An adaptive filtering method for helicopter optical atmospheric data system

Abstract

The present invention discloses an adaptive filtering method for a helicopter optical atmospheric data system, comprising the following steps: S1, peak-finding the preprocessed spectrum data to obtain a plurality of peaks that meet the signal-to-noise ratio threshold; S2, establishing a priori signal estimation model to automatically update the characteristic parameters of the state space; S3, establishing a state space model using the characteristic parameters of step S2, bringing all the peak-finding peaks in S1 into the state space model in turn and using the Kalman filter algorithm to perform noise filtering and effective signal estimation. The present invention can overcome the problem of complex and cumbersome parameter tuning of the filtering model, and can also solve the problem of the optical atmospheric data system detecting a variety of interference components in a complex environment. The present invention achieves the filtering of random noise, environmental noise, helicopter rotor interference and other noises in the signal, and has good practicality.

Description

Technical Field

[0001] The invention belongs to the technical field of airborne laser radar data processing, and in particular relates to an adaptive filtering method for a helicopter optical atmospheric data system. Background Art

[0002] With the increasing demand for three-axis atmospheric parameters and full-speed domain operations on armed helicopter platforms, the traditional atmospheric data system does not meet the requirements for low-speed measurement accuracy due to its principle limitations, and is easily affected by the downwash of the helicopter rotor. The optical atmospheric data system no longer uses sunken atmospheric data sensors such as pitot tubes and vector sensors that are affected by downwash. It measures the relative velocity vector between the macroscopic atmosphere in front of the helicopter and the helicopter through the principle of laser Doppler remote sensing, and calculates the various atmospheric vector parameters.

[0003] The optical atmospheric data system has the advantages of high temporal and spatial resolution and fast response speed, but the speed is relatively fast during the signal acquisition and processing, especially in complex environments, it is easy to introduce a variety of interference signals. Therefore, there are a variety of noise signals in the data inversion process, which mainly include random noise caused by data acquisition and signal preprocessing, noise caused by part of the laser hitting the rotor due to the installation position, and signal interference noise caused by the system hitting fixed objects (laser emission hitting other aircraft or mountains, etc.). The elimination of noise signals can improve the accuracy of signal solution, and at the same time help to increase the signal effectiveness in the full speed domain and complex multi-interference environment. Therefore, there is an urgent need for a filtering method that can comprehensively and efficiently filter out interference signals and retain the original signal components.

[0004] In the prior art, the limiting average filtering method is generally used to remove process noise, which has a good inhibitory effect on periodic interference and pulse interference caused by accidental factors. However, there are still great limitations in the actual flight process. First, the maximum threshold used for limiting needs to be iterated according to the flight data obtained each time. The process is relatively cumbersome and cannot be adaptively adjusted according to each flight data; in addition, under large helicopter maneuvers or when aerosols suddenly decrease (a necessary condition for effective detection of the system), the short-term signal feature changes too much, which will cause the filter to diverge and cannot effectively converge near the true value; finally, in the process of taking the average, the queue length needs to be effectively set. If it is too long, the real-time performance of the data will be reduced, and if it is too short, it will be difficult to effectively suppress the noise. For the noise generated by the laser hitting the helicopter rotor and the interference noise caused by the laser hitting the fixed object for a short time, the classification algorithm and other methods are used to remove the crosstalk noise and rotor noise signals respectively. The classification algorithm can effectively remove it to a certain extent, but it will lose a large amount of original signals, and it needs to be debugged separately according to the flight situation. In summary, the limited average filtering method is suitable for conventional flight conditions and situations where the aerosol backscattering signal has good consistency, while the use of the classification algorithm will lose a large number of valid original signals. Summary of the invention

[0005] The object of the present invention is to provide an adaptive filtering method for a helicopter optical atmospheric data system, aiming to solve the above-mentioned problem.

[0006] The present invention can overcome the problem of complex and tedious parameter tuning of the filtering model, and can also solve the problem of the optical atmospheric data system detecting a variety of interference components in a complex environment. The present invention can filter out random noise, environmental noise, helicopter rotor interference and other noises in the signal, and has good practicality.

[0007] The present invention is mainly achieved through the following technical solutions:

[0008] A helicopter optical atmospheric data system adaptive filtering method comprises the following steps:

[0009] S1, searching for peaks on the preprocessed spectrum data to obtain multiple peaks that meet the signal-to-noise ratio threshold;

[0010] S2, establish a priori signal estimation model to automatically update the characteristic parameters of the state space;

[0011] S3. Use the characteristic parameters of step S2 to establish a state space model, bring all the peak values ​​in S1 into the state space model in turn, and use the Kalman filter algorithm (Kalman RRA new approach to linear filtering and prediction problem [J]. Journal of Basic Eng (ASME), 1960, 82D: 95~108) to filter out noise and estimate effective signals.

[0012] In order to better implement the present invention, further, step S1 includes the following steps:

[0013] S1.1, perform Gaussian smoothing on the spectrum data, and then perform first-order and second-order derivatives on the spectrum data;

[0014] S1.2. Calculate the average value of the spectrum data, set a threshold based on the average value, and obtain the peak value.

[0015] In order to better implement the present invention, further, the peak finding method in step S1 is a derivative method.

[0016] In order to better implement the present invention, further, in step S2, multiple priori signal values ​​are substituted into the priori signal estimation model to automatically update the characteristic parameters.

[0017] In order to better implement the present invention, further, step S2 includes the following steps:

[0018] S2.1, initialize parameters according to the first m prior signal values, determine the estimated error covariance P0 and the initial signal state value x0;

[0019]

[0020] in: is the a priori estimate of the signal value;

[0021] S2.2, construct the state transfer matrix A, and use the prior estimation signal to update the excitation noise covariance matrix Q;

[0022]

[0023] Where N is the sequence number of the signal at the previous moment;

[0024] S2.3, use the first n prior signal data to obtain the estimated value of the signal at this time

[0025]

[0026] In order to better implement the present invention, further, step S3 includes the following steps:

[0027] S3.1, using the peak-finding algorithm, substitute the n observed signals z that meet the peak-finding requirements into the state space model for calculation;

[0028] z(n)={z(1),z(2),...,z(n),...}(n=1~N)

[0029] S3.2. Calculate the prior estimated covariance P prior ;

[0030] P prior =AP posterior A T +Q

[0031] Where: P posterior : posterior estimated covariance matrix;

[0032] A is a system parameter, which is a matrix for the multi-model system and 1 for the system of the present invention;

[0033] A T is the transposed matrix of A.

[0034] S3.3, determining the observation noise covariance R(n) of each iteration according to each observation signal;

[0035]

[0036] in: is the a posteriori estimate of the Kalman filter;

[0037] S3.4. Calculate the Kalman gain coefficient K(n) each time according to R(n), and H=1 at this time;

[0038]

[0039] Where: H is the parameter of the measurement system. For a multi-measurement system, it is a matrix, and for the system of the present invention, it is 1;

[0040] H T is the transposed matrix of H.

[0041] S3.5. When the Kalman gain coefficient K(n)>K(n-1), substitute the parameters of steps S3.2-S3.4 into steps S3.6 and S3.7 for calculation, and initialize K(0) to 0;

[0042] S3.6. Calculate the a posteriori estimate of the system state

[0043]

[0044] S3.7. Calculate the covariance matrix P posterior Update, at this time I=1:

[0045] P posterior =(IK H)P prior

[0046] Where: I: unit diagonal matrix;

[0047] According to the above steps S3.1 to S3.7, all n observed signals are exhaustively enumerated and multiple recursive calculations are performed using the Kalman filter algorithm. The characteristic parameters after recursion of the previous signal are used as the initial values ​​of the characteristic parameters of the next signal, and finally an effective estimation value is obtained by filtering out various complex interference noises.

[0048] The beneficial effects of the present invention are:

[0049] (1) The present invention can adaptively update characteristic parameters according to the statistical characteristics of the signal, and filter out interference signals in complex environments. The present invention uses adaptive characteristic parameters to establish state-space equations to perform adaptive multi-noise filtering on helicopter optical atmospheric data, filtering out random noise, environmental noise, helicopter rotor interference and other noises in the signal;

[0050] (2) The present invention does not use a single peak search point and the previous moment signal for recursion, but counts all the peaks of a single spectrum that meet the threshold (including interference signals, as long as the threshold conditions are met), substitutes them into the state space model and combines with the Kalman filter algorithm to obtain the effective signal at the current moment;

[0051] (3) For the process excitation noise Q, the present invention uses the prior estimation of the signal to update in real time, adapts to the change of the helicopter situation, and can obtain appropriate Q when the helicopter is flying at high and low speeds; while meeting the real-time requirements, the prior state estimation of the present invention is always calculated based on the average of multiple signal points, and the signal can still remain stable when one of the signal points is an interference signal;

[0052] (4) In the a posteriori estimation of the system state, the present invention always uses the a priori estimated state for recursion, and can always select a valid signal area when there are multiple continuous interference noises and when the signal statistics are greater than the noise statistics. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of the present invention;

[0054] Figure 2 It is a spectrum diagram of multiple interference noises and signals that meet the threshold in the spectrum at a single moment;

[0055] Figure 3 A spectrum diagram showing the signal changing with time before filtering using the present invention;

[0056] Figure 4 The figure is a spectrum diagram showing the variation of the signal over time after filtering using the present invention. DETAILED DESCRIPTION

[0057] The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0058] Embodiment 1:

[0059] An adaptive filtering method for helicopter optical atmospheric data system, such as Figure 1 As shown, the following steps are included:

[0060] S1, searching for peaks on the preprocessed spectrum data to obtain multiple peaks that meet the signal-to-noise ratio threshold;

[0061] S2, establish a priori signal estimation model to automatically update the characteristic parameters of the state space;

[0062] S3. Use the characteristic parameters of step S2 to establish a state space model, bring all the peak values ​​in S1 into the state space model in turn, and use the Kalman filter algorithm (Kalman RRA new approach to linear filtering and prediction problem [J]. Journal of Basic Eng (ASME), 1960, 82D: 95~108) to filter out noise and estimate effective signals.

[0063] The present invention introduces Kalman filtering into the optical atmospheric data system to filter out interference signals and retain original signal components, can adaptively update characteristic parameters according to data statistical characteristics, and filter out interference signals in complex environments at the same time.

[0064] Embodiment 2:

[0065] This embodiment is optimized on the basis of embodiment 1, and step S1 includes the following steps:

[0066] S1.1, perform Gaussian smoothing on the spectrum data, and then perform first-order and second-order derivatives on the spectrum data;

[0067] S1.2. Calculate the average value of the spectrum data, set a threshold based on the average value, and obtain the peak value.

[0068] Furthermore, the peak finding method in step S1 is a derivative method.

[0069] The other parts of this embodiment are the same as those of Embodiment 1, and thus will not be described in detail.

[0070] Embodiment 3:

[0071] This embodiment is an optimization based on Embodiment 2, and multiple priori signal values ​​are substituted into the priori signal estimation model to automatically update the characteristic parameters.

[0072] Furthermore, the step S2 comprises the following steps:

[0073] S2.1, initialize parameters according to the first m prior signal values, determine the estimated error covariance P0 and the initial signal state value x0;

[0074]

[0075] in: A priori estimate of the signal value.

[0076] S2.2, construct the state transfer matrix A, and use the prior estimation signal to update the excitation noise covariance matrix Q;

[0077]

[0078] Where N is the position of the signal at the previous moment;

[0079] S2.3, use the first n prior signal data to obtain the estimated value of the signal at this time

[0080]

[0081] Under the premise of meeting the real-time requirements, the prior state estimation is always calculated based on the mean of multiple signal points, and the signal can still remain stable when one of the signal points is an interference signal.

[0082] The other parts of this embodiment are the same as those of the above-mentioned embodiment 2, so they will not be described in detail.

[0083] Embodiment 4:

[0084] This embodiment is optimized on the basis of embodiment 2, and step S3 includes the following steps:

[0085] S3.1, using the peak-finding algorithm, substitute the n observed signals z that meet the peak-finding requirements into the state space model for calculation;

[0086] z(n)={z(1),z(2),...,z(n),...}(n=1~N)

[0087] S3.2. Calculate the prior estimated covariance P prior ;

[0088] P prior =AP posterior A T +Q

[0089] Where: P posterior : posterior estimated covariance matrix;

[0090] A: KF state transfer matrix;

[0091] S3.3, determining the observation noise covariance R(n) of each iteration according to each observation signal;

[0092]

[0093] in: is the a posteriori estimate of the signal;

[0094] S3.4. Calculate the Kalman gain coefficient K(n) each time according to R(n), and H=1 at this time;

[0095]

[0096] S3.5. When the Kalman gain coefficient K(n)>K(n-1), substitute the parameters of steps S3.2-S3.4 into steps S3.6 and S3.7 for calculation, and initialize K(0) to 0;

[0097] S3.6. Calculate the a posteriori estimate of the system state

[0098]

[0099] S3.7. Calculate the completed covariance matrix P posterior for the update, where I = 1 at this time:

[0100] P posterior = (I - KH)P prior

[0101] where: I: Identity diagonal matrix;

[0102] According to the above steps S3.1 to S3.7, all n observed signals are exhausted one by one and recursively calculated multiple times using the Kalman filter algorithm. The characteristic parameters after recursion of the previous signal are used as the initial values of the characteristic parameters of the next signal, and finally an effective estimated value for filtering out various complex interference noises is obtained.

[0103] For updating the excitation noise covariance matrix, it is updated in real time using the prior estimate of the signal to adapt to the changes in the helicopter's situation, and a suitable updated excitation noise covariance matrix can be obtained when the helicopter is flying at high and low speeds.

[0104] Other parts of this embodiment are the same as those of the above Embodiment 2, so they will not be elaborated here.

[0105] Embodiment 5:

[0106] This embodiment is optimized on the basis of Embodiment 2. The state space model in step S3 is as follows:

[0107] S3.1. Calculate the prior estimate covariance P prior ;

[0108] S3.2. Determine the observation noise covariance R(n) for each iteration according to n observed signals;

[0109] S3.3. Calculate the Kalman gain coefficient K(n) for each time according to R(n);

[0110] S3.4. Initialize K(0) to 0. When the Kalman gain coefficient K(n) > K(n - 1), proceed to the next calculation. When K(n) < K(n - 1), do not perform iteration;

[0111] S3.5. Calculate the posterior estimate value of the system state

[0112] S3.6. Calculate the update of the completed covariance matrix P posterior for the update.

[0113] Instead of using a single peak search point to perform recursion with the signal at the previous moment, all the peaks in a single spectrum that satisfy the threshold (including interference signals as long as they meet the threshold conditions) are statistically counted and substituted into the state space model.

[0114] Furthermore, the step S4 is specifically as follows: according to step S3, all n observed signals are exhaustively calculated using the Kalman filter algorithm for multiple recursive calculations, and the characteristic parameters after recursion of the previous signal are used as the initial values ​​of the characteristic parameters of the next signal. At this time, the signal value is the value after filtering out various complex interference noises.

[0115] The other parts of this embodiment are the same as those of the above-mentioned embodiment 2, so they will not be described in detail.

[0116] Embodiment 6:

[0117] A helicopter optical atmospheric data system adaptive filtering method is implemented as follows:

[0118] I: unit diagonal matrix

[0119] 1) If Figure 2 As shown, the circles represent effective signals and part of interference signals, and peak search for the preprocessed spectrum data will obtain multiple peaks that meet the threshold (including interference signals in this case).

[0120] 2) Substitute multiple prior signal values ​​into the prior signal estimation model to automatically update the state space characteristic parameters. The state space correlation matrix is ​​one-dimensional, which is suitable for the working environment of the helicopter optical atmospheric data system. Therefore, the following matrices are all 1x1. Specifically, the following steps are included:

[0121] (2.1) The parameters are initialized based on the first m prior signal values ​​(actual measured values) to determine the estimated error covariance P0 and the initial signal state value x0.

[0122]

[0123] (2.2) Construct the state transfer matrix A, and use the prior detection signal to update the excitation noise covariance matrix Q. N is the position of the signal at the previous moment.

[0124]

[0125] (2.3) Use the first n prior signal data to obtain the estimated value of the signal at this time

[0126]

[0127] 3) Establish a state space model and use the characteristic parameters obtained in step 2) to filter out noise and estimate effective signals. Specifically, the following steps are included:

[0128] (3.1) Using the peak-finding algorithm, we start from z(1) and substitute it into the following table to calculate the n observed signals z (including signal and interference noise peaks) that meet the threshold and peak-finding requirements in the spectrum:

[0129] z(n)={z(1),z(2),...,z(n),...}(n=1~N)

[0130] (3.2) Calculate the prior estimated covariance (initial P posterior The setting cannot be 0):

[0131] P prior =AP posterior A T +Q

[0132] (3.3) Determine the observation noise covariance R(n) of each iteration based on each observation signal:

[0133]

[0134] (3.4) Calculate the Kalman gain coefficient K(n) each time according to R(n), and H = 1 at this time:

[0135]

[0136] (3.5) When the Kalman gain coefficient K(n)>K(n-1), substitute it into (3.6) and (3.7) for calculation, and initialize K(0) to 0.

[0137] (3.6) Calculate the a posteriori estimate of the system state

[0138] (3.7) Calculate the covariance matrix P posterior Update, at this time I=1:

[0139] P posterior =(I-KH)P prior

[0140] (3.8) According to the above steps (3.1) to (3.7), all n observed signals are exhaustively enumerated and multiple recursive calculations are performed using the Kalman filter algorithm. The characteristic parameters after recursion of the previous signal are used as the initial values ​​of the characteristic parameters of the next signal. At this time, the signal value is the value after filtering out various complex interference noises.

[0141] like Figure 3 As shown, the data of this example is obtained by an airborne optical atmospheric data system. The principle is that the emitted laser interacts with the atmospheric aerosol particles to generate backscattered light, and the backscattered light is a velocity signal carrying a Doppler frequency shift. After system reception, AD acquisition, and FPGA preprocessing, the spectrum signal is obtained. The obtained signal contains a variety of complex noise signals (because the laser hits the mountain, the accompanying aircraft, rain, snow, frost and various noises). The characteristic parameters of the algorithm are adaptively adjusted by the present invention to effectively eliminate the above interference signals, such as Figure 4As shown, the dotted line is a signal containing multiple noises, and the solid line is a signal after adaptive filtering.

[0142] Description of related symbols:

[0143] A priori estimate of the signal value;

[0144] A posteriori estimate of the signal value;

[0145] z=f mea : Signal observation value, that is, finding the peak value at the current moment;

[0146] P prior : Prior estimated covariance matrix;

[0147] P posterior : posterior estimated covariance matrix;

[0148] A: state transfer matrix;

[0149] Q: process excitation noise covariance matrix;

[0150] R: covariance matrix;

[0151] K: Kalman gain.

[0152] In view of the shortcomings of the prior art, the present invention provides a method that can overcome the complex and tedious parameter tuning problem of the filtering model. At the same time, in view of the problem that the optical atmospheric data system detects multiple interference components in a complex environment, a method for adaptive multi-noise filtering of helicopter optical atmospheric data using Kalman filter characteristic parameter optimization design is proposed to filter out random noise, environmental noise, helicopter rotor interference and other noises in the signal.

[0153] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A helicopter optical atmospheric data system adaptive filtering method, characterized in that: The following steps are involved: S1, searching for peaks on the preprocessed spectrum data to obtain multiple peaks that meet the signal-to-noise ratio threshold; S2, establish a priori signal estimation model to automatically update the characteristic parameters of the state space; S3, using the characteristic parameters of step S2 to establish a state space model, bringing all the peak values ​​in S1 into the state space model in turn and using the Kalman filter algorithm to filter out noise and estimate effective signals; The step S2 comprises the following steps: S2.

1. Initialize parameters based on the first m prior signal values ​​and determine the estimated error covariance and the initial signal state value ; in: is the a priori estimated signal value; S2.2, construct the state transfer matrix A, and use the prior estimation signal to update the excitation noise covariance matrix Q; Where N is the sequence number of the signal at the previous moment; S2.3, use the first n prior signal data to obtain the estimated value of the signal at this time : ； The step S3 comprises the following steps: S3.1, using the peak-finding algorithm, sequentially substitute the n observed signals z that meet the peak-finding requirements into the state space model for calculation; S3.

2. Calculate the prior estimated covariance ; in: : posterior estimated covariance matrix; A is a system parameter, which is a matrix for the multi-model system and 1 for the system of the present invention; A T is the transposed matrix of A; S3.3, determining the observation noise covariance R(n) of each iteration according to each observation signal; in: is the a posteriori estimate of the signal; S3.

4. Calculate the Kalman gain coefficient K(n) each time according to R(n), and H=1 at this time; Where: H is the parameter of the measurement system. For a multi-measurement system, it is a matrix, and for the system of the present invention, it is 1; H T is the transposed matrix of H; S3.

5. When the Kalman gain coefficient K(n)>K(n-1), substitute the parameters of steps S3.2-S3.4 into steps S3.6 and S3.7 for calculation, and initialize K(0) to 0; S3.

6. Calculate the a posteriori estimate of the system state : S3.

7. Calculate the covariance matrix Update, when I=1: Where: I is the unit diagonal matrix; H is a parameter of the measurement system. For a multi-measurement system, it is a matrix, and for the system of the present invention, it is 1; According to the above steps S3.1~S3.7, all n observed signals are exhaustively enumerated and multiple recursive calculations are performed using the Kalman filter algorithm. The characteristic parameters after recursion of the previous signal are used as the initial values ​​of the characteristic parameters of the next signal, and finally an effective estimation value is obtained by filtering out various complex interference noises.

2. The helicopter optical air data system adaptive filtering method according to claim 1, characterized in that: The step S1 comprises the following steps: S1.1, perform Gaussian smoothing on the spectrum data, and then perform first-order and second-order derivatives on the spectrum data; S1.

2. Calculate the average value of the spectrum data, set a threshold based on the average value, and obtain the peak value.

3. A helicopter optical air data system adaptive filtering method according to claim 1 or 2, characterized in that: The peak finding method in step S1 is the derivative method.

4. The helicopter optical air data system adaptive filtering method according to claim 1, characterized in that: In the step S2, multiple priori signal values ​​are substituted into the priori signal estimation model to automatically update the characteristic parameters.

Images