A low-altitude unmanned aerial vehicle detection method and system based on event stream frequency enhancement

Abstract

This application discloses a method and system for detecting low-altitude unmanned aerial vehicles (UAVs) based on event stream frequency enhancement, relating to the fields of computer vision and UAV monitoring technology. The method includes acquiring raw event stream data of a scene to be detected; reconstructing brightness signals and performing frequency analysis on the raw event stream data to extract frequency features characterizing the rotor rotation characteristics of the UAV in the scene to be detected, and mapping the frequency features to a frequency embedding vector; generating dynamic gating parameters based on the frequency embedding vector, and using the dynamic gating parameters to modulate and fuse the event features to obtain frequency-enhanced features; and outputting the detection result of the UAV in the scene to be detected based on the frequency-enhanced features, the detection result including the UAV's category and location information. This application can effectively suppress background noise that does not conform to the rotor frequency characteristics, while enhancing the feature response of the UAV target, significantly improving the robustness and accuracy of UAV detection.

Description

Technical Field

[0001] This application relates to the fields of computer vision and drone monitoring technology, and in particular to a method and system for detecting low-altitude drones based on event stream frequency enhancement. Background Technology

[0002] With the development of the low-altitude economy, the widespread adoption of consumer unmanned aerial vehicles (UAVs) has brought increasingly serious security challenges, such as unauthorized intrusion into airports, military restricted areas, or privacy violations. Therefore, effective UAV detection technology is crucial. Traditional visual detection methods mainly rely on visible light cameras, but their performance is limited in high-speed motion blur, low-light conditions, or complex backgrounds.

[0003] Event cameras, as a novel bio-inspired visual sensor, offer advantages such as high dynamic range, low latency, and no motion blur, making them ideal for capturing high-speed moving drones. However, existing event camera-based detection algorithms typically aggregate sparse event streams directly into event frames or voxel grids, which are then input into deep neural networks. While this approach facilitates the use of existing convolutional neural network architectures, it often loses the rich high-frequency temporal information inherent in the event data.

[0004] In particular, the rotor rotation frequency unique to drones (typically between 50Hz and 500Hz) is a key physical feature distinguishing drones from background noise such as birds and rustling leaves. Existing deep neural networks struggle to implicitly learn this frequency pattern from aggregated event frames, leading to false positives in complex dynamic environments. Therefore, explicitly utilizing frequency cues to enhance event features has become a crucial issue in improving the robustness of low-altitude drone detection. Summary of the Invention

[0005] The purpose of this application is to provide a low-altitude UAV detection method and system based on event stream frequency enhancement, which can effectively suppress background noise that does not conform to the rotor frequency characteristics, while enhancing the characteristic response of UAV targets, and significantly improving the robustness and accuracy of UAV detection.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a low-altitude unmanned aerial vehicle (UAV) detection method based on event stream frequency enhancement, the low-altitude UAV detection method based on event stream frequency enhancement includes: Obtain the raw event stream data of the scene to be detected; The original event stream data is subjected to luminance signal reconstruction and frequency analysis to extract frequency features characterizing the rotor rotation characteristics of the UAV in the scene to be detected, and the frequency features are mapped into frequency embedding vectors. Dynamic gating parameters are generated based on the frequency embedding vector, and the event features are modulated and fused using the dynamic gating parameters to obtain frequency-enhanced features; wherein the event features are obtained by transforming the original event stream data; The detection results of the UAV in the scene to be detected are output based on the frequency enhancement features. The detection results include the UAV's category and location information.

[0007] Secondly, this application provides a low-altitude unmanned aerial vehicle (UAV) detection system based on event stream frequency enhancement, the low-altitude UAV detection system based on event stream frequency enhancement includes: The data acquisition module is used to acquire the raw event stream data of the scene to be detected; The frequency feature extraction module is used to reconstruct the brightness signal and analyze the frequency of the original event stream data in order to extract the frequency features that characterize the rotor rotation characteristics of the UAV in the scene to be detected, and to map the frequency features into a frequency embedding vector. The dynamic gating enhancement module is used to generate dynamic gating parameters based on the frequency embedding vector, and to modulate and fuse event features using the dynamic gating parameters to obtain frequency-enhanced features; wherein the event features are obtained by converting the original event stream data; The target detection module is used to output the detection results of the UAV in the scene to be detected based on the frequency enhancement features. The detection results include the category and location information of the UAV.

[0008] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the low-altitude unmanned aerial vehicle detection method based on event stream frequency enhancement as described above.

[0009] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the low-altitude unmanned aerial vehicle detection method based on event stream frequency enhancement as described above.

[0010] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the low-altitude unmanned aerial vehicle detection method based on event stream frequency enhancement as described above.

[0011] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method and system for detecting low-altitude unmanned aerial vehicles (UAVs) based on event stream frequency enhancement. First, the original event stream data undergoes luminance signal reconstruction and frequency analysis to extract frequency features characterizing the rotor rotation characteristics of the UAV in the scene to be detected. These frequency features are then mapped to frequency embedding vectors, enabling explicit extraction of pixel-level frequency information from the high temporal resolution event stream and converting it into frequency embeddings easily processed by neural networks. Second, dynamic gating parameters are generated based on the frequency embedding vectors. These parameters are used to modulate and fuse the event features, dynamically modulating the event features of the backbone network of the target detection network. This effectively suppresses background noise that does not conform to the rotor frequency characteristics while enhancing the feature response of the UAV target. Finally, the detection results of the UAV in the scene to be detected are output based on the frequency enhancement features. The detection results include the UAV's category and location information, significantly improving the robustness and accuracy of UAV detection in extreme environments such as low light and high dynamic backgrounds. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating a low-altitude unmanned aerial vehicle (UAV) detection method based on event stream frequency enhancement in one embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] like Figure 1As shown, this application provides a low-altitude UAV detection method based on event stream frequency enhancement. This method is executed by a computer device, specifically by a terminal or server alone, or by both. In this embodiment, the low-altitude UAV detection method based on event stream frequency enhancement includes the following steps S101 to S104. Wherein: Step S101: Obtain the raw event stream data of the scene to be detected. In this embodiment, the raw event stream data of the scene to be detected is acquired by an event camera. The raw event stream data output by the event camera is an asynchronous event stream, which can be represented as follows: ,in, Represents the first in the asynchronous event stream One event; Represents the coordinates of the event; This is a timestamp, indicating when the event occurred; The event polarity indicates whether the pixel brightness increases or decreases.

[0017] It should be noted that the difference between an event camera and a traditional camera is that each pixel of an event camera works independently, only outputting an event when a brightness change (exceeding a threshold) is detected. Its output is an asynchronous, sparse event stream, which is an asynchronous data sequence composed of several events arranged in chronological order. In the application scenario of this application, the rotation of the drone rotor is the primary event source. When the drone rotor blades rotate, they reflect or block light, causing rapid, periodic changes in brightness at the pixel location, thus continuously triggering a series of events. In addition, background dynamic interference also triggers events, such as background noise from birds, swaying leaves, and moving clouds.

[0018] Step S102: Reconstruct the brightness signal and perform frequency analysis on the original event stream data to extract frequency features that characterize the rotor rotation characteristics of the UAV in the scene to be detected, and map the frequency features into a frequency embedding vector.

[0019] In step S102 of this application embodiment, the luminance signal reconstruction and frequency analysis of the original event stream data adopts the detrended integral method, specifically including the following steps S201 to S204. Wherein: Step S201: Accumulate the event polarities in the original event stream to obtain an accumulated signal, the accumulated signal including the event polarities of multiple events arranged in chronological order.

[0020] Step S202: The accumulated signal is detrended using the moving average method. The detrended signal is then integrated and subjected to high-frequency filtering to reconstruct a brightness signal that includes the periodic changes in the rotor rotation characteristics of the UAV.

[0021] In step S202 of this application embodiment, the accumulated signal is detrended using a moving average method, and the detrended signal is integrated and high-frequency filtered to reconstruct a brightness signal containing the periodic changes in the rotor rotation characteristics of the UAV. This includes the following steps: Step 1: Definition The polarity exponential moving average up to the k-th event is the polarity exponential moving average. The calculation formula is Among them, smoothing factor ∈[0,1]; Step 2: Calculate the change in polarity after detrending. Polarity change The calculation formula is , Indicates the event polarity of the k-th event. This represents the polarity exponential moving average of the (k-1)th event; Step 3: Integrate the detrended change to obtain the intermediate brightness signal. intermediate brightness signal The calculation formula is ; Step 4: Eliminate residual low-frequency drift using a digital high-pass filter. The final reconstructed luminance signal. The brightness signal is updated using the filter coefficient β. The calculation formula is: .

[0022] Step S203: Perform zero-crossing detection on the brightness signal and record the zero-crossing times of adjacent and unidirectional points. and The zero-crossing time is determined by interpolation, and the formula is as follows: ; In the formula, Indicates past midnight. This indicates the time when the k-th event occurs.

[0023] Step S204: Calculate the time interval based on adjacent and unidirectional zero-crossing times, and calculate the instantaneous frequency of the current pixel as a frequency feature based on the time interval. The formula for calculating the instantaneous frequency is: ,in, and This indicates the zero-crossing times of adjacent points in the same direction.

[0024] The purpose of implementing steps S201 to S204 is to recover potential periodic signals from sparse events. This application is based on the rotational characteristics of UAV rotors, which cause periodic changes in brightness. The zero-crossing point captured by the signal zero-crossing point detection reflects a specific phase in the rotor rotation cycle (e.g., the moment when the blades rotate to the brightest or darkest position). Since the zero-crossing point detection algorithm is a pure time-domain method, it operates directly on these high-precision timestamps and calculates the frequency by detecting the time difference between adjacent zero-crossing points. Therefore, it does not require restoring asynchronous data into a regular time series or performing complex calculations like Fourier transform. Thus, it has the advantages of high computational efficiency and suitability for edge deployment compared to traditional Fourier transform methods.

[0025] In step S102 of this application embodiment, mapping the frequency features to frequency embedding vectors includes: mapping the frequency features within the effective frequency range to high-dimensional frequency embedding vectors according to the frequency-high-dimensional vector mapping relationship, and mapping the frequency features outside the effective frequency range to background feature vectors; wherein, the effective frequency range is set according to the rotor speed of the UAV.

[0026] In this embodiment, the frequency-high-dimensional vector mapping relationship is a frequency-color space mapping relationship; as one implementation of this application, 50Hz-500Hz (a typical UAV rotor speed range) is mapped to an RGB color space (e.g., blue represents low frequency, red represents high frequency), forming a frequency feature map with shape (H, W, 3). (High-dimensional frequency embedding vectors) Non-rotor frequency regions can be filled with specific values ​​(such as white or zero) to serve as background feature vectors. Understandably, mapping frequency features to frequency embedding vectors allows neural networks to better understand scalar frequency values.

[0027] Step S103: Generate dynamic gating parameters based on the frequency embedding vector, and use the dynamic gating parameters to modulate and fuse the event features to obtain frequency enhancement features; wherein the event features are obtained by converting the original event stream data.

[0028] In step S103 of this application embodiment, generating dynamic gating parameters based on the frequency embedding vector includes: extracting features from the frequency embedding vector to obtain high-dimensional frequency features; and dynamically generating convolution kernel parameters based on the high-dimensional frequency features as the dynamic gating parameters.

[0029] In step S103 of this application embodiment, the modulation and fusion of event features using the dynamic gating parameters to obtain frequency enhancement features includes steps S301 to S304. Wherein: Step S301: Perform format conversion on the raw event stream data to obtain event features. It should be noted that the purpose of this conversion is to transform the raw event stream data into the input format required by the backbone network of the subsequent object detection network, such as the Event Frame format. The converted data is recorded as the event features. .

[0030] Step S302: Perform convolution operation on the event features using the dynamic gating parameters to output a spatial attention map and gating coefficients. Specifically, this includes the following steps: First, use the attention module to perform convolution operation on the frequency feature map. Processing yields a dynamic kernel. to aggregate The system carries global frequency information; secondly, it uses a dynamic kernel. For the frequency feature map The spatial attention map A and gating coefficients G are obtained by performing dynamic convolution, and the process can be represented as follows: In this context, the spatial attention map A encodes the probability of the existence of a rotational structure pixel by pixel based on frequency information, and the gating coefficient G is used to measure the confidence of the prior at that frequency.

[0031] Furthermore, in the embodiments of this application, the attention module is a series structure of global average pooling (GAP), convolution kernel (Conv), activation function (ReLU) and convolution kernel (Conv) (a series structure of GAP-Conv-ReLU-Conv).

[0032] Step S303: Calculate spatial attention based on the spatial attention map and the event features; wherein the formula for calculating spatial attention is: ; In the formula, Indicates spatial attention, This represents the activation function. Representing a spatial attention map, Indicates the characteristics of an event.

[0033] Perform step S303 above to activate the function. Attention Map Each value in the compressed file is compressed to... Within the scope, generate a characteristic of the event. A weight matrix with the same spatial dimensions. Values ​​close to 1 indicate that the features at that location are highly correlated with the strong frequency characteristics of the target (UAV rotor) and should be emphasized and enhanced. Values ​​close to 0 indicate that the location may be background noise (such as swaying leaves) and should be suppressed or ignored, thereby spatially suppressing regions of frequency mismatch (such as stationary backgrounds or non-periodic moving objects).

[0034] Step S304: Perform residual fusion on the gating coefficients, the spatial attention, and the event features to obtain frequency enhancement features; wherein, the formula for residual fusion is: ; In the formula, Indicates frequency enhancement characteristics, This represents the gating coefficient.

[0035] Step S304 above allows the network to adaptively adjust the weights of frequency information. When the frequency features are not obvious (such as when the drone hovers extremely smoothly), the network can degenerate to use the original event features, thus ensuring the model's generalization ability.

[0036] Step S104: Output the detection results of the UAV in the scene to be detected based on the frequency enhancement features. The detection results include the UAV's category and location information.

[0037] In step S104 of this embodiment, outputting the detection result of the UAV in the scene to be detected based on the frequency enhancement feature includes: inputting the frequency enhancement feature into a target detection network, and outputting a bounding box and a confidence score. The bounding box represents the position information of the UAV in the scene to be detected, and the confidence score represents the category of the UAV in the scene to be detected. As one implementation, the target detection network includes the YOLO series, SSD, or a dedicated event detection network, such as RVT, MvHeat-DET, etc.

[0038] By implementing steps S101 to S104 above, this application first performs luminance signal reconstruction and frequency analysis on the original event stream data to extract frequency features characterizing the rotor rotation characteristics of the UAV in the scene to be detected, and maps the frequency features into frequency embedding vectors, thereby explicitly extracting pixel-level frequency information from the high temporal resolution event stream and converting it into frequency embeddings that are easy for neural networks to process; secondly, dynamic gating parameters are generated based on the frequency embedding vectors, and the event features are modulated and fused using the dynamic gating parameters, thereby dynamically modulating the event features of the backbone network of the target detection network, effectively suppressing background noise that does not conform to the rotor frequency characteristics, and enhancing the feature response of the UAV target; finally, the detection results of the UAV in the scene to be detected are output based on the frequency enhancement features, and the detection results include the UAV category and location information, thereby significantly improving the robustness and accuracy of UAV detection in extreme environments such as low light and high dynamic backgrounds.

[0039] Based on the same inventive concept, this application also provides a low-altitude UAV detection system based on event stream frequency enhancement. The low-altitude UAV detection system based on event stream frequency enhancement includes a data acquisition module, a frequency feature extraction module, a dynamic gating enhancement module, and a target detection module. Wherein: The data acquisition module is used to acquire the raw event stream data of the scene to be detected; The frequency feature extraction module is used to reconstruct the brightness signal and analyze the frequency of the original event stream data in order to extract the frequency features that characterize the rotor rotation characteristics of the UAV in the scene to be detected, and to map the frequency features into a frequency embedding vector. The dynamic gating enhancement module is used to generate dynamic gating parameters based on the frequency embedding vector, and to modulate and fuse event features using the dynamic gating parameters to obtain frequency-enhanced features; wherein the event features are obtained by converting the original event stream data; The target detection module is used to output the detection results of the UAV in the scene to be detected based on the frequency enhancement features. The detection results include the category and location information of the UAV.

[0040] In this embodiment, the target detection network adopts a backbone network based on event frames or voxel grids; the dynamic gating enhancement module is embedded as a plug-in before or in the feature extraction layer of the backbone network to suppress background noise interference at non-rotor frequencies.

[0041] Furthermore, in this embodiment, the dynamic gating enhancement module includes a preprocessing network and a controller sub-network, wherein the preprocessing network consists of convolutional layers, normalization layers, and activation functions; wherein: A preprocessing network is used to process the frequency embedding vector. Feature extraction is performed to obtain high-dimensional frequency features. ; The controller subnetwork is used to dynamically generate convolution kernel parameters based on the high-dimensional frequency features as the dynamic gating parameters.

[0042] In one application scenario, a low-altitude UAV detection method based on event stream frequency enhancement was validated on the EvDET200K dataset and a self-built EventUAV dataset. Experimental results show that after introducing the dynamic gating enhancement module, the detection accuracy (mAP) is significantly improved while keeping the computational cost almost unchanged (the number of parameters increases very little), especially in scenarios with a large amount of dynamic interference (such as trees swaying in the wind), where the false detection rate is greatly reduced.

[0043] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0044] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0045] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0046] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0047] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0048] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0049] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0050] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting low-altitude unmanned aerial vehicles (UAVs) based on event stream frequency enhancement, characterized in that, The low-altitude UAV detection method based on event stream frequency enhancement includes: Obtain the raw event stream data of the scene to be detected; The original event stream data is subjected to luminance signal reconstruction and frequency analysis to extract frequency features characterizing the rotor rotation characteristics of the UAV in the scene to be detected, and the frequency features are mapped into frequency embedding vectors. Dynamic gating parameters are generated based on the frequency embedding vector, and the event features are modulated and fused using the dynamic gating parameters to obtain frequency-enhanced features; wherein the event features are obtained by transforming the original event stream data; The detection results of the UAV in the scene to be detected are output based on the frequency enhancement features. The detection results include the UAV's category and location information.

2. The low-altitude UAV detection method based on event stream frequency enhancement according to claim 1, characterized in that, The process of reconstructing the luminance signal and performing frequency analysis on the original event stream data includes: The event polarities in the original event stream are accumulated to obtain an accumulated signal; The accumulated signal is detrended using the moving average method. The detrended signal is then integrated and subjected to high-frequency filtering to reconstruct a brightness signal that includes the periodic changes in the rotor rotation characteristics of the UAV. Zero-crossing detection is performed on the brightness signal, and the zero-crossing times of adjacent and unidirectional points are recorded; The time interval is calculated based on the zero-crossing times of adjacent and unidirectional points, and the instantaneous frequency of the current pixel is calculated based on the time interval as a frequency feature.

3. The low-altitude UAV detection method based on event stream frequency enhancement according to claim 2, characterized in that, The step of mapping the frequency features into a frequency embedding vector includes: According to the frequency-high-dimensional vector mapping relationship, the frequency features within the effective frequency range are mapped to high-dimensional frequency embedding vectors, and the frequency features outside the effective frequency range are mapped to background feature vectors; wherein, the effective frequency range is set according to the rotor speed of the UAV.

4. The low-altitude UAV detection method based on event stream frequency enhancement according to claim 1, characterized in that, The step of generating dynamic gating parameters based on the frequency embedding vector includes: Feature extraction is performed on the frequency embedding vector to obtain high-dimensional frequency features; Convolution kernel parameters are dynamically generated based on the high-dimensional frequency features to serve as the dynamic gating parameters.

5. The low-altitude UAV detection method based on event stream frequency enhancement according to claim 4, characterized in that, The process of modulating and fusing event features using the dynamic gating parameters to obtain frequency-enhanced features includes: Perform format conversion on the raw event stream data to obtain event characteristics; The event features are convolutionally processed using the dynamic gating parameters to output a spatial attention map and gating coefficients. Spatial attention is calculated based on the spatial attention map and the event features; wherein, the formula for calculating spatial attention is: ； In the formula, Indicates spatial attention. This represents the activation function. Representing a spatial attention map, Indicate the characteristics of the event; The gating coefficients, spatial attention, and event features are subjected to residual fusion to obtain frequency-enhanced features; wherein the formula for residual fusion is: ； In the formula, Indicates frequency enhancement characteristics, This represents the gating coefficient.

6. The low-altitude UAV detection method based on event stream frequency enhancement according to claim 1, characterized in that, The step of outputting the detection result of the UAV in the scene to be detected based on the frequency enhancement feature includes: The frequency enhancement features are input into the target detection network, which outputs bounding boxes and confidence scores. The bounding boxes represent the location information of the UAV in the scene to be detected, and the confidence scores represent the category of the UAV in the scene to be detected.

7. A low-altitude unmanned aerial vehicle (UAV) detection system based on event stream frequency enhancement, characterized in that, The low-altitude UAV detection system based on event stream frequency enhancement includes: The data acquisition module is used to acquire the raw event stream data of the scene to be detected; The frequency feature extraction module is used to reconstruct the brightness signal and analyze the frequency of the original event stream data in order to extract the frequency features that characterize the rotor rotation characteristics of the UAV in the scene to be detected, and to map the frequency features into a frequency embedding vector. The dynamic gating enhancement module is used to generate dynamic gating parameters based on the frequency embedding vector, and to modulate and fuse event features using the dynamic gating parameters to obtain frequency-enhanced features; wherein the event features are obtained by converting the original event stream data; The target detection module is used to output the detection results of the UAV in the scene to be detected based on the frequency enhancement features. The detection results include the category and location information of the UAV.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the low-altitude unmanned aerial vehicle detection method based on event stream frequency enhancement as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the low-altitude unmanned aerial vehicle detection method based on event stream frequency enhancement as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the low-altitude unmanned aerial vehicle detection method based on event stream frequency enhancement as described in any one of claims 1-6.

Images