Drone direction finding method, device, storage medium and product

Abstract

The application discloses a kind of unmanned aerial vehicle direction finding method, equipment, storage medium and product, the method includes the time-frequency diagram of adjacent receiving channel is carried out cross-correlation calculation, obtains the cross-correlation data of each receiving channel group;Cross-correlation data of each receiving channel group is accumulated, and then the modulus value and phase of cumulative cross-correlation data are obtained, and cross-correlation energy diagram and phase difference diagram are obtained;Each receiving channel group is carried out target identification to cross-correlation energy diagram, and target type and target detection frame are obtained;According to target type and phase difference diagram, determine final target detection frame;According to final target detection frame and cross-correlation energy diagram, determine the strongest channel group;According to the cross-correlation energy diagram of the strongest channel group, calculate amplitude comparison angle, and calculate phase difference according to the phase difference diagram of the strongest channel group;According to amplitude comparison angle and phase difference, the phase comparison direction finding of unmanned aerial vehicle is carried out.The application realizes unmanned aerial vehicle identification and accurate direction finding under low signal-to-noise ratio.

Description

Technical Field

[0001] This invention belongs to the field of UAV detection technology, and in particular relates to a UAV direction finding method, device, storage medium and product based on cumulative cross-correlation. Background Technology

[0002] In recent years, the use of drones in commercial and military fields has become increasingly apparent, thus creating an urgent need for effective and economical drone control measures. Generally, omnidirectional electronic reconnaissance combined with directional jamming equipment is considered the simplest and most effective method.

[0003] Currently, UAV detection primarily employs time-frequency domain analysis methods. Deep learning is used to analyze the time-frequency map data from each channel to locate UAV targets and identify signal frequency bands. The pulse width is then calculated using the original time-frequency map data to determine the azimuth. However, when the UAV is far from the receiver, its weak signal is drowned out by the electromagnetic noise of the environment, making it difficult to detect the UAV signal using only the time-frequency map, leading to target detection difficulties. Since the time-frequency map of each channel has high resolution, directly using it for training would result in an excessively large network structure. Therefore, deep learning typically sacrifices some time-domain or frequency-domain resolution to reduce the size of the network parameters. The most common method for reducing resolution is simple downsampling of the time-frequency map. While this reduces training and computation time, it does not improve system performance and has limited detection capability for low signal-to-noise ratio signals. Even when weak signals are added during training, it is prone to generating numerous false alarms. Therefore, a higher threshold is usually chosen.

[0004] Therefore, even with deep learning analysis of the time-frequency data from each channel, issues such as missed alarms, false alarms, and low direction-finding accuracy still exist. Given the difficulty in target detection and the presence of missed alarms or low direction-finding accuracy, directional jamming is unlikely to effectively counter UAVs and cannot establish a large safe airspace. Optimizing the hardware radio frequency section can achieve a longer detection range, but this incurs higher costs.

[0005] For weak signal detection by UAVs, current methods employ signal enhancement techniques from image processing to reduce the energy of Gaussian white noise, which can improve the signal-to-noise ratio (SNR) in some cases. However, this method is ineffective when the SNR is very low, involves significant computation, and the data cannot be reused in subsequent steps. It does not consider whether adjacent channels can improve the results, nor does it provide any assistance for subsequent signal pulse width calculation and direction finding. The time spent is disproportionate to the results, making it impractical in real-world engineering.

[0006] To address the issue of large direction-finding errors at low signal-to-noise ratios, multiple averaging results are often considered. However, this method ignores the influence of Gaussian white noise on the overall signal phase difference, leading to a result that stabilizes near an incorrect value even after multiple averaging. Furthermore, the entire method or system was initially designed with modular processing, each module performing its own function, without considering the whole method or system as a whole. Summary of the Invention

[0007] The purpose of this invention is to provide a method, device, storage medium, and product for UAV direction finding, in order to solve the problems of difficult target detection and low direction finding accuracy when the distance between the UAV and the receiver is long.

[0008] This invention solves the above-mentioned technical problems through the following technical solution: a method for determining the direction of an unmanned aerial vehicle (UAV), comprising the following steps:

[0009] Obtain the time-frequency diagram of the UAV signal for each receiving channel;

[0010] Cross-correlation calculations were performed on the time-frequency diagrams of UAV signals from two adjacent receiving channels to obtain cross-correlation data for each receiving channel group;

[0011] The cross-correlation data of each receiving channel group is accumulated to obtain the cumulative cross-correlation data of the corresponding receiving channel group;

[0012] For each receiving channel group, the modulus and phase of the accumulated cross-correlation data are taken to obtain the cross-correlation energy map and phase difference map of the corresponding receiving channel group, respectively.

[0013] Target identification is performed on the cross-correlation energy map of each receiving channel group to obtain the target type and target detection box of the corresponding receiving channel group;

[0014] The final target detection box for each receiving channel group is determined based on the target type and phase difference map of each receiving channel group.

[0015] The strongest channel group is determined based on the final target detection box and cross-correlation energy map of each receiving channel group;

[0016] The amplitude angle of the strongest channel group is calculated based on the cross-correlation energy diagram of the strongest channel group, and the phase difference of the strongest channel group is calculated based on the phase difference diagram of the strongest channel group.

[0017] The UAV is used for phase comparison and direction finding based on the amplitude angle and phase difference of the strongest channel group.

[0018] Furthermore, the specific implementation process for obtaining the time-frequency diagram of the UAV signal for each receiving channel is as follows:

[0019] Acquire time-domain sampling data of the UAV signal for each receiving channel;

[0020] Perform a Fast Fourier Transform on the time-domain sampled data of the UAV signal for each receiving channel to obtain the time-frequency diagram of the corresponding UAV signal.

[0021] Furthermore, the specific implementation process of performing cross-correlation calculation on the time-frequency diagrams of UAV signals from two adjacent receiving channels is as follows:

[0022] Multiply the data points in the time-frequency plot of the UAV signal from the i-th receiving channel by the conjugate of the data in the time-frequency plot of the UAV signal from the (i+1)-th receiving channel to obtain the cross-correlation data of the i-th receiving channel group; where i = 1, 2, ..., N-1, and N represents the number of receiving channels of the receiver.

[0023] Furthermore, a target recognition model is used to identify targets in the cross-correlation energy map of each receiving channel group. The target recognition model is obtained by training and validating a deep learning network model using a sample dataset. Each sample data in the sample dataset includes the energy map of a single receiving channel or the cross-correlation energy map of a single receiving channel group.

[0024] Furthermore, the final target detection box for each receiving channel group is determined based on the target type and phase difference map of each receiving channel group. The specific implementation process is as follows:

[0025] When the target type is a bandwidth signal, the target detection box obtained through target recognition is the final target detection box of the corresponding receiving channel group;

[0026] When the target type is a narrowband frequency hopping signal, the gray value of each target detection box is calculated according to the phase difference map of the receiving channel group corresponding to the narrowband frequency hopping signal; the gray values ​​of all target detection boxes in the receiving channel group are clustered, and the target detection boxes are merged according to the clustering results to obtain the final target detection box of the corresponding receiving channel group.

[0027] Furthermore, the K-means clustering algorithm is used to cluster the gray values ​​of all target detection boxes in the receiving channel group. The specific implementation process is as follows:

[0028] The target detection boxes are divided into K groups, and K objects are randomly selected as the initial cluster centers;

[0029] Based on the grayscale values ​​of the target detection boxes, calculate the distance between each object and each cluster center;

[0030] Each object is assigned to the cluster center closest to it based on the distance between each object and each cluster center. Here, the cluster center and the objects assigned to the cluster center represent a cluster.

[0031] Recalculate the cluster centers of the corresponding clusters based on the objects assigned to each cluster;

[0032] Determine if the termination condition is met. If so, obtain the clustering result; otherwise, repeat the steps of distance calculation, object assignment, and cluster center calculation.

[0033] Furthermore, the strongest channel group is determined based on the final target detection box and cross-correlation energy map of each receiving channel group. The specific implementation process is as follows:

[0034] Based on the frequency band information of the final target detection box, the time domain envelope data of the UAV signal is extracted from the cross-correlation energy map of the corresponding receiving channel group;

[0035] The temporal pulse width information is determined based on the intersection of the temporal envelope data and the set amplitude threshold.

[0036] Based on the time-domain pulse width information and frequency band information, the energy data of the corresponding receiving channel group is extracted from the cross-correlation energy map of the corresponding receiving channel group;

[0037] The strongest channel group is determined based on the energy data of each receiving channel group.

[0038] Based on the same concept, the present invention also provides a terminal device, including a memory, a processor, and a computer program / instructions stored in the memory, wherein the processor executes the computer program / instructions to implement the UAV direction finding method as described above.

[0039] Based on the same concept, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the UAV direction finding method as described above.

[0040] Based on the same concept, the present invention also provides a computer program product, including a computer program / instruction that, when executed by a processor, implements the UAV direction finding method as described above.

[0041] Beneficial effects

[0042] Compared with the prior art, the advantages of the present invention are as follows:

[0043] Before target identification, this invention first performs cross-correlation calculations on the time-frequency maps of adjacent receiving channels, and then accumulates the cross-correlation data of each receiving channel group to obtain the cross-correlation energy map and phase difference map of the receiving channel group. By utilizing the correlation and energy accumulation between adjacent receiving channels, the signal-to-noise ratio of the input signal is improved. Then, the cross-correlation energy map of the receiving channel group is used to replace the energy map of the traditional single channel for target identification, thereby improving the target identification capability. After target identification, the phase difference map of the receiving channel group is used for target merging, accurately identifying multiple weak signals and completing the separation of signals from different directions of arrival. Finally, the pulse width is solved using the cross-correlation energy map, and the azimuth is solved using the phase difference map, realizing UAV identification and accurate direction finding under low signal-to-noise ratio conditions.

[0044] This invention uses cumulative cross-correlation for UAV direction finding, which has low computational complexity, reduces the overall computational load, and has a high data reuse rate. It solves for high-precision phase difference under low signal-to-noise ratio, which improves the direction finding accuracy under low signal-to-noise ratio and realizes accurate UAV detection, model matching, and precise direction finding. Attached Figure Description

[0045] To more clearly illustrate the technical solution of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only one embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart of the UAV direction finding method in an embodiment of the present invention;

[0047] Figure 2 This is the energy diagram of the first receiving channel in this embodiment of the invention, where the horizontal axis represents frequency and the vertical axis represents time.

[0048] Figure 3 This is the energy diagram of the second receiving channel in this embodiment of the invention, where the horizontal axis represents frequency and the vertical axis represents time.

[0049] Figure 4 This is the cross-correlation energy diagram of the first receiving channel group after accumulating 4 times in this embodiment of the invention, where the horizontal axis represents frequency and the vertical axis represents time.

[0050] Figure 5 This is the phase difference diagram of the first receiving channel group after accumulating 4 times in this embodiment of the invention, where the horizontal axis represents frequency and the vertical axis represents time.

[0051] Figure 6 This is a cross-correlation energy diagram of the narrowband frequency hopping signal of the receiving channel group in an embodiment of the present invention, where the horizontal axis represents frequency and the vertical axis represents time;

[0052] Figure 7 This is a phase difference diagram of the narrowband frequency hopping signal of the receiving channel group in an embodiment of the present invention, where the horizontal axis represents frequency and the vertical axis represents time;

[0053] Figure 8 This refers to the time-domain envelope data extracted from the energy map of a single receiving channel in this embodiment of the invention;

[0054] Figure 9 This refers to the time-domain envelope data extracted from the cross-correlation energy map of the channel group in this embodiment of the invention;

[0055] Figure 10 This is a frequency domain diagram of a single receiving channel in an embodiment of the present invention, where the horizontal axis represents frequency and the vertical axis represents amplitude;

[0056] Figure 11 This is a frequency domain diagram after the cumulative cross-correlation of the channel groups in this embodiment of the invention, where the horizontal axis represents frequency and the vertical axis represents amplitude. Detailed Implementation

[0057] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] The technical solutions of this application will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0059] like Figure 1 As shown, the UAV direction-finding method provided in this embodiment of the invention includes the following steps:

[0060] Step S1: Obtain the time-frequency diagram of the UAV signal for each receiving channel.

[0061] The application scenario of this invention includes at least one unmanned aerial vehicle (UAV) and a device for detecting at least one UAV. The device includes an antenna, a receiver, and a signal processor. The antenna is used to acquire radio signals (i.e., UAV signals) emitted by the UAV during flight. The receiver includes N receiving channels, each acquiring time-domain sampling data of the UAV signals transmitted by the antenna. The signal processor performs direction finding on at least one UAV based on the received time-domain sampling data of the UAV signals. The UAV is an unmanned aerial vehicle capable of flying in any airspace, such as low altitude, medium altitude, or high altitude, including fixed-wing UAVs, rotary-wing UAVs, unmanned airships, paragliding UAVs, flapping-wing UAVs, etc.

[0062] In this embodiment, the specific implementation process for obtaining the time-frequency diagram of the UAV signal for each receiving channel is as follows:

[0063] Acquire the time-domain sampling data of the UAV signal for each receiving channel; perform a fast Fourier transform on the time-domain sampling data of the UAV signal for each receiving channel to obtain the time-frequency diagram of the corresponding UAV signal.

[0064] For example, performing a Fast Fourier Transform on every 1024 data points of the time-domain sampled data, for a total of 4096 calculations, yields a 1024*4096 time-frequency graph.

[0065] Step S2: Perform cross-correlation calculation on the time-frequency diagrams of UAV signals from two adjacent receiving channels to obtain cross-correlation data for each receiving channel group.

[0066] In this embodiment, the specific implementation process of cross-correlation calculation is as follows:

[0067] Multiply the data points in the time-frequency plot of the UAV signal from the i-th receiving channel by the conjugate of the data in the time-frequency plot of the UAV signal from the (i+1)-th receiving channel to obtain the cross-correlation data of the i-th receiving channel group; where i = 1, 2, ..., N-1, and N represents the number of receiving channels of the receiver.

[0068] For example, multiplying the data points of the time-frequency plot of the first receiving channel by the conjugate of the data points of the second receiving channel yields the cross-correlation data of the first receiving channel group consisting of the first and second receiving channels; multiplying the data points of the time-frequency plot of the second receiving channel by the conjugate of the data points of the third receiving channel yields the cross-correlation data of the second receiving channel group consisting of the second and third receiving channels, and so on. Similarly, multiplying the data points of the (N-1)th receiving channel by the conjugate of the data points of the Nth receiving channel yields the cross-correlation data of the (N-1)th receiving channel group consisting of the (N-1)th and Nth receiving channels. The conjugate of data refers to data whose real part is the same as the data's real part and whose imaginary part is opposite.

[0069] Step S3: Accumulate the cross-correlation data for each receiving channel group to obtain the accumulated cross-correlation data for the corresponding receiving channel group.

[0070] Step S4: Take the modulus and phase of the cumulative cross-correlation data for each receiving channel group to obtain the cross-correlation energy map and phase difference map of the corresponding receiving channel group.

[0071] Traditional methods directly use time-frequency maps to calculate pulse width and determine azimuth. However, when the distance between the UAV and the receiver is large, it is difficult to detect the UAV signal using time-frequency maps, leading to difficulties in target detection. When using deep learning network models for target recognition, the size of the deep learning network model is reduced by lowering the resolution of the time-frequency map. Although this reduces computation time, the improvement in the detection capability of low signal-to-noise ratio signals is limited, resulting in low direction-finding accuracy. For example, for a time-frequency map of size 1024*4096, traditional methods typically downsample the frequency or time domain of the time-frequency map by a factor of 4 to obtain a time-frequency map of size 1024*1024 as input to the target recognition model. This has limited improvement in the detection capability of low signal-to-noise ratio signals and still results in problems such as missed alarms, false alarms, or low direction-finding accuracy.

[0072] Instead of downsampling the time-frequency graph, this invention accumulates the cross-correlation data of the received channel group. For example, by accumulating 4 rows in the time domain, the 1024*4 cross-correlation data is compressed into 1024*1 accumulated cross-correlation data, and finally, 1024*1024 accumulated cross-correlation data can also be obtained.

[0073] The formula for calculating the energy after cross-correlation accumulation is:

[0074]

[0075] Where FFT stands for Fast Fourier Transform, Rx1 represents the time-domain sampled data of the first receiving channel, Rx2 represents the time-domain sampled data of the second receiving channel, s1 represents the time-domain signal of the first receiving channel, n1 represents the time-domain noise of the first receiving channel, Δφ represents the phase difference between the second and first receiving channels, n2 represents the time-domain noise of the second receiving channel, S1 represents the frequency-domain signal of the first receiving channel, N1 represents the frequency-domain noise of the first receiving channel, N2 represents the frequency-domain noise of the second receiving channel, represents the conjugate of the data, and || represents the modulus value.

[0076] According to formula (1), after cumulative cross-correlation calculation, the signal energy increases to the fourth power, while the noise energy increases to the second power. After M-fold cross-correlation accumulation, the signal-to-noise ratio of the output signal will be improved. Therefore, using the energy map after cross-correlation accumulation for target identification can directly improve the detection performance of distant targets. Figures 2-5 The diagram shows the energy plot of a single receiving channel after four iterations, the cross-correlation energy plot of the receiving channel group, and the phase difference plot. Figure 4 The signal-to-noise ratio has been significantly improved, and Figure 5 The phase difference diagram can also be used to identify signals.

[0077] Step S5: Perform target identification on the cross-correlation energy map of each receiving channel group to obtain the target type and target detection box of the corresponding receiving channel group.

[0078] In this embodiment, a target recognition model is used to identify targets in the cross-correlation energy maps of each receiving channel group. This target recognition model is obtained by training and validating a deep learning network model using a sample dataset. Each sample in the dataset includes either the energy map of a single receiving channel or the cross-correlation energy map of a single receiving channel group. That is, either an existing target recognition model (trained using the energy maps of the receiving channels) can be used to identify targets in the cross-correlation energy maps of each receiving channel group, or a new target recognition model (trained using the cross-correlation energy maps of the receiving channel groups) can be used to identify targets in the cross-correlation energy maps of each receiving channel group.

[0079] Since the noise itself is uncorrelated, this invention uses cross-correlation energy maps for target recognition, which can suppress Gaussian white noise and improve the signal-to-noise ratio.

[0080] Step S6: Determine the final target detection box for each receiving channel group based on the target type and phase difference map of each receiving channel group.

[0081] The target types in the receiving channel group include wideband signals and narrowband frequency-hopping signals. For wideband signals, which have fixed frequencies and bandwidths, multiple different targets can be directly identified through the target recognition model. Therefore, when the target type is a wideband signal, the target detection box obtained through target recognition is the final target detection box for the corresponding receiving channel group, and there is no need to merge the target detection boxes. Figure 6 and Figure 7 As shown, for narrowband frequency hopping signals, they hop rapidly within a large frequency band according to a certain pattern. Once multiple targets are mixed together, the target recognition model will be unable to separate them through the energy map and can only identify them as a single target. However, the phase difference map of the receiving channel group directly provides the direction of arrival information of the target. Based on the direction of arrival information, the narrowband frequency hopping target can be separated.

[0082] The cross-correlation energy map of each receiving channel group is input into the target recognition model. When the target type is a wideband signal, the target recognition model outputs a large target detection box, i.e., the final target detection box, completing target recognition. When the target type is a narrowband frequency-hopping signal, each frequency-hopping bar will generate one target detection box, and further target segmentation is required using the phase difference map. The specific implementation process is as follows:

[0083] The grayscale value of each target detection box is calculated based on the phase difference map of the corresponding receiving channel group;

[0084] Cluster the grayscale values ​​of all target detection boxes, and merge the target detection boxes according to the clustering results to obtain the final target detection boxes of the corresponding receiving channel group.

[0085] Since the frequency hopping bars of the same target have the same direction of arrival, they have a very close phase difference of the receiving channel group. The gray values ​​of their target detection boxes should be very close. By using this feature, the target detection boxes corresponding to frequency hopping bars with different directions of arrival can be split, and the target detection boxes corresponding to frequency hopping bars with the same direction can be merged, thus realizing the accurate identification of multiple weak signals.

[0086] In this embodiment, the K-means clustering algorithm is used to cluster the gray values ​​of all target detection boxes in each receiving channel group. The specific implementation process is as follows:

[0087] Step S6.1: Divide the target detection boxes into K groups, and randomly select K objects as the initial cluster centers;

[0088] Step S6.2: Calculate the distance between each object and each cluster center based on the gray value of the target detection box;

[0089] Step S6.3: Based on the distance between each object and each cluster center, assign each object to the cluster center that is closest to it, where the cluster center and the object assigned to that cluster center represent a cluster;

[0090] Step S6.4: Recalculate the cluster centers of the corresponding clusters based on the objects assigned to each cluster;

[0091] Step S6.5: Determine whether the termination condition is met. If yes, obtain the clustering result; otherwise, proceed to step S6.2. In this embodiment, the termination condition is: all objects have been correctly assigned to their corresponding clusters.

[0092] Step S7: Determine the strongest channel group based on the final target detection box and cross-correlation energy map of each receiving channel group.

[0093] In this embodiment, the strongest channel group is determined based on the final target detection box and cross-correlation energy map of each receiving channel group. The specific implementation process is as follows:

[0094] Step S7.1: Based on the frequency band information of the final target detection box, extract the time domain envelope data of the UAV signal from the cross-correlation energy map of the corresponding receiving channel group;

[0095] Step S7.2: Determine the time-domain pulse width information based on the intersection point between the time-domain envelope data and the set amplitude threshold;

[0096] Step S7.3: Extract the energy data of the corresponding receiving channel group from the cross-correlation energy map of the corresponding receiving channel group based on the time-domain pulse width information and frequency band information;

[0097] Step S7.4: Determine the strongest channel group based on the energy data of each receiving channel group.

[0098] By using a target recognition model to perform deep learning on the energy map, a target detection box is generated after a target is detected, and the target's frequency band information is provided. Based on the target's frequency band information, the time-domain envelope data of the UAV signal can be extracted from the energy map. However, directly extracting the time-domain envelope data from the energy map of a single receiver channel may result in numerous glitches, such as… Figure 8 As shown. This invention extracts time-domain envelope data from the cross-correlation energy map based on frequency band information, thus obtaining clear time-domain envelope data, such as... Figure 9 As shown.

[0099] After obtaining the time-domain envelope data, the amplitude threshold is calculated. The intersection of the amplitude threshold and the time-domain envelope data determines the time-domain pulse width information, which is the start and end position of each pulse. Figure 9 The red dots indicate the start position of the pulse width, and the blue dots indicate the end position. To facilitate the extraction of effective data for calculation and eliminate pure noise, the energy data of each receiving channel group can be extracted from the cross-correlation energy map based on the time-domain pulse width information and frequency band information. The receiving channel group with the largest energy data among the N-1 receiving channel groups is the strongest channel group, i.e., the direction of signal arrival. The cross-correlation energy of the strongest channel group is greater than the cross-correlation energy of other receiving channel groups.

[0100] Step S8: Calculate the amplitude angle of the strongest channel group based on the cross-correlation energy diagram of the strongest channel group, and calculate the phase difference of the strongest channel group based on the phase difference diagram of the strongest channel group.

[0101] The calculation of amplitude angle and phase difference is a current technology. Amplitude angle and phase difference are important input parameters required for UAV phase comparison direction finding.

[0102] Step S9: Perform phase comparison and direction finding of the UAV based on the amplitude angle and phase difference of the strongest channel group.

[0103] In phase-comparison direction finding, when the signal-to-noise ratio (SNR) is too low, the phase difference of a single data point may fluctuate significantly. Even after averaging, there will still be a large deviation from the true phase difference of the signal. This is because the phase difference of Gaussian white noise follows a normal distribution with a mean of 0, which will skew the overall phase difference of the signal. Therefore, simply averaging multiple times cannot solve the problem of low SNR. This invention utilizes cross-correlation accumulated data to directly improve the SNR, allowing for the direct calculation of a more accurate phase difference, thus achieving precise direction finding.

[0104] The frequency domain diagram of the received signal is as follows Figure 10 As shown, the frequency domain plot after cross-correlation accumulation is as follows: Figure 11 As shown. From Figure 10 and Figure 11 It can be seen that the signal-to-noise ratio is significantly improved after cross-correlation accumulation.

[0105] This invention utilizes the cross-correlation of adjacent receiving channels to improve the signal-to-noise ratio (SNR) and performs time-domain cross-correlation accumulation. This not only reduces the network size of deep learning but also improves long-range detection performance. The accumulated cross-correlation data can be directly used to solve for accurate time-domain pulse width and high-precision phase difference, reducing the overall computational load and significantly improving both time and performance. This invention has low computational cost, high data reuse rate, and can be directly applied to existing hardware and deep learning network models, saving time and improving SNR, thus demonstrating promising application prospects.

[0106] Example 2

[0107] This invention also provides a terminal device, which includes a memory, a processor, and a computer program / instructions stored in the memory. The processor executes the computer program / instructions to implement the UAV direction finding method described in Embodiment 1.

[0108] Although not shown, the terminal device includes a processor that can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) or loaded from a storage portion into random access memory (RAM). The processor can be a multi-core processor or may contain multiple processors. In some embodiments, the processor may include a general-purpose main processor and one or more specialized coprocessors, such as a central processing unit, graphics processing unit (GPU), neural network processor (NPU), digital signal processor (DSP), etc. Various programs and data required for the operation of the terminal device are also stored in the RAM. The processor, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0109] The processor and memory described above are used together to execute programs / instructions stored in the memory. When the program / instructions are executed by the computer, they can implement the methods, steps, or functions described in the above embodiments.

[0110] Although not shown, embodiments of the present invention also provide a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the UAV direction finding method described in Embodiment 1.

[0111] Storage media in embodiments of the present invention include articles that are permanent or non-permanent, removable or non-removable, and can store information by any method or technology. Examples of storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information that can be accessed by a computing device.

[0112] Readable storage media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient media, such as modulated data signals and carrier waves.

[0113] Although not shown, embodiments of the present invention also provide a computer program product, including: a computer program / instructions that, when executed by a processor, implement the UAV direction finding method as described in Embodiment 1.

[0114] The above description only discloses specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or modifications that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A direction-finding method for unmanned aerial vehicles (UAVs), characterized in that, The direction finding method includes the following steps: Obtain the time-frequency diagram of the UAV signal for each receiving channel; Cross-correlation calculations were performed on the time-frequency plots of UAV signals from two adjacent receiving channels to obtain cross-correlation data for each receiving channel group; The cross-correlation data of each receiving channel group is accumulated to obtain the cumulative cross-correlation data of the corresponding receiving channel group; For each receiving channel group, the modulus and phase of the accumulated cross-correlation data are taken to obtain the cross-correlation energy map and phase difference map of the corresponding receiving channel group; Target identification is performed on the cross-correlation energy map of each receiving channel group to obtain the target type and target detection box of the corresponding receiving channel group; The final target detection box for each receiving channel group is determined based on the target type and phase difference map of each receiving channel group. The strongest channel group is determined based on the final target detection box and cross-correlation energy map of each receiving channel group; The amplitude angle of the strongest channel group is calculated based on the cross-correlation energy diagram of the strongest channel group, and the phase difference of the strongest channel group is calculated based on the phase difference diagram of the strongest channel group. Phase comparison and direction finding of the UAV are performed based on the amplitude angle and phase difference of the strongest channel group; The final target detection box for each receiving channel group is determined based on the target type and phase difference map of each receiving channel group. The specific implementation process is as follows: When the target type is a bandwidth signal, the target detection box obtained through target recognition is the final target detection box of the corresponding receiving channel group; When the target type is a narrowband frequency hopping signal, the gray value of each target detection box is calculated according to the phase difference map of the receiving channel group corresponding to the narrowband frequency hopping signal; the gray values ​​of all target detection boxes in the receiving channel group are clustered, and the target detection boxes are merged according to the clustering results to obtain the final target detection box of the corresponding receiving channel group.

2. The UAV direction finding method according to claim 1, characterized in that, The specific implementation process for obtaining the time-frequency diagram of the UAV signal for each receiving channel is as follows: Acquire time-domain sampling data of the UAV signal for each receiving channel; Perform a Fast Fourier Transform on the time-domain sampled data of the UAV signal for each receiving channel to obtain the time-frequency diagram of the corresponding UAV signal.

3. The UAV direction finding method according to claim 1, characterized in that, The specific implementation process of performing cross-correlation calculation on the time-frequency diagrams of UAV signals from two adjacent receiving channels is as follows: The first i The data point multiplied by the first data point in the time-frequency graph of the UAV signal from the receiving channel. i The conjugate of the data in the time-frequency plot of the UAV signal from the +1 receiving channel is used to obtain the first... i Cross-correlation data of each receiving channel group; among which... i =1,2,…,N-1, where N represents the number of receiver channels.

4. The UAV direction finding method according to claim 1, characterized in that, The target recognition model is used to identify targets in the cross-correlation energy map of each receiving channel group. The target recognition model is obtained by training and validating a deep learning network model using a sample dataset. Each sample data in the sample dataset includes the energy map of a single receiving channel or the cross-correlation energy map of a single receiving channel group.

5. The UAV direction finding method according to claim 1, characterized in that, The K-means clustering algorithm is used to cluster the gray values ​​of all target detection boxes in the receiving channel group. The specific implementation process is as follows: The target detection boxes are divided into K groups, and K objects are randomly selected as the initial cluster centers; Based on the grayscale values ​​of the target detection boxes, calculate the distance between each object and each cluster center; Each object is assigned to the cluster center closest to it based on the distance between each object and each cluster center. Here, the cluster center and the objects assigned to the cluster center represent a cluster. Recalculate the cluster centers of the corresponding clusters based on the objects assigned to each cluster; Determine if the termination condition is met. If so, obtain the clustering result; otherwise, repeat the steps of distance calculation, object assignment, and cluster center calculation.

6. The UAV direction finding method according to any one of claims 1 to 5, characterized in that, The strongest channel group is determined based on the final target detection box and cross-correlation energy map of each receiving channel group. The specific implementation process is as follows: Based on the frequency band information of the final target detection box, the time domain envelope data of the UAV signal is extracted from the cross-correlation energy map of the corresponding receiving channel group; The temporal pulse width information is determined based on the intersection of the temporal envelope data and the set amplitude threshold. Based on the time-domain pulse width information and frequency band information, the energy data of the corresponding receiving channel group is extracted from the cross-correlation energy map of the corresponding receiving channel group; The strongest channel group is determined based on the energy data of each receiving channel group.

7. A terminal device, comprising a memory, a processor, and a computer program / instructions stored in the memory, characterized in that, The processor executes the computer program / instructions to implement the UAV direction finding method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instruction is executed by the processor, it implements the UAV direction finding method according to any one of claims 1 to 6.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the UAV direction finding method according to any one of claims 1 to 6.

Images