High-precision sensor array data interpolation fitting method and system based on time sequence prediction

Abstract

The invention discloses a high-precision sensor array data interpolation fitting method and system based on time sequence prediction. The method comprises the steps of obtaining sensor time sequence data before a to-be-detected time period, within the to-be-detected time period and after the to-be-detected time period; performing prediction processing on the sensor array data before and after the to-be-measured time period by using a forward and reverse time sequence expansion algorithm, and expanding the time sequence into the to-be-measured time period to obtain expanded sensor array data in the measuring line; performing interpolation processing on the expanded sensor array data in the measuring line to obtain interpolated sensor array data in the measuring line; and performing polynomial fitting on the interpolated sensor array data in the measuring line, selecting a physical field function corresponding to the optimal polynomial fitting frequency as an optimal physical field function in the measuring line, and obtaining a physical field change trend in the measuring line. The method is stable in function and good in practicability, and the fitting precision of the sensor array in the physical field fitting process can be remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-precision physical field fitting of sensors, and relates to a sensor array data interpolation fitting method and system, and in particular to a high-precision sensor array data interpolation fitting method and system based on time series prediction. Background Art

[0002] In recent years, with the increasing requirements for data accuracy in smart agriculture, industrial production, environmental monitoring and other fields, how to improve the accuracy of sensor fitting to physical fields is the current focus. Especially in the field of environmental monitoring, with the development of advanced surveying and mapping technologies such as total stations, rangefinders, and laser ranging technology, new options have been provided for the ranging field, and these devices have gradually replaced traditional measurement methods. These advanced measurement devices have significantly improved measurement accuracy, but at the same time, the sensitivity of their measurement results to fluctuations in atmospheric environmental parameters has also increased. Therefore, realizing a high-precision sensor array data interpolation fitting method is the key to further improving ranging accuracy in the current ranging field.

[0003] Due to the complexity of the ranging environment, the method of obtaining the atmospheric refractive index during the ranging process mainly relies on the atmospheric environment parameter measurement method, that is, directly obtaining the physical field environment parameters within the measurement line through high-precision sensor array equipment. This method is simple in principle and has an accuracy of up to 10 -8 Theoretically, it can meet the distance measurement accuracy requirements in most cases. However, due to the atmospheric parameter jitter, parameter distribution unevenness and insufficient atmospheric parameter processing algorithm in the actual environment, the accuracy of long-distance measurement is only 10 -7 The order of magnitude is still difficult.

[0004] In response to this situation, in recent years, many people have studied the multi-point calibration method based on the classic two-point method of data acquisition, that is, to improve the accuracy of the physical field within the measurement line by increasing the density of the sensor array. However, due to the influence of terrain in different ranging environments and the cost constraints of high-precision sensors and auxiliary measurement equipment, the feasibility of improving the accuracy of environmental parameter measurements by increasing the accuracy or number of sensors is limited.

[0005] Therefore, in response to the current demand for high-precision sensor array data interpolation and fitting methods and the defects of existing methods, the present invention proposes a high-precision sensor array data interpolation and fitting method based on time series prediction. This method expands the sensor array based on forward and reverse time series analysis and prediction, and fuses the hybrid proxy model of polynomial fitting through radial basis interpolation, so that the physical field fitted by the sensor array is more accurate. Summary of the invention

[0006] To solve the above problems, the present invention provides a high-precision sensor array data interpolation fitting method and system based on time series prediction. The sensor array data processing method is based on forward and reverse time series prediction technology to achieve the expansion of the sensor array. At the same time, high-precision physical field data is obtained by interpolating and fitting the expanded sensor array data. It has been verified in practice that the scheme is feasible and has good practicality.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A high-precision sensor array data interpolation fitting method based on time series prediction includes the following steps:

[0009] S1. Obtain sensor time series data before, during and after the time period to be measured;

[0010] S2. Use the forward and reverse time series expansion algorithm to predict the sensor array data before and after the time period to be measured, so that the time series is expanded to the time period to be measured, and the sensor array data within the expanded measurement line is obtained;

[0011] S3. performing interpolation processing on the expanded sensor array data within the survey line to obtain interpolated sensor array data within the survey line;

[0012] S4. Perform polynomial fitting on the sensor array data within the measurement line after interpolation, select the physical field function corresponding to the best polynomial fitting number as the best physical field function within the measurement line, and obtain the change trend of the physical field within the measurement line.

[0013] Furthermore, step S1 specifically includes: collecting time series data of sensors in the measurement line in three time periods: before the time period to be measured, during the time period to be measured, and after the time period to be measured, and the sensors in the measurement line are at different points in the three time periods.

[0014] Furthermore, step S2 is specifically as follows: using a forward time series expansion algorithm to predict the sensor time series data before the time period to be measured, and expanding the time series to the time period to be measured; using a reverse time series expansion algorithm to transpose and predict the sensor time series data after the time period to be measured, and transpose the processed time series again to expand it to the time period to be measured, so that the data at different positions are expanded to the same time dimension through time series prediction, thereby expanding the number of sensor observation points at the same time in the measurement line to three times the original.

[0015] Furthermore, in step S3, the radial basis function interpolation method is used to interpolate the sensor array data in the expanded measurement line, so that the data in the measurement line is denser, which is convenient for subsequent polynomial fitting of the physical field.

[0016] Furthermore, step S4 is specifically as follows: perform polynomial fitting on the sensor array data within the survey line after interpolation, determine the change and trend of the physical field function on the survey line, and compare the fitted physical field with the actual sensor time series data to obtain the root mean square error under different polynomial fitting times, compare the root mean square error values ​​under different polynomial fitting times, select the physical field function corresponding to the polynomial fitting times that makes the root mean square error the lowest as the best physical field function in the survey line, and obtain the change trend of the physical field in the survey line.

[0017] A high-precision sensor array data interpolation fitting system based on time series prediction, comprising:

[0018] Data acquisition module: used to obtain sensor time series data before, during and after the time period to be measured;

[0019] Data expansion module: used to use the forward and reverse time series expansion algorithm to predict and process the sensor array data before and after the time period to be measured, so as to expand the time series to the time period to be measured and obtain the expanded sensor array data in the measurement line;

[0020] Data interpolation module: used to perform interpolation processing on the sensor array data in the expanded survey line to obtain the interpolated sensor array data in the survey line;

[0021] Polynomial optimization fitting module: used to perform polynomial fitting on the sensor array data within the measurement line after interpolation, select the physical field function corresponding to the best polynomial fitting number as the best physical field function within the measurement line, and obtain the change trend of the physical field within the measurement line.

[0022] A computer device, comprising:

[0023] one or more processors;

[0024] A memory for storing one or more programs;

[0025] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned high-precision sensor array data interpolation fitting method based on time series prediction.

[0026] A computer-readable storage medium storing computer instructions, when the computer instructions are executed by one or more processors, causes the one or more processors to perform the steps in the above method.

[0027] The beneficial effects of the present invention are:

[0028] 1) An algorithm for expanding the density of sensor arrays based on forward and reverse time series. The sensor array is tested within the required fitting time period, and the sensor array is placed at different points before and after the time period. The data is expanded to the required time period through forward and reverse time series expansion. This deep learning strategy improves the density of sensor array observation points while ensuring that the number of sensors does not increase.

[0029] 2) Physical field fitting based on radial basis interpolation fusion polynomials. On the basis of expanding the sensor array through forward and reverse time series, the sensor data in the measurement line is further expanded through radial basis interpolation to facilitate the subsequent polynomial fitting of the physical field.

[0030] 3) Optimization strategy for polynomial fitting of sensor array data. Based on the radial basis interpolation of sensor array data to expand the observed data points, the root mean square error data of the physical field is fitted according to different polynomial orders, and the best polynomial fitting parameters are selected through the optimization algorithm to obtain a higher-precision physical field fitting result.

[0031] The present invention is aimed at the demand for a high-precision sensor array data interpolation fitting method based on time series prediction, and performs a practical design of a sensor array data processing method. After the sensor array data processing method is implemented, a functional test is performed. It can be found through actual testing that the high-precision sensor array data interpolation fitting method based on time series prediction of the present invention and its function are feasible and have good practicality, and can excellently complete the task of high-precision sensor array data processing in the baseline field. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The figure is an overall hardware block diagram of a high-precision sensor array data interpolation fitting method based on time series prediction in an embodiment of the present invention.

[0033] Figure 2 It is a software functional block diagram of a high-precision sensor array data interpolation fitting method based on time series prediction in an embodiment of the present invention.

[0034] Figure 3 This is a diagram of the sensor array model for forward and reverse time series expansion in an embodiment of the present invention.

[0035] Figure 4 Schematic diagram of the sensor array layout in an embodiment of the present invention.

[0036] Figure 5 Schematic diagram of expanding observation points of the sensor array in an embodiment of the present invention.

[0037] Figure 6 It is a flow chart of the algorithm of the interpolation fitting optimization module in an embodiment of the present invention.

[0038] Figure 7 Schematic diagram of reverse time series partitioning of a data set in an embodiment of the present invention.

[0039] Figure 8 Flow chart of the reverse time series partitioning algorithm in an embodiment of the present invention.

[0040] Fig. 9 This is a measured diagram of the hardware of the sensor array compensation system in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings and specific examples.

[0042] A high-precision sensor array data interpolation fitting method based on time series prediction is used in technologies in fields with high requirements for physical field fitting accuracy, such as smart agriculture, industrial production, and environmental monitoring, and includes the following steps:

[0043] S1. Obtain sensor time series data before, during and after the time period to be measured;

[0044] S2. Use the forward and reverse time series expansion algorithm to predict the sensor array data before and after the time period to be measured, so that the time series is expanded to the time period to be measured, and the sensor array data within the expanded measurement line is obtained;

[0045] S3. performing interpolation processing on the expanded sensor array data within the survey line to obtain interpolated sensor array data within the survey line;

[0046] S4. Perform polynomial fitting on the sensor array data within the measurement line after interpolation, select the physical field function corresponding to the best polynomial fitting number as the best physical field function within the measurement line, and obtain the change trend of the physical field within the measurement line.

[0047] Step S1 specifically includes: collecting time series data of sensors in the measurement line in three time periods before the time period to be measured, in the time period to be measured, and after the time period to be measured, and the sensors in the measurement line are at different points in the three time periods.

[0048] Step S2 is specifically as follows: a forward time series expansion algorithm is used to predict the sensor time series data before the time period to be measured, and the time series is expanded to the time period to be measured; a reverse time series expansion algorithm is used to transpose and predict the sensor time series data after the time period to be measured, and the processed time series is transposed again to expand it to the time period to be measured, so that the data at different positions are expanded to the same time dimension through time series prediction, thereby expanding the number of sensor observation points at the same time in the measurement line to three times the original.

[0049] In step S3, the radial basis function interpolation method is used to interpolate the sensor array data in the expanded measurement line, so that the data in the measurement line is denser, which is convenient for subsequent polynomial fitting of the physical field.

[0050] Step S4 is specifically as follows: perform polynomial fitting on the sensor array data within the survey line after interpolation, determine the change and trend of the physical field function on the survey line, and compare the fitted physical field with the actual sensor time series data to obtain the root mean square error under different polynomial fitting times, compare the root mean square error values ​​under different polynomial fitting times, select the physical field function corresponding to the polynomial fitting times that makes the root mean square error the lowest as the best physical field function in the survey line, and obtain the change trend of the physical field in the survey line.

[0051] Example

[0052] The high-precision sensor array data interpolation fitting method based on time series prediction in this embodiment has been applied to a baseline field atmospheric compensation correction system of a micro-satellite center of a university.

[0053] like Figure 1 As shown, the hardware of the high-precision sensor array data interpolation fitting method based on time series prediction of the present invention includes: a high-precision sensor unit, which is used to measure sensor data at different points of the measurement line; and a data communication module, which is used for communication between the host computer and the high-precision sensor unit.

[0054] like Figure 2 As shown in the figure, the algorithm for processing sensor array data in the host computer mainly includes: forward and reverse time series expansion and interpolation fitting algorithm. First, the sensor array data of the sensor before the time period to be measured, during the time period to be measured, and after the time period to be measured are processed separately. The data before the time period to be measured is expanded to the time period to be measured by the forward time series expansion algorithm, and the data after the time period to be measured is expanded to the time period to be measured by the reverse time series expansion algorithm. Through this forward and reverse time series expansion algorithm, the actual sensor array observation points in the measurement line are expanded to three times; after completing the forward and reverse time series expansion, the appropriate radial basis interpolation function and polynomial degree fitting range are selected, and the physical field changes in the measurement line are fitted through the interpolation fitting agent model; at the same time, according to different interpolation fitting agent models, their interpolation functions and polynomial degrees are different, and the root mean square errors of the corresponding observation points and fitting functions are different. According to the change trend of the root mean square error, the optimization algorithm is used to finally select the appropriate physical field function.

[0055] like Figure 3As shown, the sensor array data processing method of the present invention adopts a forward and reverse time series to expand the sensor array model, and the construction of its neural network architecture is realized by Python. Taking the three-point sensor as an example, the data of the model in the time period t2 to t3 is the expanded data set, the time period t1 to t2 is the advanced data, and the time period t3 to t4 is the lagging data. The time series of sensors 1 to 3 are expanded at t1 to t2, and the time series of sensors 1' to 3' are expanded at t3 to t4. Both time series are learned and expanded to the time period t2 to t3 through the neural network model. In this way, data collection is achieved in the two time periods t1 to t2 and t3 to t4 when the sensor does not need data collection, thereby achieving the purpose of expanding the sensor array data in the time period t2 to t3.

[0056] like Figure 4 As shown, the sensor array of the sensor array data processing method of the present invention needs to change the sampling points in the three time periods of t1-t2, t2-t3, and t3-t4. When changing the sensor array points, it is necessary to ensure that the sensors are evenly arranged after expansion, and the sensor array should minimize the displacement time of the sensors when shifting. The time period of t1-t2 is the first data measured in the entire time dimension. After this period of time, the sensor is shifted. The data measured in t2-t3 is the time period data to be processed by the high-precision sensor array data interpolation fitting method based on time series prediction. After completing the measurement of this period of time, the sensor continues to be shifted. The time period of t3-t4 after this period of time is the last measured data of the sensor array, that is, the reverse time series data segment in the entire method. This data is used to invert the data of the time period of t2-t3 at this point.

[0057] like Figure 5 As shown, the sensor array data processing method of the present invention can eventually expand the sensor array data of three different dimensions to the same time dimension through forward and reverse time series expansion, so that the observation points of the physical field in the survey line are more dense. The data of sensors 1 to 3 are used for forward time series processing, the time series data of sensors I to III are temporarily stored, waiting for the forward and reverse time series data processing of the remaining point sensors to be completed, and the data of sensors 1' to 3' are used for reverse time series processing. After processing, the sensor array data of these three positions are expanded to the same time dimension, and the number of sensor array observation points is increased while keeping the number and accuracy of sensors unchanged.

[0058] like Figure 6As shown, the interpolation fitting algorithm of the sensor array data processing method of the present invention requires an optimization module to select a suitable fitting physical field. The core of this method is to use the root mean square error of different functions after interpolation fitting for screening. Before interpolation fitting, the sensor array data after the forward and reverse time series expansion must first be imported into the radial basis interpolation function to generate a mapping function. According to the physical field dimension, discrete observation points after interpolation expansion are generated. After the radial basis interpolation is completed, the variable polynomial fitting order is set to i, and loss(i-1)=inf is defined. The discrete points after interpolation expansion are imported into the i-th order polynomial for fitting, and the loss value of the polynomial fitting function and the discrete points is calculated. The optimal physical field fitting polynomial is obtained according to the loss value results of different polynomial orders. After the interpolation fitting algorithm is combined with the forward and reverse time series algorithm flow, the steps for implementing the overall method are described in detail as follows:

[0059] 1) This sensor array timing prediction method requires that the sensors be arranged at other points before and after the time period to be measured, and the time series data of the three periods before, during and after the time period to be measured are transmitted to the host computer through the data communication module.

[0060] 2) Using the sensor data before the test time, the host computer uses the forward time series expansion algorithm to analyze and predict and expand it to the test time; after the sensor data after the test time is transposed, the host computer uses the reverse time series expansion algorithm to analyze and predict and expand it to the test time.

[0061] 3) After the sensor array data is processed through forward and reverse time series, the sensor array data within the survey line is expanded, so that the number of sensor observation points within the survey line at the same time is expanded to three times the original.

[0062] 4) Perform interpolation processing on the sensor array after forward and reverse time series processing to make the data in the measurement line more dense, which is convenient for subsequent polynomial fitting of the physical field.

[0063] 5) Perform polynomial fitting on the sensor array data after interpolation processing, and compare the fitted physical field with the actual measurement point data to obtain the root mean square error under different polynomial fitting orders.

[0064] 6) Compare the changes in the root mean square error values ​​under different polynomial fitting times, select the polynomial fitting times with the lowest root mean square error, and obtain the optimal physical field function within the measurement line under this condition.

[0065] like Figure 7 As shown in FIG. 1 , the core of the reverse time series expansion algorithm of the present invention is to reverse the data according to the reverse time series. s As the dividing point, t sThe subsequent observation points are used as the training set, and the previous observation points are used as the test set. The deep learning model after training and testing can be applied to the actual set for time series prediction experiments.

[0066] like Figure 8 As shown in the figure, the biggest difference between the reverse time series expansion algorithm of the present invention and the forward time series expansion algorithm is that the original time series needs to be transposed and further processed. First, the input forward time series X and the serialization ratio value γ are determined, and the division time t is initialized according to the forward time series and the serialization ratio value. s =floor(γ*X), and perform reverse processing on the time series X=reverse(X), according to the divided time t s The time series is divided into training set and test set, and the sliding window N is set according to the range of the training set and the test set, and finally the reverse time series test set and training set after the sliding window processing are obtained. After the reverse time series transposition algorithm is processed, the reverse time series can be expanded to the specified time period to be tested like the forward time series.

[0067] like Fig. 9 The figure shows the hardware measurement diagram of the sensor array compensation system of the present invention. In order to conduct hardware measurement on the function and performance of the high-precision sensor array data interpolation fitting method based on time series prediction of the present invention, the hardware includes a ranging system mainly composed of a transmitting antenna and a receiving antenna, a high-precision sensor unit and a data communication module. The present invention uses the communication technology of the data communication module to transmit the sensor data of all time periods to the host computer, and uses the Python language to implement the software function of the high-precision sensor array data interpolation fitting method based on time series prediction in the host computer. The high-precision sensor array data interpolation fitting method based on time series prediction of the present invention, through the forward and reverse time series, expands the sensor array of the same time dimension to three times without changing the number and accuracy of the sensors; the accuracy and stability of the physical field fitting during ranging are further improved by the radial basis interpolation fusion polynomial fitting method. In order to verify the feasibility and performance of the present invention, the Monte Carlo method is used to simulate and observe the effect of the high-precision sensor array data interpolation fitting method based on time series prediction for many times. After verification, taking the temperature field in meteorological parameters as an example, the simulation results show that the high-precision sensor array data interpolation fitting method based on time series prediction has improved the accuracy of environmental parameters by 71.8% and reduced the standard deviation by 73.1%. The results of Monte Carlo simulation analysis further prove that this method has stronger stability than the traditional multi-point method.

[0068] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 Process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0070] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0072] The above specific implementation modes are used to explain the present invention rather than to limit the present invention. Any modification and change made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A high-precision sensor array data interpolation fitting method based on time series prediction, characterized in that: The following steps are involved: S1. Obtain sensor time series data before, during and after the time period to be measured; S2. Use the forward and reverse time series expansion algorithm to predict the sensor array data before and after the time period to be measured, so that the time series is expanded to the time period to be measured, and the sensor array data within the expanded measurement line is obtained; S3. performing interpolation processing on the expanded sensor array data within the survey line to obtain interpolated sensor array data within the survey line; S4. Perform polynomial fitting on the sensor array data within the measurement line after interpolation, select the physical field function corresponding to the best polynomial fitting number as the best physical field function within the measurement line, and obtain the change trend of the physical field within the measurement line.

2. The high-precision sensor array data interpolation fitting method based on time series prediction according to claim 1 is characterized in that: Step S1 specifically includes: collecting time series data of sensors in the measurement line in three time periods before the time period to be measured, in the time period to be measured, and after the time period to be measured, and the sensors in the measurement line are at different points in the three time periods.

3. The high-precision sensor array data interpolation fitting method based on time series prediction according to claim 1 is characterized in that: Step S2 is specifically as follows: using a forward time series expansion algorithm to predict the sensor time series data before the time period to be measured, and expanding the time series to the time period to be measured; using a reverse time series expansion algorithm to transpose and predict the sensor time series data after the time period to be measured, and transpose the processed time series again to expand it to the time period to be measured.

4. The high-precision sensor array data interpolation fitting method based on time series prediction according to claim 1 is characterized in that: In step S3, the radial basis function interpolation method is used to interpolate the sensor array data in the expanded measurement line.

5. The high-precision sensor array data interpolation fitting method based on time series prediction according to claim 1 is characterized in that: Step S4 is specifically as follows: perform polynomial fitting on the sensor array data within the survey line after interpolation, and compare the fitted physical field with the actual sensor time series data to obtain the root mean square error under different polynomial fitting times, compare the root mean square error values ​​under different polynomial fitting times, select the physical field function corresponding to the polynomial fitting times that makes the root mean square error the lowest as the optimal physical field function within the survey line, and obtain the change trend of the physical field within the survey line.

6. A high-precision sensor array data interpolation fitting system based on time series prediction, characterized in that: include: Data acquisition module: used to obtain sensor time series data before, during and after the time period to be measured; Data expansion module: used to use the forward and reverse time series expansion algorithm to predict and process the sensor array data before and after the time period to be measured, so as to expand the time series to the time period to be measured and obtain the expanded sensor array data in the measurement line; Data interpolation module: used to perform interpolation processing on the sensor array data in the expanded survey line to obtain the interpolated sensor array data in the survey line; Polynomial optimization fitting module: used to perform polynomial fitting on the sensor array data within the measurement line after interpolation, select the physical field function corresponding to the best polynomial fitting number as the best physical field function within the measurement line, and obtain the change trend of the physical field within the measurement line.

7. A computer device, characterized in that: The computer device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the high-precision sensor array data interpolation fitting method based on time series prediction as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing computer instructions, characterized in that: When the computer instructions are executed by one or more processors, the one or more processors are caused to execute the steps in the method according to any one of claims 1 to 5.

Images