CN111523667A

Movatterモバイル変換

Info

Publication number: CN111523667A
Application number: CN202010365734.5A
Authority: CN
Inventors: 吴浩; 石海涛
Original assignee: Tianjin University
Current assignee: Tianjin University
Priority date: 2020-04-30
Filing date: 2020-04-30
Publication date: 2020-08-11
Anticipated expiration: 2040-04-30
Also published as: CN111523667B

Abstract

Translated fromChinese

本发明公开了一种基于神经网络的RFID定位方法，包括读卡器、处理器，所述处理器通过程序代码实现如下步骤：定位算法模块读取读卡器坐标数据、参考标签坐标数据和待测点真实坐标数据，通过采用LANDMARC算法、BVIRE算法和VIRE算法获得待测点测量坐标和定位误差数据；误差分析模块提取待测点测量坐标和定位误差数据中的时间戳、误差主要特征构建训练集；定位修正模块对最小误差位置点坐标和最小误差位置点时间进行修正输出待测点坐标修正值和对应最大出现可能的位置坐标，该方法通过使用神经网络的方法对误差分析，减少原始噪声与数据量，降低算法误差、计算量和计算复杂度。

The invention discloses an RFID positioning method based on a neural network, comprising a card reader and a processor. The processor implements the following steps through program codes: a positioning algorithm module reads the card reader coordinate data, the reference tag coordinate data and the to-be-to-be-reader coordinate data. The real coordinate data of the measuring point is obtained by using the LANDMARC algorithm, the BVIRE algorithm and the VIRE algorithm to obtain the measured coordinates and positioning error data of the point to be measured; the error analysis module extracts the time stamp and the main features of the error in the measured coordinates and positioning error data of the point to be measured. Set; the positioning correction module corrects the coordinates of the minimum error position point and the time of the minimum error position point, and outputs the correction value of the coordinates of the point to be measured and the corresponding maximum possible position coordinates. This method uses the neural network method to analyze the error and reduce the original noise. With the amount of data, the algorithm error, the amount of calculation and the computational complexity are reduced.

Description

Translated fromChinese

一种基于神经网络的RFID定位方法A Neural Network-Based RFID Localization Method

技术领域technical field

本发明主要涉及基于RFID网络的室内定位技术领域,尤其涉及一种基于神经网络的RFID定位方法。The invention mainly relates to the technical field of indoor positioning based on an RFID network, in particular to an RFID positioning method based on a neural network.

背景技术Background technique

目前在室内环境下，射频识别(Radio Frequency Identification，RFID)技术由于其具有信息携载功能和传输可靠识别等特点得到了广泛的重视和应用。RFID定位技术利用标签对物体的唯一标识特性，依据读卡器接收到的电子标签发送的信号来获得电子标签的位置信息。RFID室内定位的依据是结合RFID信号的接收信号强度、相位等参数，利用定位算法完成距离和方位的计算，即通过在室内预先放置若干读卡器，当具有电子标签的移动物体进入读卡器的识别范围后，就可以将接收的信号上传至上位机，上位机通过计算标签的信号衰减程度以及相邻已知标签的位置信息即可实现定位算法。At present, in the indoor environment, radio frequency identification (RFID) technology has been widely valued and applied due to its characteristics of information carrying function and reliable transmission identification. RFID positioning technology uses the unique identification characteristics of tags to objects, and obtains the position information of electronic tags according to the signals sent by electronic tags received by card readers. The basis of RFID indoor positioning is to combine the received signal strength, phase and other parameters of the RFID signal, and use the positioning algorithm to complete the calculation of distance and orientation. After the recognition range is reached, the received signal can be uploaded to the host computer, and the host computer can realize the positioning algorithm by calculating the signal attenuation degree of the tag and the position information of the adjacent known tags.

基于RFID的室内定位算法可分为两大类：非测距定位算法和测距定位算法。测距定位算法中的测RSSI定位法相比较于其它定位方法更易操作，测RSSI定位法的思路是通过测量信号强度值RSSI来估算物体位置，较为先进的定位算法包括LANDMARC算法、BVIRE算法和VIRE算法。RFID-based indoor positioning algorithms can be divided into two categories: non-ranging positioning algorithms and ranging positioning algorithms. Compared with other positioning methods, the RSSI positioning method in the ranging positioning algorithm is easier to operate. The idea of the RSSI positioning method is to estimate the position of the object by measuring the signal strength value RSSI. The more advanced positioning algorithms include LANDMARC algorithm, BVIRE algorithm and VIRE algorithm .

LANDMARC(Location Identification Based on Dynamic ActiveRFIDCalibration)算法因其简单且定位精度高得到了许多肯定，在LANDMARC算法的基础上用相似的方插入网格虚拟参考标签和边界参考标签得到边界虚拟标签算法(BoundaryVirtual Label Algorithm，BVIRE)。在边界虚拟标签算法中采用两个权值，比LANDMARC算法多1个，而且在邻近标签的选择上采用阈值TH来排除小概率的大误差标签，使得BVIRE算法在定位精度上大幅提高。VIRE算法在利用LANDMARC的原理的同时，引入了虚拟参考标签和邻近地图的概念，提出了虚拟参考标签的概念，利用插值法来对虚拟参考标签的信号强度值进行估算，并将这些虚拟的参考标签作为实际参考标签进行后期的计算定位，提高了精确度和定位计算可行性。The LANDMARC (Location Identification Based on Dynamic ActiveRFIDCalibration) algorithm has been affirmed a lot because of its simplicity and high positioning accuracy. On the basis of the LANDMARC algorithm, similar squares are used to insert grid virtual reference labels and boundary reference labels to obtain the boundary virtual label algorithm (Boundary Virtual Label). Algorithm, BVIRE). Two weights are used in the boundary virtual label algorithm, one more than the LANDMARC algorithm, and the threshold TH is used in the selection of adjacent labels to exclude the small probability large error labels, which makes the BVIRE algorithm greatly improve the positioning accuracy. While using the principle of LANDMARC, the VIRE algorithm introduces the concept of virtual reference labels and proximity maps, and proposes the concept of virtual reference labels. The interpolation method is used to estimate the signal strength value of the virtual reference labels, and these virtual reference labels are used to estimate the signal strength values. The tag is used as the actual reference tag for later calculation and positioning, which improves the accuracy and the feasibility of positioning calculation.

传统的方案需要将所有的测试样本存储在数据库中。这将严重影响定位效率和精度，因为在复杂的室内环境中，大部分测试样本都是有噪声的。并且需要采集大量的数据作为比对来高精度。Traditional solutions require all test samples to be stored in a database. This will seriously affect the localization efficiency and accuracy, because in complex indoor environments, most of the test samples are noisy. And it needs to collect a large amount of data as a comparison to achieve high precision.

近年来，机器学习在rfid的处理中得到了越来越多的应用。但是，由于所采集到的RSSI信号通常是有噪声的，现有的特征的选取都是基于原始数据，需要大量的运算以及调优工作才能使得模型精度提高。因此，误差修正与定位计算有机的结合，实现减小定位误差、提高定位准确率、增加定位计算效率成为亟待解决的问题。In recent years, machine learning has been increasingly applied in the processing of RFID. However, since the collected RSSI signals are usually noisy, the selection of existing features is based on the original data, which requires a lot of computation and tuning work to improve the model accuracy. Therefore, the organic combination of error correction and positioning calculation to reduce positioning error, improve positioning accuracy, and increase positioning calculation efficiency has become an urgent problem to be solved.

发明内容SUMMARY OF THE INVENTION

本发明旨在设计一种基于神经网络的RFID定位方法，通过使用神经网络的方法对误差分析，减少原始噪声与数据量，降低算法误差、计算量和计算复杂度。The present invention aims to design an RFID positioning method based on neural network. By using the method of neural network to analyze the error, the original noise and data amount are reduced, and the algorithm error, calculation amount and calculation complexity are reduced.

针对上述议题，本发明提供了一种基于神经网络的RFID定位方法，通过计算基于随机位置的读卡器读取某一位置的移动物体位置数据的最小误差值。根据三种定位算法的优化方案进行定位计算，对定位计算的结果数据进行数据分析得出三种算法的误差数据，以该误差数据为建模对象，利用神经网络算法进行建模，得出最大出现可能的位置坐标以及对应时间。本发明集合LANDMARC算法、BVIRE算法和VIRE算法，结合神经网络算法，得到更加高效准确的结果，从而实现对RFID定位的精度提高。In view of the above issues, the present invention provides an RFID positioning method based on a neural network, by calculating the minimum error value of the position data of a moving object in a certain position read by a card reader based on a random position. The positioning calculation is carried out according to the optimization scheme of the three positioning algorithms, and the error data of the three algorithms is obtained by data analysis of the result data of the positioning calculation. Possible location coordinates and corresponding times appear. The invention integrates the LANDMARC algorithm, the BVIRE algorithm and the VIRE algorithm, and combines with the neural network algorithm to obtain more efficient and accurate results, thereby improving the accuracy of RFID positioning.

本发明的成果通过以下步骤实现来实现：The achievement of the present invention is realized through the following steps:

一种基于神经网络的RFID定位方法，包括读卡器、处理器，所述处理器通过程序代码实现如下步骤：A neural network-based RFID positioning method, comprising a card reader and a processor, wherein the processor implements the following steps through program codes:

定位算法模块读取读卡器坐标数据、参考标签坐标数据和待测点真实坐标数据，通过采用LANDMARC算法、BVIRE算法和VIRE算法获得待测点测量坐标和定位误差数据；The positioning algorithm module reads the coordinate data of the card reader, the coordinate data of the reference label and the real coordinate data of the point to be measured, and obtains the measured coordinates and positioning error data of the point to be measured by using the LANDMARC algorithm, the BVIRE algorithm and the VIRE algorithm;

误差分析模块提取待测点测量坐标和定位误差数据中的时间戳、误差主要特征构建训练集；The error analysis module extracts the time stamp and main error features in the measurement coordinates of the point to be measured and the positioning error data to construct a training set;

训练集通过网格加密和神经网络算法对时间戳、误差主要特征进行处理输出最小误差位置点坐标和最小误差位置点时间；The training set is processed by grid encryption and neural network algorithm to the main features of timestamp and error, and the coordinates of the minimum error position and the time of the minimum error position are output;

定位修正模块对最小误差位置点坐标和最小误差位置点时间进行修正输出待测点坐标修正值和对应最大出现可能的位置坐标。The positioning correction module corrects the coordinates of the minimum error position point and the time of the minimum error position point, and outputs the correction value of the coordinates of the point to be measured and the corresponding maximum possible position coordinates.

进一步，所述神经网络算法是将LANDMARC算法、BVIRE算法和VIRE算法的定位误差进行效率计算和数据稳定性评估后，有选择输入RNN、CNN、和LSTM神经网络算法中获得最小误差位置点坐标和最小误差位置点时间。Further, the neural network algorithm is to perform efficiency calculation and data stability evaluation on the positioning errors of the LANDMARC algorithm, the BVIRE algorithm and the VIRE algorithm, and then select the input RNN, CNN, and LSTM neural network algorithm to obtain the minimum error position point coordinates and Minimum error position point time.

进一步，所述定位修正模块对最小误差位置点坐标和最小误差位置点时间进行修正是通过误差最小值的数据筛选寻找对应的待测点测量坐标，将待测点测量坐标与待测点真实坐标通过移动物体的路径拟合进行数据分析，实现对待测点测量坐标的定位修正Further, when the positioning correction module corrects the minimum error position point coordinates and the minimum error position point time, it searches for the corresponding measurement coordinates of the point to be measured through the data screening of the minimum error value, and compares the measurement coordinates of the point to be measured with the real coordinates of the point to be measured. Perform data analysis through the path fitting of the moving object to realize the positioning correction of the measured coordinates of the point to be measured

有益效果beneficial effect

1、本发明提出了基于RFID室内定位算法的神经网络分析算法，提高了计算精确度和定位计算复杂度。1. The present invention proposes a neural network analysis algorithm based on the RFID indoor positioning algorithm, which improves the calculation accuracy and the positioning calculation complexity.

2、本发明提出了一种基于时间戳、误差值作为特征集的神经网络分析方法，利用反馈神经网络将定位问题转化为误差问题进行分析。与现有方法相比，该方法将多种算法进行结合，有效的节约了数据资源，不直接对原始数据分析，提高数据运算分析速度。2. The present invention proposes a neural network analysis method based on time stamps and error values as feature sets, and uses a feedback neural network to transform the positioning problem into an error problem for analysis. Compared with the existing method, the method combines multiple algorithms, effectively saves data resources, does not directly analyze the original data, and improves the speed of data operation and analysis.

附图说明Description of drawings

图1定位算法成果示意图。Figure 1. Schematic diagram of the results of the positioning algorithm.

图2为基于三种RFID定位算法与神经网络的分析方法的逻辑结构图；Fig. 2 is the logical structure diagram of the analysis method based on three kinds of RFID positioning algorithms and neural network;

图3为基于反馈神经网络的逻辑结构图。Figure 3 is a logical structure diagram based on a feedback neural network.

具体实施方式Detailed ways

以下结合附图，对本发明设计的数据分析、网络训练和数据计算详细说明如下。整个定位原理框图如图1、图2、图3所示。The data analysis, network training and data calculation designed by the present invention will be described in detail below with reference to the accompanying drawings. The whole positioning principle block diagram is shown in Figure 1, Figure 2 and Figure 3.

步骤1：利用三种定位算法模块进行数据分析，输入数据为读卡器坐标数据、参考标签坐标数据和待测点真实坐标数据，输出数据为待测点测量坐标和定位误差，定位结果示意图见下图1。Step 1: Use three positioning algorithm modules for data analysis. The input data are the coordinate data of the card reader, the coordinate data of the reference label and the real coordinate data of the point to be measured. The output data are the measured coordinates of the point to be measured and the positioning error. The schematic diagram of the positioning result is shown in Figure 1 below.

即，通过LANDMARC算法、BVIRE算法和VIRE算法三种定位算法模块进行并行计算，利用Matlab进行模拟仿真，将读卡器坐标数据、参考标签坐标数据和待测点真实坐标数据进行输入，通过计算参考标签与待测标签的接受信号强度实现定位计算，并将三种算法根据分布式原理同时开启计算，输出数据为待测点测量坐标和定位误差。That is, the three positioning algorithm modules of LANDMARC algorithm, BVIRE algorithm and VIRE algorithm are used for parallel calculation, and Matlab is used for simulation. The received signal strength of the tag and the tag to be tested realizes the positioning calculation, and the three algorithms are simultaneously started to calculate according to the distributed principle, and the output data are the measured coordinates of the point to be tested and the positioning error.

步骤2：误差分析模块利用神经网络算法进行网络训练，输入数据为三种算法的定位误差，同时输入RNN、CNN、和LSTM神经网络算法，输出数据为误差最小值及其所对应的时间戳。Step 2: The error analysis module uses the neural network algorithm for network training. The input data is the positioning error of the three algorithms, and the RNN, CNN, and LSTM neural network algorithms are input at the same time, and the output data is the minimum error value and its corresponding timestamp.

即，通过对定位误差数据进行预处理，利用网格计算和神经网络算法，将三种算法的定位误差进行输入，同时输入RNN、CNN、和LSTM神经网络算法，根据计算效率和数据稳定性选择神经网络算法，使用时间戳、误差值作为其中的主要特征集来进行训练，通过训练被选中的神经网络算法将输出定位误差的最小值和其出现所对应的时间戳。That is, by preprocessing the positioning error data, using grid computing and neural network algorithms, the positioning errors of the three algorithms are input, and the RNN, CNN, and LSTM neural network algorithms are input at the same time, and the selection is based on computational efficiency and data stability. The neural network algorithm uses the timestamp and the error value as the main feature set for training, and the selected neural network algorithm will output the minimum value of the positioning error and the timestamp corresponding to its occurrence.

步骤3：利用定位修正算法进行数据计算，输入数据为误差的最小值，输出数据为待测点坐标修正值和对应最大出现可能的位置坐标以及对应时间。Step 3: Use the positioning correction algorithm to calculate the data, the input data is the minimum value of the error, and the output data is the correction value of the coordinates of the point to be measured and the corresponding maximum possible position coordinates and the corresponding time.

即，通过对定位误差最小值的选择实现定位修正算法，通过误差最小值的数据筛选寻找对应的待测点测量坐标，将待测点测量坐标与待测点真实坐标通过移动物体的路径拟合进行数据分析，实现对待测点测量坐标的定位修正，输出待测点坐标修正值和对应最大出现可能的位置坐标。That is, the positioning correction algorithm is realized by selecting the minimum value of the positioning error, the corresponding measurement coordinates of the point to be measured are found through the data screening of the minimum value of the error, and the measurement coordinates of the point to be measured and the real coordinates of the point to be measured are fitted through the path of the moving object Carry out data analysis, realize the positioning correction of the measured coordinates of the point to be measured, and output the correction value of the coordinates of the point to be measured and the corresponding maximum possible position coordinates.

Claims

1. An RFID positioning method based on a neural network comprises a card reader and a processor, and is characterized in that: the processor realizes the following steps through program codes:

the positioning algorithm module reads the coordinate data of the card reader, the coordinate data of the reference label and the real coordinate data of the point to be measured, and obtains the measured coordinate and the positioning error data of the point to be measured by adopting a LANDMRC algorithm, a BVIRE algorithm and a VIRE algorithm;

an error analysis module extracts the time stamp and the main error characteristics in the measured coordinates of the point to be measured and the positioning error data to construct a training set;

the training set processes the time stamp and the error main characteristics through a grid encryption algorithm and a neural network algorithm to output a minimum error position point coordinate and a minimum error position point time;

and the positioning correction module corrects the coordinate of the minimum error position point and the time of the minimum error position point and outputs a coordinate correction value of the point to be measured and the corresponding position coordinate with the maximum occurrence possibility.

2. The neural network-based RFID location method of claim 1, wherein:

the neural network algorithm is characterized in that after the positioning errors of the LANDMARC algorithm, the BVIRE algorithm and the VIRE algorithm are subjected to efficiency calculation and data stability evaluation, the positioning errors are selectively input into the RNN, CNN and LSTM neural network algorithms to obtain the coordinates of the minimum error position point and the time of the minimum error position point.

3. The neural network-based RFID location method of claim 1, wherein:

the positioning correction module corrects the coordinate of the minimum error position point and the time of the minimum error position point, searches the corresponding measured coordinate of the point to be measured through data screening of the minimum error value, performs data analysis on the measured coordinate of the point to be measured and the real coordinate of the point to be measured through path fitting of a moving object, and realizes positioning correction of the measured coordinate of the point to be measured.