Spectral graph local phase coding feature extraction method in multi-rotor unmanned aerial vehicle recognition

Abstract

The present application belongs to the technical field of multi-rotor unmanned aerial vehicle target recognition, and particularly relates to a spectrum local phase coding feature extraction method in multi-rotor unmanned aerial vehicle recognition. Firstly, the present application carries out short-time Prouhet-Thue-Morse transformation on the radar echo data sequence of the unmanned aerial vehicle target to obtain a time-frequency spectrum image, then divides the time-frequency spectrum image into a plurality of non-overlapping sub-blocks, calculates the phase value of each sub-block element, transforms the element phase value into binary according to a certain rule, XORs the element phase binary code with the code of the center element of the sub-block, and forms a coding sequence in counterclockwise order, all the sub-block coding sequences form a vector, which is taken as the spectrum phase local coding feature to complete the recognition of the multi-rotor unmanned aerial vehicle. Since the phase information in the sub-block structure is fully utilized, the structural shape feature of the target can be extracted from the radar echo data, thereby improving the target recognition rate.

Description

Technical Field

[0001] This invention belongs to the field of multi-rotor unmanned aerial vehicle (UAV) target recognition technology, specifically relating to a method for extracting spectral local phase coding features in multi-rotor UAV recognition. Background Technology

[0002] Currently, transformation methods mainly extract target classification features by establishing a transformation matrix based on a training dataset of UAV targets according to certain criteria. However, conventional transformation methods can only extract global structural features from an overall perspective, while ignoring local structural features that reflect target details and failing to fully utilize the differences in target details. Therefore, there is room for improvement in the target recognition rate of conventional multi-rotor UAV transformation recognition methods. Summary of the Invention

[0003] The purpose of this invention is to propose a method for extracting local phase coding features from a spectrogram. By dividing the time-spectrum into non-overlapping sub-blocks and encoding the phase of elements in the sub-blocks according to certain rules, the method can better characterize the structural shape information of the target, increase the degree of difference in structural details of the target, and thus improve the target recognition rate.

[0004] The technical solution of this invention is as follows:

[0005] A method for extracting spectral local phase coding features in multi-rotor UAV identification includes the following steps:

[0006] S1. Define the acquired radar echo training data sequence of the multi-rotor UAV as an n-dimensional column vector x. ij i = 1, 2, ... g, j = 1, 2, ... N i Where i represents the category of drone, g represents the number of categories, and N i Let represent the number of training samples for the i-th type of UAV target, then the total number of training samples is .

[0007] S2, training sample data x for the i-th type of UAV target. ij Perform a short-time Fourier transform to obtain the time-frequency spectrum S. ij :

[0008] S ij =[s ij,km ] K×M

[0009] Among them, s ij,km Represents the time spectrum diagram S ij In the array, k = 1, 2, ..., K, m = 1, 2, ..., M, k is the row subscript indicating the direction of frequency change, m is the column subscript indicating the direction of time change, and K is the S... ij The number of rows, M is S ijThe number of columns;

[0010] S3, the time spectrum diagram S ij Divide into sub-blocks and calculate the phase values ​​of the sub-block elements:

[0011] Using a 3x3 window in the time-spectrum plot S ij Slide the elements in the middle to obtain multiple sub-blocks, forming a sub-block set P. ij :

[0012] P ij =[Q ij,1 Q ij,2 … Q ij,L ]

[0013] Q ij,l =[q ij,l,ef ] 3×3 e,f=1,2,3

[0014] l = 1, 2, ..., L

[0015] Among them, Q ij,l Represents the subset P ij The l-th sub-block matrix in, q ij,l,ef The sub-block matrix Q is represented in the middle. ij,l In the middle element, L represents the total number of sub-blocks;

[0016] Calculate sub-block element q ij,l,ef phase

[0017]

[0018] Where Im(·) and Re(·) represent taking the imaginary and real parts of the complex number, respectively;

[0019] S4. Encode the phase of the sub-block column elements of the time-spectrum graph to obtain local phase coding features:

[0020] According to the magnitude of the element phase, the element q ij,l,ef phase Binarization:

[0021]

[0022] in, Indicates phase The value after binarization;

[0023] Binarize the phase of the center element of the sub-block. XOR the binarized phase values ​​of other elements in the sub-block:

[0024]

[0025] Where xor represents the exclusive OR operation. This represents the phase value after XOR;

[0026] Will The following sequences are used to form an encoding sequence:

[0027]

[0028] in, Indicates sub-block Q ij,l The corresponding local phase encoding sequence;

[0029] S5. Assemble the local phase coding sequences of all sub-blocks into a vector:

[0030]

[0031] The obtained vector c ij For data sequence x ij The local phase coding features of the spectrum.

[0032] After obtaining the features, a minimum distance classifier can be used to identify multi-rotor UAV targets.

[0033] The beneficial effects of this invention are as follows: First, a short-time Fourier transform is performed on the radar echo data sequence of a UAV target to obtain a time-spectrum image. Then, the time-spectrum image is divided into multiple non-overlapping sub-blocks. The phase value of each element in the sub-block is calculated, and the element phase values ​​are binarized according to certain rules. The binary code of the element phase is XORed with the code of the central element of the sub-block, and a coding sequence is formed in counterclockwise order. All sub-block coding sequences form a vector, which serves as the local coding feature of the spectrum phase, thus completing the identification of multi-rotor UAVs. By fully utilizing the phase information in the sub-block structure, structural shape features of the target can be extracted from the radar echo data, thereby improving the target recognition rate. Simulation results on four types of multi-rotor UAVs verify the effectiveness of the method. Detailed Implementation

[0034] The following simulations demonstrate the effectiveness of this invention and the progress achieved:

[0035] The simulation experiment designed four types of UAVs, including a tri-rotor UAV, a quadcopter UAV, a hexacopter UAV, and an octacopter UAV. Their simulation parameters are shown in Table 1. The simulated radar parameters included: radar carrier frequency of 24 GHz; pulse repetition frequency of 100 kHz; target-radar distance of 200 m; and UAV elevation angle of 10° and azimuth angle of 30° relative to the radar.

[0036] Table 1 Simulation parameters for four types of UAVs

[0037]

[0038]

[0039] For each target class, radar echo signals were recorded for 10 seconds and divided into segments of fixed length 0.05 seconds (containing at least one rotation cycle). The overlap between segments was 50%. Each segment contained 0.05 × 100,000 = 5000 radar echo sampling data points, resulting in 400 segments per class. 200 segments were randomly selected from these 400 segments as the training dataset, and the remaining 200 segments as the test dataset. Therefore, the training dataset for the four target classes comprised a total of 800 segments, and the test dataset also comprised 800 segments. For the selected training dataset, spectral local phase coding features were extracted using the method presented in this paper to establish a template library. Similarly, spectral local phase coding features were extracted from the test samples. A minimum distance classifier was used for classification, achieving an average correct recognition rate of 98% for the four types of multi-rotor UAVs. The signal-to-noise ratio was 10 dB.

Claims

1. A method for extracting spectral local phase coding features in multi-rotor UAV identification, characterized in that, Includes the following steps: S1. Define the acquired radar echo training data sequence of the multi-rotor UAV as an n-dimensional column vector x. ij i = 1, 2, ... g, j = 1, 2, ... N i Where i represents the category of drone, g represents the number of categories, and N i Let represent the number of training samples for the i-th type of UAV target, then the total number of training samples is . S2, training sample data x for the i-th type of UAV target. ij Perform a short-time Fourier transform to obtain the time-frequency spectrum S. ij : S ij =[s ij,km ] K×M Among them, s ij,km Represents the time spectrum diagram S ij In the array, k = 1, 2, ..., K, m = 1, 2, ..., M, k is the row subscript indicating the direction of frequency change, m is the column subscript indicating the direction of time change, and K is the time index. ij The number of rows, M is S ij The number of columns; S3, the time spectrum diagram S ij Divide into sub-blocks and calculate the phase values ​​of the sub-block elements: Using a 3x3 window in the time-spectrum plot S ij Slide the elements in the middle to obtain multiple sub-blocks, forming a sub-block set P. ij : P ij =[Q ij,1 Q ij,2 … Q ij,L ] Q ij,l =[q ij,l,ef ] 3×3 ,e,f=1,2,3 l＝1,2,…L Among them, Q ij,l Represents the subset P ij The l-th sub-block matrix in, q ij,l,ef The sub-block matrix Q is represented in the middle. ij,l In the middle element, L represents the total number of sub-blocks; Calculate sub-block element q ij,l,ef phase Where Im(·) and Re(·) represent taking the imaginary and real parts of the complex number, respectively; S4. Encode the phase of the sub-block column elements of the time-spectrum graph to obtain local phase coding features: Based on the magnitude of the element phase, assign elements q... ij,l,ef phase Binarization: in, Indicates phase The value after binarization; Binarize the phase of the center element of the sub-block. XOR the binarized phase values ​​of other elements in the sub-block: e,f=1,3 Where xor represents the exclusive OR operation. This represents the phase value after XOR; Will The following sequences are used to form an encoding sequence: in, Indicates sub-block Q ij,l The corresponding local phase encoding sequence; S5. Assemble the local phase coding sequences of all sub-blocks into a vector: The obtained vector c ij For the data sequence x ij The local phase encoding features of the spectrum.

Images