Satellite rotation period estimation method based on radar radar cross section (RCS) data

Abstract

The invention discloses a satellite rotation period estimation method based on radar radar cross section (RCS) data. Comprising the steps of carrying out noise reduction and zero mean processing on satellite RCS original signals, carrying out statistics on an improved autocorrelation sequence and an average amplitude difference sequence, constructing a periodic statistical vector, extracting a peak time sequence of the periodic statistical vector, eliminating a false peak value, calculating a coarse average period, removing double peak values, and calculating an average period to obtain a satellite rotation period. The problems of frequency division and frequency multiplication misjudgment, burr phenomenon, amplitude attenuation and the like of a traditional method can be effectively solved, and the satellite rotation period estimation based on the radar RCS data is more accurate.

Description

Technical Field

[0001] The invention belongs to the technical field of radar RCS data processing, and in particular relates to a satellite rotation period estimation method based on radar RCS data. Background Art

[0002] Satellites (or other space targets, referred to as satellites in this invention) must maintain a certain stable attitude when operating in space orbits. The main attitude control methods include three-axis stabilization and spin stabilization. A satellite with out-of-control attitude will move in orbit in a "tumbling" attitude. The period is an important parameter for measuring the on-orbit state of a spin-stable and tumbling satellite.

[0003] Spin stabilization is the process of a satellite rotating around its own central axis to obtain gyroscopic axial stability, so that the direction of the spin axis is oriented in inertial space. Spin-stabilized targets have a relatively fast spin angular velocity, generally tens to hundreds of revolutions per minute. Tumbling is a special form of satellite motion, which is characterized by the target tumbling around a certain spatial axis, and the tumbling plane remains unchanged. Failed satellites undergo complex rotational motions under the influence of the earth's gravity and perturbation, and eventually tend to tumble; the final stage of a rocket after the separation of the satellite and the rocket often evolves into a tumbling motion. The tumbling period is an important feature of a space tumbling target.

[0004] At present, the characteristic data obtained by non-cooperative measurement is an important means to estimate satellite periodicity. The radar cross section (RCS) of a satellite is a physical quantity that characterizes the scattering ability of a radar target to the irradiated electromagnetic waves.

[0005] Normally operating satellites need to adjust their components and orbits, and these adjustments will cause changes in radar observation angles, target attitudes, etc., and RCS will be a non-random function of these factors. However, at a certain observation angle, the inaccuracy of target adjustment and the influence of observation noise will make the target's RCS random. Therefore, for targets such as three-axis stabilized satellites, their RCS observation value should be a constant under normal circumstances. However, due to the influence of the above factors, the target RCS observation value will be a superposition of a constant and random noise, thus showing randomness; and for spin-stabilized satellites or rolling targets, since they are constantly rotating according to a certain period, their RCS sequence will show periodicity. Although there is also a superposition of random noise, this periodicity and trend will generally not be covered by randomness. Therefore, the RCS sequence can be directly determined and extracted for periodicity, thereby realizing the determination of satellite spin and roll.

[0006] For rotating, especially rolling targets, the RCS sequence presents multiple fluctuations and non-stationary characteristics. It is usually difficult to automatically estimate the rolling period. Usually, manual experience is used to judge the rolling period of the target. Traditional methods use variance analysis to estimate the period of the RCS sequence of space targets. This method has the problem of frequency doubling and frequency division, and is sensitive to the overall rising and falling trend of the RCS sequence. There are also methods that use the autocorrelation function and the average amplitude difference function or the cyclic autocorrelation and the cyclic average amplitude difference function to estimate the precession period of the cone target, but the autocorrelation function and the average amplitude difference function have the phenomenon of continuously decreasing peak amplitude and burrs; the cyclic average amplitude difference function is used to estimate the period. When the sequence length is an integer multiple of the actual period, the frequency doubling misjudgment will occur. In short, it is difficult for the current narrow-band radar to extract the motion period of the on-orbit satellite only from its own RCS data. There is still a lack of an efficient and universal motion period estimation method, which limits the application of narrow-band radar in the field of target recognition. Summary of the invention

[0007] In view of the above problems, the present invention combines measures such as improving the autocorrelation sequence and the average amplitude difference sequence (also called the autocorrelation function and the average amplitude difference function), eliminating false peaks, and removing the peak bipolar values, etc., which can effectively reduce the impact of problems such as frequency division and frequency multiplication misjudgment, burr phenomenon, amplitude attenuation, etc. in estimating the satellite rotation period using traditional radar RCS data, thereby making the estimation of the satellite's rotation state more accurate and facilitating satellite status monitoring.

[0008] The present invention designs a satellite rotation period estimation method based on radar RCS data. The method at least comprises the following steps: sequentially performing noise reduction processing, zero mean processing, statistical calculation of improved autocorrelation sequence and average amplitude difference sequence (also called autocorrelation function and average amplitude difference function) on radar RCS data (radar cross-sectional area, i.e., original satellite RCS data acquired by radar, also called original signal), constructing a period statistical vector, extracting a peak time series of the period statistical vector, removing false peaks, calculating a rough average period, removing peak bipolar values, and calculating the average period to obtain the satellite rotation period;

[0009] The statistical calculation of the improved average amplitude difference sequence and autocorrelation sequence does not distinguish between the order of precedence;

[0010] The noise reduction process includes: i ,i=1,2,…N} denoising, calculate the RCS data sequence after denoising Where N is the sequence length;

[0011] The zero mean processing includes: Statistical mean me, construct zero mean RCS data sequence

[0012] Calculating the improved autocorrelation sequence includes using the zero-mean RCS data sequence Calculating the autocorrelation sequence {R τ , τ = 1, 2, … M}, where int() is the rounding function;

[0013]

[0014] Calculating the improved average magnitude difference sequence includes using the zero-mean RCS data sequence Calculating the average magnitude difference sequence {D τ , τ = 1, 2, … M};

[0015]

[0016] Constructing the periodic statistical vector includes using the autocorrelation sequence R τ and the average magnitude difference sequence D τ Calculating the periodic statistical vector {RD τ , τ = 1, 2, … M};

[0017]

[0018] Extracting the peak time series of the periodic statistical vector includes extracting the peak time series {(t τ , RD i ), i = 1, 2, … P} from the periodic statistical vector {RD i , τ = 1, 2, … M}; where P < M, is the length of the peak time series, and t i is the moment corresponding to each sequence value of the time series after removing false peaks;

[0019] Removing false peaks includes removing false peaks from the peak time series {(t i , RD i ), i = 1, 2, … P} to form the time series after removing false peaks {(t i , RD i ), i = 1, 2, … L}; where L ≤ P, is the length of the time series after removing false peaks; the false peaks include false peaks formed due to measurement noise and other reasons;

[0020] The rough average period T1:

[0021] For the time series after removing false peaks {(t i , RD i ), i = 1, 2, … L}, denote the peak sequence after removing the double extrema of the wave crest as: {(t i,RD i ), i = 1, 2, ..., h}, the method for removing the peak bipolar value includes:

[0022]

[0023] Among them, h is the peak sequence length after removing the peak bipolar value, and K3 is the threshold parameter for removing the bipolar value;

[0024] The averaging period T2:

[0025] Furthermore, the noise reduction method includes:

[0026]

[0027] Among them, s is the sliding window.

[0028] Furthermore, the zero mean processing method includes:

[0029]

[0030] Furthermore, the peak time series method for extracting the periodic statistical vector includes: satisfying {RD i >RD i-1 And RD i >RD i+1 , i=2, 3, ..., M-1} all (t i ,RD i );

[0031] Among them, t i For the corresponding time scale.

[0032] Furthermore, the peak time series method for extracting the periodic statistical vector includes: satisfying

[0033] {RD i >RD i-1 ≥RD i-2 And RD i >RD i+1 ≥RD i+2 , i=

[0034] 3, 4, ..., M-2} all (t i ,RD i ).

[0035] Furthermore, the false peak satisfies:

[0036]

[0037] Wherein, K2 is the false peak threshold parameter.

[0038] Further, the steps of the satellite rotation period estimation method include:

[0039] S1. Acquisition of the original RCS signal;

[0040] S2. Denoising of the original RCS signal;

[0041] S3. Zero-mean processing of the denoised RCS signal;

[0042] S4. Calculate the improved autocorrelation function R τ ;

[0043] S5. Calculate the improved average magnitude difference function D τ ;

[0044] S6. Construct the period statistical vector RD τ ;

[0045] S7. Calculate the peak time series {(t i , RD i ), i = 1, 2, … P}

[0046] S8. Eliminate the false peaks of RD τ to obtain a new peak time series {(t i , RD i ), i = 1, 2, … L};

[0047] S9. Calculate the rough average period T1 of the peak time series;

[0048] S10. Remove the double extrema of the wave peaks to obtain a new peak time series {(t i , RD i ), i = 1, 2, …, h};

[0049] S11. Calculate the average period T2 of the peak series after removing the double extrema of the wave peaks and make a judgment:

[0050] If ABS(T2 - T1) / T1 < K4, it is considered normal, and output the satellite rotation period T = T2; otherwise, let T1 = T2 and repeat S10; where ABS() is the absolute value function and K4 is the period threshold parameter.

[0051] The advantages and beneficial effects of the present invention are as follows: the radar RCS data satellite rotation period estimation designed by the present invention calculates the satellite rotation period by statistically analyzing the improved autocorrelation function and average amplitude difference function of the satellite RCS data, constructing the period statistical vector, calculating the peak sequence, and using the statistical method to calculate the satellite rotation period. The identification process is simple and effective, and the result is more accurate, which solves the problems of frequency division and frequency multiplication misjudgment, burr phenomenon, amplitude attenuation, etc. of the traditional method. The application of the method of the present invention can improve the function of the radar system and improve its data processing ability, thereby improving the comprehensive performance of the radar equipment or product. In particular, in aerospace measurement and control, a narrowband radar is used to obtain the radar cross section (RCS) of the target, and the RCS data is used to accurately estimate the satellite rotation period, which is beneficial to the satellite on-orbit status monitoring.

[0052] Although the present invention is effective in estimating the rotation period of RCS data of narrow-band radar, it can also be applied to estimating the rotation period of RCS data of wide-band radar. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a principle and steps diagram of radar RCS data classification method;

[0054] Figure 2 This is a comparison of the target RCS original signal (upper half) and the noise-reduced signal (lower half);

[0055] Figure 3 The comparison diagram of the traditional autocorrelation function (a) and the improved autocorrelation function (b) of the target RCS data;

[0056] Figure 4 The comparison diagram of the traditional average amplitude difference function (a) and the improved average amplitude difference function (b) of the target RCS data;

[0057] Figure 5 This is a comparison chart before (a) and after (b) removing the false peaks. DETAILED DESCRIPTION

[0058] Since the traditional method uses variance analysis to estimate the period of the RCS sequence of space targets, this method has the problem of frequency doubling and frequency division, and is sensitive to the overall rising and falling trend of the RCS sequence. There are also methods that use the autocorrelation function and the average amplitude difference function or the cyclic autocorrelation and the cyclic average amplitude difference function to estimate the precession period of the conical target, but the autocorrelation function and the average amplitude difference function have the phenomenon of continuously decreasing peak amplitude and burrs; when the cyclic average amplitude difference function is used to estimate the period, when the sequence length is an integer multiple of the actual period, it will produce frequency doubling misjudgment. There is still a lack of an efficient, universal, and simple method for estimating the satellite rotation period for RCS data.

[0059] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0060] Embodiment 1:

[0061] like Figure 1 As shown, the present invention designs a satellite rotation period estimation method based on radar RCS data, characterized in that the method at least includes sequentially performing noise reduction processing, zero mean processing, statistical calculation of improved autocorrelation sequence and average amplitude difference sequence (also called autocorrelation function and average amplitude difference function) on radar RCS data (radar scattering cross section, i.e., original satellite RCS data acquired by radar, also called original signal), constructing a period statistical vector, extracting the peak time series of the period statistical vector, removing false peaks, calculating a rough average period, removing peak bipolar values, and calculating the average period to obtain the satellite rotation period;

[0062] The statistical calculation of the improved average amplitude difference sequence and autocorrelation sequence does not distinguish between the order of precedence;

[0063] The noise reduction process includes: i ,i=1,2,…N} denoising, calculate the RCS data sequence after denoising Among them, N is the sequence length; here the sequence length N only means that the maximum length is N, and in engineering applications it should be determined based on the actual sequence length after processing; in fact, the length of the new sequence obtained after denoising may be smaller than the original sequence length, such as when using sliding window denoising.

[0064] The zero mean processing includes: Statistical mean me, construct zero mean RCS data sequence

[0065] Calculating the improved autocorrelation sequence includes using the zero-mean RCS data sequence Calculate the autocorrelation sequence {R τ ,τ=1,2,…M}, where, int() is the rounding function;

[0066]

[0067] Calculating the improved mean amplitude difference sequence includes using the zero mean RCS data sequence Calculate the average amplitude difference sequence {D τ ,τ=1,2,…M};

[0068]

[0069] The traditional autocorrelation function and the average amplitude difference function are described in the literature "A method for distinguishing spin-stable targets" and "A method for estimating the precession period based on improved CAMDF". The traditional CAMDF uses a convolution-like method to redefine the traditional average amplitude difference function. At this time, the sequence length is the key factor affecting the accuracy of the period estimation of the cyclic average amplitude difference function. When the sequence length is an integer multiple of the actual period, a frequency doubling misjudgment will occur; and when the sequence length is not an integer multiple of the actual period, no frequency doubling misjudgment will occur. The improvement of the present invention mainly lies in that when calculating the cyclic autocorrelation function and the cyclic average amplitude difference function, the entire RCS sequence is not calculated, but only the first half of the sequence is calculated. In this way, even if the sequence length is an integer multiple of the actual period, the overlapping part is only the first half of the original sequence after translation, and the length is only half of the original sequence. The average amplitude obtained will not produce obvious frequency doubling problems. After multiple actual measurements, it is shown that the estimation accuracy of the improved method is better than the previous method.

[0070] Through the analysis of existing technologies, it can be seen that for the period estimation of the RCS sequence of the precessing target, the traditional spectrum analysis method has great limitations. The reason is that the autocorrelation function (ACF) measures the correlation between the signal and itself after a certain time delay. If a signal is periodic, the autocorrelation function will show a peak at the delay of each period. The cyclic autocorrelation (CACF) is a special form of autocorrelation designed for periodic signals. It assumes that the signal is periodic and uses this assumption in calculation, so it can better highlight the periodic characteristics of the signal. The calculation of the cyclic autocorrelation function is very important for the initial estimation of the periodic time series, because different initial period lengths may lead to different results. Therefore, when calculating the cyclic autocorrelation function, it is very critical to choose the appropriate period length. On the other hand, directly using the results of the average amplitude difference function (AMDF) method for period estimation faces two problems: first, when the delay time increases, the number of summation terms of the AMDF will gradually decrease, resulting in a continuous decrease in the peak amplitude of the AMDF; second, in the last section of the signal waveform, the peak intervals are uneven, and burrs will appear. These two problems cause errors in the detection of peak points and subsequent period estimation. The use of normalized average amplitude difference function can suppress the peak amplitude attenuation problem of the traditional average amplitude difference function, so that the peak can always be maintained at a certain height, thereby improving the accuracy of peak detection, but it cannot solve the second problem, that is, there is still an error in the period estimation.

[0071] In order to solve the problem of period estimation error caused by signal "burrs", the cyclic mean amplitude difference function (CAMDF) can be used as an effective solution. Compared with the direct calculation method of AMDF, CAMDF first determines the expected period of the signal during calculation, and then expands the periodicity based on this. The cyclic mean amplitude difference function adds consideration to the periodicity of the signal, which makes it more advantageous when processing signals with clear periodicity. Its core idea is to use an autoregressive-sliding model to extend the period of the RCS sequence. CAMDF redefines the traditional average amplitude difference function in a convolution-like manner. The sequence length is the key factor affecting the accuracy of the period estimation of the cyclic mean amplitude difference function. When the sequence length is an integer multiple of the actual period, a frequency multiplication error will occur; when the sequence length is not an integer multiple of the actual period, a frequency multiplication error will not occur. Although CAMDE can improve the problem of peak amplitude attenuation and solve the burr phenomenon between waveforms, the cyclic mean amplitude difference function will introduce the problem of frequency multiplication error.

[0072] In response to various problems existing in the four functions of the prior art, including but not limited to burr phenomenon, frequency misjudgment, amplitude attenuation, etc., the cyclic autocorrelation and cyclic average amplitude difference are improved on the basis of the autocorrelation and average amplitude difference functions, effectively solving the error problem.

[0073] The present invention designs a new method to improve the autocorrelation function and the average amplitude difference function, that is, only the first half of the sequence is translated, so that the overlapping part is always half of the sequence length. After multiple actual tests, it is shown that the estimation accuracy of the improved method is better than the previous method.

[0074] This combined function (periodic statistical vector) not only still guarantees that the periodicity requirements are met, but also combines the advantages of the autocorrelation function and the average amplitude difference function, making the peak more prominent, suppressing false peaks and valleys, and enhancing the periodicity of the signal, obtaining the current optimal result, solving the amplitude attenuation problem that occurs when only the autocorrelation function is used, as well as the burr phenomenon and frequency doubling misjudgment problems that occur when only the average amplitude difference function is used, further reducing the impact of the amplitude attenuation problem on the results when only the average autocorrelation function is used and the frequency doubling error when only the average amplitude difference function is used, making the results more accurate.

[0075] The construction of the periodic statistical vector includes using the autocorrelation sequence R τ and the average amplitude difference sequence D τ Calculate the cycle statistics vector {RD τ , τ=1,2,…M};

[0076]

[0077] The peak time series of the extracted period statistical vector includes extracting the peak time series \((t τ , RD i )\), \(i = 1, 2, \ldots, P\) from the period statistical vector \(\{RD i , \tau = 1, 2, \ldots, M\}\). In this embodiment, the extraction operation is performed based on the findpeaks function; where \(P \lt M\) is the length of the peak time series, and \(t i \) is the moment corresponding to each sequence value of the time series after removing false peaks;

[0078] Removing false peaks includes removing false peaks from the peak time series \((t i , RD i )\), \(i = 1, 2, \ldots, P\) to form the time series \((t i , RD i ), i = 1, 2, \ldots, L\) after removing false peaks; where \(L \leq P\) is the length of the time series after removing false peaks; the false peaks include false peaks formed due to measurement noise;

[0079] The rough average period \(T1\):

[0080] For the time series \((t i , RD i ), i = 1, 2, \ldots, L\) after removing false peaks, the peak sequence after removing the double extreme values of the wave peak is denoted as: \((t i , RD i ), i = 1, 2, \ldots, h\). The method for removing the double extreme values of the wave peak includes:

[0081]

[0082] where \(h\) is the length of the peak sequence after removing the double extreme values of the wave peak, and \(K3\) is the threshold parameter for removing double extreme values, generally with a value range of 2 to 5. In this embodiment, \(K3 = 2\);

[0083] Due to factors such as signal complexity, measurement noise, and data processing methods, the phenomenon of two maximum points appearing in one wave peak causes the true peak of the signal to shift, affecting the extraction of period characteristics. Removing the double extreme values of the wave peak is mainly to eliminate the influence of noise, etc., and improve the accuracy of peak detection.

[0084] The average period \(T2\):

[0085] The steps of the satellite rotation period estimation method include:

[0086] S1. RCS original signal acquisition;

[0087] S2. RCS original signal noise reduction;

[0088] S3, RCS noise reduction signal zero mean processing;

[0089] S4. Calculate the improved autocorrelation function R τ ;

[0090] S5. Calculate the average magnitude difference D of the improvement τ ;

[0091] S6. Construct periodic statistical vector RD τ ;

[0092] S7, calculate the peak time series {(t i ,RD i ), i=1,2,…P}

[0093] S8, remove RD τ The false peak value is obtained to obtain the new peak time series {(t i ,RD i ),i=1,2,…L};

[0094] S9, calculating the rough average period T1 of the peak time series;

[0095] S10, remove the peak bipolar value and get the new peak time series {(t i ,RD i ),i=1,2,…,h};

[0096] S11, calculate the average period T2 of the peak sequence after removing the peak bipolar value.

[0097] The noise reduction processing, zero mean processing, statistically improved autocorrelation sequence and average amplitude difference sequence, peak time series of periodic statistical vectors, and methods for removing false peaks can be based on existing technologies. For example, the noise reduction processing can adopt the conventional data noise removal method with excessive adjacent difference values; the zero mean processing is to process the mean of the data sequence to zero; the calculation of the autocorrelation sequence and the average amplitude difference sequence can refer to the relevant methods of speech processing or the design principles of the relevant library functions of matlab; the peak time series of the periodic statistical vector can be directly extracted according to its physical meaning, such as when the measured value exceeds the adjacent values ​​on the left and right, it is considered to be a peak value, and its periodic characteristics can be statistically determined after the peak value is determined; the removal of false peaks should generally be based on the defects of the peak judgment method, and the corresponding strategy should be adopted to remove false peaks, including that although the measured value exceeds the adjacent values ​​on the left and right, it is essentially caused by a small fluctuation in measurement noise, which is a false peak value, or exceeds the adjacent values ​​on both sides too much, which obviously belongs to the category of wild values, etc.

[0098] Through the comprehensive processing of the above steps, after noise reduction and zero mean processing, the autocorrelation sequence and average amplitude difference sequence (also called autocorrelation function and average amplitude difference function) improved by statistical calculation are used to construct a periodic statistical vector, and the peak time series of the periodic statistical vector is extracted. After eliminating false peaks, calculating the rough average period, removing the peak bipolar value, and then calculating the average period, the influence of problems such as frequency division and frequency multiplication misjudgment, burr phenomenon, amplitude attenuation, etc. can be significantly reduced, so that the rotational motion state of the space target can be more accurately estimated.

[0099] Attached Figures 2 to 5 The experimental results are shown in the attached Figure 2 It can be seen that the original target RCS signal is smoother overall after noise reduction processing;

[0100] From the attached Figure 3 It can be seen that for the target RCS data, the traditional average autocorrelation function subgraph (a) has an obvious amplitude attenuation problem, while the improved autocorrelation function subgraph (b) has a relatively stable amplitude;

[0101] From the attached Figure 4 It can be seen that for the target RCS data, the traditional average amplitude difference function sub-graph (a) has both obvious amplitude attenuation and burrs, which easily leads to misjudgment of frequency doubling, while the improved average amplitude difference function sub-graph (b) has a relatively stable amplitude without obvious burrs;

[0102] From the attached Figure 5 It can be seen that before removing the false peaks, the peak amplitude values ​​of sub-image (a) are quite different and the period is unstable. After removing the false peaks, the peak amplitude values ​​of sub-image (b) are basically consistent and the period is stable.

[0103] Preferably, the noise reduction method comprises:

[0104]

[0105] Among them, s is a sliding window, and its value range is generally 2 2 ~2 10 In this embodiment, s=2 4 =16.

[0106] Preferably, the zero mean processing method includes:

[0107]

[0108] Preferably, the peak time series method for extracting the periodic statistical vector includes: satisfying {RD i >RD i-1 And RD i >RD i+1 , i=2, 3, ..., M-1} all (t i, RD i );

[0109] wherein, t i is the corresponding time stamp.

[0110] Preferably, the false peak satisfies:

[0111]

[0112] wherein, K2 is the false peak threshold parameter, and its general value range is 1 / 3 to 1 / 6. In this embodiment, K2 = 1 / 3 is taken.

[0113] Embodiment 2:

[0114] The difference from Embodiment 1 is that step S11 of the satellite rotation period estimation method in this embodiment is:

[0115] S11. Calculate the average period T2 of the peak sequence after removing the double extrema of the wave peaks and judge:

[0116] If ABS(T2 - T1) / T1 < K4, it is considered normal, and the satellite rotation period T = T2 is output. Otherwise, let T1 = T2, and repeat S10; wherein, ABS() is the absolute value function, and K4 is the period threshold parameter. Its general value range is between 0 and 0.5. In this embodiment, K4 = 0.2 is taken. When performing cyclic iteration, the total number of iteration times is limited to avoid entering an infinite loop, and generally it is set to not exceed 20 times. The main purpose of this step of cyclic judgment is to make the average period obtained approach the true period through cycling, and to avoid entering an infinite loop when the satellite is out of control and "tumbles", etc.

[0117] Embodiment 3:

[0118] The difference from Embodiment 1 is that the method for extracting the peak time series of the period statistical vector in this embodiment includes: satisfying {RD i > RD i-1 ≥ RD i-2 and RD i > RD i+1 ≥ RD i+2 , i = 3, 4,..., M - 2} for all (t i , RD i ).

[0119] Embodiment 4:

[0120] The difference from Embodiment 1 is that in this embodiment, s = 2 5 = 32, K2 = 1 / 4, K3 = 3, K4 = 0.5.

[0121] The main principle of the present invention is: in view of various problems existing in the current satellite rotation period estimation method based on radar RCS data, including but not limited to burr phenomenon, frequency multiplication misjudgment, amplitude attenuation, etc., a new method is designed to improve the autocorrelation function and the average amplitude difference function, that is, only the first half of the sequence is translated, so that the overlapping part is always half of the sequence length. After multiple actual measurements, it is shown that the estimation accuracy of the improved method is better than the method used before. By improving the autocorrelation function and the average amplitude difference function, it is conducive to subsequent period estimation, and the estimation accuracy is better under certain conditions, but the problem of frequency division and frequency multiplication misjudgment has not been completely solved. By observing the signal waveform generated by the improved method, it is found that in addition to the peak points and trough points at the signal period, there are also some false peak and valley points. When these peak and valley points are relatively large, they will interfere with the subsequent period estimation results. In order to improve the robustness of period estimation, by combining the improved autocorrelation function and the average amplitude difference function, the false peak point is proposed to generate a periodic statistical combination function. This combination function not only still guarantees the periodicity, but also combines the advantages of the autocorrelation function and the average amplitude difference function, making the peak more prominent, suppressing the false peak and valley points, and enhancing the periodicity of the signal. The current optimal result is obtained, which solves the amplitude attenuation problem that occurs when only the autocorrelation function is used, as well as the burr phenomenon and the frequency misjudgment problem that occur when only the average amplitude difference function is used. It further reduces the influence of the amplitude attenuation problem on the result when only the average autocorrelation function is used and the frequency misjudgment problem when only the average amplitude difference function is used, making the result more accurate.

[0122] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical principles of the present invention, several improvements and modifications can be made, such as different values ​​and combinations of threshold parameters s, K2, K3, K4, using different noise reduction or denoising methods, peak extraction methods, etc. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A satellite rotation period estimation method based on radar RCS data, characterized in that: The method at least includes performing noise reduction processing, zero-mean processing, statistically calculating an improved autocorrelation sequence and an average magnitude difference sequence, constructing a periodic statistical vector, extracting the peak time series of the periodic statistical vector, removing false peaks, calculating a rough average period, removing double extrema of peaks, and calculating the average period to obtain the satellite rotation period; The statistical calculation of the improved average magnitude difference sequence and autocorrelation sequence does not distinguish the order; The noise reduction process includes: i ,i=1,2,…N} denoising, calculate the RCS data sequence after denoising Where N is the sequence length; The zero mean processing includes: Statistical mean me, construct zero mean RCS data sequence Calculating the improved autocorrelation sequence includes using the zero-mean RCS data sequence Calculate the autocorrelation sequence {R τ ,τ=1,2,…M}, where, int() is the rounding function; Calculating the improved mean amplitude difference sequence includes using the zero mean RCS data sequence Calculate the average amplitude difference sequence {D τ ,τ=1,2,…M}; The construction of the periodic statistical vector includes using the autocorrelation sequence R τ and the average amplitude difference sequence D τ Calculate the cycle statistics vector {RD τ , τ=1,2,…M}; The peak time series of the extracted period statistical vector includes extracting the peak time series \((t τ , RD i )\), \(i = 1, 2, \ldots, P\) from the period statistical vectors \(\{RD τ , \tau = 1, 2, \ldots, M\}\); where \(P < M\) is the length of the peak time series, and \(t i \) is the moment corresponding to each sequence value of the time series after removing false peaks; The method of removing false peaks includes removing the false peaks from the peak time series {(t i ,RD i ), i=1,2,…P}, remove the false peaks, and form the time series {(t i ,RD i ),i=1,2,…L}; where L≤P is the length of the time series after removing the false peaks; The rough averaging period T1: For the time series after removing the false peaks {(t i ,RD i ),i=1,2,…L}, the peak sequence after removing the peak bipolar value is recorded as: {(t i ,RD i ), i = 1, 2, ..., h}, the method for removing the peak bipolar value includes: where h is the length of the peak sequence after removing the double extrema of peaks, and K3 is the threshold parameter for removing double extrema; The averaging period T2:

2. A satellite rotation period estimation method based on radar RCS data according to claim 1, characterized in that: The noise reduction method includes: where s is the sliding window.

3. The satellite rotation period estimation method based on radar RCS data according to claim 1, characterized in that: The zero-mean processing method includes:

4. The satellite rotation period estimation method based on radar RCS data according to claim 1, characterized in that: The peak time series method for extracting periodic statistical vectors includes: satisfying {RD i >RD i-1 And RD i >RD i+1 , i=2, 3, ..., M-1} all (t i ,RD i ); Among them, t i For the corresponding time scale.

5. The satellite rotation period estimation method based on radar RCS data according to claim 1, characterized in that: The peak time series method for extracting periodic statistical vectors includes: satisfying {RD i >RD i-1 ≥RD i-2 And RD i >RD i+1 ≥RD i+2 , i=3, 4, ..., M-2} all (t i ,RD i ).

6. The satellite rotation period estimation method based on radar RCS data according to claim 1, characterized in that: The false peak satisfies: where K2 is the false peak threshold parameter.

7. The satellite rotation period estimation method based on radar RCS data according to claim 1, characterized in that: The method steps include: S1. Acquisition of the original RCS signal; S2. Noise reduction of the original RCS signal; S3. Zero-mean processing of the RCS noise-reduced signal; S4. Calculate the improved autocorrelation function R τ ; S5. Calculate the average magnitude difference D of the improvement τ ; S6. Construct periodic statistical vector RD τ ; S7, calculate the peak time series {(t i ,RD i ), i=1,2,…P} S8, remove RD τ The false peak value is obtained to obtain the new peak time series {(t i ,RD i ),i=1,2,…L}; S9. Calculating the rough average period T1 of the peak time series; S10, remove the peak bipolar value and get the new peak time series {(t i ,RD i ),i=1,2,…,h}; S11. Calculating the average period T2 of the peak sequence after removing the double extrema of peaks and judging: If ABS(T2 - T1) / T1 < K4, it is considered normal, and the satellite rotation period T = T2 is output; otherwise, let T1 = T2, and repeat S10; where ABS() is the absolute value function, and K4 is the period threshold parameter.

Images