Low-altitude unmanned aerial vehicle dual-mode detection method based on micro-doppler radar and infrared thermal imaging

Abstract

The application relates to the technical field of unmanned aerial vehicle detection and security and protection, and discloses a low-altitude unmanned aerial vehicle dual-mode detection method of micro-Doppler radar and infrared thermal imaging. The method continuously scans a monitoring area through micro-Doppler radar, and when a suspected target with a signal strength lower than a preset threshold is identified, triggers an infrared thermal imaging device to start. The infrared device directly adjusts the direction and focuses on a specific area according to the target azimuth and distance information provided by the radar, and obtains a high-resolution thermal imaging sequence. Then, the radiation temperature distribution, temperature gradient and thermal profile shape characteristics of the target are analyzed to complete the final determination. The application realizes on-demand starting and rapid and accurate guidance of the infrared sensor, reduces the overall power consumption of the system, and improves the cooperative detection response speed and identification accuracy of the fast-moving target.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) detection and security technology, specifically a dual-mode detection method for low-altitude UAVs using micro-Doppler radar and infrared thermal imaging. Background Technology

[0002] In the field of low-altitude security, a single sensor is insufficient for reliable detection of "low, slow, and small" drones. Micro-Doppler radar has the ability to detect the speed and micro-motion characteristics of moving targets, but it suffers from high false alarm rates and difficulties in feature extraction when facing weak signals, environmental clutter, or small drones, making accurate judgments difficult when used alone. Infrared thermal imaging captures images by sensing the temperature difference between the target and the background, offering good concealment, but its resolution is insufficient in wide-field-of-view search modes, making it difficult to detect small targets at long distances. Furthermore, continuous large-area scanning results in high power consumption and a heavy data processing burden.

[0003] Existing radar and infrared collaborative detection schemes mostly employ simple combinations such as spatiotemporal synchronization or master-slave alternation. These schemes typically allow the two sensors to operate independently in parallel or switch monitoring areas according to a fixed time sequence, lacking intelligent decision-making and resource scheduling capabilities. This mode leads to infrared sensors often running continuously when there is no clear threat, resulting in significant energy waste. Furthermore, when radar detects a suspicious target, the infrared system cannot immediately obtain the target's precise spatial location and must still re-search within its own field of view, causing confirmation delays and making it unable to cope with the rapid penetration of drones.

[0004] A collaborative detection method is needed that can dynamically allocate sensor resources based on the detection situation, enabling a rapid transition from wide-area early warning to precise identification. This method must address how to intelligently activate infrared resources when radar detection confidence is insufficient, and how to utilize prior radar information to guide infrared detection for instantaneous and precise staring, thereby reducing system power consumption while improving the speed and reliability of joint identification of weak signals or camouflaged targets. Summary of the Invention

[0005] The purpose of this invention is to provide a dual-mode detection method for low-altitude unmanned aerial vehicles using micro-Doppler radar and infrared thermal imaging, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a dual-mode detection method for low-altitude unmanned aerial vehicles (UAVs) using micro-Doppler radar and infrared thermal imaging, the method comprising:

[0007] The low-altitude surveillance area is continuously scanned and detected by micro-Doppler radar to obtain radar echo data streams containing information on distance, speed and micro-motion characteristics.

[0008] The radar echo data stream is analyzed in real time. When a suspected UAV target is identified and its echo signal strength is lower than the preset radar confirmation threshold, an infrared trigger command is generated. The infrared trigger command includes the angle, azimuth and distance information of the suspected UAV target.

[0009] According to the infrared trigger command, the infrared thermal imaging device is controlled to adjust its direction and focus on the spatial region corresponding to the angle, azimuth and distance information, and a high-resolution infrared thermal imaging data frame sequence of the spatial region is obtained;

[0010] The infrared thermal imaging data frame sequence is subjected to target thermal feature analysis processing to calculate the radiation temperature distribution, temperature gradient characteristics and thermal profile morphology characteristics of the target area;

[0011] Based on the radiation temperature distribution, the temperature gradient characteristics, and the thermal profile morphology characteristics, it is determined whether the suspected drone target is a real drone target, and a preliminary infrared determination result is generated.

[0012] Preferably, the target thermal feature analysis processing of the infrared thermal imaging data frame sequence includes:

[0013] Extract a set of pixels representing a potential target region from the infrared thermal imaging data frame sequence. The potential target region consists of consecutive pixels with a temperature higher than a set value of background noise.

[0014] Temperature calibration processing is performed on the pixel set to generate the radiation temperature distribution, which includes the highest temperature point, average temperature and temperature standard deviation data within the target area.

[0015] The temperature gradient feature is calculated based on the radiation temperature distribution. The temperature gradient feature is obtained by calculating the temperature change rate between adjacent pixels in the pixel set and generating a two-dimensional temperature gradient map.

[0016] Morphological analysis is performed on the potential target region to extract the thermal profile morphological features, which include the profile perimeter, area, aspect ratio, and circumscribed rectangle fitting parameters.

[0017] Preferably, determining whether the suspected drone target is a real drone target based on the radiation temperature distribution, the temperature gradient characteristics, and the thermal profile morphology characteristics includes:

[0018] The highest temperature point is compared with the preset temperature range of typical working parts of the drone. If the highest temperature point is within the temperature range of typical working parts of the drone, a valid temperature matching event is recorded.

[0019] The two-dimensional temperature gradient map is matched with a preset UAV heat source gradient model. If the similarity matching result exceeds the preset gradient matching threshold, a valid gradient matching event is recorded.

[0020] The perimeter of the outline, the area, the aspect ratio, and the fitting parameters of the bounding rectangle are input into a pre-trained classifier. If the output confidence of the classifier exceeds the preset morphological judgment threshold, a valid morphological matching event is recorded.

[0021] If the sum of the cumulative counts of the effective temperature matching events, the effective gradient matching events, and the effective shape matching events reaches a preset confirmation count threshold, then a preliminary infrared determination result confirming the real UAV target is generated; otherwise, a preliminary infrared determination result that cannot be confirmed is generated.

[0022] Preferably, the method further includes:

[0023] Independent inspections of low-altitude surveillance areas are conducted using infrared thermal imaging equipment to acquire wide-field-of-view infrared image data streams.

[0024] The system analyzes the wide field-of-view infrared image data stream in real time. When a moving suspected thermal target is identified and it is impossible to distinguish the suspected thermal target as a source of thermal interference from a drone, bird, or hot air balloon based on its thermal characteristics, a radar trigger command is generated. The radar trigger command includes the azimuth and estimated distance information of the moving suspected thermal target.

[0025] Based on the radar trigger command, the micro-Doppler radar is controlled to adjust its beam and perform high-precision detection on the spatial region corresponding to the azimuth and estimated distance information to obtain the target echo signal in the spatial region.

[0026] Preferably, the control of the micro-Doppler radar to adjust the beam for high-precision detection of the spatial region corresponding to the azimuth and estimated range information includes:

[0027] Based on the azimuth and estimated distance information, the servo mechanism of the micro-Doppler radar is driven to point the radar beam axis toward the spatial region.

[0028] Configure the waveform parameters and repetition frequency of the radar transmitted signal, wherein the waveform parameters and repetition frequency are suitable for analyzing the micro-motion characteristics of low-speed small targets;

[0029] In pointing mode, high data rate echo signal acquisition is performed for a predetermined duration to obtain high-precision detection results including target radial velocity spectrum, range profile and time series echo data.

[0030] Preferably, the method further includes:

[0031] The high-precision detection results are subjected to time-frequency analysis to extract the micro-Doppler characteristic spectrum of the target reflection signal;

[0032] Identify periodic modulation components from the microDoppler characteristic spectrum, and calculate the rotational frequency characteristics and harmonic characteristics corresponding to the periodic modulation components;

[0033] The rotation frequency feature is matched with a preset UAV propeller feature frequency database. If the match is successful, a final radar determination result confirming the UAV target is generated.

[0034] The harmonic features are compared with a preset harmonic feature library for non-UAV moving targets. If the harmonic features do not match the features in the harmonic feature library for non-UAV moving targets, the final radar determination result is further confirmed.

[0035] Preferably, the step of performing time-frequency analysis processing on the high-precision detection results to extract the micro-Doppler feature spectrum of the target reflection signal includes:

[0036] The time-series echo data is processed by short-time Fourier transform to generate a time-spectrum graph;

[0037] In the time-spectrum graph, spectral ridges with sinusoidal or periodic characteristics are searched along the time axis to extract the curve of the frequency change of the periodic modulation component over time.

[0038] The extracted frequency-time curve is subjected to Hilbert transform to demodulate the instantaneous frequency change information.

[0039] Based on the instantaneous frequency change information, the mean, variance, and periodicity of the micro-Doppler frequency shift caused by the rotation of the target component are calculated, and the micro-Doppler characteristic spectrum is generated.

[0040] Preferably, identifying periodic modulation components from the microDoppler feature spectrum includes:

[0041] Automatic peak detection processing is performed on the micro-Doppler characteristic spectrum to identify the main peak frequency point and the secondary peak frequency point in the characteristic spectrum;

[0042] Based on the interval relationship between the main peak frequency point and the secondary peak frequency point, calculate the possible fundamental frequency value and harmonic order;

[0043] The autocorrelation function of the microDoppler characteristic spectrum is calculated to detect the significance of the signal periodicity;

[0044] If the autocorrelation function calculation shows significant periodicity, then the dominant frequency and harmonic structure of the periodic modulation component are determined based on the calculated fundamental frequency value.

[0045] The signal components corresponding to the dominant frequency and harmonic structure are separated from the micro-Doppler characteristic spectrum to obtain the periodic modulation components.

[0046] Preferably, the method further includes:

[0047] When the preliminary infrared determination result is that it cannot be confirmed, and the final radar determination result is that the drone target is confirmed, the suspected drone target is determined to be a real threat target, and target confirmation information is generated.

[0048] When the preliminary infrared determination result is that it cannot be confirmed, and the rotation frequency feature fails to match the UAV propeller feature frequency database, the moving suspected thermal target is determined to be a thermal interference source, and target exclusion information is generated.

[0049] If the initial infrared determination result confirms a real UAV target, then there is no need for secondary radar determination; the target confirmation information can be generated directly.

[0050] Preferably, the method further includes:

[0051] A dual-mode detection decision fusion log is established, which records the event sequence, corresponding triggering conditions, processing procedures and final judgment results for each instance of radar triggering infrared detection or infrared triggering radar detection.

[0052] Based on historical data from the dual-mode detection decision fusion log, the radar confirmation threshold, the gradient matching threshold, and the morphological determination threshold are dynamically adjusted to optimize the sensitivity and accuracy of dual-mode collaborative detection.

[0053] The target confirmation information and the target exclusion information are integrated into a unified low-altitude target situation map, and the alarm and tracking system is driven.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] By pre-setting a signal strength confirmation threshold in the radar detection process, the infrared sensor is triggered only when a suspected target is identified and its echo intensity is below the threshold. This scheme establishes a decision-making mechanism based on radar detection confidence, transforming the infrared device's operating mode from continuous operation to event-driven. The infrared system only activates when the radar cannot make a reliable determination on its own, thus avoiding ineffective operation when the radar can clearly identify the target or when there is no target in the environment. This reduces the average workload of the infrared sensor and the overall system energy consumption, while also reducing the total amount of infrared data that needs to be processed, achieving on-demand and efficient allocation of computing resources and energy consumption.

[0056] Guidance commands are generated using real-time angle, azimuth, and distance information provided by radar to control the optical axis pointing and focusing distance of the infrared thermal imaging device. Following these commands, the infrared device directly aligns with and stares at the specific airspace containing the target, acquiring a high-resolution thermal imaging sequence of that local area. This scheme utilizes the precise ranging and angle measurement capabilities of radar to provide clear spatial coordinate guidance for the infrared device, allowing it to skip the lengthy process of autonomous search and target acquisition. This enables a seamless and rapid switch from wide-area radar search to precise infrared point-to-point investigation, shortening the system's time for detailed feature confirmation of suspicious targets and improving the collaborative detection response speed and identification accuracy for fast-moving targets. Attached Figure Description

[0057] Figure 1 This is a schematic diagram illustrating the working principle of the dual-mode detection method for low-altitude UAVs using micro-Doppler radar and infrared thermal imaging as described in this invention.

[0058] Figure 2 Flowchart for target thermal characteristic analysis and processing;

[0059] Figure 3 A flowchart for infrared-triggered radar detection;

[0060] Figure 4 Grouped bar chart of accuracy for dual-mode detection by UAVs;

[0061] Figure 5 Grouped bar charts showing the processing time for each stage of dual-mode UAV detection. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some 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.

[0063] Please see Figure 1This invention provides a dual-mode detection method for low-altitude unmanned aerial vehicles (UAVs) using micro-Doppler radar and infrared thermal imaging. The method includes: a micro-Doppler radar deployed in a low-altitude surveillance area, whose antenna continuously scans the airspace in a preset scanning mode to acquire a continuous radar echo data stream. This data stream contains target range information, radial velocity information, and micro-Doppler modulation information generated by the micro-motion of target components. An infrared thermal imaging device deployed at the same station has independent wide-angle inspection capabilities, and its optical axis pointing and radar beam pointing can be correlated and driven by a collaborative control unit. During system operation, the radar signal processing unit analyzes the input radar echo data stream in real time. When a suspected UAV target is initially identified through conventional moving target detection and micro-motion feature assessment, and the signal-to-noise ratio or amplitude of the target's echo signal is lower than a preset radar confirmation threshold, meaning that a single radar mode cannot reliably confirm the target, the radar signal processing unit generates an infrared trigger command. This command is sent to the collaborative control unit via a data bus, and the command encodes the azimuth angle, elevation angle, and estimated range information of the suspected target relative to the radar. The collaborative control unit parses the instruction, calculates the required pitch and azimuth angles for the infrared thermal imaging device, and controls its servo mechanism to rotate, quickly aligning the field of view center of the infrared thermal imaging device with the spatial area specified in the instruction. The infrared thermal imaging device then focuses and acquires a high-resolution infrared thermal imaging data frame sequence for that area. The infrared image processing unit receives this data frame sequence and performs target thermal feature analysis, calculating quantitative parameters such as the radiation temperature distribution, temperature gradient characteristics, and thermal profile morphology of the target area. Based on these extracted thermal feature parameters and according to a preset logic or classification model, the fusion decision unit determines whether a suspected UAV target is a real UAV target and generates a preliminary infrared determination result.

[0064] In one embodiment of the present invention, see [reference] Figure 2After triggering infrared thermal imaging equipment for collaborative detection via micro-Doppler radar, the acquired infrared thermal imaging data frame sequence is processed for target thermal feature analysis. From the current frame of the infrared thermal imaging data frame sequence, based on the difference between pixel grayscale values ​​and the background statistical model, a pixel set of potential target regions is extracted. The potential target regions consist of consecutive pixels with temperatures higher than the background noise set value. Temperature calibration processing is performed on the extracted pixel set. According to the calibration parameters of the infrared thermal imaging equipment, the pixel grayscale values ​​are converted into absolute or relative temperature values, generating a radiation temperature distribution. The radiation temperature distribution includes the highest temperature point value within the target region, the average temperature value of all pixels, and the temperature standard deviation data. Based on the generated radiation temperature distribution, temperature gradient features are calculated. By calculating the rate of temperature change of adjacent pixels in the horizontal and vertical directions, a two-dimensional temperature gradient map reflecting the rate and direction of temperature change is generated. Simultaneously, morphological analysis processing is performed on the potential target region. An edge detection algorithm is used to extract the target's thermal contour, and then the thermal contour morphological features are calculated. The thermal contour morphological features include the contour perimeter, the area enclosed by the contour, the aspect ratio of the contour's minimum bounding rectangle, and the fitting parameters between the contour and its bounding rectangle.

[0065] After acquiring the radiation temperature distribution, temperature gradient features, and thermal profile morphology features, target identification is performed. The highest temperature point in the radiation temperature distribution is compared with the preset temperature range of typical working parts of the UAV. If the highest temperature point is within the temperature range of typical working parts of the UAV, a valid temperature matching event is recorded. The calculated two-dimensional temperature gradient map is compared with the preset UAV thermal source gradient model for similarity matching. The structural similarity index or correlation coefficient between the two is calculated. If the similarity matching result exceeds the preset gradient matching threshold, a valid gradient matching event is recorded. The contour perimeter, area, aspect ratio, and circumscribed rectangle fitting parameters are input into a pre-trained classifier. The classifier outputs a confidence score indicating that the target is a UAV. If the output confidence score exceeds the preset morphological determination threshold, a valid morphological matching event is recorded. The cumulative number of valid temperature matching events, valid gradient matching events, and valid morphological matching events is combined. If the sum of the cumulative number of events reaches the preset confirmation threshold in the analysis of several consecutive frames of data, a preliminary infrared determination result confirming a real UAV target is generated; otherwise, a preliminary infrared determination result that cannot be confirmed is generated.

[0066] In practical implementation, after the micro-Doppler radar triggers the infrared thermal imaging equipment for collaborative detection, the acquired infrared thermal imaging data frame sequence undergoes target thermal feature analysis and judgment. From the current frame of the infrared thermal imaging data frame sequence, binarization segmentation is performed based on the difference between the grayscale value of each pixel and the background statistical model established based on historical frames to extract the pixel set of potential target areas. The potential target areas consist of connected pixel domains with temperatures higher than the background noise set value. The extracted pixel set of potential target areas undergoes temperature calibration processing. The temperature calibration processing converts the pixel grayscale values ​​into absolute temperature values ​​based on the non-uniformity correction coefficient and radiation temperature measurement curve calibrated by the infrared thermal imaging equipment at the factory, generating a radiation temperature distribution. The radiation temperature distribution includes the highest temperature point value in the target area, the average temperature value of all pixels in the potential target area, and the temperature standard deviation data of the pixels in the potential target area.

[0067] In practical implementation, temperature gradient features are calculated based on the generated radiation temperature distribution. The temperature difference between each pixel in the potential target region's pixel set and its eight neighboring pixels in the horizontal and vertical directions is calculated to obtain the lateral and vertical temperature change rates, respectively. These rates are then combined to generate a two-dimensional temperature gradient map reflecting the intensity and direction of spatial temperature changes. Morphological analysis is performed on the potential target region. The Canny edge detection algorithm is used to extract the thermal contour boundary point sequence of the potential target region. Based on the boundary point sequence, thermal contour morphological features are calculated. These features include the contour perimeter, the area enclosed by the contour, the aspect ratio of the contour's minimum bounding rectangle, and the area fit parameter between the contour and its minimum bounding rectangle. The area fit parameter is characterized by the ratio of the contour area to the area of ​​the minimum bounding rectangle.

[0068] In specific implementation, the highest temperature point in the radiation temperature distribution is compared with the preset temperature range of typical working components of the UAV. The temperature range of typical working components of the UAV is determined by thermal radiation experimental data of the motor and electronic speed controller under typical operating conditions. If the highest temperature point is within the temperature range of typical working components of the UAV, a valid temperature matching event is recorded. Optionally, the calculated two-dimensional temperature gradient map is matched with a preset UAV thermal source gradient model. The UAV thermal source gradient model is a standard temperature gradient template of a typical multi-rotor UAV under infrared thermal imaging. The similarity matching adopts the structural similarity index calculation method. If the calculated structural similarity index result exceeds the preset gradient matching threshold, a valid gradient matching event is recorded. It can be understood that the contour perimeter, area, aspect ratio, and circumscribed rectangle fitting parameters are input into a pre-trained support vector machine (SVM) classifier. The SVM classifier is trained offline using historically collected thermal contour morphological feature samples of UAV and non-UAV targets. If the output confidence score of the SVM classifier exceeds the preset morphological judgment threshold, a valid morphological matching event is recorded.

[0069] Optionally, by combining the cumulative counts of effective temperature matching events, effective gradient matching events, and effective shape matching events during the analysis of five consecutive frames of infrared thermal imaging data, if the sum of the cumulative counts reaches a preset confirmation threshold, a preliminary infrared determination result confirming the real UAV target is generated; if the sum of the cumulative counts does not reach the preset confirmation threshold, a preliminary infrared determination result indicating no confirmation is generated. In specific implementation, the formula for calculating the area fitting parameter is:

[0070]

[0071] in: This represents the area fit parameter. This represents the area enclosed by the contour calculated from the sequence of boundary points. This represents the area of ​​the smallest bounding rectangle of the outline.

[0072] In one embodiment of the present invention, see [reference] Figure 3 In specific implementation, the infrared thermal imaging equipment independently inspects the low-altitude monitoring area. The equipment performs periodic horizontal scans at a fixed elevation angle to acquire a wide-field-of-view infrared image data stream covering a broad airspace. This data stream is input to the infrared image processing unit at a fixed frame rate for real-time analysis. In some embodiments, the real-time analysis of the wide-field-of-view infrared image data stream is processed using a background subtraction algorithm and an inter-frame difference method. The background subtraction algorithm uses a Gaussian mixture model to establish a background model, while the inter-frame difference method calculates the changes in pixel grayscale values ​​between consecutive frames to identify moving suspected thermal targets. Moving suspected thermal targets are represented by connected hot pixel regions whose positions change continuously across consecutive frames. In practice, after identifying a moving suspected thermal target, a preliminary thermal feature analysis is performed on the moving suspected thermal target. The preliminary thermal feature analysis includes calculating the average radiation temperature of pixels and the pixel area of ​​the region of the moving suspected thermal target. If the average radiation temperature value of pixels falls within the preset temperature range where birds and drones coexist and the pixel area value of the region is close to the preset typical pixel area value of small birds, then it is determined that the moving suspected thermal target cannot be distinguished from the thermal feature as a source of thermal interference from drones, birds, or hot air balloons.

[0073] Understandably, when the determination result is indistinguishable, the infrared image processing unit generates a radar trigger command. This command includes the azimuth information and estimated distance information of the moving suspected thermal target. The azimuth information is calculated by combining the azimuth encoder reading of the infrared thermal imaging device's gimbal with the pixel coordinate offset of the moving suspected thermal target in the infrared image plane. The estimated distance information is estimated based on the pixel size of the moving suspected thermal target in the infrared image and the preset physical size of a typical UAV target through perspective projection. Optionally, based on the received radar trigger command, the co-control unit parses the azimuth and estimated distance information in the command and generates control commands to drive the servo mechanism of the micro-Doppler radar. The servo mechanism adjusts the pointing angle of the radar antenna, precisely aligning the radar beam axis with the spatial area specified by the radar trigger command. In some embodiments, the signal processing unit of the micro-Doppler radar configures the waveform parameters and pulse repetition frequency of the radar transmitted signal according to the estimated range information in the radar trigger command. The waveform parameters are selected as linear frequency modulated continuous wave waveforms to obtain high range resolution, and the pulse repetition frequency is set to a value greater than twice the Doppler frequency corresponding to the expected maximum rotation frequency of the UAV propeller. The configuration of the waveform parameters and pulse repetition frequency is suitable for resolving the micro-motion characteristics of low-speed small targets.

[0074] In practical implementation, after the micro-Doppler radar beam is stably pointed towards the target space region, the radar transmitter and receiver perform high-data-rate echo signal acquisition for a predetermined duration. The sampling rate of the high-data-rate echo signal acquisition meets the Nyquist sampling theorem requirement for the expected signal bandwidth, obtaining high-precision detection results including the target radial velocity spectrum, high-resolution range profile, and time-series echo data. The formula for calculating the estimated range information can be understood as follows:

[0075]

[0076] in: This indicates the estimated distance of a moving suspected hot target. This indicates the focal length of the optical lens in an infrared thermal imaging device. This indicates the actual size of a pre-defined typical drone target. This represents the pixel width of a moving, suspected thermal target in the infrared image along the horizontal direction. This represents the physical size of a single pixel in the infrared detector. Optionally, the generation of the radar trigger command also relies on the continuity check of the trajectory of a moving suspected hot target. The continuity check is achieved by calculating the displacement vector of the moving suspected hot target in multiple consecutive frames of infrared images and checking the stability of the direction and magnitude of the displacement vector.

[0077] In one embodiment of the present invention, after obtaining high-precision detection results for a specific area through infrared triggering mode, the results are analyzed in depth to confirm the nature of the target. Time-frequency analysis processing is performed on the acquired time-series echo data to extract the micro-Doppler characteristic spectrum of the target reflection signal. Periodic modulation components are identified from the micro-Doppler characteristic spectrum, and the rotational frequency characteristics and harmonic characteristics corresponding to the periodic modulation components are calculated. The calculated rotational frequency characteristics are matched with a preset UAV propeller characteristic frequency database, which contains typical rotational frequencies and ranges of different UAV propeller models. If the match is successful, a final radar determination result confirming the UAV target is generated. Simultaneously, the identified harmonic characteristics are compared with a preset harmonic characteristic library for non-UAV moving targets. This library includes typical harmonic modes of movements such as bird wing flapping. If the harmonic characteristics do not match the typical modes in this library, the reliability of the final radar determination result is further confirmed.

[0078] In practical implementation, the high-precision detection results for specific spatial regions acquired through infrared triggering mode are analyzed in depth. These high-precision detection results include the target radial velocity spectrum, high-resolution range profile, and time-series echo data. The radar signal processing unit performs time-frequency analysis on the time-series echo data, employing a short-time Fourier transform method to generate a time-spectrum image and extract micro-Doppler characteristic spectra reflecting the micro-motion characteristics of the target components. In some embodiments, periodic modulation components are identified from the extracted micro-Doppler characteristic spectra. This identification process includes peak search and spectral line tracing in the spectral domain to separate the periodic signal components generated by rotating components such as propellers. The rotational frequency characteristics and harmonic characteristics corresponding to the periodic modulation components are calculated. The rotational frequency characteristics refer to the fundamental frequency of the modulation component, and the harmonic characteristics refer to the amplitude and phase relationship of the harmonic components at integer multiples of the fundamental frequency.

[0079] In practical implementation, the calculated rotational frequency characteristics are matched with a pre-set database of UAV propeller characteristic frequencies. This database stores the range of propeller rotational frequencies for various commercial UAV models under different throttle conditions. The matching process compares the calculated rotational frequency characteristic value with the frequency range in the database one by one. If the rotational frequency characteristic value falls within the typical rotational frequency range of any UAV propeller model, the match is considered successful, and a final radar determination result confirming the UAV target is generated. This can be understood as comparing the identified harmonic features with a pre-set harmonic feature library for non-UAV moving targets. This library includes typical harmonic modes generated by the flapping wings of birds, the main rotor of rotorcraft helicopters, etc. Harmonic modes include quantitative descriptions such as the amplitude ratio of the fundamental frequency to harmonics and phase relationships. The harmonic feature comparison calculates the degree of difference between the current harmonic feature and each typical harmonic mode in the feature library.

[0080] In some embodiments, if the harmonic features differ significantly from the typical amplitude attenuation patterns recorded in the harmonic feature library of bird flapping motion (where amplitude attenuation refers to the rate at which harmonic amplitude decreases with increasing harmonic order), and the features in the harmonic feature library of non-UAV moving targets contain specific amplitude attenuation thresholds, then it is determined that the current harmonic features do not match the features in the harmonic feature library of non-UAV moving targets. This result serves as supporting evidence to further confirm the reliability of the final radar determination result. Optionally, the matching of rotation frequency features and the comparison of harmonic features can be performed independently and in parallel. The logic for generating the final radar determination result is set such that a high-confidence final radar determination result is generated only when the rotation frequency feature is successfully matched and the harmonic feature comparison does not match. In a specific implementation, the formula for calculating the difference between the harmonic features and the k-th typical mode in the feature library is expressed as:

[0081]

[0082] in: This represents the overall degree of difference from the k-th typical pattern. This indicates the total number of harmonic orders considered. This represents the measured amplitude of the nth harmonic. This represents the reference value of the nth harmonic amplitude corresponding to the kth typical mode in the feature library. The weighting coefficients represent the amplitude differences of the nth harmonic. This indicates the measured fundamental harmonic phase. This represents the fundamental harmonic phase reference value corresponding to the k-th typical mode in the feature library. This represents the weighting coefficient for phase difference. It can be understood that if the calculated minimum difference... If the difference exceeds the preset threshold, the harmonic features are determined to be inconsistent with the features in the harmonic feature library of non-UAV moving targets.

[0083] In one embodiment of the present invention, time-frequency analysis is performed on the time-series echo data in the high-precision detection results to extract the micro-Doppler characteristic spectrum. The time-series echo data is a complex baseband signal sequence received by the radar from a specific spatial region. The radar signal processing unit performs short-time Fourier transform processing on the time-series echo data. The short-time Fourier transform processing uses a finite-length sliding time window to segment and window the signal. Discrete Fourier transform is performed on the signal segment within each window to generate a time-spectrum diagram with time as the horizontal axis and frequency as the vertical axis. In some embodiments, refer to Table 1 for the specific parameter configuration of the short-time Fourier transform processing.

[0084] Table 1: Specific parameter configuration table for short-time Fourier transform processing

[0085] Parameter name Setting value or type illustrate Window function type Hanning Window Used to reduce spectrum leakage Window length 256 sampling points The trade-off between time resolution and frequency resolution Overlap rate 75% The proportion of overlapping sampling points between adjacent windows Fourier transform points 512 Determines the refinement level of the frequency axis.

[0086] It can be understood that in the generated time-spectrum graph, spectral ridges with sinusoidal or periodic characteristics are searched along the time axis. The search algorithm achieves this by detecting local maxima of the spectral amplitude at each time point and connecting the maxima points of adjacent time points to form continuous trajectories. The frequency of the periodic modulation components represented by these trajectories is extracted as a function of time. In specific implementation, the extracted frequency-time curves are subjected to Hilbert transform processing. Hilbert transform processing obtains the analytical signal of the original curve by convolving the original curve with a specific kernel function, and then demodulates the instantaneous frequency change information from the analytical signal. Based on the demodulated instantaneous frequency change information, the statistical characteristics of the micro-Doppler frequency shift caused by the rotation of the target component are calculated. The calculation process includes taking the arithmetic mean of the instantaneous frequency change information sequence to obtain the mean, taking the average of the squares of the differences between each element of the sequence and the mean to obtain the variance, and using Fourier transform to analyze the spectral components of the sequence to evaluate the periodicity, generating a micro-Doppler characteristic spectrum characterizing the micro-motion characteristics of the target.

[0087] In some embodiments, periodic modulation components are identified from the generated micro-Doppler characteristic spectrum. Automatic peak detection processing is performed on the micro-Doppler characteristic spectrum. This automatic peak detection processing scans the amplitude array of the micro-Doppler characteristic spectrum by setting an amplitude threshold and a minimum peak interval to identify the main peak frequency points and secondary peak frequency points in the characteristic spectrum. Based on the frequency interval relationship between the identified main peak frequency points and secondary peak frequency points, the common divisor or least common multiple of the intervals is analyzed to calculate the possible fundamental frequency value and harmonic order. The fundamental frequency value is an estimate of the greatest common divisor of these frequency intervals. Optionally, an autocorrelation function is calculated on the micro-Doppler characteristic spectrum. The autocorrelation function calculation correlates the characteristic spectrum amplitude sequence with its own time-shifted sequence. The significance of the periodicity of the original micro-Doppler characteristic spectrum is determined by detecting the interval pattern of peak occurrences in the autocorrelation function result. If the autocorrelation function calculation result shows significant peaks with equal intervals at multiple delay positions, the periodicity is determined to be significant. In practical implementation, after determining the periodicity to be significant, the dominant frequency and harmonic structure of the periodic modulation component are determined based on the previously calculated fundamental frequency value. The dominant frequency is determined as the calculated fundamental frequency value, and the harmonic structure is determined by whether there are corresponding peak values ​​at integer multiples of the dominant frequency. The signal components corresponding to the dominant frequency and harmonic structure are separated from the original micro-Doppler characteristic spectrum. The separation process is achieved by designing a bandpass filter bank centered on the dominant frequency and its harmonics to filter the micro-Doppler characteristic spectrum, obtaining pure periodic modulation components for subsequent analysis. The significance of periodicity can be quantified by calculating the kurtosis index of the autocorrelation function, and the calculation formula is:

[0088]

[0089] in: Kurtosis is a metric representing the kurtosis of the autocorrelation function, used to quantify the sharpness of periodic peaks. Indicates the amplitude sequence of the microDoppler characteristic spectrum in the delay The autocorrelation function value at that location, and This indicates the start and end values ​​for calculating the delay. It can be understood that when... When the value exceeds the preset kurtosis threshold, the result of the autocorrelation function calculation is determined to show significant periodicity.

[0090] See Figure 4 This is a grouped bar chart showing the accuracy of dual-mode detection and judgment for UAVs. Its core content is to demonstrate the accuracy performance of infrared, radar, and fusion judgment under different detection scenarios. In the "direct infrared confirmation" scenario, infrared accuracy is significantly higher than radar; in the "radar secondary confirmation" scenario, radar accuracy is higher than infrared, demonstrating radar's advantage in secondary verification; in most scenarios, the fusion judgment accuracy is between infrared and radar, or close to the higher of both, indicating that dual-mode fusion can balance the shortcomings of a single method; the accuracy in all scenarios is above 85%, reflecting the strong reliability of this dual-mode detection method. This type of chart is commonly used for performance evaluation of low-altitude UAV dual-mode detection systems. By comparing the accuracy in different scenarios, the triggering conditions and threshold parameters for dual-mode collaboration can be optimized, improving the sensitivity and accuracy of target recognition.

[0091] In one embodiment of the present invention, the system performs a fusion decision on the judgment results from two collaborative modes: radar-triggered infrared detection and infrared-triggered radar detection. When the infrared thermal imaging device is initially triggered by micro-Doppler radar, and the initial infrared judgment result generated by the infrared image processing unit is "unable to confirm," but subsequently, after high-precision detection by the micro-Doppler radar triggered by the suspected target, the final radar judgment result generated by the radar signal processing unit is "confirmed UAV target," the fusion decision unit determines the initial suspected UAV target as a real threat target and generates complete target confirmation information including the target's three-dimensional coordinates, radial velocity, timestamp, and target classification information.

[0092] In practical implementation, when the infrared thermal imaging device detects the target for the first time after the micro-Doppler radar triggers it, and the initial infrared judgment result generated by the infrared image processing unit is "unable to confirm," and the rotation frequency characteristics obtained by the subsequent radar signal processing unit from the high-precision detection result analysis fail to match the UAV propeller characteristic frequency database, the fusion decision unit will classify the corresponding moving suspected thermal target as a thermal interference source such as a bird or hot air balloon, generate target exclusion information including the target's initial position, the reason for the judgment, and a timestamp, and remove the target from the system's continuous tracking queue. It can be understood that if the initial infrared judgment result generated by the infrared image processing unit is directly "confirmed as a real UAV target," the fusion decision unit does not need to initiate or wait for the radar's secondary judgment process; it can directly generate target confirmation information based on the infrared judgment result. The target confirmation information includes the target's position and thermal characteristic parameters calculated from the infrared thermal imaging data.

[0093] In some embodiments, the system establishes a dual-mode detection decision fusion log. This log records each event sequence triggered by radar for infrared detection or vice versa in a structured database table. The event sequence records include the type of triggering sensor, trigger time, initial parameters of the target at trigger time, key parameters of the subsequent processing by the infrared thermal imaging device or micro-Doppler radar, and the preliminary infrared determination result or the final radar determination result. Optionally, based on historical data accumulated in the dual-mode detection decision fusion log, the radar confirmation threshold, gradient matching threshold, and morphological determination threshold are dynamically adjusted. This dynamic adjustment is based on statistical analysis of historical false alarm events and historical missed detection events to optimize the sensitivity and accuracy of dual-mode collaborative detection. When adjusting the radar confirmation threshold, if historical data shows an increase in the false alarm rate within a specific signal-to-noise ratio range, the radar confirmation threshold is appropriately increased. When adjusting the gradient matching threshold and morphological determination threshold, if historical data shows an increase in missed detections of a certain type of UAV, the corresponding threshold is appropriately decreased.

[0094] In practical implementation, the target confirmation and target exclusion information generated by the fusion decision unit are integrated into a unified low-altitude target situation map. The low-altitude target situation map uses different icons and trajectory lines on the display interface to distinguish confirmed real threat targets from excluded false alarm targets. The low-altitude target situation map drives external audible and visual alarm systems and automatic tracking systems. When new target confirmation information is added to the low-altitude target situation map, the audible and visual alarm system is activated, and the automatic tracking system controls the photoelectric tracking device to lock onto and continuously track the target based on its coordinates. The calculation formula for adjusting the morphological judgment threshold based on historical logs by the fusion decision unit can be expressed as:

[0095]

[0096] in: This indicates the adjusted morphological threshold. This indicates the morphological threshold before adjustment. This indicates adjusting the step size coefficient. This indicates the proportion of false alarms in recent historical data caused by morphological judgment thresholds. This represents the proportion of missed events in recent historical data due to morphological threshold determination. It can be understood that this formula allows for adaptive adjustment of the threshold: raising the threshold when the false alarm rate is higher than the missed alarm rate, and lowering the threshold when the missed alarm rate is higher than the false alarm rate.

[0097] See Figure 5 This is a grouped bar chart showing the processing time of each stage in a dual-mode UAV detection process, illustrating the time overhead of different stages. "Radar time-frequency analysis" is the longest stage (approximately 180ms), reflecting the higher computational complexity of radar signal processing. The "threshold adjustment" and "situation update" stages are the shortest (approximately 15-20ms), representing lightweight post-decision processing. The "infrared feature extraction" stage involves only the infrared module (radar time is 0ms), and the "radar time-frequency analysis" stage involves only the radar module, demonstrating the "stage-by-stage collaboration" logic of dual-mode detection. The "fusion decision" stage has the same processing time for both infrared and radar (approximately 45ms), indicating that this stage involves parallel fusion computation of dual-mode data. These types of charts are primarily used for performance optimization of dual-mode detection systems, assessing resource consumption at each stage, and providing a basis for hardware computing power allocation.

[0098] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0099] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dual-mode detection method for low-altitude unmanned aerial vehicles using micro-Doppler radar and infrared thermal imaging, characterized in that, The method includes: The low-altitude surveillance area is continuously scanned and detected by micro-Doppler radar to obtain radar echo data streams containing information on distance, speed and micro-motion characteristics. The radar echo data stream is analyzed in real time. When a suspected UAV target is identified and its echo signal strength is lower than the preset radar confirmation threshold, an infrared trigger command is generated. The infrared trigger command includes the angle, azimuth and distance information of the suspected UAV target. According to the infrared trigger command, the infrared thermal imaging device is controlled to adjust its direction and focus on the spatial region corresponding to the angle, azimuth and distance information, and a high-resolution infrared thermal imaging data frame sequence of the spatial region is obtained; The infrared thermal imaging data frame sequence is subjected to target thermal feature analysis processing to calculate the radiation temperature distribution, temperature gradient characteristics and thermal profile morphology characteristics of the target area; Based on the radiation temperature distribution, the temperature gradient characteristics, and the thermal profile morphology characteristics, determine whether the suspected UAV target is a real UAV target and generate a preliminary infrared determination result. The method further includes: When the preliminary infrared determination result is that it cannot be confirmed, the low-altitude monitoring area is independently inspected by infrared thermal imaging equipment to obtain a wide field-of-view infrared image data stream. The system analyzes the wide field-of-view infrared image data stream in real time. When a moving suspected thermal target is identified and it is impossible to distinguish the suspected thermal target as a source of thermal interference from a drone, bird, or hot air balloon based on its thermal characteristics, a radar trigger command is generated. The radar trigger command includes the azimuth and estimated distance information of the moving suspected thermal target. According to the radar trigger command, the micro-Doppler radar is controlled to adjust the beam and perform high-precision detection on the spatial region corresponding to the azimuth and estimated distance information to obtain the target echo signal in the spatial region. The control micro-Doppler radar adjusts its beam to perform high-precision detection of the spatial region corresponding to the azimuth and estimated range information, including: Based on the azimuth and estimated distance information, the servo mechanism of the micro-Doppler radar is driven to point the radar beam axis toward the spatial region. Configure the waveform parameters and repetition frequency of the radar transmitted signal, wherein the waveform parameters and repetition frequency are suitable for analyzing the micro-motion characteristics of low-speed small targets; In pointing mode, high data rate echo signal acquisition is performed for a predetermined duration to obtain high-precision detection results including target radial velocity spectrum, range profile and time series echo data; The high-precision detection results are subjected to time-frequency analysis to extract the micro-Doppler characteristic spectrum of the target reflection signal; Identify periodic modulation components from the microDoppler characteristic spectrum, and calculate the rotational frequency characteristics and harmonic characteristics corresponding to the periodic modulation components; Furthermore, if the final radar determination result confirms the drone target, the suspected drone target will be determined as a real threat target, and target confirmation information will be generated. Furthermore, if the rotation frequency feature fails to match the UAV propeller feature frequency database, the moving suspected thermal target will be identified as a thermal interference source, and target exclusion information will be generated. If the initial infrared determination result confirms a real UAV target, then there is no need for secondary radar determination; the target confirmation information can be generated directly.

2. The dual-mode detection method for low-altitude UAVs using micro-Doppler radar and infrared thermal imaging according to claim 1, characterized in that, The target thermal feature analysis processing of the infrared thermal imaging data frame sequence includes: Extract a set of pixels representing a potential target region from the infrared thermal imaging data frame sequence. The potential target region consists of consecutive pixels with a temperature higher than a set value of background noise. Temperature calibration processing is performed on the pixel set to generate the radiation temperature distribution, which includes the highest temperature point, average temperature and temperature standard deviation data within the target area. The temperature gradient feature is calculated based on the radiation temperature distribution. The temperature gradient feature is obtained by calculating the temperature change rate between adjacent pixels in the pixel set and generating a two-dimensional temperature gradient map. Morphological analysis is performed on the potential target region to extract the thermal profile morphological features, which include the profile perimeter, area, aspect ratio, and circumscribed rectangle fitting parameters.

3. The dual-mode detection method for low-altitude UAVs using micro-Doppler radar and infrared thermal imaging according to claim 2, characterized in that, The determination of whether a suspected drone target is a real drone target based on the radiation temperature distribution, the temperature gradient characteristics, and the thermal profile morphology characteristics includes: The highest temperature point is compared with the preset temperature range of typical working parts of the drone. If the highest temperature point is within the temperature range of typical working parts of the drone, a valid temperature matching event is recorded. The two-dimensional temperature gradient map is matched with a preset UAV heat source gradient model. If the similarity matching result exceeds the preset gradient matching threshold, a valid gradient matching event is recorded. The perimeter of the outline, the area, the aspect ratio, and the fitting parameters of the bounding rectangle are input into a pre-trained classifier. If the output confidence of the classifier exceeds the preset morphological judgment threshold, a valid morphological matching event is recorded. If the sum of the cumulative counts of the effective temperature matching events, the effective gradient matching events, and the effective shape matching events reaches a preset confirmation count threshold, then a preliminary infrared determination result confirming the real UAV target is generated; otherwise, a preliminary infrared determination result that cannot be confirmed is generated.

4. The dual-mode detection method for low-altitude unmanned aerial vehicles using micro-Doppler radar and infrared thermal imaging according to claim 3, characterized in that, The method further includes: The rotation frequency feature is matched with a preset UAV propeller feature frequency database. If the match is successful, a final radar determination result confirming the UAV target is generated. The harmonic features are compared with a preset harmonic feature library for non-UAV moving targets. If the harmonic features do not match the features in the harmonic feature library for non-UAV moving targets, the final radar determination result is further confirmed.

5. The dual-mode detection method for low-altitude unmanned aerial vehicles using micro-Doppler radar and infrared thermal imaging according to claim 4, characterized in that, The step of performing time-frequency analysis on the high-precision detection results to extract the micro-Doppler feature spectrum of the target reflection signal includes: The time-series echo data is processed by short-time Fourier transform to generate a time-spectrum graph; In the time-spectrum graph, spectral ridges with sinusoidal or periodic characteristics are searched along the time axis to extract the curve of the frequency change of the periodic modulation component over time. The extracted frequency-time curve is subjected to Hilbert transform to demodulate the instantaneous frequency change information. Based on the instantaneous frequency change information, the mean, variance, and periodicity of the micro-Doppler frequency shift caused by the rotation of the target component are calculated, and the micro-Doppler characteristic spectrum is generated.

6. The dual-mode detection method for low-altitude unmanned aerial vehicles using micro-Doppler radar and infrared thermal imaging according to claim 5, characterized in that, The step of identifying periodic modulation components from the microDoppler characteristic spectrum includes: Automatic peak detection processing is performed on the micro-Doppler characteristic spectrum to identify the main peak frequency point and the secondary peak frequency point in the characteristic spectrum; Based on the interval relationship between the main peak frequency point and the secondary peak frequency point, calculate the possible fundamental frequency value and harmonic order; The autocorrelation function of the microDoppler characteristic spectrum is calculated to detect the significance of the signal periodicity; If the autocorrelation function calculation shows significant periodicity, then the dominant frequency and harmonic structure of the periodic modulation component are determined based on the calculated fundamental frequency value. The signal components corresponding to the dominant frequency and harmonic structure are separated from the micro-Doppler characteristic spectrum to obtain the periodic modulation components.

7. The dual-mode detection method for low-altitude unmanned aerial vehicles using micro-Doppler radar and infrared thermal imaging according to claim 6, characterized in that, The method further includes: A dual-mode detection decision fusion log is established, which records the event sequence, corresponding triggering conditions, processing procedures and final judgment results for each instance of radar triggering infrared detection or infrared triggering radar detection. Based on historical data from the dual-mode detection decision fusion log, the radar confirmation threshold, the gradient matching threshold, and the morphological determination threshold are dynamically adjusted to optimize the sensitivity and accuracy of dual-mode collaborative detection. The target confirmation information and the target exclusion information are integrated into a unified low-altitude target situation map, and the alarm and tracking system is driven.

Images