CN111830451A - A kind of inspection method of non-imaging sensor - Google Patents

Abstract

A non-imaging sensor inspection method, comprising: acquiring non-imaging sensor detection data, including to-be-analyzed data and background data, and preprocessing the acquired data, wherein the to-be-analyzed data and the background data are detection data of the same area; Perform grid processing on the preprocessed data to be analyzed and background data to obtain time series data to be analyzed and time series background data; filter and normalize the time series data to be analyzed and time series background data; calculate the normalized time series The average value of the difference, the standard deviation of the difference and the correlation coefficient between the data to be analyzed and the time series background data; check whether the non-imaging sensor works normally according to the calculation result. The method is simple, efficient and versatile. It can realize the on-orbit detection of non-imaging sensors when the external field conditions are insufficient and the standard target is limited.

Description

A kind of inspection method of non-imaging sensor

technical field

The invention relates to the field of satellite monitoring, in particular to a non-imaging sensor inspection method.

Background technique

Electromagnetic monitoring satellites are usually equipped with a variety of space electromagnetic detection sensors, which can provide space detection data support for space environment monitoring, earthquake prediction and prediction and earth science research. Due to the influence of radiation from the space environment, the working state and physical properties of the sensor will change during the satellite operation cycle, which in turn affects the accuracy of the sensor detection data and its reliability in various applications. Therefore, periodic tracking of sensor performance is a necessary link to ensure the efficient operation of electromagnetic satellite sensor detection.

At present, due to the limitations of satellite sensors, space, energy consumption, technology, etc., electromagnetic monitoring satellites do not yet have on-board calibration capabilities. The detection methods for non-imaging sensors are still mostly laboratory detection and limited field tests. Supplemented, such as the DEMETER satellite. The former simulates the space environment in the laboratory to simulate the detection of the sensor, and cannot transmit the detection results to meet the needs of the non-imaging sensor performance change detection after the satellite is in operation. The latter uses the test field under specific outdoor background conditions to evaluate the operation status of the on-board sensors through real observation data, and can compensate for various drift deviations generated in the actual work of the sensors. The accuracy of electromagnetic satellite space observations. However, using the test field to carry out inspection has strict requirements on the site and the measurement process, and the field measurement requires a lot of manpower and material resources, which cannot meet the high-frequency detection and analysis requirements of non-imaging sensors.

In recent years, for imaging sensors, a field-free sensor detection technology, namely cross-calibration, has been developed. This technology does not require a ground test site, and can perform on-orbit detection and analysis between remote sensing imaging sensors using the same standard target on the ground. . However, the electromagnetic monitoring satellite sensor is oriented to the observation of discrete points in space, and the data sampling frequency is high, which is different from conventional satellite data and does not have imaging characteristics. Therefore, it is necessary to design a sensor on-orbit inspection method based on the detection data itself based on the detection characteristics of non-imaging sensors of electromagnetic monitoring satellites, and use its high-frequency discrete point observation characteristics to realize the periodicity of electromagnetic detection sensors during satellite operation. , High-frequency performance testing and analysis capabilities.

SUMMARY OF THE INVENTION

(1) Technical problems to be solved

In view of the existing technical problems, the present invention provides a non-imaging sensor inspection method, which is used to at least partially solve the above technical problems.

(2) Technical solutions

The present invention provides a non-imaging sensor inspection method, comprising:

S1: Acquire the data to be analyzed and the background data detected by the non-imaging sensor, and preprocess the data to be analyzed and the background data, wherein the data to be analyzed and the background data are the detection data of the same area; S2: The preprocessed data to be analyzed Perform grid processing with the background data to obtain the time series data to be analyzed and the time series background data; S3: filter and normalize the time series data to be analyzed and the time series background data; S4: calculate the normalized time series data to be analyzed and The difference value average value, the difference value standard deviation and the correlation coefficient of the time series background data; S5: Check whether the non-imaging sensor works normally according to the difference value average value, the difference value standard deviation and the correlation coefficient.

Optionally, the storage format of the data to be analyzed and the background data is a text file format. In step S1, the preprocessing of the data to be analyzed and the background data includes: S1-1: read and write the data to be analyzed and the background data, extract " Time", "track number", "longitude", "latitude", and "detection value" key parameters to form a pure data file; S1-2; abnormal data elimination processing is performed on the data to be analyzed and the background data.

Optionally, performing abnormal data elimination processing on the detection data includes: delimiting magnetic storm magnetic disturbance conditions according to two geomagnetic indices, Kp index and Dst index; and removing abnormal data in the data to be analyzed and background data according to the magnetic storm magnetic disturbance conditions.

Optionally, S2 includes: S2-1: According to the characteristics of the satellite reentry cycle and the sampling frequency of the non-imaging sensor, perform spatial grid and time window settings on the preprocessed data to be analyzed, wherein, in the spatial grid, the time window Move on the time axis; S2-2: For the time window corresponding to the time on the time axis in the spatial grid, average the data to be analyzed in the time window to obtain the current spatial grid and the current time period. Average the data to be analyzed; S2-3: Repeat step S2-2, for the preprocessed data to be analyzed, sequentially calculate the average value of the data to be analyzed in the time window corresponding to each time on the time axis to obtain a plurality of average data to be analyzed, consisting of Time series data to be analyzed; S2-4: Repeat steps S2-1 to S2-3 to process the preprocessed background data to obtain time series background data.

对时序待分析数据和时序背景数据进行归一化，其中，x和x^*分别为归一化前后的时序待分析数据和时序背景数据，x_max和x_min分别为时序待分析数据和时序背景数据对应最大值和最小值。Optionally, S3 includes: S3-1: Use at least one method in convolution smoothing or simple moving average or linear weighted smoothing or Gaussian filtering smoothing or Butterworth filtering smoothing to filter the time series data to be analyzed and the time series background data ;S3-2: Use the mapping function

Normalize the time series data to be analyzed and the time series background data, where x and x ^* are the time series data to be analyzed and the time series background data before and after normalization, respectively, and x _max and x _min are the time series data to be analyzed and the time series background, respectively The data corresponds to the maximum and minimum values.

计算归一化后的时序待分析数据，得到待分析差值平均值μ_analysis；利用公式

计算归一化后的时序待分析数据，得到待分析相关系数R_analysis；Optionally, the time series data to be analyzed and the time series background data both include electron concentration data and ion concentration data, and S4 includes: S4-1: using the formula

Calculate the normalized time series data to be analyzed, and obtain the average value of the difference to be analyzed μ _analysis ; use the formula

Calculate the normalized time series data to be analyzed to obtain the correlation coefficient R _analysis to be analyzed;

计算归一化后的时序背景数据，得到背景差值平均值μ_bg；利用公式

计算归一化后的时序背景数据，得到背景差值标准差σ_bg；利用公式

计算归一化后的时序背景数据，得到背景相关系数R_bg；S4-2: Utilize formula

Calculate the time series background data after normalization, obtain the background difference mean value μ _bg ; use the formula

Calculate the normalized time series background data to obtain the background difference standard deviation σ _bg ; use the formula

Calculate the normalized time series background data to obtain the background correlation coefficient R _bg ;

Among them, X and Y represent electron concentration data and ion concentration data, respectively, μ _XY is the average value of the difference between the two, σ _XY is the standard deviation of the difference between the two, Z is the difference between the two, and R _XY is the two The correlation coefficient of , Cov(X, Y) is the covariance of the two, D(X), D(Y) are the variance of the two respectively.

Optionally, the average value of the difference to be analyzed, the correlation coefficient to be analyzed, the average value of the background difference, the standard deviation of the background interpolation, and the background correlation coefficient meet the conditions:

μ _bg -σ _bg ≤μ _analysis ≤μ _bg +σ _bg

R _analysis ≥R _bg

Then it is determined that the non-imaging sensor is working normally, otherwise it is prompted that the non-imaging sensor is abnormal.

(3) Beneficial effects

The invention provides a non-imaging sensor inspection method, which is simple, efficient and universal, and can realize on-orbit detection of the non-imaging sensor under the condition of insufficient external field conditions and limited standard targets. At the same time, considering the characteristics of the sensor to be tested, the sensor operating state analysis can be carried out directly by using its own detection data at a high frequency.

Description of drawings

FIG. 1 is a schematic diagram of a non-imaging sensor inspection method based on time series data according to an embodiment of the present invention.

FIG. 2 is a flowchart of a non-imaging sensor inspection method based on time series data according to an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

The embodiment of the present invention proposes a non-imaging sensor on-orbit inspection method based on time series data. The method only uses the sensor's own detection data on the same satellite platform during the quiet period of the space environment, does not need to build a ground test field, and is not subject to detection conditions, manpower Constrained by factors such as material resources, detection and analysis can be carried out anytime, anywhere. The problems of the existing non-imaging sensor detection technology solutions, such as the limited layout of the test field, the lack of standard targets, and the inability to transmit laboratory test results, are solved. Moreover, by directly based on the time-series detection data of the sensors on the same satellite platform, the influence of insufficient number of satellites in orbit and low coverage overlap rate of the same type of satellites can be effectively reduced to meet the needs of practical applications.

The basic principle of the method is shown in Figure 1. First, the data detected by the non-imaging sensors (plasma analyzer and Langmuir probe) on the same satellite platform, namely the data to be analyzed and the background data (basic data) are acquired. The data detected by the two sensors on the DEMETER (Detection of Electro-Magnetic Emission Transmitted from Earthquake Regions, DEMETER) satellite that has been stably operated and matured in application obtained in this embodiment, secondly, the data to be analyzed and the background data are preprocessed, and Eliminate abnormal data. Thirdly, grid processing is performed on the data to be analyzed and the background data, and the grid point data is converted into gridded time series data, thereby obtaining the time series data to be analyzed and the time series background data. Then, the time series data to be analyzed and the time series background data are smoothed and normalized. Finally, on the basis of the background field data, correlation and threshold analysis are carried out, and through correlation boundary processing, the rationality of the fluctuation of the data to be analyzed can be directly detected, so as to analyze the performance status of the sensors on the same platform in orbit.

FIG. 2 is a flowchart of a non-imaging sensor inspection method based on time series data according to an embodiment of the present invention.

As shown in Figure 2, the non-imaging sensor inspection method may include the following steps:

S1, acquire the data to be analyzed and the background data detected by the non-imaging sensor, and preprocess the data.

In the above operation S1, based on the two non-imaging sensors of the Langmuir probe and the plasma analyzer of the DEMETER satellite platform, the current detection data in the same area/point is selected as the data to be analyzed; Historical detection data, as background data. According to the data storage characteristics of the non-imaging sensor, the data to be analyzed and the background data are selected from the electron concentration data and ion concentration data of the night segment.

The data obtained above are directly detected by non-imaging sensors, belong to Level-1 scientific data, and belong to the text file format. In order to facilitate subsequent operations, data reading processing is required to form a pure data file, and the pure data to be analyzed and the data are obtained. Background data. The specific reading operation is: use matlab software to write a text file reading and writing program, according to the Level-1 scientific data format, use the fscanf statement to read the directly detected data, and use fprintf to convert the "time", "track number", "latitude and longitude", Data such as "probe value" are written to text.

Since changes in the space environment such as magnetic storms and magnetic disturbances will cause abnormal disturbances of space electrons and space ions, the following abnormal data may exist in the acquired data to be analyzed and background data. Therefore, these abnormal data need to be eliminated. In this embodiment, two geomagnetic indices, the Kp index (Kp index) and the Dst index (Dst index), are used to eliminate data in a special space environment. Specific method of exclusion conditions: Based on Kp≥4 and Dst≤-50nT, the magnetic disturbance conditions of the magnetic storm are delineated, and the detection data obtained by the sensor during this time period are eliminated, so as to remove the ionospheric disturbance caused by the abnormal space environment, so as to analyze the ionization in the quiet period. Timing variation characteristics of layer parameters. Among them, the geomagnetic index data can be downloaded from the Japan World Geomagnetic Data Center website (http://wdc.kugi.kyoto-u.ac.jp/wdc/Sec3.html).

S2: Perform grid processing on the preprocessed data to be analyzed and the background data, respectively, to obtain time-series data to be analyzed and time-series background data.

It is mentioned in the above operation S1 that both the data to be analyzed and the background data are selected from the electron concentration data and the ion concentration data of the night period, that is, the data to be analyzed and the background data both include electron concentration data and ion concentration data.

Since the detection data of non-imaging sensors is collected in milliseconds and stored in the form of a single track, the data sampling frequency is high, and the sampling frequency of points between different sensors may be inconsistent, which cannot be matched one-to-one. Therefore, it is necessary to set a spatial grid, and perform grid processing on the data to be analyzed and the background data.

The grid processing method is as follows: according to the characteristics of the satellite reentry cycle and the sampling frequency of the sensor, a spatial grid of N°×M° and a time window of X days are set, wherein, in the spatial grid, the time window is on the time axis move. Calculate the average value of the point detection data (data to be analyzed and background data) within the current time period of the spatial grid, and use the average value as the detection data of the grid in the current time period; move the time window on the time axis , continue to obtain the average value of the data within X days of the grid, as the detection data of the next time period of the grid; and so on, use this method to perform a time series network on the data to be analyzed and the background data in the set spatial grid. Grid processing to obtain time series data to be analyzed and time series background data. In this embodiment, a 2°×2° spatial grid is used, and a time window is set with a time scale of 30 days (one month), and the data is aggregated to obtain time series data to be analyzed and time series background data.

S3, filtering and normalizing the time series data to be analyzed and the time series background data.

In order to reduce the influence of random fluctuation of data on the analysis and test results, it is necessary to perform smoothing filtering on the time series data to be analyzed and the time series background data respectively. The time series data to be analyzed and the time series background data may be filtered by at least one method of convolution smoothing, simple moving average, linear weighted smoothing, Gaussian filtering smoothing, or Butterworth filtering smoothing. This embodiment adopts the Savitzky-Golay filtering algorithm, which can smooth the time series data to be analyzed and the time series background data under the condition that the change trend and characteristics of the time series data remain unchanged. The specific formula is:

为拟合值，X_j+i为原始值，C_i为时序化背景数据(时序化待分析数据)中的第i个数据时的系数，m为半个滤波窗口，N为滤波器长度，N≤2m+1；T_k为第k次迭代后的序列拟合结果指数，

为未迭代和第k次迭代后序列中时序待分析数据(时序背景数据)中第i个数据。in,

is the fitted value, X _j+i is the original value, C _i is the coefficient of the i-th data in the time series background data (the time series data to be analyzed), m is half the filter window, N is the filter length, N≤2m+1; Tk is the index of the sequence fitting result after the _kth iteration,

is the i-th data in the time-series data to be analyzed (time-series background data) in the sequence without iteration and after the k-th iteration.

Due to the large difference in the magnitude of the detection data of electromagnetic detection type non-imaging sensors, in order to make the data analysis more contrastive, it is necessary to normalize the electron concentration data and ion concentration data of the time series data to be analyzed and the time series background data. The values are uniformly normalized to [0, 1]. The linear mapping function is:

Among them, x and x ^* are the time series data to be analyzed and the time series background data before and after normalization, respectively, and x _max and x _min are the corresponding maximum and minimum values of the time series data to be analyzed and the time series background data, respectively.

S4, calculate the difference average value and the correlation coefficient of the normalized time series data to be analyzed, and obtain the difference value average value to be analyzed and the correlation coefficient to be analyzed; calculate the difference average value and the difference value of the normalized time series background data The standard deviation of the value and the correlation coefficient were obtained to obtain the mean value of the background difference, the standard deviation of the background difference and the background correlation coefficient.

计算时序待分析数据的待分析差值平均值和时序背景数据的背景差值平均值；Specifically, using the formula

Calculate the average value of the difference to be analyzed of the time series data to be analyzed and the average value of the background difference of the time series background data;

计算时序背景数据的背景差值标准差；Use the formula

Calculate the background difference standard deviation of time series background data;

计算时序待分析数据的相关系数和时序背景数据的相关系数；Using Pearson correlation coefficient (Pearson correlation coefficient) expression

Calculate the correlation coefficient of the time series data to be analyzed and the correlation coefficient of the time series background data;

Wherein, X and Y represent the respective electron concentration data and ion concentration data of the time series data to be analyzed and the time series background data, μ _XY is the difference average value of the electron concentration data and the ion concentration data, and σ _XY is the difference standard between the two Difference, Z is the difference between the two, R _XY is the correlation coefficient of the two, Cov(X, Y) is the covariance of the two, D(X), D(Y) are the variance of the two, respectively.

S5, check whether the non-imaging sensor works normally according to the average value of the difference to be analyzed, the correlation coefficient to be analyzed, the average value of the background difference, the standard deviation of the background difference, and the background correlation coefficient.

Specifically, if the conditions are met:

μ _bg -σ _bg ≤μ _analysis ≤μ _bg +σ _bg

R _analysis ≥R _bg

Then it is determined that the performance and working state of the non-imaging sensor is good, otherwise, it indicates that the non-imaging sensor is abnormal, and the performance of the non-imaging sensor needs to be further analyzed in detail. Wherein, μ _bg is the mean value of the background difference, σ _bg is the standard deviation of the background difference, R _bg is the background correlation coefficient, μ _analysis is the mean value of the difference to be analyzed, and R _analysis is the correlation coefficient to be analyzed.

In addition, the above-mentioned definitions of each step and method are not limited to the various specific formulas or methods mentioned in the embodiments, and those of ordinary skill in the art can simply modify or replace them, such as:

1) In terms of data preprocessing, considering the difference in the on-orbit running time of different satellites, when applied to other non-imaging sensors, the background time series data can also use the data of the previous month or the previous two or three years; data files can also be read and written through Implementation in other programming languages such as Python, C, C++;

2) In the process of data grid processing, the time series data in the same grid can also be obtained by using different time windows such as ten-day average and weekly average according to the change of data volume;

3) Time series data smoothing can also use simple moving average, linear weighted smoothing, Gaussian filtering smoothing, Butterworth filtering smoothing, etc. to perform data smoothing processing.

To sum up, the embodiments of the present invention provide a non-imaging sensor inspection method based on time series data. The correlation calculation is performed on the time series data, and the correlation calculation results are used to check the working state of the non-imaging sensor. The method is simple, efficient and versatile, and can realize on-orbit detection of non-imaging sensors in the case of insufficient external field conditions and limited standard targets. At the same time, considering the characteristics of the sensor to be tested, the sensor operating state analysis can be carried out directly by using its own detection data at a high frequency.

The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in further detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (7)

1. a kind of inspection method of non-imaging sensor, is characterized in that, comprises: S1: Acquire the data to be analyzed and the background data detected by the non-imaging sensor, and preprocess the data to be analyzed and the background data, wherein the data to be analyzed and the background data are detection data of the same area; S2: Perform grid processing on the preprocessed data to be analyzed and background data to obtain time series data to be analyzed and time series background data; S3: Filter and normalize the time series data to be analyzed and the time series background data; S4: Calculate the average value of the difference, the standard deviation of the difference, and the correlation coefficient between the normalized time series data to be analyzed and the time series background data; S5: Check whether the non-imaging sensor works normally according to the average value of the difference, the standard deviation of the difference and the correlation coefficient. 2. The performance testing method of a non-imaging sensor according to claim 1, wherein the storage format of the data to be analyzed and the background data is a text file format, and in step S1, the data to be analyzed and the background data are stored in a format of a text file. Preprocessing includes: S1-1: Perform read-write processing on the data to be analyzed and background data, extract key parameters of "time", "track number", "longitude", "latitude", and "detection value" to form a pure data file; S1-2: Perform abnormal data elimination processing on the data to be analyzed and the background data. 3. The method for checking the performance of a non-imaging sensor according to claim 2, wherein the processing of removing abnormal data from the detection data comprises: According to the two geomagnetic indices Kp index and Dst index, the magnetic disturbance conditions of the magnetic storm are delineated; Abnormal data in the to-be-analyzed data and the background data is eliminated according to the magnetic storm magnetic disturbance condition. 4. The method for checking the performance of a non-imaging sensor according to claim 1, wherein the step S2 comprises: S2-1: According to the characteristics of the satellite re-entry period and the sampling frequency of the non-imaging sensor, perform spatial grid and time window settings on the preprocessed data to be analyzed, wherein, in the spatial grid, the time window is on the time axis move up; S2-2: For the time window corresponding to the time on the time axis in the spatial grid, average the data to be analyzed in the time window to obtain the current spatial grid and the average pending data in the current time period. analyze data; S2-3: Repeat step S2-2, for the preprocessed data to be analyzed, sequentially calculate the average value of the data to be analyzed in the time window corresponding to each time on the time axis to obtain a plurality of average data to be analyzed, and form a time series to be analyzed analyze data; S2-4: Repeat steps S2-1 to S2-3 to process the preprocessed background data to obtain time-series background data. 5. The performance inspection method of a non-imaging sensor according to claim 1, wherein the step S3 comprises: S3-1: Filter the time series data to be analyzed and the time series background data by adopting at least one method of convolution smoothing, simple moving average, linear weighted smoothing, Gaussian filtering smoothing, or Butterworth filtering smoothing; S3-2: Utilize the mapping function

The time series data to be analyzed and the time series background data are normalized, wherein x and x ^* are the time series data to be analyzed and the time series background data before and after normalization, respectively, and x _max and x _min are the time series, respectively The data to be analyzed and the time series background data correspond to the maximum and minimum values. 6. The performance testing method of a non-imaging sensor according to claim 1, wherein the time series data to be analyzed and the time series background data both include electron concentration data and ion concentration data, and the step S4 comprises: S4-1: Utilize formula

Calculate the time series data to be analyzed after the normalization, obtain the difference mean value μ _analysis to be analyzed; Utilize the formula

Calculate the time series data to be analyzed after the normalization, obtain the correlation coefficient R _analysis to be analyzed; S4-2: Utilize formula

Calculate the time series background data after the normalization, obtain the background difference mean value μ _bg ; Utilize the formula

Calculate the normalized time series background data to obtain the background difference standard deviation σ _bg ; use the formula

Calculate the normalized time series background data to obtain the background correlation coefficient R _bg ; Wherein, X and Y represent the electron concentration data and the ion concentration data respectively, _μXY is the average value of the difference between the two, _σXY is the standard deviation of the difference between the two, Z is the difference between the two, and R _XY is the correlation coefficient of the two, Cov(X, Y) is the covariance of the two, and D(X) and D(Y) are the variance of the two, respectively. 7 . The performance testing method of a non-imaging sensor according to claim 6 , wherein the mean value of the difference to be analyzed, the correlation coefficient to be analyzed, the mean value of the background difference, the standard deviation of the background interpolation and the background correlation coefficient satisfy 7 . condition: μ _bg -σ _bg ≤μ _analysis ≤μ _bg +σ _bg R _analysis ≥R _bg Then it is determined that the non-imaging sensor is working normally, otherwise it is prompted that the non-imaging sensor is abnormal.

Images