the RFID fingerprint positioning method, how to efficiently establish and maintain the fingerprint database is a difficult point of RFID positioning, because the indoor environment and the change of weather in time, the received intensity value at the same place is not a stable and unchangeable value, the change of the signal intensity value has a great influence on the result of fingerprint positioning, and it is a very costly task to collect the fingerprint data again for the RFID positioning system. Therefore, a simple and accurate method and system for updating the fingerprint database is needed.
The prior art represented by the invention patent 'a dynamic construction method and a system of an indoor positioning fingerprint library' (patent number: CN103747519A) introduces a dynamic construction method of the indoor positioning fingerprint library based on a professional device, and collects label signal strength values for many times on a preset path by arranging a robot or a mobile car fingerprint collecting device capable of automatically moving. However, this type of method based on professional devices only sets one signal intensity value acquisition device, and when the fingerprint changes, the fingerprint acquisition needs to be performed again, without reducing the acquisition time and requiring a certain amount of manpower.
The prior art represented by an article 'local updating fingerprint positioning algorithm based on region division' updates a fingerprint database in a mode of partitioning a positioning region, mainly divides a fingerprint map into a plurality of sub-regions by a clustering algorithm or region partitioning, selects a representative point from each sub-region to represent the fingerprint effectiveness of the sub-region, and updates the region fingerprint by detecting the effectiveness of the representative point. However, the method does not solve the problem of updating all fingerprint libraries, and the workload is not reduced when all fingerprints need to be updated if the navigation mark points are changed.
Disclosure of Invention
In order to solve the defects of the prior art, the present disclosure provides an update system and method for an RFID indoor positioning fingerprint database, which solves the problems of low efficiency, high time and labor cost, inaccurate fingerprint update and the like in the prior art when updating the fingerprint database, and realizes accurate and rapid update of the RFID indoor positioning fingerprint database.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
the first aspect of the disclosure provides an update system of an RFID indoor positioning fingerprint database.
An RFID indoor positioning fingerprint database updating system comprises a plurality of RFID readers, a plurality of target tags, a plurality of anchor tags, a front processor, an upper computer and a data processing terminal;
the RFID reader configured to: acquiring signal intensity values of the anchor label and the target label in real time and transmitting the signal intensity values to the pre-processor in real time;
the pre-processor configured to: transmitting the received signal strength values of the anchor tag and the target tag to an upper computer in real time, monitoring an RFID reader in real time, and alarming when the RFID reader stops sending signals is detected;
the upper computer is configured to: the received signal strength signals of the anchor labels and the target labels are subjected to noise reduction and filtering and are transmitted to a data processing terminal for storage, and the upper computer also displays real-time signal strength values, historical signal strength values and signal strength value trend curves of all the anchor labels in real time and provides signal strength query service;
the data processing terminal is configured to: the method comprises the steps that an initial fingerprint library is arranged and loaded with a fingerprint updating model based on a gradient lifting decision tree, and an anchor label signal intensity value received in real time, an Euclidean distance matrix of an anchor label and an original target label and a mapping coordinate value difference value are used as input data to be input into the trained fingerprint library updating model to obtain an updated target label signal intensity value; and replacing and storing corresponding values in the initial fingerprint library by the acquired anchor label signal intensity value and the calculated target label signal intensity value, so as to update the fingerprint library.
As possible implementation manners, the RFID reader, the pre-processor, the upper computer and the data processing terminal are communicated by RS232 serial ports.
And as possible implementation modes, the upper computer is also used for receiving the updated fingerprint library in real time, storing the fingerprint library in the upper computer and providing real-time display and real-time query services of the updated fingerprint library.
As some possible implementation manners, the data processing terminal is provided with an initial fingerprint library, and the initialization stage of the fingerprint library specifically includes: arranging anchor labels and target labels in a positioning area, measuring two-dimensional coordinates of each anchor label and each target label, simultaneously acquiring static signal intensity values of each anchor label and each target label, setting the static signal intensity values as an initial fingerprint library and storing the initial fingerprint library in a data processing terminal;
and further, in an online updating stage, the target label is removed, only the signal intensity value of the anchor label is collected, and the fingerprint database is updated.
The second aspect of the disclosure provides an update method of an RFID indoor positioning fingerprint database.
An updating method of an RFID indoor positioning fingerprint database comprises the following steps:
(1) arranging anchor tags and target tags in a positioning area, establishing a plane rectangular coordinate system, and collecting and storing coordinate values of each anchor tag and each target tag;
(2) arranging a plurality of RFID readers around the positioning area, and respectively carrying out static acquisition and storage on the signal strength values of the anchor tag and the target tag by utilizing the RFID readers at different positions;
(3) respectively collecting signal intensity values of an anchor label and a target label in the same time interval to form an offline anchor label signal intensity sequence and an offline target label signal intensity sequence;
(4) calculating Euclidean distance between each anchor label and each target label in the positioning area, and calculating a coordinate difference value of each anchor label and each target label on an X, Y axis in the positioning area;
(5) taking the signal intensity sequences of all the acquired anchor labels and the calculated Euclidean distances and coordinate difference values as input features, taking the fingerprint signal intensity sequence of each target label as an output target value, and performing regression training on a gradient lifting decision tree to obtain a trained fingerprint updating model based on the gradient lifting decision tree;
(6) acquiring a signal intensity value of an anchor label in a positioning area in real time, and inputting the signal intensity value of the anchor label, an Euclidean distance matrix between the anchor label and an original target label and a mapping coordinate value difference value into a trained fingerprint library updating model by taking the signal intensity value of the anchor label and the Euclidean distance matrix and the mapping coordinate value difference value as input data to obtain a prediction result, wherein the prediction result is the signal intensity value of the target label after updating;
(7) and replacing the corresponding value in the initial fingerprint library by the acquired anchor label signal intensity value and the calculated target label signal intensity value to update the fingerprint library.
As some possible implementations, in step (1), a planar rectangular coordinate system is established with an arbitrary fixed point of the rectangular space in the indoor positioning region as a coordinate origin and with two adjacent sides of the coordinate origin as an X axis and a Y axis.
As some possible implementation manners, in the step (3), after forming an offline anchor tag signal intensity sequence and an offline target tag signal intensity sequence, processing an abnormal value in the sequences by using a bit-splitting method to obtain an offline anchor tag signal intensity sequence set and an offline target tag signal intensity sequence set after the abnormal value is processed.
As a further limitation, processing the abnormal value in the sequence by using a quantile method, specifically: in each signal strength sequence, using Q3Representing the third quartile, Q, in the signal strength sequence1Denotes the first quartile, and furthermore IRQ ═ Q3-Q1And is recorded as Q3+1.5 IQR, with the lower limit Q3+1.5 IQR, the signal intensity values not within the upper and lower bounds of the signal intensity sequence are replaced by the average of the left and right adjacent signal intensity values, respectively, to obtain the signal intensity sequence after abnormal value processing.
As some possible implementation manners, in the step (5), performing regression training of the gradient boost decision tree, specifically:
(5-1) initializing a classification regression tree weak learning device, and inputting a training data set M { (x)1,y1),(x2,y2),......,(xN,yN) And f, wherein X is data of training characteristics (the signal intensity value of each anchor label on the reader and the actual coordinate, Euclidean distance, X-axis and Y-axis mapping coordinate difference values of each target label), Y is data of a training label (the signal intensity value of the target label on the reader), L represents a loss function, c represents a fitting residual error (the difference value of the target label and the model label), m represents iteration times, and f represents the number of iterations0(x) Representation initializationWeak learner regression tree:
(5-2) for iteration round T1, 2, … T, for sample i 1, 2.. N, a negative gradient is calculated:
wherein r istiRepresents the negative gradient value of the loss function at the ith sample in the t iteration,represents the loss function L (y)i,f(xi) )) pair f (x)i) Derivation, f (x) represents the weak learner regression tree of the t round, f (t-1) represents the regression tree of the t-1 iteration;
(5-3) Using (x)i,rti) Fitting a CART regression tree to obtain a t regression tree, wherein the corresponding leaf node region of the t regression tree is RtjJ is 1,2, J is the number of leaf nodes of the regression tree t, wherein i is 1,2, m;
(5-4) for leaf area J ═ 1,2 · · J, calculate the best fit:
wherein, ctjRepresenting the best output value of the fitted leaf node;
updating the strong learner:
f is thent(x) A strong learner representing a t-th round of fitting;
(5-5) finally obtaining an expression of the gradient boosting regression tree f (x) of the strong learner, namely the expression is used for updating the fingerprint algorithm training model:
wherein f is0(x) Indicating the initialization of the weak learner, T indicating the total T rounds of iteration, J indicating the number of leaf nodes of each round of regression tree T, ctjRepresenting the best output value to fit the leaf node.
As some possible implementation manners, in the step (6), the position of the anchor tag is kept unchanged, and after the target tag in the positioning area is removed, the signal intensity value of the anchor tag in the positioning area is acquired in real time.
Compared with the prior art, the beneficial effect of this disclosure is:
according to the system and the method for updating the signal intensity value of the target label by using the RFID anchor label, the automatic updating of the RFID fingerprint database is realized, and the accuracy of the fingerprint database is greatly improved.
According to the method, the actual coordinates, Euclidean distances, X-axis and Y-axis mapping coordinate difference data of the anchor labels and the target labels are used as input data for the first time, the signal intensity of the target labels is used as the target value to conduct GBDT model training, the real-time change relation of the anchor labels and the target labels is established, and the signal sampling working intensity is greatly reduced under the condition that the fingerprint database is accurately updated on line.
Drawings
Fig. 1 is a schematic diagram illustrating an arrangement of indoor tags in an RFID indoor positioning fingerprint library updating system according to embodiment 1 of the present disclosure.
Fig. 2 is a schematic structural diagram of an RFID indoor location fingerprint database updating system according to embodiment 1 of the present disclosure.
Fig. 3 is a flowchart of an update method of an RFID indoor location fingerprint library according to embodiment 1 of the present disclosure.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example 1:
as shown in fig. 1-2, an embodiment 1 of the present disclosure provides an update system of an RFID indoor positioning fingerprint library, where in fig. 1, a planar rectangular coordinate system is established with an arbitrary fixed point of an indoor positioning rectangle as an origin, and two adjacent sides of the origin as an X axis and a Y axis.
The RFID readers R1, R2, R3 and R4 can collect the signal intensity values of the anchor tag AT and the target tag DT in real time, and transmit the signal intensity values into the STM32 single chip microcomputer module through RS232 serial ports in JSON format files. In the fingerprint library initialization stage, signal intensity values of an anchor label AT and a target label DT are collected AT the same time, axis coordinate values of the anchor label AT and the target label DT are measured AT the same time and used as an initial fingerprint library, and only the anchor label is set and no target label is set in the fingerprint library real-time updating stage.
STM32 single chip microcomputer module has signal transmission and monitoring function in FIG. 2, reads the ware through RS232 serial ports and RFID and is connected, reads the ware through STM32 serial ports and communicates with the host computer, and at any stage, with the real-time signal transmission who gathers of RFID reading ware R1, R2, R3, R4. Meanwhile, the STM32 has a real-time monitoring and alarming function, and alarms when the RFID reader is detected to stop sending signals, so that the normal work of the system is facilitated.
The upper computer in the figure 2 has the functions of calculation, real-time display, query and transmission, and mainly completes the filtering and noise reduction function of the RFID signal intensity sequence to filter and reduce the noise of the RFID signal intensity signal transmitted by the STM32 singlechip module. Meanwhile, the upper computer can display the real-time signal intensity value of each anchor label in real time, the historical signal intensity value and the signal intensity value trend curve. Historical signal intensity can be conveniently inquired through the upper computer. Meanwhile, the upper computer is connected with the server through a local area network, the signal intensity value sequence is transmitted to the server, and a new fingerprint library predicted by the server can also be transmitted to a retrieval local storage for real-time display.
The server in fig. 2 mainly has functions of storage calculation, model training and model prediction. Training the anchor label and target label signal intensity sequence transmitted by the local area network and the upper computer and the local stored actual coordinate, Euclidean distance, X-axis and Y-axis mapping coordinate difference data of the anchor label and the target label, predicting to obtain a new target label signal intensity sequence value, and storing the updated anchor label and target label signal intensity value again to obtain an updated fingerprint library.
Example 2:
as shown in fig. 3, an embodiment 2 of the present disclosure provides an update method of an RFID indoor location fingerprint library, which includes the following steps:
step 1: the method comprises the following steps of establishing a plane rectangular coordinate system based on a rectangular space of a positioning area, arranging an anchor label and a target label in the positioning area, updating the anchor label setting in an offline fingerprint collection and online fingerprint library and keeping the coordinates in the positioning area unchanged, setting the target label only in an offline stage, and measuring and storing the plane rectangular coordinate values of the anchor label and the target label, wherein the specific method comprises the following steps:
a plane rectangular coordinate system is established on the basis of a rectangular space of the positioning area, an arbitrary fixed point is taken as an origin O, and two adjacent sides of the origin are taken as X, Y axes. Taking the RFID anchor mark signal strength value label set in the positioning area as a reference point set, and recording the reference point set as { AT }1,AT2,…ATi,…ATmWhere ATi represents the set of signal strength values for the ith anchor tag, and m represents the total number of anchor tags. Taking the target label information intensity value set in the positioning area as a target point set, and recording { DT1,DT2,…DTj,…DTmDTj, which represents the information strength value set of the jth target label. The coordinates of notes ATi and DTj on the X and Y axes are
Step 2: entering an off-line data acquisition stage, and setting n readers around the positioning area, so that the signal intensity value set of the anchor recording label is as follows:
the set of signal strength values for the tag of interest is:
preferably, 4 readers are arranged around the positioning area, and are used for initializing and collecting signal intensity values of the anchor tag and the target tag in 4 positioning positions and recording data.
And step 3: the method comprises the following steps of carrying out static acquisition on signal intensity values of an anchor label and a target label in a positioning area, and forming an offline anchor label signal intensity sequence and a target label signal intensity sequence by respectively acquiring the signal intensities of the anchor label and the target label within the same interval time, wherein the method specifically comprises the following steps:
and performing RSSI static acquisition on the anchor tag and the target tag in the positioning area, and recording the signal intensity sequence of the acquired anchor tag at the same interval time as:
the signal intensity sequence of the target tag collected at the same time interval is recorded as:
where k represents the kth anchor tag, j represents the jth reader, and i represents the ith target tag.
And 4, step 4: processing abnormal values in the signal intensity sequences of the anchor tag and the target tag, performing abnormal value processing by using a bit-splitting method, and performing abnormal value processing on the signal intensity sequences of the ith anchor tag and the kth target tag on the jth reader Abnormal values are processed by using a quantile method. In each signal strength sequence, using Q3Representing the third quartile, Q, in the signal strength sequence1Denotes the first quartile, and furthermore IRQ ═ Q3-Q1And is recorded as Q3+1.5 IQR, with the lower limit Q3+1.5 IQR, for signal intensity values in the signal intensity sequence that are not within the upper and lower boundsUsing the mean of the intensity values of the left and right adjacent signals, respectivelyInstead, a set of signal strength sequences after outlier processing is obtainedSequentially extracting collections And the signal intensity value of the corresponding time period and the coordinate values of the label on the X axis and the Y axis form an initial fingerprint database.
And 5: calculating Euclidean distance between each anchor label and each target label in the positioning area, and calculating a coordinate difference value of each anchor label and each target label on an X, Y axis in the positioning area, wherein the coordinate difference value is used as a part of input features of a fingerprint updating model, and the method specifically comprises the following steps:
calculating Euclidean distances between the m anchor tags and the k target tags:
the k-th target tag is expressed as the Euclidean distance in combination with the anchor tagSimultaneously calculating the coordinate difference of mapping of the anchor label and the target label on X, Y axesExpressing the k-th target label and the m anchor labels as a X, Y-axis coordinate difference set
Step 6: taking the signal intensities of all anchor labels in the initial fingerprint library extracted in the step 4, the Euclidean distance characteristics extracted in the step 5 and the X, Y axis difference value characteristics as input characteristics of a gradient lifting decision tree model, taking the fingerprint signal intensity of each target label extracted in the initial fingerprint library as an output target value of the gradient lifting decision tree model, and performing regression training on the gradient lifting decision tree to obtain a trained fingerprint updating model based on the gradient lifting decision tree; the method specifically comprises the following steps:
sequentially extracting the training data extracted in multiple time periods to obtain a signal intensity sequence set RSS' of m anchor labels on n readersAT={RSS″AT1,…,RSS″ATtElement in } Note as RSS'ATThe Euclidean distance matrix set L of the m anchor labels and the target label, and the difference value (LX, LY) of the mapping coordinate values take the element values in the three sets as input features. In turn liftTaking the signal intensity sequence set RSS' of t target tags on n readersDT={RSS″DT1,…,RSS″DTmElement in } Note as RSS'DTAs the target value, a training set { X ═ RSS ″, 'is composed'DT1,L,LX,LY),Y=(RSS″′DT). Inputting the two models into a GBDT regression algorithm model, and respectively training the two models
And 7: and in the online positioning stage, only setting an anchor label in the positioning area, and removing the target label. The method comprises the steps of collecting signal intensity values of fixed anchor labels in a positioning area in real time, inputting Euclidean distance matrixes of the anchor labels and original target labels and mapping coordinate value difference values serving as input data into a trained fingerprint library updating model based on a gradient lifting decision tree to obtain a prediction result, namely the updated target label signal intensity values, and replacing corresponding values in an initial fingerprint library with the collected anchor label signal intensity values and the calculated target label signal intensity values, so that the fingerprint library is updated.
In step 6, the step of performing the signal strength difference update model training by using the GBDT regression algorithm is as follows:
step (6-1): sequentially extracting the training data extracted in multiple time periods to obtain a signal intensity sequence set RSS' of m anchor labels on n readersAT={RSS″AT1,…,RSS″ATtElement in } Note as RSS'ATThe Euclidean distance matrix set L of the m anchor labels and the target label, and the difference value (LX, LY) of the mapping coordinate values take the element values in the three sets as input features. Sequentially extracting signal intensity sequence sets RSS' of t target tags on n readersDT={RSS″DT1,…,RSS″DTmElement in } Note as RSS'DTAs the target value, a training set { X ═ RSS ″, 'is composed'AT1,L,LX,LY),Y=(RSS″′DT). And recording the length of the signal intensity sequence as S, and then the dimension of the training set is S x T.
Step (6-2): initializing a classification regression tree weak learning device, and inputting a training data set M { (x)1,y1),(x2,y2),……,(xN,yN) And f, wherein X is data of training characteristics (the signal intensity value of each anchor label on the reader and the actual coordinate, Euclidean distance, X-axis and Y-axis mapping coordinate difference values of each target label), Y is data of a training label (the signal intensity value of the target label on the reader), L represents a loss function, c represents a fitting residual error (the difference value of the target label and the model label), m represents iteration times, and f represents the number of iterations0(x) Represent initializing the weak learner regression tree:
(6-3) for iteration round T1, 2, … T, and for sample i 1, 2.. N, a negative gradient is calculated:
rtirepresents the negative gradient value of the loss function at the ith sample in the t iteration,represents the loss function L (y)i,f(xi) )) pair f (x)i) Derivation, f (x) represents the weak learner regression tree of the t round, f (t-1) represents the regression tree of the t-1 iteration;
(6-4) Using (x)i,rti) Fitting a CART regression tree to obtain a t regression tree, wherein the corresponding leaf node region of the t regression tree is RtjJ is 1,2, J is the number of leaf nodes of the regression tree t, wherein i is 1,2, N;
(6-5) for leaf region J ═ 1,2 · J, calculate the best fit:
ctjrepresenting the best output value of the fitted leaf node;
updating the strong learner:
f is thent(x) A strong learner representing a t-th round of fitting;
(6-6) finally obtaining an expression of the gradient boosting regression tree f (x) of the strong learner, namely the expression is used for updating the fingerprint algorithm training model:
wherein f is0(x) Indicating the initialization of the weak learner, T indicating the total T rounds of iteration, J indicating the number of leaf nodes of each round of regression tree T, ctjRepresenting the best output value to fit the leaf node.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Claims (8)
1. An RFID indoor positioning fingerprint database updating system is characterized by comprising a plurality of RFID readers, a plurality of target tags, a plurality of anchor tags, a preprocessor, an upper computer and a data processing terminal;
the RFID reader configured to: acquiring signal intensity values of the anchor label and the target label in real time and transmitting the signal intensity values to the pre-processor in real time;
the pre-processor configured to: transmitting the received signal strength values of the anchor tag and the target tag to an upper computer in real time, monitoring an RFID reader in real time, and alarming when the RFID reader stops sending signals is detected;
the upper computer is configured to: the received signal strength signals of the anchor labels and the target labels are subjected to noise reduction and filtering and are transmitted to a data processing terminal for storage, and the upper computer also displays real-time signal strength values, historical signal strength values and signal strength value trend curves of all the anchor labels in real time and provides signal strength query service;
the data processing terminal is configured to: the method comprises the steps of setting an initial fingerprint library, arranging anchor labels and target labels in a positioning area, measuring two-dimensional coordinates of the anchor labels and the target labels, simultaneously acquiring static signal intensity values of the anchor labels and the target labels, setting the initial fingerprint library and storing the initial fingerprint library in a data processing terminal, further, removing the target labels, only acquiring signal intensity values of the anchor labels, updating the fingerprint library, carrying a fingerprint updating model based on a gradient lifting decision tree, and inputting the signal intensity values of the anchor labels received in real time, Euclidean distance matrixes of the anchor labels and the original target labels and mapping coordinate value difference values serving as input data into the trained fingerprint library updating model to obtain updated signal intensity values of the target labels; replacing and storing corresponding values in an initial fingerprint library by the acquired anchor label signal intensity value and the calculated target label signal intensity value, and realizing the updating of the fingerprint library, wherein the steps are as follows:
(1) arranging anchor tags and target tags in a positioning area, establishing a plane rectangular coordinate system, and collecting and storing coordinate values of each anchor tag and each target tag;
(2) arranging a plurality of RFID readers around the positioning area, and respectively carrying out static acquisition and storage on the signal strength values of the anchor tag and the target tag by utilizing the RFID readers at different positions;
(3) respectively collecting the signal intensity of an anchor label and the signal intensity of a target label in the same time interval to form an offline anchor label signal intensity sequence and an offline target label signal intensity sequence;
(4) calculating Euclidean distance between each anchor label and each target label in the positioning area, and calculating a coordinate difference value of each anchor label and each target label on an X, Y axis in the positioning area;
(5) taking the signal intensity sequences of all the acquired anchor labels and the calculated Euclidean distances and coordinate difference values as input characteristics, taking the fingerprint signal intensity sequence of each target label as an output target value, performing regression training of the gradient lifting decision tree to obtain a trained fingerprint updating model based on the gradient lifting decision tree,
the method specifically comprises the following steps:
(5-1) initializing a classification regression tree weak learning device, and inputting a training data set M { (x)1,y1),(x2,y2),......,(xN,yN) Where x is training feature data, y is a training label, L is a loss function, c is a fitting residual, m is an iteration number, f is a training residual, and0(x) Represent initializing the weak learner regression tree:
(5-2) for iteration round T1, 2, … T, for sample i 1, 2.. N, a negative gradient is calculated:
wherein r istiRepresenting the negative gradient value of the loss function of the ith sample in the t iteration, f (x) representing the weak learner regression tree of the t iteration, and f (t-1) representing the regression tree of the t-1 iteration;
(5-3) Using (x)i,rti) Fitting a CART regression tree to obtain a t regression tree, wherein the corresponding leaf node region of the t regression tree is RtjJ is 1,2, J is the number of leaf nodes of the regression tree t, wherein i is 1,2, N;
(5-4) for leaf area J ═ 1,2 · · J, calculate the best fit:
ctjrepresenting the best output value of the fitted leaf node;
updating the strong learner:
f is thent(x) A strong learner representing a t-th round of fitting;
(5-5) finally obtaining an expression of the gradient boosting regression tree f (x) of the strong learner, namely the expression is used for updating the fingerprint algorithm training model:
wherein f is0(x) Indicating the initialization of the weak learner, T indicating the total T rounds of iteration, J indicating the number of leaf nodes of each round of regression tree T, ctjRepresenting the best output value of the fitted leaf node;
(6) acquiring a signal intensity value of an anchor label in a positioning area in real time, and inputting the signal intensity value of the anchor label, an Euclidean distance matrix between the anchor label and an original target label and a mapping coordinate value difference value into a trained fingerprint library updating model by taking the signal intensity value of the anchor label and the Euclidean distance matrix and the mapping coordinate value difference value as input data to obtain a prediction result, wherein the prediction result is the signal intensity value of the target label after updating;
(7) and replacing the corresponding value in the initial fingerprint library by the acquired anchor label signal intensity value and the calculated target label signal intensity value to update the fingerprint library.
2. The RFID indoor location fingerprint library updating system of claim 1, wherein the RFID reader, the pre-processor, the upper computer and the data processing terminal are all communicated with each other by RS232 serial ports.
3. The RFID indoor positioning fingerprint database updating system of claim 1, wherein the host computer is further configured to receive and store the updated fingerprint database in real time in the host computer, and provide real-time display and real-time query services of the updated fingerprint database.
4. An updating method of an RFID indoor positioning fingerprint database is characterized by comprising the following steps:
(1) arranging anchor tags and target tags in a positioning area, establishing a plane rectangular coordinate system, and collecting and storing coordinate values of each anchor tag and each target tag;
(2) arranging a plurality of RFID readers around the positioning area, and respectively carrying out static acquisition and storage on the signal strength values of the anchor tag and the target tag by utilizing the RFID readers at different positions;
(3) respectively collecting the signal intensity of an anchor label and the signal intensity of a target label in the same time interval to form an offline anchor label signal intensity sequence and an offline target label signal intensity sequence;
(4) calculating Euclidean distance between each anchor label and each target label in the positioning area, and calculating a coordinate difference value of each anchor label and each target label on an X, Y axis in the positioning area;
(5) taking the signal intensity sequences of all the acquired anchor labels and the calculated Euclidean distances and coordinate difference values as input characteristics, taking the fingerprint signal intensity sequence of each target label as an output target value, performing regression training of the gradient lifting decision tree to obtain a trained fingerprint updating model based on the gradient lifting decision tree,
the method specifically comprises the following steps:
(5-1) initializing a classification regression tree weak learning device, and inputting a training data set M { (x)1,y1),(x2,y2),......,(xN,yN) Where x is training feature data, y is a training label, L is a loss function, c is a fitting residual, m is an iteration number, f is a training residual, and0(x) Represent initializing the weak learner regression tree:
(5-2) for iteration round T1, 2, … T, for sample i 1, 2.. N, a negative gradient is calculated:
wherein r istiRepresenting the negative gradient value of the loss function of the ith sample in the t iteration, f (x) representing the weak learner regression tree of the t iteration, and f (t-1) representing the regression tree of the t-1 iteration;
(5-3) Using (x)i,rti) Fitting a CART regression tree to obtain a t regression tree, wherein the corresponding leaf node region of the t regression tree is RtjJ is 1,2, J is the number of leaf nodes of the regression tree t, wherein i is 1,2, N;
(5-4) for leaf area J ═ 1,2 · · J, calculate the best fit:
ctjrepresenting the best output value of the fitted leaf node;
updating the strong learner:
f is thent(x) A strong learner representing a t-th round of fitting;
(5-5) finally obtaining an expression of the gradient boosting regression tree f (x) of the strong learner, namely the expression is used for updating the fingerprint algorithm training model:
wherein f is0(x) Indicating the initialization of the weak learner, T indicating the total T rounds of iteration, J indicating the number of leaf nodes of each round of regression tree T, ctjRepresenting the best output value of the fitted leaf node;
(6) acquiring a signal intensity value of an anchor label in a positioning area in real time, and inputting the signal intensity value of the anchor label, an Euclidean distance matrix between the anchor label and an original target label and a mapping coordinate value difference value into a trained fingerprint library updating model by taking the signal intensity value of the anchor label and the Euclidean distance matrix and the mapping coordinate value difference value as input data to obtain a prediction result, wherein the prediction result is the signal intensity value of the target label after updating;
(7) and replacing the corresponding value in the initial fingerprint library by the acquired anchor label signal intensity value and the calculated target label signal intensity value to update the fingerprint library.
5. The method for updating an RFID indoor positioning fingerprint database according to claim 4, wherein in the step (1), a planar rectangular coordinate system is established with any fixed point of a rectangular space in the indoor positioning area as an origin of coordinates, and with two adjacent sides of the origin of coordinates as an X axis and a Y axis.
6. The method for updating the RFID indoor positioning fingerprint database as claimed in claim 4, wherein in the step (3), after the offline anchor tag signal strength sequence and the target tag signal strength sequence are formed, the abnormal values in the sequences are processed by a bit-splitting method, so as to obtain the offline anchor tag and target tag signal strength sequence set after the abnormal values are processed.
7. The method for updating the RFID indoor positioning fingerprint database as claimed in claim 6, wherein the abnormal values in the sequence are processed by a quantile method, specifically: in each signal strength sequence, using Q3Representing the third quartile, Q, in the signal strength sequence1Denotes the first quartile, and furthermore IRQ ═ Q3-Q1And is recorded as Q3+1.5 IQR, with the lower limit Q1And 1.5 IQR, respectively replacing signal intensity values which are not in the range of the upper and lower limits in the signal intensity sequence by using the average values of the left and right adjacent signal intensity values to obtain a signal intensity sequence set after abnormal value processing.
8. The method for updating the RFID indoor location fingerprint library of claim 4, wherein in the step (6), the position of the anchor tag is kept unchanged, and after the target tag in the location area is removed, the signal intensity value of the anchor tag in the location area is acquired in real time.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911075626.8A CN110996280B (en) | 2019-11-06 | 2019-11-06 | RFID indoor positioning fingerprint database updating system and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911075626.8A CN110996280B (en) | 2019-11-06 | 2019-11-06 | RFID indoor positioning fingerprint database updating system and method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110996280A CN110996280A (en) | 2020-04-10 |
CN110996280B true CN110996280B (en) | 2020-12-25 |
Family
ID=70083251
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911075626.8A Active CN110996280B (en) | 2019-11-06 | 2019-11-06 | RFID indoor positioning fingerprint database updating system and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110996280B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112243193B (en) * | 2020-09-29 | 2022-08-12 | 成都长虹网络科技有限责任公司 | Indoor positioning method and device, computer equipment and readable storage medium |
CN114449435B (en) * | 2020-10-19 | 2023-12-15 | 中国移动通信集团重庆有限公司 | Fingerprint positioning method based on measurement learning of large-interval neighbor |
CN112804642B (en) * | 2021-04-08 | 2022-03-04 | 上海迹寻科技有限公司 | Fingerprint data updating judgment method and device for target area |
CN113660606B (en) * | 2021-08-16 | 2022-12-27 | 中国建设银行股份有限公司 | Indoor positioning method and device, electronic equipment and storage medium |
CN113945215A (en) * | 2021-10-11 | 2022-01-18 | 浙江工业大学 | RFID indoor positioning method based on stacking model |
CN115209341A (en) * | 2022-06-30 | 2022-10-18 | 南京捷希科技有限公司 | Weighted random forest indoor positioning method based on channel state information |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9008685B1 (en) * | 2011-04-26 | 2015-04-14 | Andrew Joseph Gold | Method and system for indoor RF mapping |
CN104853317A (en) * | 2014-11-24 | 2015-08-19 | 北京航空航天大学 | WiFi indoor positioning fingerprint database construction and update method |
CN108871332A (en) * | 2018-04-26 | 2018-11-23 | 广西大学 | A kind of RFID indoor locating system and method based on XGBoost |
CN110087188A (en) * | 2019-04-25 | 2019-08-02 | 中山大学 | The virtual finger print data base construction method of indoor positioning based on RFID label tag |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104105196B (en) * | 2013-04-09 | 2017-07-21 | 广东美晨通讯有限公司 | The method and system positioned based on radio-frequency fingerprint |
US20160150029A1 (en) * | 2014-11-26 | 2016-05-26 | Avaya Inc. | Service discovery using a location database |
CN105137389B (en) * | 2015-09-02 | 2018-04-27 | 安宁 | A kind of video auxiliary radio frequency positioning method and device |
CN105242239B (en) * | 2015-10-19 | 2017-06-16 | 华中科技大学 | A kind of indoor subregion localization method based on the sub-clustering of mass-rent fingerprint and matching |
CN105682039B (en) * | 2016-03-15 | 2019-03-22 | 广州杰赛科技股份有限公司 | A kind of RF fingerprint positioning method and system |
-
2019
- 2019-11-06 CN CN201911075626.8A patent/CN110996280B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9008685B1 (en) * | 2011-04-26 | 2015-04-14 | Andrew Joseph Gold | Method and system for indoor RF mapping |
CN104853317A (en) * | 2014-11-24 | 2015-08-19 | 北京航空航天大学 | WiFi indoor positioning fingerprint database construction and update method |
CN108871332A (en) * | 2018-04-26 | 2018-11-23 | 广西大学 | A kind of RFID indoor locating system and method based on XGBoost |
CN110087188A (en) * | 2019-04-25 | 2019-08-02 | 中山大学 | The virtual finger print data base construction method of indoor positioning based on RFID label tag |
Also Published As
Publication number | Publication date |
---|---|
CN110996280A (en) | 2020-04-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110996280B (en) | RFID indoor positioning fingerprint database updating system and method | |
CN109444813B (en) | RFID indoor positioning method based on BP and DNN double neural networks | |
CN110866528B (en) | Model training method, energy consumption use efficiency prediction method, device and medium | |
CN109212476B (en) | RFID indoor positioning algorithm based on DDPG | |
CN108871332A (en) | A kind of RFID indoor locating system and method based on XGBoost | |
CN114581855B (en) | Information collection method and system based on big data | |
CN108600965B (en) | Passenger flow data prediction method based on guest position information | |
CN113052295B (en) | Training method of neural network, object detection method, device and equipment | |
CN115482465A (en) | Crop disease and insect pest prediction method and system based on machine vision and storage medium | |
CN111652195A (en) | Land utilization change analysis and prediction method | |
CN116416884B (en) | Testing device and testing method for display module | |
CN203759763U (en) | Intelligent power equipment patrol system based on RFID technique | |
Sun et al. | Remote estimation of grafted apple tree trunk diameter in modern orchard with RGB and point cloud based on SOLOv2 | |
CN111539910B (en) | Rust area detection method and terminal equipment | |
CN112702408A (en) | Internet of things system and method based on multi-sensing function | |
CN110852795A (en) | Mobile phone valuation system and valuation method based on computer vision | |
CN114066070A (en) | Weather forecasting method based on artificial intelligence and numerical calculation | |
CN113743547A (en) | Asset supervision method and platform based on RFID (radio frequency identification) tag and storage medium | |
CN108462939A (en) | A kind of indoor orientation method of earth magnetism Time-Series analysis | |
CN116933626A (en) | Data monitoring method and device based on digital twinning | |
CN116563841A (en) | Detection method and detection device for power distribution network equipment identification plate and electronic equipment | |
CN114004138A (en) | Building monitoring method and system based on big data artificial intelligence and storage medium | |
CN109063735A (en) | A kind of classification of insect Design Method based on insect biology parameter | |
CN111462087B (en) | Image detection method, device and system based on artificial intelligence and storage medium | |
CN104331836B (en) | Crops decision support method and device |
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 |