TWI498582B - A 3-dimensions space positioning method - Google Patents

Description

Three-dimensional spatial positioning method

The present invention relates to a positioning method; and more particularly to a three-dimensional spatial positioning method.

With the development of technology, infrared, ultrasonic, IEEE 802.11 or radio frequency identification (RFID) technology has gradually become the mainstream of positioning technology, among which RFID positioning technology has the advantages of high stability, good environmental tolerance and low cost. It can be used for positioning by means of signal arrival angle (AoA), signal arrival time (ToA), signal arrival time difference (TDoA) and received signal strength (RSSI). Among them, the technology of receiving signal strength is suitable for large-scale deployment, and the cost is low, and can be applied to mobile computing devices, and gradually becomes the research and development focus of RFID positioning technology, as illustrated below.

A typical RFID positioning technique is the LANDMARC algorithm (see "Pahlavan, K., L. Xinrong, JPMakela, "Indoor geolocation science and technology," IEEE Communications Magazine, Vol. 40, No. 2, 2002, pp. 112). -118."). GYJin and other scholars (2006) used LANDMARC as a reference to propose a screening mechanism for reference tags, thereby improving the accuracy of indoor positioning and reducing the number of RFID tags. STShih et al. (2006) cited the theory of LANDMARC, and The signal strength determines the weight, and the weight is used to represent the object's distance, and then the triangle is used to divide the region and slowly approach the target point. Xuejing Jiang and other scholars (2009) locate the reference tag and the closest tracking tag (Tracking Tag) after positioning through the LANDMARC system. ) Interchange, after repositioning, compared with previous experimental results, can improve the positioning accuracy of the system; Chai-Hao Chang and other scholars (2009) Then combined with RFID and ultrasonic technology for positioning experiments to reduce the impact of RFID signals caused by external interference. In addition, the "RFID positioning method and system thereof" of the Republic of China Patent Publication No. 201032138 A1 discloses that a computer can calculate an RFID tag by a positioning algorithm (for example, a triangulation algorithm) by using a plurality of relays to identify the RSSI value of a message. Coordinate position. However, the above research content is only applicable to the positioning of two-dimensional (2-D) space, and cannot meet the positioning requirements of three-dimensional (3-D) space.

In addition, M Ayoub Khan and other scholars (2009) have shown experimental results that the LANDMARC algorithm is feasible in three-dimensional space. However, its positioning accuracy is not good, and it still needs to be adjusted and improved in the way of taking adjacent points, so that it can be practically applied to three-dimensional positioning.

In view of this, it is necessary to propose a three-dimensional spatial positioning method to improve the above-mentioned shortcomings based on the previous technology of LANDMARC, conform to the positioning requirements of the three-dimensional space, and improve its practicability.

The main object of the present invention is to provide a three-dimensional spatial positioning method to improve the accuracy of RFID application in three-dimensional spatial positioning.

The present invention provides a three-dimensional spatial positioning method, comprising: setting eight readers at eight corners of a cubic space, and uniformly arranging a plurality of reference labels in the cubic space, the readers collecting the reference labels The signal strength index is calculated by a computer system to calculate an average value and a standard deviation of the signal strength indexes, and after filtering out the signal intensity index deviating from the standard deviation, calculating a probability density function of each signal intensity index, the probability density function The calculation is as follows: Where f(x) is the probability density function, x is the signal strength index for the computer rate density function, σ is the standard deviation, μ is the average value, and each signal strength index and the coordinates of the reference label are trained by a type of neural network. For generating the spacing of the reference label; collecting, by the reader, a signal strength index of the tracking label, and submitting to the computer system to calculate a difference between the tracking label and the signal strength index of each reference label, and selecting a smaller difference The eight reference tags are used as eight positioning tags; and the computer system calculates a correlation coefficient between the tracking tag and the signal strength index between the positioning tags, and calculates a weight value of each positioning tag according to the correlation coefficient, according to each weight value and The coordinates of the positioning tag are calculated to calculate the coordinates of the tracking tag.

Preferably, the average is calculated as follows: Among them, μ, For the average, n is the number of these signal strength indices, k( k [1, n ]) is the number of each reference label, and x _k is the signal strength index of each reference label.

Preferably, the standard deviation is calculated as follows: Where σ is the standard deviation, μ is the average value, and n is the number of the signal strength indices, k( k [1, n ]) is the number of each reference label, and x _k is the signal strength index of each reference label.

Preferably, the correlation coefficient is calculated as follows: Where i is the number of the plurality of positioning tags, E _i is a correlation coefficient between the i-th positioning tag and the signal strength index of the tracking tag, θ _i is a signal strength index of each positioning tag, and S is a signal of the tracking tag Strength index, where n is the number of the positioning tags.

Preferably, the weight value is calculated as follows: Where i and j are the numbers of the plurality of positioning tags, w _i is the weight value of the jth positioning tag, k is the number of the positioning tags, and E _i is the signal strength index between the i th positioning tag and the tracking tag Correlation coefficient.

Preferably, the coordinates of the tracking tag are calculated as follows: Where (x, y, z) is the coordinate of the tracking tag, k is the number of the positioning tags, w _j is the weight value of the jth positioning tag, and (x _i , y _i , z _i ) is the ith Position the coordinates of the label.

〔this invention〕

1‧‧‧ physical layer

11‧‧‧Reader

11a~11h‧‧‧Reader

12‧‧‧ Reference Label

12a’~12h’‧‧‧ Positioning Label

2‧‧‧Positioning layer

21‧‧‧Signal collection module

22‧‧‧Scope cutting module

23‧‧‧Coordinate positioning module

24‧‧‧Database

3‧‧‧Application layer

31‧‧‧Book Bibliographic Management System

32‧‧‧Logistics warehousing management system

33‧‧‧Household Goods Management System

C‧‧‧ corner

P‧‧‧ cubic space

R‧‧‧ computer system

T‧‧‧ Tracking Label

X, Y, Z‧‧‧ axes

S1‧‧‧ training steps

S2‧‧‧ cutting steps

S3‧‧‧ positioning steps

1 is a system architecture diagram of an embodiment of a three-dimensional spatial positioning method of the present invention.

Figure 2 is a flow chart showing the operation of an embodiment of the three-dimensional spatial positioning method of the present invention.

3 is a front view and a perspective view of a label position of an embodiment of the three-dimensional spatial positioning method of the present invention.

Figure 4 is a schematic diagram of the label coordinates of an embodiment of the three-dimensional spatial positioning method of the present invention.

The above and other objects, features and advantages of the present invention will become more <RTIgt; "Radio Frequency Identification" (RFID) refers to a technology in which a reader can identify different tags by radio frequency signals, and can be divided into active according to whether the tag has transmission capability. And passive RFID tags are understood by those of ordinary skill in the art to which the present invention pertains.

The "coupled connection" as used throughout the present invention refers to the mutual transfer of data between two devices by means of wireless communication (e.g., RFID technology), as will be understood by those of ordinary skill in the art to which the present invention pertains.

The "cubic space" as used throughout the text of the present invention means a A cubic spatial extent is presented having eight corners, each corner being located at each vertex of the cuboid, as will be understood by those of ordinary skill in the art to which the present invention pertains.

The "Back-propagation neural network" (BPN) described in the full text of the present invention refers to a neural network that combines a multilayer perceptron (MLP) with an Error Back Propagation (EBP) algorithm. Network technology (see "Zhang Feizhang, Zhang Liqiu," Neural Network", Donghua Book Company, 2005"), in which the inverted transmission neural network will output the error value from the output during the learning phase. The layer is passed back to the hidden layer, to the input layer, and the bond values between the networks are corrected to obtain a more closely expected output, as will be understood by those of ordinary skill in the art to which the present invention pertains.

Please refer to FIG. 1 , which is a system architecture diagram of an embodiment of a three-dimensional spatial positioning method according to the present invention. The three-dimensional spatial positioning system includes a physical layer 1 and a positioning layer 2, and the physical layer 1 is provided with a plurality of readers (RFID Readers) 11 and a plurality of reference tags 12, the reference tags 12 Preferably, the reader 11 is coupled to the reference tag 12 for collecting the Received Signal Strength Index (RSSI) of the reference tag 12 and transmitting it to the positioning layer 2. In this embodiment, the RFID tag 12 is implemented as a passive RFID tag, but is not limited thereto.

Referring to FIG. 1 again, the positioning layer 2 can execute a software program by a computer system R that can receive the signal strength index. The computer system R can be coupled to the reader 11 and the software program can be planned into a signal. A module 21, a range cutting module 22, a standard positioning module 23, and a database 24 for calculating a tracking tag set in a target according to the signal strength index. The coordinates of the coordinates. In this embodiment, the signal collection module 21 of the positioning layer 2 has a signal strength 〞, a Probability Density Function (PDF), and a 神经-like neural network (eg, a reverse-transfer-like neural network). The function is used to determine the relationship between the range of the RSSI values that have been screened and the distance, and is sent to the trained neural network; the range cutting module 22 has a range segmentation, such as: Al (algorithm) 〝 and 〝 range confirmation function, etc., for selecting a plurality of reference tags 12 that are closer to the tracking tag, and performing range cutting according to the reference tag 12 by the Divide-and-Conquer algorithm to estimate the Tracking the range of the tag; the coordinate positioning module 23 has a function of estimating the RSSI relationship set 〞, 〝 estimating the reference tag weight 〞, and 〝 estimating the tracking tag coordinate , to calculate the RSSI value between the tracking tag and the reference tag. a relationship set, the weight of each reference label is calculated by the relationship between the tracking label and the reference label, and the coordinates of the tracking label are calculated according to the weight of each reference label; the database 24 is used to store the value required by the calculation process; This is limited.

Referring to FIG. 1 again, the positioning layer 2 can also be connected to an application layer 3 by using a wired or wireless network. The application layer 3 can be a computer system with special application functions, such as a book bibliographic management system and logistics. The warehouse management system or the household item management system is used to perform special application functions such as book bibliographic management, logistics storage management or household item management based on the above location information, but not limited thereto. In this embodiment, the application layer 3 includes a book bibliographic management system 31, a logistics warehousing management system 32, and a household item management system 33, but is not limited thereto.

Referring to FIG. 2, it is an operational flowchart of an embodiment of the three-dimensional spatial positioning method of the present invention, wherein the method includes a training step S1, a cutting step S2, and a positioning step S3, respectively, as described later.

The training step S1 is to set eight readers in eight corners of a cubic space, and uniformly arrange a plurality of reference labels in the cubic space, and the readers collect the signal strength indexes of the reference labels, and Calculating the average value and standard deviation of the signal strength indices by a computer system, filtering out the signal strength index deviating from the standard deviation, and training each signal strength index and the coordinates of the reference label with a type of neural network to generate The spacing of the reference labels. Wherein, after filtering the signal strength index deviating from the standard deviation, it is better to calculate the probability density function of each signal intensity index to determine the range of the signal strength index of each reference label, if there is a subsequent step If the signal strength index of an unidentified tag (such as the tag of the target object) is detected, it can be compared with the signal strength index of each reference tag to determine which reference tag the unknown tag is adjacent to. In detail, referring to Figures 1 to 3, first, eight readers 11a to 11h are disposed in eight corners C of the cubic space P (as shown in Fig. 3), and in the cube A plurality of reference tags 12 are evenly arranged in the shape space P, for example, uniformly distributed along the X, Y, and Z axes between the readers 11a to 11h, and coordinates of the reference tags 12 are recorded to facilitate subsequent processes. Then, the signal intensity index (RSSI) of each reference tag 12 is collected by the readers 11a-11h, and the computer system R (as shown in FIG. 1) is used to calculate the average value and standard of the signal strength indexes. The difference is calculated by the following equations (1) and (2):

Among them, μ, For the average, σ is the standard deviation, n is the number of these signal strength indices, k( k [1, n ]) is the number of each reference label, and x _k is the signal strength index of each reference label. Then, the computer system R can filter out the signal strength index that deviates from the standard deviation (for example, the signal strength index is greater than the average plus standard deviation, or less than the average minus the standard deviation) to exclude the excessively large signal strength index. Thereafter, the computer system R calculates a probability density function of each signal strength index (ie, the signal strength index that is not excluded), and the probability density function is calculated as shown in the following equation (3): Where f(x) is the probability density function, x is the signal strength index for the computer rate density function, σ is the standard deviation, and μ is the average.

Then, the computer system R stores the coordinates of the reference tag 12 to which each signal strength index belongs, and trains the residual signal strength index and the coordinates of the reference tag 12 by using such a neural network (eg, an inverted transmission type neural network, etc.). For generating the residual signal strength index The spacing of the tags 12 is referenced as a basis for subsequent calculation of the relative distance of each of the reference tags 12. In this embodiment, the neural network preferably uses an inverted transfer neural network, and the learning process of the inverse transfer neural network includes: (1) setting a transfer function and a network value; (2) setting The initial weight of the network (such as: 14) and the bias value (such as: -14); (3) the input training group (such as: 0) and the target output value (such as: 40); (4) calculate the hidden layer Output value and output value of the output layer; (5) Calculate the allowable gap between the output layer and the hidden layer; (6) Calculate the gap between the output layer and the hidden layer; (7) Determine the difference between the output layer and the hidden layer Whether it is greater than the allowable gap; (8) Correct the weighted value and the partial weight of the output layer and the hidden layer; repeat (4)~(8) until the difference between the output layer and the hidden layer is less than the allowable gap, that is, complete the training. process.

In the cutting step S2, the reader collects a signal strength index of the tracking tag, and the computer system calculates a difference between the tracking tag and the signal intensity index of each reference tag, and selects eight references with smaller differences. The tag acts as eight positioning tags. In detail, please refer to the first to third figures. First, the signal intensity index of the tracking tag T (located in the target object) is collected by the readers 11a to 11h, and then the signal intensity of the tracking tag T is collected. The index is given to the computer system R, and the computer system R calculates the difference between the tracking index T and the signal intensity index of each reference tag 12, and selects eight reference tags 12 with smaller differences to perform range cutting. Positioning the label, as shown in FIG. 4, a positioning range (rectangular space) is formed by the positioning tags 12a'~12h', and the positioning tags 12a', 12b', 12c' and 12d' constitute a plane, and the positioning tags 12e', 12f ', 12g' and 12h' constitute another plane. In this embodiment, the range cutting algorithm may be selected as a Divide-and-Conquer algorithm, as shown in the following formula (4): Where E is a plane equation, ( x _i , y _i , z _i ) is the (x, y, z) coordinate of the i-th point on the plane (eg: 12a', 12b', 12c', 12d'), Is a vector that is perpendicular to the plane. Therefore, if the eight reference labels with smaller differences are not located near the tracking tag T, the Divide-and-Conquer algorithm can be used to obtain the range of the tracking tag T for related processes of subsequent positioning.

In the positioning step S3, the computer system calculates a correlation coefficient between the tracking tag and the signal strength index of each positioning tag, calculates a weight value of each positioning tag according to the correlation coefficient, and calculates according to each weight value and a coordinate of the positioning tag to which the tag belongs. The coordinates of the tracking tag. In detail, please refer to Figures 1 to 3, the computer system R calculates the correlation coefficient between the tracking tag T and the signal intensity index of each positioning tag (such as 12a'~12h' shown in Fig. 4). As shown in the following formula (5): Where i is the number of the plurality of positioning tags, E _i is the correlation coefficient between the i-th positioning tag (eg, 12a'~12h') and the tracking tag T, and θ _i is the positioning tag (eg: The signal strength index of 12a'~12h'), S is the signal strength index of the tracking tag T, and n is the number of the positioning tags (n=8). Then, the computer system R calculates the weight value of each positioning tag according to the correlation coefficient, as shown in the following formula (6): Where i and j are the numbers of the plurality of positioning tags (i=j), w _j is the weight value of the jth positioning tag (eg: 12a′~12h′), and k is the number of the positioning tags (k= 8), E _i is the correlation coefficient between the i-th positioning tag (eg: 12a'~12h') and the T-signal strength index between the tracking tags. Thereafter, the computer system R calculates the coordinates of the tracking tag T according to the coordinates of each weight value w _j and its associated positioning tag (eg, 12a'~12h'), as shown in the following formula (7): Where (x, y, z) is the coordinate of the tracking tag T, k is the number of the positioning tags (k=8), and w _j is the weight value of the jth positioning tag, (x _i , y _i , z _i ) is the coordinate of the ith positioning tag.

For example, as shown in FIG. 4, the coordinates and signal strength index of each reference tag 12 are as shown in Table 1 below. If the tracking strength of the target tag is 29, according to the signal strength index, The coordinates of the tracking tag are close to the coordinates (1,0,0) and (1,1,0), so the computer system needs to calculate the plane formed by the coordinates (1,0,0) and (1,1,0). Equation, let the coordinates (1,1,0) be the reference point, and find the vector perpendicular to the plane as the difference between the coordinates (0,1,0) and (1,1,0) (-1,0, 0), the coordinates (1,1,0) and the difference (-1,0,0) are brought into the above equation (4), and the plane equation E=-1(x-1)+0(y-1) is obtained. ) +0 (z-0).

Then, according to the signal strength index, the target object can be obtained in the plane range. Since the foregoing process can also calculate the probability density function of the signal intensity index of each reference label, the known probability density function can be used to confirm the proximity of the tracking label. Which reference label and the range of information, and the difference between the tracking label and the signal strength index of each reference label, it can be seen that the eight positioning coordinates constituting the cutting range are (0, 0, 0), (0,0,1), (0,1,0), (0,1,1), (1,0,0), (1,0,1), (1,1,0), (1 , 1,1), substituting this eight coordinates into the above formula (5), available .

Then, the computer system calculates the weight value of each positioning tag according to the above formula (6), and obtains w ₁ =1, , , , , , , . Therefore, the computer system substitutes the weight values w1~w8 and the coordinates of the positioning tag to which the positioning tag belongs to the above formula (7), and the coordinates of the tracking tag are as follows: ( x, y, z )=w ₁ (1, 1,0)+w ₂ (1,0,0)+w ₃ (0,1,1)+w ₄ (0,0,1)+w ₅ (1,1,1)+w ₆ (2, 0,0)+w ₇ (0,1,0)+w ₈ (0,0,0).

The main features of the embodiment of the three-dimensional spatial positioning method of the present invention are as follows: First, eight readers are arranged in eight corners of a cubic space, and several are evenly arranged in the cubic space. Referring to the label, the readers collect the signal strength index of each reference label, and the computer system calculates the average value and the standard deviation of the signal strength indexes, and filters the signal intensity index deviating from the standard deviation to calculate The probability density function of each signal strength index trains each signal strength index and the coordinates of the reference tag with a type of neural network to generate the spacing of the reference tag. Then, the reader collects a signal strength index of the tracking tag, and the computer system calculates a difference between the tracking tag and the signal strength index of each reference tag, and selects eight reference tags with a small difference as eight. Position the label. Then, the computer system calculates a correlation coefficient between the tracking tag and the signal strength index of each positioning tag, calculates a weight value of each positioning tag according to the correlation coefficient, and calculates the tracking tag according to each weight value and a coordinate of the positioning tag to which the tag belongs. coordinate. In this way, the shortcomings of conventional positioning based on the LANDMARC algorithm can be improved, and the positioning requirements of the three-dimensional space are met, and the effect of "improving the accuracy of RFID applied to three-dimensional positioning" is achieved.

While the invention has been described in connection with the preferred embodiments described above, it is not intended to limit the scope of the invention. The technical scope of the invention is protected, and therefore the scope of the invention is defined by the scope of the appended claims.

S1‧‧‧ training steps

S2‧‧‧ cutting steps

S3‧‧‧ positioning steps

Claims (6)

A three-dimensional spatial positioning method includes: setting eight readers in eight corners of a cubic space, and uniformly arranging a plurality of reference labels in the cubic space, the readers collecting signal strength indexes of each reference label Calculated by a computer system to calculate the average and standard deviation of the signal strength indices, and filter the signal intensity index deviating from the standard deviation, and calculate the probability density function of each signal intensity index. The probability density function is calculated as follows: As shown in the formula: Where f(x) is the probability density function, x is the signal strength index for the computer rate density function, σ is the standard deviation, μ is the average value, and each signal strength index and the coordinates of the reference label are trained by a type of neural network. For generating the spacing of the reference label; collecting, by the reader, a signal strength index of the tracking label, and submitting to the computer system to calculate a difference between the tracking label and the signal strength index of each reference label, and selecting a smaller difference The eight reference tags are used as eight positioning tags; and the computer system calculates a correlation coefficient between the tracking tag and the signal strength index between the positioning tags, and calculates a weight value of each positioning tag according to the correlation coefficient, according to each weight value and The coordinates of the positioning tag are calculated to calculate the coordinates of the tracking tag. According to the three-dimensional spatial positioning method described in claim 1, wherein the average value is calculated as follows: Among them, μ, For the average, n is the number of these signal strength indices, k( k [1, n ]) is the number of each reference label, and x _k is the signal strength index of each reference label. According to the three-dimensional spatial positioning method described in claim 1, wherein the standard deviation is calculated as follows: Where σ is the standard deviation, μ is the average value, and n is the number of the signal strength indices, k( k [1, n ]) is the number of each reference label, and x _k is the signal strength index of each reference label. According to the three-dimensional spatial positioning method described in claim 1, wherein the correlation coefficient is calculated as follows: Where i is the number of the plurality of positioning tags, E _i is a correlation coefficient between the i-th positioning tag and the signal strength index of the tracking tag, θ _i is a signal strength index of each positioning tag, and S is a signal of the tracking tag Strength index, where n is the number of the positioning tags. According to the three-dimensional spatial positioning method described in claim 1, wherein the weight value is calculated as follows: Where i and j are the numbers of the plurality of positioning tags, w _j is the weight value of the jth positioning tag, k is the number of the positioning tags, and E _i is the signal strength index between the i th positioning tag and the tracking tag Correlation coefficient. According to the three-dimensional spatial positioning method described in claim 1, wherein the coordinate calculation method of the tracking label is as follows: Where (x, y, z) is the coordinate of the tracking tag, k is the number of the positioning tags, w _j is the weight value of the jth positioning tag, and (x _i , y _i , z _i ) is the ith Position the coordinates of the label.