Target pose angle estimation method and device based on hidden markov model

Abstract

The present application relates to the technical field of attitude angle estimation, in particular to a target attitude angle estimation method and device based on a hidden Markov model. The method comprises: taking the attitude angle of a target as the hidden state of a hidden Markov model and taking the RCS value as the observation state of the hidden Markov model to establish the initial state probability matrix, the state transition probability matrix and the observation probability matrix of the hidden Markov model; for each RCS time sequence to be estimated, the following is performed: after inputting the current RCS time sequence into the hidden Markov model, the RCS time sequence is decoded using the Viterbi decoding algorithm to obtain the attitude angle sequence of the target. This scheme regards the motion process of the target as a hidden Markov process, and after modeling the hidden Markov model, the Viterbi decoding algorithm can be used to calculate the attitude angle sequence with the maximum probability corresponding to each RCS time sequence to be estimated, thereby effectively solving the problem that the target attitude angle is difficult to estimate.

Description

Technical Field

[0001] This invention relates to the field of attitude angle estimation technology, and in particular to a target attitude angle estimation method and apparatus based on a hidden Markov model. Background Technology

[0002] Most cylindrical and conical targets are roll-symmetric structures about their central axis, allowing attitude angle estimation to be simplified from two dimensions to one, thus simplifying the estimation process. However, radar target RCS is highly attitude-sensitive; small changes in attitude angle can cause drastic changes in RCS. Furthermore, RCS exhibits a multi-valued effect, meaning that multiple attitude angles correspond to the same or similar RCS values, all of which complicate attitude angle estimation. Most critically, there is a discrepancy between the radar-measured RCS during flight and the static RCS calculated through anechoic chamber measurements or simulations—the inconsistency between dynamic and static data—which is the direct cause of the difficulty in accurately estimating attitude angles.

[0003] Therefore, there is an urgent need for a target attitude angle estimation method based on a hidden Markov model. Summary of the Invention

[0004] To address the current lack of effective attitude angle estimation methods, this invention provides a target attitude angle estimation method and apparatus based on a hidden Markov model.

[0005] In a first aspect, embodiments of the present invention provide a target attitude angle estimation method based on a hidden Markov model, the method comprising:

[0006] The target's attitude angle is used as the hidden state of the Hidden Markov Model, and the RCS value is used as the observed state of the Hidden Markov Model to establish the initial state probability matrix, state transition probability matrix, and observation probability matrix of the Hidden Markov Model.

[0007] For each RCS time series to be estimated, perform the following:

[0008] After inputting the current RCS time series into the Hidden Markov Model, the Viterbi decoding algorithm is used to decode the RCS time series to obtain the target's attitude angle sequence.

[0009] Secondly, embodiments of the present invention also provide a target attitude angle estimation device based on the method described in any embodiment of this specification, the device comprising:

[0010] A unit is established to use the target's attitude angle as the hidden state of the Hidden Markov Model and the RCS value as the observed state of the Hidden Markov Model to establish the initial state probability matrix, state transition probability matrix and observation probability matrix of the Hidden Markov Model.

[0011] The estimation unit performs the following steps for each RCS time series to be estimated: after inputting the current RCS time series into the Hidden Markov Model, it decodes the RCS time series using the Viterbi decoding algorithm to obtain the target's attitude angle sequence.

[0012] Thirdly, embodiments of the present invention also provide a computing device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method described in any embodiment of this specification.

[0013] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods described in any embodiment of this specification.

[0014] This invention provides a method and apparatus for estimating target attitude angles based on a Hidden Markov Model (HMM). Based on the correspondence between the radar target's attitude angle and its Relative Cross Section (RCS) value, the target's motion process is viewed as a HMM process, the target's attitude angle is considered the hidden state of the HMM, and the RCS is considered the observed state of the HMM. This allows for the establishment of the initial state probability matrix, state transition probability matrix, and observation probability matrix within the HMM. The state transition probability matrix describes the target's motion pattern, the observation probability matrix describes the target's statistical characteristics within a certain angle range, and the initial state probability matrix describes the probability of the target's attitude angle at the initial moment. After modeling the three probability distributions, the Viterbi decoding algorithm is used to calculate the attitude angle sequence with the highest probability corresponding to each RCS time series to be estimated. This solution effectively addresses the problem of difficulty in estimating target attitude angles. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a target attitude angle estimation method based on a hidden Markov model provided in an embodiment of the present invention;

[0017] Figure 2 This is a hardware architecture diagram of a computing device provided in an embodiment of the present invention;

[0018] Figure 3This is a structural diagram of a target attitude angle estimation device based on a hidden Markov model provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] The following describes the specific implementation of the above concept.

[0021] Please refer to Figure 1 This invention provides a target attitude angle estimation method based on a hidden Markov model, the method comprising:

[0022] Step 100: Use the target's attitude angle as the hidden state of the Hidden Markov Model and the RCS value as the observed state of the Hidden Markov Model to establish the initial state probability matrix, state transition probability matrix and observation probability matrix of the Hidden Markov Model.

[0023] Step 102: For each RCS time series to be estimated, the following steps are performed: After inputting the current RCS time series into the Hidden Markov Model, the Viterbi decoding algorithm is used to decode the RCS time series to obtain the target's attitude angle sequence.

[0024] In this embodiment of the invention, based on the correspondence between the radar target's attitude angle and RCS value, the target's motion process is considered as a Hidden Markov Process (HMM), the target's attitude angle is considered as the hidden state of the HMM, and the RCS is considered as the observed state of the HMM. This allows for the establishment of the initial state probability matrix, state transition probability matrix, and observation probability matrix in the HMM. The state transition probability matrix describes the target's motion pattern, the observation probability matrix describes the target's statistical characteristics within a certain angle range, and the initial state probability matrix describes the probability of the target's attitude angle at the initial moment. After modeling the three probability distributions, the Viterbi decoding algorithm can be used to calculate the attitude angle sequence with the highest probability corresponding to each estimated RCS time series. This scheme can effectively solve the problem of difficulty in estimating the target's attitude angle.

[0025] For step 100:

[0026] In some embodiments, step 100 may include:

[0027] Based on the angle range and sampling rate of the static RCS, the number of hidden states in the Hidden Markov Model is determined; where each hidden state is an attitude angle.

[0028] Assign the quotient of 1 and the number of hidden states to each element in the initial state probability matrix to obtain the initial state probability matrix of the Hidden Markov Model.

[0029] Obtain several attitude angle sequence samples, perform statistics on the attitude angle sequence samples, and establish the state transition probability matrix of the hidden Markov model;

[0030] The RCS signal and signal-to-noise ratio measured at each attitude angle are obtained, and the RCS signal and signal-to-noise ratio are substituted into the pre-determined detection distribution formula of the signal when the RCS target is detected to obtain the observation probability distribution corresponding to each attitude angle, and thus the observation probability matrix of the Hidden Markov Model.

[0031] In this embodiment, the initial state probability matrix describes the probability distribution of the target's attitude angle at the initial moment of radar illumination. However, since the initial moment of radar illumination of the target is unknown, that is, the range of the target's initial attitude angle is uncertain and it may be at any attitude angle, the establishment of the initial state probability matrix is ​​relatively difficult.

[0032] Here we can assume that the target's attitude angle ranges from 0 to 180° at the initial moment. Therefore, the target's attitude angle at the initial moment may be in any state between 0 and 180°, and the probability distribution of the initial state is an average distribution.

[0033] Assuming the static RCS angle interval is 1°, then the number of hidden states N′ in the Hidden Markov Model is 181, and the initial probability distribution is:

[0034]

[0035] π i Let be the value of each element in the initial state probability matrix of the Hidden Markov Model. It can be understood that the dimension of the initial state probability matrix is ​​1*N′.

[0036] In some implementations, the step "acquiring several attitude angle sequence samples, and statistically analyzing the attitude angle sequence samples to establish the state transition probability matrix of the hidden Markov model" may include:

[0037] Count the number of times each hidden state appears in the attitude angle sequence samples;

[0038] Based on each pair of adjacent hidden states in the attitude angle sequence sample, count the number of times each hidden state transitions to each other.

[0039] Based on the number of times each hidden state transitions to each hidden state and the number of times each hidden state appears, a state transition probability matrix of the Hidden Markov Model is established.

[0040] In this embodiment, the state transition probability matrix describes the probability distribution of a target transitioning from one attitude angle to another during flight. Since a target needs to maintain stability during high-speed motion, its attitude angles are often relatively stable. At the next moment, the attitude angle state will either transition to an adjacent attitude angle or remain unchanged. Based on this principle, the state transition probability matrix can be established. Furthermore, it can be seen that the state transition probability matrix is ​​related to the radar sampling rate. If the radar sampling rate is high, the attitude angle difference between adjacent time intervals will be smaller; if the radar sampling rate is low, the attitude angle difference between adjacent time intervals may be relatively larger. Therefore, the radar sampling rate factor needs to be considered when establishing the state transition probability matrix.

[0041] Given sufficient training data, statistical methods can be used to generate the state transition probability matrix directly. The obtained attitude angle sequence samples should include as many attitude angles as possible. Only by including as many attitude angles as possible can the possible state transitions be included, and the resulting state transition probability matrix can have better generalization performance.

[0042] We can count the number of times each hidden state, i.e., each attitude angle i, appears in the sequence of attitude angle samples, N. i Then, based on every two adjacent hidden states in the pose angle sequence samples, the number of times N is counted for each hidden state i to transition to each hidden state j. ij Then the probability of changing from attitude angle i to attitude angle j is:

[0043]

[0044] This formula can be used to calculate the value of each element in the state transition probability matrix of a Hidden Markov Model. It can be understood that the dimension of the state transition probability matrix is ​​N′*N′.

[0045] Furthermore, for non-cooperative targets, the target's attitude angles are often unavailable or only a very limited sequence is obtained. This makes it impossible to obtain a state transition probability matrix with good generalization performance through statistical methods. Therefore, it is necessary to establish the state transition probability matrix by identifying the target's motion patterns. With a sufficient sampling rate, the range of attitude angle changes between the current and next time steps during stable flight is relatively small. Based on existing attitude angle sequence samples, a relatively scientific state transition probability matrix can be fitted.

[0046] In some implementations, the signal detection distribution formula for RCS target detection is:

[0047]

[0048] in,

[0049] rcs=σ0 2

[0050] SNR=σ0 / N

[0051] In the formula, σ is the detected signal, σ0 is the static RCS signal, rcs is the static RCS value, SNR is the signal-to-noise ratio, N is the noise, and I0(z) is the zero-order modified Bessel function with parameter z.

[0052] The observation probability distribution describes the RCS distribution under each attitude angle state. The most important part of this part is to establish the relationship between attitude angle and RCS. The relationship between attitude angle and RCS can be shown from the target static data. Assuming the angle interval is 0.1°, there are 1801 points from 0 to 180°, and each angle corresponds to an RCS value. In this way, the attitude angle and RCS have a one-to-one correspondence. However, in the radar detection process, due to the effect of noise, the target's attitude angle and RCS are not a one-to-one correspondence, but rather present a probability distribution.

[0053] Based on target detection theory, when a sinusoidal signal of amplitude A and noise coexist, the probability density function of the envelope for target detection is a Rice distribution, which can determine the detection distribution formula of the signal for RCS target detection as shown above.

[0054] Based on the RCS signal and signal-to-noise ratio measured at each attitude angle of the target under static conditions, the relationship between attitude angle and RCS can be established. From the perspective of signal detection, the RCS distribution corresponding to each attitude angle can be established. Substituting the RCS signal and signal-to-noise ratio measured at each attitude angle into the predetermined RCS target detection signal detection distribution formula, the observation probability distribution corresponding to each attitude angle can be obtained, and the observation probability matrix of the Hidden Markov Model can be obtained.

[0055] At this point, the initial state probability matrix, state transition probability matrix, and observation probability matrix of the Hidden Markov Model can be established.

[0056] Regarding step 102:

[0057] In some implementations, step 102 may include:

[0058] Determine the maximum time T of the current RCS time series;

[0059] The probability of being in each hidden state at each time step is expressed as:

[0060] δ t (i)=P{σ1,σ2,…,σ t ,s1,s2,…,s t =i|(A,B,π)}

[0061] In the formula, A is the state transition probability matrix of the Hidden Markov Model, B is the observation probability matrix of the Hidden Markov Model, π is the initial state probability matrix of the Hidden Markov Model, σ is the RCS value at each time step in the current RCS time series, and s t Hide the state at each moment;

[0062] At the initial moment, the probability is calculated from the initial state probability matrix and the observation probability matrix, i.e.

[0063] δ1(i)=π i b i (σ1)

[0064] In the formula, δ1(i) represents the probability of each hidden state at the initial time, and π i Let b be the element corresponding to each hidden state in the initial state probability matrix. i (σ1) represents the observation probability distribution in each hidden state at the initial time.

[0065] Understandable, b i (σ1) is the result obtained by substituting the square root of the RCS value σ1 at the initial time in the current RCS time series as σ0 into the detection distribution formula for each hidden state.

[0066] Perform time recursion for t = 2, 3, ..., T:

[0067]

[0068]

[0069] In the formula, δ t (i) represents the maximum probability at each time step t = 2, 3, ..., T, a ji Ψ represents the probability value of transitioning from hidden state j to hidden state i in the state transition probability matrix, N′ represents the number of hidden states in the Hidden Markov Model, and Ψ represents the number of hidden states. t (i) represents the hidden state corresponding to the maximum probability at each time step t = 2, 3, ..., T;

[0070] Obtain the hidden state with the highest probability at the maximum time T of the current RCS time series:

[0071]

[0072]

[0073] In the formula, P * i represents the maximum probability at time T. T * This represents the hidden state with the highest probability at time T.

[0074] Perform optimal path backtracking and calculate the hidden state corresponding to the maximum probability at each time step t = T-1, T-2, ..., 1 using the following formula:

[0075]

[0076] Obtain the target's attitude angle sequence

[0077] In this embodiment, the probability of each hidden state (attitude angle) at the initial moment can be calculated first using the initial state probability matrix and the observation probability matrix. Then, the maximum probability δ at each subsequent moment can be calculated through time recursion. t (i) and determine the hidden state corresponding to the maximum probability at each time point t = 2, 3, ..., T. In this way, determine the hidden state with the highest probability at the maximum time point T of the current RCS time series. Then, through optimal path backtracking, determine the hidden state corresponding to the maximum probability at each time point t = T-1, T-2, ..., 1, and obtain the optimal attitude angle sequence corresponding to the current RCS time series.

[0078] like Figure 2 , Figure 3 As shown, this invention provides a target attitude angle estimation device based on a hidden Markov model. The device can be implemented in software, hardware, or a combination of both. From a hardware perspective, as... Figure 2 The diagram shown is a hardware architecture diagram of a computing device housing a target attitude angle estimation device based on a hidden Markov model, provided in an embodiment of the present invention. (Except for...) Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown, the computing device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing packets. Taking software implementation as an example, such as... Figure 3 As shown, a device, in a logical sense, is formed by the CPU of its computing device reading the corresponding computer program from non-volatile memory into memory and running it. This embodiment provides a target attitude angle estimation device based on a Hidden Markov Model, the device comprising:

[0079] Unit 301 is established to use the target's attitude angle as the hidden state of the Hidden Markov Model and the RCS value as the observed state of the Hidden Markov Model to establish the initial state probability matrix, state transition probability matrix and observation probability matrix of the Hidden Markov Model.

[0080] The estimation unit 302 is used to perform the following for each RCS time series to be estimated: after inputting the current RCS time series into the Hidden Markov Model, the Viterbi decoding algorithm is used to decode the RCS time series to obtain the target's attitude angle sequence.

[0081] The information interaction and execution process between the various units in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.

[0082] This invention also provides a computing device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a target attitude angle estimation method based on a hidden Markov model according to any embodiment of this invention.

[0083] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program causes the processor to perform a target attitude angle estimation method based on a hidden Markov model according to any embodiment of this invention.

[0084] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0085] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0086] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0087] It should be clear that not only can the program code read by the computer be executed, but also the operating system or other components on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby achieving the function of any of the embodiments described above.

[0088] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.

[0089] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0090] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as ROM, RAM, magnetic disk, or optical disk.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A target attitude angle estimation method based on a hidden Markov model, characterized in that, include: The target's attitude angle is used as the hidden state of the Hidden Markov Model, and the RCS value is used as the observed state of the Hidden Markov Model to establish the initial state probability matrix, state transition probability matrix, and observation probability matrix of the Hidden Markov Model. For each RCS time series to be estimated, perform the following: After inputting the current RCS time series into the Hidden Markov Model, the Viterbi decoding algorithm is used to decode the RCS time series to obtain the target's attitude angle sequence. The method of establishing the initial state probability matrix, state transition probability matrix, and observation probability matrix of the Hidden Markov Model (HMM) by using the target's attitude angle as the hidden state and the RCS value as the observed state of the HMM includes: Based on the angle range and sampling rate of the static RCS, the number of hidden states in the Hidden Markov Model is determined; wherein each hidden state is an attitude angle. Assign the quotient of 1 and the number of hidden states to each element of the initial state probability matrix to obtain the initial state probability matrix of the hidden Markov model. Acquire several attitude angle sequence samples, and perform statistics on the attitude angle sequence samples to establish the state transition probability matrix of the hidden Markov model; The RCS signal and signal-to-noise ratio measured at each attitude angle are obtained, and the RCS signal and signal-to-noise ratio are substituted into the pre-determined detection distribution formula of the signal when the RCS target is detected to obtain the observation probability distribution corresponding to each attitude angle, and thus the observation probability matrix of the hidden Markov model is obtained. The detection distribution formula for the signal during RCS target detection is as follows: in, In the formula, For the detected signal, It is a static RCS signal. This is the static RCS value. Where N is the signal-to-noise ratio and N is the noise level. Let z be the zeroth-order modified Bessel function with parameter z.

2. The method according to claim 1, characterized in that, The step of acquiring several attitude angle sequence samples and statistically analyzing these samples to establish the state transition probability matrix of the hidden Markov model includes: Count the number of times each hidden state appears in the attitude angle sequence sample; Based on each pair of adjacent hidden states in the attitude angle sequence sample, the number of times each hidden state transitions to each other is counted. The state transition probability matrix of the Hidden Markov Model is established based on the number of times each hidden state transitions to each hidden state and the number of times each hidden state appears.

3. The method according to claim 1, characterized in that, The step of inputting the current RCS time series into the Hidden Markov Model and then decoding the RCS time series using the Viterbi decoding algorithm to obtain the target's attitude angle sequence includes: Determine the maximum moment of the current RCS time series. ; The probability of each hidden state at each time step is expressed as: In the formula, A is the state transition probability matrix of the hidden Markov model. Let be the observation probability matrix of the hidden Markov model. Let be the initial state probability matrix of the hidden Markov model. This represents the RCS value at each time point in the current RCS time series. Hide the state at each moment; At the initial moment, the probability is calculated from the initial state probability matrix and the observation probability matrix, i.e. In the formula, Let be the probability of each of the hidden states at the initial time. For each of the hidden states in the initial state probability matrix, represents an element. This represents the observation probability distribution for each of the hidden states at the initial time. Perform time recursion, for : In the formula, for The maximum probability at each moment. Let be the probability value in the state transition probability matrix corresponding to the transition from hidden state j to hidden state i. This represents the number of hidden states in the Hidden Markov Model. for The hidden state corresponding to the maximum probability at each time step; Get the maximum moment of the current RCS time series The hidden state with the highest probability: In the formula, For the maximum moment The maximum probability at that time, For the maximum moment The hidden state corresponding to the highest probability at that time; Optimal path backtracking is performed using the following formula. The hidden state corresponding to the maximum probability at each time step: Obtain the target's attitude angle sequence .

4. A target attitude angle estimation device based on the method of any one of claims 1-3, characterized in that, include: A unit is established to use the target's attitude angle as the hidden state of the Hidden Markov Model and the RCS value as the observed state of the Hidden Markov Model to establish the initial state probability matrix, state transition probability matrix and observation probability matrix of the Hidden Markov Model. The estimation unit performs the following steps for each RCS time series to be estimated: after inputting the current RCS time series into the Hidden Markov Model, it decodes the RCS time series using the Viterbi decoding algorithm to obtain the target's attitude angle sequence.

5. A computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the method as described in any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-3.

Images