A real-time identification method and system for drone models based on lidar point cloud features

Abstract

This application provides a real-time UAV model identification method and system based on lidar point cloud features. The method first acquires a raw point cloud sequence, then separates the inherent scattering signal associated with the UAV model from the transient noise signal associated with complex meteorological interference. The method then parses the inherent scattering signal to extract a dynamic structural response spectrum, which is then synchronously compared with multiple known UAV model benchmark response spectra to generate an optimal fit determination result. Finally, based on the optimal fit determination result, the UAV model is output. The technical solution provided by this application significantly enhances the stability and accuracy of UAV identification in complex meteorological environments.

Description

Technical Field

[0001] The present application relates to the field of drone identification technology, and in particular to a method and system for real-time drone model identification based on laser radar point cloud features. Background Art

[0002] With the increasing popularity of drones, real-time identification of medium- and long-range commercial drone models in harsh scenarios such as dense urban electromagnetic interference, complex weather conditions such as dust and haze, and high-speed dive maneuvers requires rapid response capabilities, high recognition accuracy, and all-weather anti-interference characteristics.

[0003] To meet the above needs, the current recognition solution adopts the fusion of lidar point cloud and thermal imaging information. The spatial configuration is constructed by scanning the three-dimensional point cloud of the target surface, and the thermal distribution map is collected to reflect the characteristics of the power components. The fusion features are extracted with the help of pre-trained deep neural network to match the model database.

[0004] This solution has significant limitations. First, the point cloud distortion caused by the high-speed movement of the target makes the contour features obviously distorted, the thermal imaging lag causes prominent deviations in the thermal identification of core components, and the asynchronous acquisition of multi-source sensors causes time and space mismatch problems, causing the system's recognition accuracy to deteriorate sharply under severe weather conditions, making it difficult to ensure the recognition reliability of complex scenes. Summary of the Invention

[0005] The present application provides a real-time identification method and system for drone models based on lidar point cloud features, which is used to solve the problems of thermal imaging lag distortion, spatiotemporal mismatch error, and point cloud feature distortion under high-speed motion caused by reliance on multi-source sensor fusion in the existing technology.

[0006] In a first aspect, the present application provides a method for real-time identification of drone models based on lidar point cloud features, comprising:

[0007] Acquire an original point cloud sequence generated by the laser radar scanning the UAV in real time under the target flight area, wherein the original point cloud sequence includes disturbed three-dimensional space reflection points;

[0008] Based on the physical reflection characteristics of the original point cloud sequence, separating the inherent scattered signal related to the UAV configuration and the instantaneous noise signal related to complex meteorological interference;

[0009] Resolving the inherent scattering signal to obtain a dynamic structural response spectrum representing the external components of the UAV;

[0010] Synchronously comparing the dynamic structural response spectrum with a plurality of pre-stored benchmark response spectra of known UAV models, and generating an optimal fit determination result based on the comparison results;

[0011] Based on the optimal fit determination result, the drone model is output.

[0012] Optionally, obtaining an original point cloud sequence generated by the laser radar scanning the drone in real time under the target flight area, wherein the original point cloud sequence contains disturbed three-dimensional space reflection points, including:

[0013] Controlling the laser radar to transmit a scanning beam group toward the flight area, wherein the scanning beam group generates a continuous laser pulse sequence in a scanning plane at a fixed pulse repetition frequency;

[0014] capturing a group of multiple scattered echo signals generated by the continuous laser pulse sequence striking the surface of the UAV and meteorological suspended particles after passing through the complex meteorological interference layer;

[0015] Performing time domain signal slicing on the multiple scattered echo signal group to generate a time window unit synchronized with the pulse period;

[0016] Extract the energy concentration distribution characteristics and pulse return time series of scattered echoes in each time window unit;

[0017] Identifying, based on the persistent correlation of the main energy clusters in the energy concentration distribution characteristics, the structural reflection echo belonging to the drone body and the ambient noise echo belonging to meteorological suspended particles;

[0018] Determining the spatial reflection point position on the surface of the UAV body based on the pulse return time sequence of the structure reflected echo;

[0019] The spatial reflection point positions are combined with the environmental noise interference intensity mark of the corresponding time window unit to construct a three-dimensional reflection point set with interference perception attributes, and the three-dimensional reflection point set with interference perception attributes is combined to form an original point cloud sequence.

[0020] Optionally, based on the physical reflection characteristics of the original point cloud sequence, separating the inherent scattered signal associated with the UAV configuration and the transient noise signal associated with complex meteorological interference includes:

[0021] Extracting the laser wavelength channel reflection intensity ratio, polarization state change degree and pulse transmission time change rate corresponding to each three-dimensional spatial reflection point from the original point cloud sequence;

[0022] Generating wavelength reflection intensity distribution bands under different meteorological interference types based on the laser wavelength channel reflection intensity ratio;

[0023] constructing a dynamic interference filtering model based on the coupling relationship between the polarization state change degree and the pulse transmission time change rate;

[0024] Marking a set of reflection points in the wavelength reflection intensity distribution band that satisfy the metal material reflection stability law as a first-level candidate signal;

[0025] In the dynamic interference filtering model, based on the linear constraint threshold of the coupling relationship, the echo points with continuous pulse time series in the first-level candidate signal are screened;

[0026] The three-dimensional spatial position of the screened echo points is traced to form the inherent scattering signal related to the UAV configuration;

[0027] A set of reflection points not covered by the wavelength reflection intensity distribution band and the dynamic interference filtering model is classified as the instantaneous noise signal.

[0028] Optionally, parsing a dynamic structural response spectrum representing the external component of the UAV from the inherent scattering signal includes:

[0029] Dividing a plurality of spatial reflection points in the inherent scattered signal into position areas corresponding to the external components of the UAV;

[0030] Fitting the motion trajectory vector of each position area based on the displacement vector sequence of the spatial reflection point changing with time;

[0031] Matching the motion trajectory vector of each position area with the standard component stiffness parameters in a preset UAV manufacturing material property library to obtain the matched standard component stiffness parameters;

[0032] generating a local vibration wave propagation direction chain of the external component of the UAV based on the matched standard component stiffness parameters;

[0033] On the local vibration wave propagation direction chain, mapping the vibration energy envelope of the spatial reflection point corresponding to each propagation direction node;

[0034] generating a time-space response intensity mapping relationship of the position area according to an amplitude variation relationship of the vibration energy envelope in a scanning time sequence corresponding to the position area;

[0035] The time-space response intensity mapping relationship of all position areas is integrated to form the dynamic structural response map.

[0036] Optionally, the dynamic structural response spectrum is synchronously compared with a plurality of pre-stored reference response spectra of known UAV models, and an optimal fit determination result is generated based on the comparison results, including:

[0037] Extracting the characteristic change frequency of the time-space response intensity mapping relationship of each position area in the dynamic structural response map;

[0038] Aligning the characteristic change frequency with the reference frequency band corresponding to the pre-stored UAV model reference response spectrum in time and space coverage area;

[0039] After the spatiotemporal coverage areas are aligned, calculating the degree of continuous distribution overlap of the characteristic change frequencies within the reference frequency band;

[0040] Based on the continuous distribution overlap, identifying the similarity of spectral energy transfer paths between the dynamic structural response spectrum and each benchmark response spectrum;

[0041] generating a frequency offset phase difference set according to a sequence of overlapping positions of the characteristic change frequency and the reference frequency band;

[0042] Suppressing the contribution of phase anomaly points associated with a meteorological interference intensity of a preset threshold value in the frequency offset phase difference set to obtain a phase anomaly point suppression result;

[0043] The continuous distribution coincidence, the spectrum energy transfer path similarity and the phase abnormal point suppression result are combined to generate the optimal fit determination result.

[0044] In a second aspect, the present application provides a real-time drone model identification system based on lidar point cloud features, comprising:

[0045] An acquisition module is used to acquire an original point cloud sequence generated by the laser radar scanning the UAV in real time under the target flight area, wherein the original point cloud sequence includes disturbed three-dimensional space reflection points;

[0046] a separation module, configured to separate the inherent scattered signal associated with the UAV configuration and the transient noise signal associated with complex meteorological interference based on the physical reflectance characteristics of the original point cloud sequence;

[0047] An analysis module, configured to analyze the inherent scattering signal to obtain a dynamic structural response spectrum representing the external components of the UAV;

[0048] a comparison module, configured to synchronously compare the dynamic structural response spectrum with a plurality of pre-stored reference response spectra of known UAV models, and generate an optimal fit determination result based on the comparison results;

[0049] The output module is used to output the drone model based on the optimal fit determination result.

[0050] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a real-time drone model identification method based on lidar point cloud features as described in the first aspect above.

[0051] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a real-time identification method for drone models based on lidar point cloud features as described in the first aspect.

[0052] This embodiment of the application builds a complete anti-interference recognition system through in-depth analysis of the physical properties of a single-source laser point cloud, replacing the traditional technology approach that relies on multi-sensor fusion. Based on the coupled analysis of the laser wavelength reflection characteristics, polarization state, and the time-varying relationship of pulse transmission, it achieves precise separation of meteorological noise and target body reflection, effectively avoiding the lag distortion and spatiotemporal mismatch defects of thermal imaging and multi-source signal. At the same time, by leveraging the dynamic structural response spectrum construction mechanism, the spatial displacement characteristics of the point cloud are converted into a vibration wave propagation path and spatiotemporal energy response model dominated by material mechanics. This completely resolves the technical bottleneck of target contour distortion in high-speed maneuvering scenarios and significantly enhances recognition stability and accuracy in complex meteorological environments.

[0053] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0055] Figure 1 A flowchart of a method for real-time identification of drone models based on lidar point cloud features provided by the present application is shown;

[0056] Figure 2 The present invention provides a schematic diagram of a system for real-time identification of drone models based on lidar point cloud features.

[0057] Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0058] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0059] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0060] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0061] Figure 1 The present invention provides a flowchart of a method for real-time identification of drone models based on laser radar point cloud features, as shown in FIG. Figure 1 As shown, the method includes:

[0062] Step 101: Obtain an original point cloud sequence generated by the laser radar in real-time scanning of the drone in the target flight area, wherein the original point cloud sequence includes disturbed three-dimensional space reflection points.

[0063] In this step, the original point cloud sequence refers to the data stream sequence containing three-dimensional coordinates and reflection intensity information generated by the laser radar scanning target, and its three-dimensional spatial reflection points are discrete spatial point sets with time and space attributes; the interfered three-dimensional spatial reflection points refer to a mixed data set containing target body reflection points and environmental interference reflection points, and the environmental interference reflection points specifically refer to the scattering signals of meteorological suspended particles to the laser; the scanning beam group refers to an array of laser pulses emitted at a fixed pulse repetition frequency, which forms scanning coverage in the horizontal / vertical direction through a beam deflection device; the multiple scattered echo signal group refers to the composite echo formed by the laser pulse scattered by the target surface and multiple reflections of suspended particles, which contains time-aliased intensity and phase information.

[0064] In an embodiment of the present application, a scanning beam group modulated by a fixed pulse repetition frequency is first emitted to the flight area through a laser emission module, and each pulse covers a preset azimuth angle in the scanning plane; secondly, a receiving module captures a multi-scattered echo signal group formed by scattering of atmospheric suspended particles and reflection on the surface of the UAV, which carries a phase distortion feature of time aliasing; then a signal processing unit performs a time domain slicing operation on the echo group, divides the time window unit based on the pulse period, and extracts the energy concentration distribution characteristics and pulse return time series of the echo in each window; then, based on the spatial overlap rate and intensity attenuation continuity of the main energy cluster in the continuous time window, the UAV body structure reflection echo and the meteorological suspended particle environmental noise echo are distinguished; finally, the three-dimensional coordinates of the spatial reflection point are calculated according to the pulse time series of the structure reflection echo, and the mean value of the environmental noise intensity in the time window is superimposed as an interference mark to construct an original point cloud sequence integrating interference perception attributes.

[0065] For example, in the low-altitude monitoring scenario of cities with dust weather interference, the lidar emits a scanning beam group to the flight area at a fixed pulse repetition frequency. After passing through the dust interference layer, the laser beam hits the fuselage of the quadcopter and suspended particles, forming a mixed echo signal group containing wing metal reflection and dust scattering; the system slices the echo in the time domain to generate a synchronous time window unit, and extracts the energy concentration distribution characteristics in the azimuth angle range of [-15.2°, +5.8°] within a certain window, where the wing reflection point shows a continuous attenuation characteristic of intensity, while the dust scattering point shows a discrete intensity mutation; based on the spatial overlap rate of the main energy cluster in the continuous window, the energy concentration distribution characteristics are determined. The quadrant [-12.5°, +3.1°] represents the reflection area of ​​the drone's main structure, and the temporal and spatial consistency of the pulse return time series in this area is synchronously detected. The spatial coordinates of the left wingtip are calculated as (X: -12.5m, Y: +3.1m, Z: 50.8m) based on the time-of-flight principle. The average intensity of dust scattering at the azimuth angle [-28.7°, -10.3°] within the same window is calculated to generate an interference marker. Finally, the three-dimensional reflection points containing the coordinate information and the interference marker are combined into a point cloud sequence with environmental perception attributes. The left wingtip reflection point is stored as "coordinates + interference marker value 42" to form the complete original point cloud data.

[0066] Step 102 : Separate the inherent scattered signal related to the UAV configuration and the instantaneous noise signal related to complex meteorological interference based on the physical reflection characteristics of the original point cloud sequence.

[0067] In this step, the wavelength channel reflection intensity ratio refers to the reflection intensity ratio of the same spatial reflection point in different wavelength channels of the lidar, which is used to identify the optical properties of the material; the polarization state change refers to the change in the polarization angle of the laser beam after reflection from the target, which characterizes the surface roughness characteristics; the dynamic interference filtering model refers to the noise filtering logic channel constructed based on the coupling relationship of physical reflection characteristics. The three-dimensional spatial position tracing is used to reconstruct the three-dimensional topological association of the target configuration based on the reflection point pulse return time series.

[0068] In an embodiment of the present application, three physical parameters of each reflection point are first extracted from the original point cloud sequence: the reflection intensity ratio of the dual-wavelength channels, the change in linear polarization angle, and the transmission time difference of adjacent pulse echoes; secondly, the distribution confidence interval of the reflection intensity ratio of the metal material is established according to different meteorological scenarios to generate a wavelength reflection intensity distribution band; then, a joint constraint rule of the polarization state change degree and the pulse transmission time change rate is constructed; then, the reflection points with smooth fluctuations in the reflection intensity ratio of multiple consecutive frames in the distribution band and that conform to the metal law are screened and marked as first-level candidate signals; then, the physical coupling consistency of the polarization-time change rate in the candidate signal is verified based on the dynamic interference filtering model, and the qualified point set is screened; finally, the reflection points that pass the double verification are aggregated through three-dimensional spatial position tracing to form an inherent scattering signal, and the remaining points are classified as instantaneous noise signals.

[0069] Following the labeled point cloud data output from step 101, for the left wingtip reflection point, first extract its dual-wavelength channel reflection intensity ratio and the polarization state change of adjacent scanning frames, and simultaneously calculate the changing trend of the pulse transmission time; on this basis, verify that the point conforms to the reflection law of the aluminum alloy component of the UAV through the preset metal material reflection intensity stability range under the sandstorm weather scene, completing the first level of screening; then analyze the coupling relationship between its polarization state change and the pulse transmission time change rate based on the dynamic interference filtering model, verify that the two show a continuous linear correlation characteristic, and confirm that the point is not affected by turbulent noise; finally, through three-dimensional spatial position tracing, the left wing tip point and the continuous reflection points in the adjacent wing beam area are aggregated to form an inherent scattering signal subset that characterizes the integrity of the wing structure. At the same time, the reflection point with a sudden change at a certain azimuth angle is classified as the instantaneous noise signal set caused by meteorological interference, achieving high-confidence signal separation.

[0070] Step 103: parse the inherent scattering signal to obtain a dynamic structural response spectrum representing the external components of the UAV.

[0071] In this step, the displacement vector sequence refers to the three-dimensional trajectory segment composed of the position movement of the spatial reflection point within the continuous scanning cycle, which describes the micro-vibration state of the component; the motion trajectory vector refers to the smooth motion direction and velocity characteristic vector generated by displacement vector fitting, which reflects the overall motion trend of the component; the vibration wave propagation direction chain refers to the logical topology of the vibration energy transfer path derived based on the material stiffness properties, which includes conduction nodes and path segments; the vibration energy envelope refers to the vibration intensity distribution curve of the reflection point at each node along the propagation direction chain, which quantifies the amplitude of the stimulated response of the component.

[0072] In the embodiment of the present application, the inherent scattering signal is first divided into position areas corresponding to components such as wings and fuselage according to the physical configuration of the drone; secondly, the displacement change of the spatial reflection point in each area in the continuous time frame is calculated, and the trajectory vector describing the overall motion of the area is generated by least squares fitting; then the inclination angle and acceleration characteristics of the trajectory vector are matched with the stiffness parameters of the standard components in the material library to determine the main axis of vibration wave conduction; based on the material stiffness and the continuity constraints of the boundaries of adjacent areas, the conduction path connecting the key nodes is extended and constructed; the path offset angle at the conduction node is dynamically corrected according to the real-time direction change rate of the motion trajectory vector; the vibration energy time series distribution of the corresponding reflection point is mapped at the node position of the corrected propagation direction chain; finally, the energy-time mapping relationship of each position area is integrated to form a global dynamic response map.

[0073] For example, following the wing inherent scattering signal data output from step 102, the spatial reflection points in the left wing area are first divided into position area clusters containing key feature points according to the physical configuration; based on the displacement change sequence of the continuous time frame of each reflection point, a trajectory vector model describing the overall motion of the wing surface is generated, which reflects the current pitch attitude and maneuvering trend; then, the spatial inclination angle and acceleration characteristics of the trajectory vector are accurately matched with the carbon fiber wing stiffness parameters in the pre-stored material library, the main conduction path of the vibration wave along the wingspan direction is determined, and a continuous conduction path chain from the wing root connector to the wing tip is constructed; when the trajectory vector direction is suddenly offset by atmospheric turbulence disturbance, the path refraction angle of the wingtip conduction node is dynamically corrected according to the direction change rate, and the time-varying distribution characteristics of the vibration energy of the spatial reflection point corresponding to the node are synchronously mapped to capture the micro-vibration response curve of the wing surface under aerodynamic load; finally, the energy sequence time series distribution of the multi-level nodes on the conduction path chain is integrated to form a spatiotemporal map representing the structural dynamic response of the wing in a turbulent environment.

[0074] Step 104 : synchronously compare the dynamic structural response spectrum with a plurality of pre-stored reference response spectra of known UAV models, and generate an optimal fit determination result based on the comparison results.

[0075] In this step, the spatiotemporal coverage area alignment refers to the operational means of synchronizing the physical positions of the dynamic spectrum and the reference spectrum on the time axis and spatial area; the continuous distribution overlap describes the time proportion characteristics of the continuous existence of the characteristic frequency in the reference frequency band; the similarity of the spectrum energy transfer path refers to the degree of consistency of the energy transition sequence between the dynamic spectrum and the reference spectrum between frequency bands; the suppression of the contribution of phase anomaly points specifically refers to the control mechanism for weight attenuation of phase mutation points associated with meteorological interference.

[0076] In the embodiment of the present application, the characteristic change frequency sequence corresponding to the time-space response intensity mapping relationship of each position area of ​​the dynamic structure response spectrum is first extracted; then the frequency sequence is physically partitioned and the time scanning window is doubly aligned with the reference frequency band of the pre-stored drone model reference spectrum; the continuous distribution overlap of the real-time characteristic change frequency in the reference frequency band is calculated under the time-space alignment mapping framework; based on the overlap distribution, the motion similarity characteristics of the dynamic spectrum and the reference spectrum in the energy cross-band transfer path are identified; a frequency offset phase difference set is constructed according to the phase angle offset of the frequency overlap point of the two spectra; the contribution attenuation control is performed on the phase anomaly points in the set that are significantly correlated with the atmospheric turbulence intensity fluctuations; and finally, the continuous distribution overlap, the spectrum energy transfer path similarity and the phase correction results are combined to generate the optimal fit judgment result.

[0077] For example, following the spatiotemporal response map of the left wing generated in step 103, first extract the frequency sequence of characteristic changes in the region; align it with the frequency band of the pre-stored quadrotor type A reference map in spatiotemporal order; calculate the continuous overlap distribution of real-time frequencies within the reference band; analyze the energy transfer path; detect the phase offset of the wingtip node to form a phase difference set; apply contribution suppression to the phase anomaly point; and combine the high proportion characteristics of the continuous overlap distribution with the path similarity parameters to generate the optimal fit judgment of model A.

[0078] Step 105: Output the drone model based on the optimal fit determination result.

[0079] In this step, the response feature kernel refers to the vibration modal feature envelope unique to a specific UAV model in the pre-stored model database, including the distribution of the main frequency of structural resonance and the identification of the energy transfer path; the spatial compatibility verification describes the process of detecting the physical matching between the feature kernel and the current environmental parameters; the feature backtracking consistency check is a systematic offset control mechanism by comparing the real-time recognition results with the historical records; the dynamic fluctuation tolerance threshold is the deviation tolerance limit dynamically adjusted according to the meteorological environment.

[0080] In the embodiment of the present application, the aggregated characteristic value of the multi-source indicator parameters in the optimal fit judgment result is first extracted; then the aggregated value is mapped and matched with the local structural vibration source position area of ​​the pre-stored model feature database to generate the potential associated area distribution result of the target model; the response feature core data packet of the corresponding UAV model is activated according to the area mapping result; the activated feature core is verified for compatibility with the spatial distribution of the current atmospheric turbulence intensity parameter; based on the joint analysis of the compatibility verification result and the multi-source indicator weight distribution, the most matching target model identification is locked; finally, the target model identification is compared with the time series feature offset of the historical identification record, and the model result is output when the offset amplitude does not exceed the preset dynamically adjusted tolerance threshold.

[0081] For example, following the optimal fit determination result of model A generated in step 104, the aggregated feature vector is first extracted and matched with the feature space of the quadcopter; the response feature kernel of model A is activated through the wing vibration source area mapping result; the spatial compatibility of the feature kernel in the dust turbulence environment is verified; the model A identifier is locked according to the verification result and the weight distribution of the aggregated value; the historical data of the previous three frames of recognition (all of which are model A) are reviewed to calculate the root mean square error of the feature frequency deviation; and after confirming that the deviation is lower than the dynamic tolerance threshold of the dust scene, model A is output as the final recognition result.

[0082] To address the problem of point cloud reflection aliasing caused by meteorological interference, a three-dimensional reflection point set with interference-aware properties is constructed based on the persistent correlation of the main energy clusters. In some embodiments, according to step 101, an original point cloud sequence generated by the lidar scanning the drone in real time under the target flight area is obtained. The original point cloud sequence contains the interfered three-dimensional space reflection points, including:

[0083] Step 201: Control the laser radar to transmit a scanning beam group to the flight area, wherein the scanning beam group generates a continuous laser pulse sequence in a scanning plane at a fixed pulse repetition frequency.

[0084] In this step, the fixed pulse repetition frequency refers to the setting parameter that laser pulses are emitted at a constant time interval per unit time; the scanning plane describes the two-dimensional spatial coverage formed by the laser beam group in the horizontal azimuth angle and the vertical pitch angle; the continuous laser pulse sequence represents a combination of discrete optical pulses emitted continuously at a fixed time interval.

[0085] In the embodiment of the present application, a scanning beam group is first emitted to the target flight area through the laser radar control module. The beam group generates equally spaced laser pulses in a rectangular scanning plane formed by the horizontal and vertical directions at a preset fixed time interval; each pulse is emitted at a specific azimuth-pitch angle coordinate position to form a continuous pulse sequence covering the monitored airspace.

[0086] Step 202 : capturing a group of multiple scattered echo signals generated when the continuous laser pulse sequence hits the surface of the UAV and meteorological suspended particles after passing through the complex meteorological interference layer.

[0087] In this step, the complex meteorological interference layer refers to the vertical spatial region where the concentration of suspended particulate matter in the atmosphere is significantly increased; the multiple scattered echo signal group represents the superimposed waveform set formed by the laser pulse reflected by the target surface and scattered by the environmental particles.

[0088] The receiving device captures the mixed scattered signals formed when the continuous laser pulse sequence is transmitted and hits the metal surface of the drone and the surrounding environment particles after passing through the atmospheric suspended particle layer. The reflected signal of the drone body shows the energy concentration characteristic, while the scattered signal of the suspended particles shows the energy dispersion characteristic.

[0089] Step 203: performing time domain signal slicing on the multiple scattered echo signal group to generate a time window unit synchronized with the pulse period.

[0090] In this step, time domain signal slicing refers to an operation method of cutting the echo signal into segments according to fixed time lengths; the time window unit refers to a signal segmentation processing unit that is strictly aligned with the laser pulse emission period.

[0091] The captured multiple scattered echo signal group is segmented in the time domain, and is cut into equal time segments based on the laser pulse emission interval to ensure that the starting time of each time window unit is accurately synchronized with the pulse emission time.

[0092] Step 204: extract the energy concentration distribution characteristics and pulse return time sequence of the scattered echo in each time window unit.

[0093] In this step, the energy concentration distribution feature refers to the quantitative parameter of the energy intensity of the echo signal within the time window unit in the spatial azimuth-elevation plane; the pulse return time series describes a time series data set composed of the time difference from transmission to reception of multiple pulse echo signals captured by the same receiver.

[0094] In the embodiment of the present application, the time window unit data set generated in step 203 and synchronized with the pulse period is first received; secondly, the following operations are performed for each window unit: all azimuth-elevation grid areas covered by the window are scanned, and the energy intensity integral value of the received echo signal in each grid area is counted; the spatial concentration characteristic value is calculated based on the energy intensity distribution; the precise time difference of each pulse echo signal in the window from the laser emission to the receiving detector is synchronously recorded, and a time difference sequence is constructed according to the pulse triggering order; and finally a window unit analysis report containing the energy concentration distribution characteristic vector and the pulse time difference sequence is generated.

[0095] Step 205 : Identify the structural reflection echo belonging to the drone body and the environmental noise echo belonging to meteorological suspended particles based on the continuous correlation of the main energy cluster in the energy concentration distribution feature.

[0096] In this step, the continuous correlation of the main energy cluster refers to the continuity characteristics of the spatial distribution position of the high-energy echo signal within the continuous time window; the structural reflection echo describes the echo signal generated by the regular geometric structure on the surface of the drone body; the environmental noise echo represents the randomly distributed echo signal group formed by the scattering of meteorological suspended particles.

[0097] In the embodiment of the present application, the energy aggregation distribution characteristics within the time window unit are first analyzed, and the aggregation area where the energy is significantly higher than the background noise is extracted as the main energy cluster; then the spatial position overlap and energy attenuation continuity of the main energy cluster in the continuous time window sequence are detected; according to the judgment principle that the position overlap rate remains high and the energy attenuation is continuous, the reflected echo of the drone body structure is identified; at the same time, the echo with discrete spatial position and irregular intensity mutation is classified as meteorological suspended particle environmental noise echo.

[0098] Step 206 : determining the spatial reflection point position on the surface of the UAV body based on the pulse return time sequence of the structure reflected echo.

[0099] In this step, the pulse return time series refers to the time series data composed of the time difference from the transmission to the reception of the structure reflection echo; the spatial reflection point position represents the three-dimensional spatial coordinates of the UAV body surface solved by the echo propagation time.

[0100] For the identified structure reflection echo, the time difference between the pulse emission time and the reception time is calculated by the time flight principle; the laser pulse propagation path length value is obtained based on the product of the time difference and the speed of light constant; combined with the emission azimuth-pitch angle parameters of the scanning beam group, the spherical coordinate system analytical algorithm is used to solve the three-dimensional spatial coordinate position of the reflection point on the surface of the UAV body.

[0101] In step 207 , the spatial reflection point positions are combined with the environmental noise interference intensity mark of the corresponding time window unit to construct a three-dimensional reflection point set with interference perception attributes, and the three-dimensional reflection point set with interference perception attributes is combined to form an original point cloud sequence.

[0102] In this step, the environmental noise interference intensity mark is the average intensity quantization value of the environmental noise echo within the time window unit; the three-dimensional reflection point set with interference perception attributes contains a comprehensive data unit of spatial coordinate position and corresponding noise intensity.

[0103] First, the average energy intensity value of all environmental noise echoes in the current time window unit is counted to generate an environmental noise interference intensity marker; the marker is associated and bound with the three-dimensional spatial coordinate position of the structure reflection echo determined in step 206; a combined data unit containing the spatial coordinate position and the noise intensity marker is constructed; finally, all combined data units generated by the continuous time window units are aggregated in time series to form an original point cloud data set.

[0104] To address the issue of inaccurate scattered signal extraction in complex environments, in some embodiments, according to step 102, the inherent scattered signal associated with the UAV configuration and the transient noise signal associated with complex meteorological interference are separated based on the physical reflectance characteristics of the original point cloud sequence, including:

[0105] Step 301 : extracting the laser wavelength channel reflection intensity ratio, polarization state change degree, and pulse transmission time change rate corresponding to each three-dimensional spatial reflection point from the original point cloud sequence.

[0106] In this step, the laser wavelength channel reflection intensity ratio refers to the ratio parameter of the reflection intensity of the same spatial reflection point under different laser wavelength channels; the polarization state change degree describes the quantitative value of the change in polarization angle of the laser beam after reflection from the target; and the pulse transmission time change rate represents the dynamic change characteristics of the difference in transmission time of adjacent pulse echoes.

[0107] In the embodiment of the present application, the multi-spectral channel reflection intensity data of each three-dimensional spatial reflection point in the original point cloud sequence is first extracted, and the reflection intensity ratio parameter of the preset wavelength combination is calculated; the polarization angle change of the reflection point in adjacent scanning cycles is synchronously obtained; finally, the transmission time difference change trend of adjacent pulse echoes at the same reflection point is calculated to form a complete physical feature data set.

[0108] Step 302 : generating wavelength reflection intensity distribution bands under different meteorological interference types based on the laser wavelength channel reflection intensity ratio.

[0109] In this step, the wavelength reflection intensity distribution band under different meteorological interference types refers to the wavelength reflection intensity ratio confidence interval model preset according to meteorological scenes such as rain, fog, dust, etc., which is used to identify the stable reflection characteristics of the material in a specific environment.

[0110] Based on the wavelength reflection intensity ratio data set extracted in step 301, the preset parameter model is called according to the meteorological type identification result; corresponding reflection intensity ratio confidence intervals are generated for different interference scenes such as rain, fog, dust, etc.; this distribution band serves as a benchmark model for material stability verification.

[0111] Step 303: construct a dynamic interference filtering model based on the coupling relationship between the polarization state change degree and the pulse transmission time change rate.

[0112] In this step, the coupling relationship between the polarization state change degree and the pulse transmission time change rate describes the physical correlation model between the target surface polarization characteristics and the motion state change; the dynamic interference filtering model is a noise point screening logic channel constructed based on this coupling relationship.

[0113] Based on the polarization state change degree and pulse transmission time change rate dataset extracted in step 301, the linear constraint relationship between the two is analyzed; a coupling feature confidence interval model is established (for example, when the polarization changes slightly but the time difference fluctuates violently, it is marked as an anomaly); and an environmental adaptive filtering model framework is constructed based on this physical coupling rule.

[0114] Step 304: Mark the set of reflection points in the wavelength reflection intensity distribution band that meet the metal material reflection stability law as first-level candidate signals.

[0115] In this step, it refers to the physical property that the wavelength intensity ratio of the metal reflection point continues to be in the confidence interval and the fluctuation amplitude maintains a narrow range under a specific meteorological interference environment; the first-level candidate signal is a set of potential effective reflection points that are initially screened through the wavelength reflection intensity distribution band and are subject to further verification.

[0116] In the embodiment of the present application, the wavelength reflection intensity distribution band data model generated in step 302 is first extracted; then, all reflection points whose wavelength intensity ratios are within the distribution band confidence interval are screened in the original point cloud sequence; the fluctuation amplitude of the wavelength intensity ratio of these reflection points in the continuous time frame sequence is further detected; and the reflection points whose fluctuation amplitudes are always maintained below the preset fluctuation limit are marked as first-level candidate signals.

[0117] Step 305 : In the dynamic interference filtering model, based on the linear constraint threshold of the coupling relationship, filter the echo points with continuous pulse time series in the first-level candidate signal.

[0118] In this step, the linear constraint threshold of the coupling relationship is the acceptable linear correlation interval boundary defined based on the correlation model between the polarization state change degree and the pulse transmission time change rate established in step 303; the pulse time sequence is continuous and specifically refers to the pulse transmission time of the echo point in each time frame showing a regular gradual change trend.

[0119] In the embodiment of the present application, first, for the first-level candidate signal set marked in step 304, the polarization state change degree and pulse transmission time change rate data set corresponding to each point are extracted; based on the linear correlation rule in the dynamic interference filtering model constructed in step 303, it is verified whether the candidate signal point data is within the preset linear constraint threshold range; synchronously detect whether the pulse transmission time of the candidate signal point maintains a monotonic change characteristic in continuous time frames; and the reflection points that pass the linear constraint verification and have a continuous time series are determined as qualified valid echo points.

[0120] Step 306: perform three-dimensional spatial position tracing on the filtered echo points to form inherent scattering signals related to the UAV configuration.

[0121] In this step, three-dimensional spatial position tracing refers to the operational process of reconstructing the topological association of the UAV's physical structure based on the spatial distribution characteristics of the point cloud coordinates; the inherent scattering signal related to the UAV's configuration represents a set of scattering points that describe the structural characteristics of the target entity, and its spatial distribution conforms to the physical connection relationship of the UAV's mechanical components.

[0122] In the embodiment of the present application, the three-dimensional spatial coordinate data of all valid echo points that have passed the verification in step 305 are first extracted; based on the relative position relationship of each coordinate point in the spatial distribution and the principle of physical connection continuity (the spacing between key points conforms to the mechanical structure constraints), the structural topology network of the drone body components is reconstructed; the reflection points located in the key structures such as the wing skin area and the motor bracket area are clustered and aggregated into a subset of scattered signals representing specific components according to the physical position correlation; finally, an inherent scattered signal set including core structures such as the fuselage frame and the rotor assembly is formed.

[0123] Step 307: Classify the reflection point set not covered by the wavelength reflection intensity distribution band and the dynamic interference filtering model as the instantaneous noise signal.

[0124] In this step, the uncovered reflection point set refers to the scattering points that have neither passed the wavelength distribution band verification in step 304 nor the dynamic interference model verification in step 305; the instantaneous noise signal represents the random scattering point set with no structure correlation generated by environmental interference.

[0125] In the embodiment of the present application, reflection points in the original point cloud sequence that are not marked as first-level candidate signals are first screened (step 304, filtering points); reflection points that fail the dynamic interference filtering model verification are simultaneously extracted (step 305, filtering points); these two types of reflection points are merged into a set of noise points to be processed; the spatial distribution discreteness and physical topological fracture characteristics of each point in the set are detected; and finally, they are classified as transient noise signals unrelated to the drone structure.

[0126] In order to overcome the problem of characteristic drift caused by high-speed maneuvering deformation, in some embodiments, according to step 103, the dynamic structural response spectrum representing the external components of the UAV is parsed from the inherent scattering signal, including:

[0127] Step 401: Divide the location areas of multiple spatial reflection points in the inherent scattered signal corresponding to the external components of the drone.

[0128] In this step, the position area refers to the continuous spatial range of the spatial reflection points in the inherent scattering signal divided according to the physical configuration of the UAV, and its boundary corresponds to the actual geometric contours of mechanical components such as the wing skin area and the motor bracket area.

[0129] In the embodiment of the present application, a spatial boundary template of the core component (such as the maximum outward extension contour line of the wing skin) is first preset based on the three-dimensional digital model of the drone; then the spatial reflection points in the inherent scattering signal are mapped to the template coordinate system; the shortest distance from each reflection point to the component contour line and the density distribution of the neighborhood points are calculated; the clusters of reflection points within the distance threshold and whose density conforms to the surface characteristics of the component are classified into the same position area; finally, a set of discrete component areas such as the leading edge of the wing and the middle of the fuselage are formed.

[0130] Step 402: Fitting the motion trajectory vector of each position area based on the displacement vector sequence of the spatial reflection point that changes with time.

[0131] In this step, the displacement vector sequence describes the spatial trajectory fragment composed of the three-dimensional coordinate movement of the same reflection point in the position area in continuous time frames; the motion trajectory vector is a set of direction and acceleration features that characterize the overall motion trend of the area generated by fitting the vector sequence.

[0132] In the embodiment of the present application, the three-dimensional coordinate change data of each reflection point in the position area in continuous time frames are first extracted to generate a displacement vector sequence set of each point; based on the spatial vector distribution characteristics of the sequence set, the motion trajectory of the center of mass of the area is calculated by the least squares fitting method; the pitch direction angle, yaw direction angle and linear acceleration components of the trajectory are further decomposed; and finally, a feature vector describing the global motion state of the position area is generated.

[0133] Step 403 : Match the motion trajectory vector of each position area with the standard component stiffness parameters in a preset UAV manufacturing material property library to obtain matched standard component stiffness parameters.

[0134] In this step, the standard component stiffness parameters refer to the set of physical property data pre-stored in the material property library that describes the deformation resistance of the materials of each component of the drone; the matched standard component stiffness parameters represent the standardized stiffness index corresponding to the real-time component dynamic response obtained by screening the motion trajectory vector characteristics.

[0135] In the embodiment of the present application, the motion trajectory vector (including angular and acceleration components) of the position area fitting is first input into the material property library; the candidate stiffness parameter set that meets the angular fluctuation range and acceleration change interval is retrieved from the library; based on the energy propagation correlation model between the trajectory vector and the stiffness parameter, the matching confidence of each candidate parameter is calculated; and finally, the stiffness parameter with a confidence level within a preset acceptance threshold is selected as the matching result output.

[0136] Step 404: Generate a local vibration wave propagation direction chain of the external component of the UAV based on the matched standard component stiffness parameters.

[0137] In this step, the local vibration wave propagation direction chain describes the topological structure of the conduction path of vibration energy on the external components of the UAV, which is derived based on the stiffness characteristics of the components, including the main conduction axis and key refraction nodes.

[0138] In the embodiment of the present application, the main conduction axis (high stiffness direction) of the vibration wave in the component material is first determined based on the matching stiffness parameters; the conduction path is extended according to the boundary geometric continuity constraints of adjacent position areas; the turning points of the wave conduction path are marked at the component connection; the refraction angle correction coefficient is calculated based on the ratio of the stiffness parameter to the material density; and finally, a propagation direction chain including conduction direction, path nodes and refraction characteristics is generated.

[0139] Step 405 : Mapping the vibration energy envelope of the spatial reflection point corresponding to each propagation direction node on the local vibration wave propagation direction chain.

[0140] In this step, the propagation direction node is the coordinate point of the path turning position marked on the vibration wave propagation direction chain; the vibration energy envelope specifically refers to the continuous distribution curve of the energy intensity of the spatial reflection point corresponding to the node over time during the complete scanning cycle.

[0141] In the embodiment of the present application, the three-dimensional spatial coordinates of all key nodes on the link are first extracted based on the local vibration wave propagation direction chain generated in step 404; then, based on the nearest neighbor matching algorithm, the spatial reflection point with the closest Euclidean distance to the node coordinate is located in the inherent scattering signal point cluster; then, all energy sampling values ​​of the time window unit where the reflection point is located are retrieved from the laser radar original echo database; the energy sampling sequence is smoothed by Gaussian filtering to eliminate pulse noise; finally, a cubic spline interpolation algorithm is used to generate a continuous energy intensity change envelope, and a mapping relationship between the node position and the envelope data set is established.

[0142] Step 406 : generating a time-space response intensity mapping relationship of the position area according to the amplitude variation relationship of the vibration energy envelope in the scanning time sequence corresponding to the position area.

[0143] In this step, the time-space response intensity mapping relationship is a matrix model constructed through the spatiotemporal joint analysis of the energy envelopes of all nodes in the location area. Its column vector is the time series, the row vector is the node position, and the matrix element is the normalized energy intensity value.

[0144] In the embodiment of the present application, a time base axis is first established based on the laser scanning time window sequence t0, t1, ..., tn defined in step 203; discrete energy values ​​are sampled according to the time base axis for all node energy envelopes generated in step 405; an M×N matrix is ​​constructed (rows are node numbers, columns are time serial numbers), and the matrix element Aij represents the normalized energy intensity of the i-th node at time tj; the conduction path of the energy peak between nodes is identified by calculating the energy gradient change of the column vector and the spatial correlation of the row vector; and finally a three-dimensional response mapping structure is generated that includes time conduction delay characteristics and spatial energy distribution characteristics.

[0145] Step 407 : Integrate the time-space response intensity mapping relationship of all position areas to form the dynamic structural response map.

[0146] In this step, the dynamic structural response map refers to the global structural dynamic feature expression model formed by integrating the spatiotemporal response intensity mapping relationship of all position areas of the UAV. It encodes the energy-time evolution correlation of different component areas in a three-dimensional voxel matrix, reflecting the vibration conduction law of the entire machine under airflow excitation.

[0147] In the embodiment of the present application, a global space coordinate system is first established with the center of gravity of the drone as the origin; the spatiotemporal response intensity mapping relationship of each position area is extracted; spatial alignment is achieved through the energy gradient consistency test of the regional boundary connection points; conduction delay compensation is performed on the time axis of each region; the compensated regional energy matrix is ​​feature fused according to the spatial topological relationship; and finally, a global dynamic structural response map is generated with the global coordinate system as the reference and the time axis synchronously calibrated.

[0148] To address the problem of inaccurate spectrum comparison under weather interference, in some embodiments, according to step 104, the dynamic structural response spectrum is synchronously compared with a plurality of pre-stored reference response spectra of known UAV models, and an optimal fit determination result is generated based on the comparison results, including:

[0149] Step 501: extracting the characteristic change frequency of the time-space response intensity mapping relationship of each position area in the dynamic structural response spectrum.

[0150] In this step, the characteristic change frequency refers to the spectral characteristics of the energy intensity evolving with time extracted from the time-space response intensity mapping relationship, which is expressed as a combination of fundamental and harmonic frequencies; the position area corresponds to the independent vibration partition of the external components of the UAV, such as the wing skin area, fuselage frame area and other physical structural units.

[0151] In an embodiment of the present application, a voxel matrix of the spatiotemporal response intensity mapping relationship of each position area in the dynamic structural response map is first loaded; secondly, a group of main nodes with significant vibration energy is screened for each position area; then, the energy time series waveform of the selected node within the complete scanning cycle is extracted; then, a fast Fourier transform is performed on the time series of each node to obtain the spectrum distribution; then, the top three most significant spectral peak frequencies and their amplitude values ​​in the spectrum are extracted; finally, the characteristic frequencies of all main nodes in the area are weighted averaged according to the amplitude intensity to generate a characteristic change frequency that characterizes the vibration characteristics of the position area.

[0152] Step 502 : aligning the characteristic change frequency with the reference frequency band corresponding to the pre-stored reference response spectrum of the UAV model in terms of time and space coverage.

[0153] In this step, the reference frequency band is the pre-stored vibration frequency response range of a specific drone model under standard working conditions; the spatiotemporal coverage area alignment refers to the operation of establishing a synchronous mapping relationship between the real-time characteristic frequency and the reference frequency band in spatial partitioning and time scanning period.

[0154] In an embodiment of the present application, first, the reference frequency band data of the pre-stored target model benchmark response map is called; secondly, a spatial partition mapping relationship is established: the centroid coordinates of the real-time position area are nearest neighbor matched with the benchmark map partition; then the vibration wave conduction delay compensation amount is calculated: the time compensation value is derived based on the actual distance difference from the real-time node to the vibration source and the material wave velocity characteristics; then delay compensation is applied to the real-time characteristic frequency timestamp to synchronize it with the benchmark frequency band time axis; finally, a spatiotemporal alignment deviation matrix is ​​constructed, whose rows correspond to the matched spatial partitions, columns correspond to the synchronized time series, and element values ​​record the absolute deviation between the real-time and benchmark frequencies.

[0155] Step 503: After the spatial and temporal coverage areas are aligned, the degree of continuous distribution overlap of the characteristic change frequencies within the reference frequency band is calculated.

[0156] In the embodiment of the present application, the spatiotemporal alignment deviation matrix generated in step 502 is first obtained; secondly, the valid comparison position points in the matrix that meet the preset frequency deviation threshold are extracted; then, the continuous residence period of the characteristic frequency in each spatial partition within the reference band is analyzed: the length of the time segment in which the frequency fluctuation range is always within the reference band boundary and the frequency change rate is lower than the critical value is detected; then the proportion of this period in the scanning cycle is calculated to generate the single partition continuous overlap; finally, the overlap values ​​of all partitions are aggregated to generate the global continuous distribution overlap parameter.

[0157] Step 504 : Based on the continuous distribution overlap, identify the similarity of the spectral energy transfer paths between the dynamic structural response spectrum and each reference response spectrum.

[0158] In this step, the continuous distribution overlap refers to the proportion of the continuous residence time of the real-time feature change frequency in the reference frequency band; the similarity of the spectrum energy transfer path describes the degree of consistency between the dynamic spectrum and the reference spectrum in the vibration energy cross-frequency band transition path.

[0159] In an embodiment of the present application, first, based on the continuous distribution overlap, continuous periods of high overlap on the time axis are extracted; then, for each period of high overlap, the spectrum energy distribution characteristics of the real-time dynamic spectrum and the reference spectrum are extracted respectively; then, the energy transfer direction and timing relationship between adjacent spectrum segments are analyzed; finally, the direction matching rate and timing offset variance of the two energy transfer paths are calculated to generate a path similarity coefficient that quantitatively describes the similarity.

[0160] Step 505: Generate a frequency offset phase difference set according to the sequence of overlapping positions of the characteristic change frequency and the reference frequency band.

[0161] In this step, the frequency offset phase difference set is a phase angle deviation data set at corresponding moments between the characteristic change frequency and the reference frequency, reflecting the relative phase offset caused by meteorological interference.

[0162] In an embodiment of the present application, first, based on the spatiotemporal alignment deviation matrix, the position sequence of the overlapping points of the real-time characteristic change frequency and the reference frequency band is extracted; then, for each overlapping point position, the instantaneous phase angle of the real-time characteristic frequency signal and the instantaneous phase angle of the reference frequency signal are calculated respectively; then, the phase offset at the corresponding moment is calculated; finally, a triple set containing timestamp, spatial position, and phase offset is constructed.

[0163] Step 506 : Suppressing the contribution of phase anomaly points in the frequency offset phase difference set that are associated with a meteorological interference intensity of a preset threshold value, to obtain a phase anomaly point suppression result.

[0164] In this step, the meteorological interference intensity correlation refers to the significant correlation between the phase offset amplitude and the environmental noise interference mark; contribution suppression is a technical operation to reduce the weight of abnormal phase points in the matching decision.

[0165] In an embodiment of the present application, first, based on the frequency offset phase difference set, the time and space coordinate information of each phase anomaly point is extracted; then, the environmental noise interference intensity mark database is associated to obtain the interference intensity value of the corresponding time and space coordinates; then, a correlation model between the phase offset amplitude and the interference intensity is established; a weight penalty is imposed on the anomaly points that meet the requirements; finally, a suppression result triplet containing the original phase difference and the weight reduction coefficient is generated.

[0166] Step 507: Combining the continuous distribution overlap, the spectrum energy transfer path similarity, and the phase abnormal point suppression result to generate the optimal fit determination result.

[0167] In this step, the optimal fit determination result is a model matching decision parameter quantified by integrating multiple physical characteristics, and its numerical range is [0,1], reflecting the matching confidence level.

[0168] In the embodiment of the present application, the degree of overlap of the continuous distribution is first normalized; then the similarity coefficient of the spectrum energy transfer path is logarithmically enhanced; then the weighted suppression result set is extracted to calculate the average suppression factor; finally, the optimal fit result is generated through the multi-source fusion decision formula.

[0169] In order to avoid the propagation of single recognition errors, in some embodiments, according to step 105, outputting the drone model based on the optimal fit determination result includes:

[0170] Step 601: extracting the aggregated value of multiple source indicators in the optimal fit determination result, wherein the multiple source indicators include continuous distribution coincidence, spectrum energy transfer path similarity, and phase abnormal point suppression result.

[0171] In this step, the aggregation value refers to the decision parameter complex that integrates the continuous distribution overlap, spectral energy transfer path similarity and phase suppression results; the local structure vibration source position area mapping represents the matching process of spatially associating the decision parameters with the vibration response characteristics of specific components of the UAV.

[0172] In the embodiment of the present application, the optimal fit judgment result data is first extracted; secondly, the three core indicator values ​​contained therein are analyzed respectively, namely, the continuous distribution overlap, the spectrum energy transfer path similarity coefficient, and the phase anomaly suppression factor; then, the weighted sum value is calculated based on the weight allocation strategy of dynamic adjustment of the meteorological interference intensity; and then the weighted sum value is standardized to generate an aggregate value in the interval [0,1].

[0173] Step 602 : Performing regional mapping of the local structural vibration source position on the aggregated value and the pre-stored model feature space to generate a regional mapping result.

[0174] In this step, the model feature space is a pre-stored UAV model benchmark dataset, which contains the typical vibration response characteristic distribution of each model; the local structural vibration source location area refers to the spatial distribution hot zone of the components that produce the core vibration characteristics of the UAV, such as the motor base area, wingtip flutter area and other physical locations.

[0175] In the embodiment of the present application, the pre-stored model feature space database is first loaded; secondly, the matching degree of the aggregation value and the threshold interval of each model feature space is calculated; then, the matching area weight distribution is dynamically selected according to the current meteorological interference intensity; then, the thermal map of the core position of the vibration source in the feature space is extracted and aligned with the real-time scanning space coordinate system; finally, a regional mapping result is generated including the model candidate matching degree and the core vibration source space position coordinates.

[0176] Step 603: Based on the area mapping result, activate the response feature core corresponding to the drone model.

[0177] In this step, the response feature kernel is the vibration feature envelope unique to a specific UAV model, including physical fingerprints such as the resonant frequency band distribution and energy transfer path identification; the activation operation refers to the process of calling and initializing the corresponding model feature kernel data based on the spatial mapping results.

[0178] In the embodiment of the present application, the regional mapping result is first received; secondly, the model identification and core vibration source coordinates with the highest matching degree are extracted; then, the model response feature kernel data set is retrieved from the pre-stored model feature library; then, the feature kernel space reference coordinate system is initialized according to the core vibration source coordinates; finally, a mapping channel between the feature kernel parameters and the real-time point cloud space coordinates is established to generate a callable feature kernel instance.

[0179] Step 604: Verify the spatial compatibility between the response characteristic kernel and the current atmospheric turbulence correlation factor to obtain a spatial compatibility verification result.

[0180] In this step, the atmospheric turbulence correlation factor refers to a comprehensive parameter that characterizes the current environmental turbulence intensity; spatial compatibility verification is a matching analysis process that compares the attenuation characteristics of the response characteristic kernel under different turbulence intensities with the real-time observation values; the spatial compatibility verification result outputs a confidence score that quantitatively describes the environmental adaptability of the characteristic kernel.

[0181] In the embodiment of the present application, the current atmospheric turbulence correlation factor is first loaded; secondly, the standard attenuation characteristic curve of the response characteristic kernel in different turbulence intensity ranges is extracted; then, the energy attenuation observation value of the real-time scanning point cloud at the core vibration source position is collected; then, the root mean square error between the standard attenuation curve and the real-time observation value is calculated; and finally, the spatial compatibility confidence score is output: the score is negatively correlated with the error value, and a full score indicates complete compatibility.

[0182] Step 605: Lock the target model identifier according to the spatial compatibility verification result and the weight distribution of the aggregation value.

[0183] In this step, the target model identification locking is a two-factor decision-making process that integrates compatibility verification and preliminary weight distribution; the aggregate value weight distribution refers to the weight ratio of each indicator dynamically allocated in the calculation of step 601.

[0184] In the embodiment of the present application, the spatial compatibility verification result is first received; secondly, the weight distribution record is extracted; then a two-factor decision matrix is ​​established: if the compatibility score is >0.6, the current model is directly confirmed; if the score is ≤0.6, the decision is made based on the aggregate value weight distribution: when the γ weight ratio is >40%, the alternative model retrieval is started; finally, the locked model identification and decision confidence mark are output.

[0185] Step 606: Perform a feature backtracking consistency check between the target model identifier and the historical recognition result. When the deviation value of the feature backtracking consistency check is less than a preset dynamic fluctuation tolerance threshold, output the target model identifier as a drone model.

[0186] In this step, feature backtracking consistency check refers to the detection of systematic deviations by comparing the current identification features with historical identification records; the dynamic fluctuation tolerance threshold is the deviation acceptance limit adjusted in real time according to the environmental turbulence intensity.

[0187] In the embodiment of the present application, the characteristic fingerprints (fundamental frequency, harmonic distribution) of the three most recent valid identification records are first retrieved; secondly, the cosine similarity deviation between the characteristic fingerprint of the currently locked model and the historical record is calculated; then, the dynamic tolerance threshold is calculated based on the current turbulence factor: the stronger the turbulence, the higher the threshold; then, it is determined whether the similarity deviation is less than the threshold; the final decision is: if the threshold is met, the model is output, otherwise a rescan is triggered.

[0188] In order to improve the physical consistency of the vibration propagation direction chain, in some embodiments, according to step 404, based on the matched standard component stiffness parameters, generating the local vibration wave propagation direction chain of the external component of the UAV includes:

[0189] Step 701: Determine the main direction of axial stress conduction in the position area according to the matched standard component stiffness parameters.

[0190] In this step, the main direction of axial stress conduction refers to the dominant direction of vibration wave energy transmission determined by the stiffness parameters of the component.

[0191] In the embodiment of the present application, the stiffness parameters of the standard components are matched first; secondly, the isotropic stiffness distribution data in the parameters are used to calculate the ratio of the Young's modulus of the material in the orthogonal direction; then, the direction with the largest Young's modulus and the smallest Poisson's ratio is selected as the potential main conduction direction; then, the directional stability of the historical motion trajectory vector of the area is combined to verify the final determination of the main direction vector of axial stress conduction to ensure its compatibility with the real-time motion trend.

[0192] Step 702: Based on the boundary constraints of adjacent areas of the location area, extend the main direction of axial stress conduction to form a continuous conduction path.

[0193] In this step, the boundary constraints of adjacent regions are spatially limited conditions for the material continuity requirements at the connections between different component position regions; the continuous conduction path describes the energy transfer channel across the connection boundary when the vibration wave extends in the axial main direction.

[0194] In the embodiment of the present application, the axial main direction vector is first obtained; secondly, the boundaries of the adjacent position areas on the extension line of the direction are detected; then, the elastic modulus difference ratio and Poisson's ratio compatibility index of the materials on both sides of the boundary are calculated; then, the feasibility of the conduction path extension is judged based on the compatibility threshold, and finally, a B-spline curve is used to smoothly connect the conduction path segments across the boundary.

[0195] Step 703: Mark the connection nodes at the boundaries of adjacent position areas on the continuous conduction path as vibration wave refraction nodes.

[0196] In this step, the vibration wave refraction node is the turning point of the continuous conduction path at the material mutation boundary; the connection node specifically refers to the physical contact point on the interface of different material areas, and its geometric characteristics cause the wave conduction direction to be refracted and offset.

[0197] In the embodiment of the present application, first, all boundary crossing points are located on the continuous conduction path generated in step 702; second, the interface normal vector of the crossing point and the acoustic impedance ratio of the materials on both sides are calculated; then, the theoretical refraction angle is calculated according to the law of acoustic refraction; and then, the point on the path is marked as a refraction node.

[0198] Step 704 : Based on the direction change rate of the motion trajectory vector, correct the conduction angle offset of the continuous conduction path at the vibration wave refraction node to obtain a corrected conduction path segment.

[0199] In this step, the directional change rate of the motion trajectory vector describes the real-time change rate of the motion direction of the position area; the conduction angle offset is the original directional deviation of the vibration wave caused by material mutation at the refraction node.

[0200] In an embodiment of the present application, the motion trajectory vector direction change rate data is first extracted; secondly, the refraction node parameters are read; then, the projection component of the motion direction change vector on the normal plane of the node interface is calculated; then, the angle compensation amount is calculated based on the vector deviation between the motion projection component and the refraction direction; finally, the original refraction direction is rotationally corrected to generate a corrected conduction path segment and update the node storage parameters.

[0201] Step 705: Connect all corrected conductive path segments to generate a local vibration wave propagation direction chain.

[0202] In this step, the local vibration wave propagation direction chain refers to the complete vibration energy transfer topology network formed by integrating all conduction path segments.

[0203] In an embodiment of the present application, the parameters of all corrected conduction path segments are first summarized; secondly, a spatial connection relationship diagram of the path segments is established: the node positions are vertices and the path segments are directed edges; then the break points in the connection diagram are detected; then an adaptive curve interpolation algorithm is used to fill the break; and finally, a local vibration wave propagation direction chain is generated.

[0204] Figure 2 The present invention provides a schematic diagram of a real-time drone model recognition system based on laser radar point cloud features. Figure 2 As shown, the system includes:

[0205] An acquisition module 21 is configured to acquire an original point cloud sequence generated by the laser radar when scanning the UAV in real time under the target flight area, wherein the original point cloud sequence includes disturbed three-dimensional space reflection points;

[0206] a separation module 22 for separating the inherent scattered signals associated with the UAV configuration and the transient noise signals associated with complex meteorological interference based on the physical reflectance characteristics of the original point cloud sequence;

[0207] An analysis module 23 is configured to analyze the inherent scattering signal to obtain a dynamic structural response spectrum representing the external components of the UAV;

[0208] a comparison module 24 for synchronously comparing the dynamic structural response spectrum with a plurality of pre-stored reference response spectra of known UAV models, and generating an optimal fit determination result based on the comparison results;

[0209] The output module 25 is used to output the drone model based on the optimal fit determination result.

[0210] Figure 2 The real-time identification system for drone models based on laser radar point cloud features can be executed Figure 1 The implementation principles and technical effects of the method for real-time drone model identification based on LiDAR point cloud features described in the illustrated embodiment are not further elaborated. The specific manner in which each module and unit performs operations in the system for real-time drone model identification based on LiDAR point cloud features in the aforementioned embodiment has been described in detail in the relevant embodiments of the method and will not be further elaborated here.

[0211] In one possible design, Figure 2 The real-time identification system of drone models based on laser radar point cloud features of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0212] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0213] The processing component 32 is used for the above Figure 1 The embodiment provides a real-time identification method for drone models based on lidar point cloud features.

[0214] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A real-time identification method for UAV models based on laser radar point cloud features, characterized in that: include: Acquire an original point cloud sequence generated by the laser radar scanning the UAV in real time under the target flight area, wherein the original point cloud sequence includes disturbed three-dimensional space reflection points; Based on the physical reflection characteristics of the original point cloud sequence, separating the inherent scattered signal related to the UAV configuration and the instantaneous noise signal related to complex meteorological interference; Resolving the inherent scattering signal to obtain a dynamic structural response spectrum representing the external components of the UAV; Synchronously comparing the dynamic structural response spectrum with a plurality of pre-stored benchmark response spectra of known UAV models, and generating an optimal fit determination result based on the comparison results; Based on the optimal fit determination result, the drone model is output.

2. The method according to claim 1, characterized in that Obtain an original point cloud sequence generated by the laser radar scanning the drone in real time in the target flight area, wherein the original point cloud sequence contains disturbed three-dimensional space reflection points, including: Controlling the laser radar to transmit a scanning beam group toward the flight area, wherein the scanning beam group generates a continuous laser pulse sequence in a scanning plane at a fixed pulse repetition frequency; capturing a group of multiple scattered echo signals generated by the continuous laser pulse sequence striking the surface of the UAV and meteorological suspended particles after passing through the complex meteorological interference layer; Performing time domain signal slicing on the multiple scattered echo signal group to generate a time window unit synchronized with the pulse period; Extract the energy concentration distribution characteristics and pulse return time series of scattered echoes in each time window unit; Identifying, based on the persistent correlation of the main energy clusters in the energy concentration distribution characteristics, the structural reflection echo belonging to the drone body and the ambient noise echo belonging to meteorological suspended particles; Determining the spatial reflection point position on the surface of the UAV body based on the pulse return time sequence of the structure reflected echo; The spatial reflection point positions are combined with the environmental noise interference intensity mark of the corresponding time window unit to construct a three-dimensional reflection point set with interference perception attributes, and the three-dimensional reflection point set with interference perception attributes is combined to form an original point cloud sequence.

3. The method according to claim 1, characterized in that Separating the inherent scattered signal associated with the UAV configuration and the transient noise signal associated with complex meteorological interference based on the physical reflectance characteristics of the original point cloud sequence, including: Extracting the laser wavelength channel reflection intensity ratio, polarization state change degree and pulse transmission time change rate corresponding to each three-dimensional spatial reflection point from the original point cloud sequence; Generating wavelength reflection intensity distribution bands under different meteorological interference types based on the laser wavelength channel reflection intensity ratio; constructing a dynamic interference filtering model based on the coupling relationship between the polarization state change degree and the pulse transmission time change rate; Marking a set of reflection points in the wavelength reflection intensity distribution band that satisfy the metal material reflection stability law as a first-level candidate signal; In the dynamic interference filtering model, based on the linear constraint threshold of the coupling relationship, the echo points with continuous pulse time series in the first-level candidate signal are screened; The three-dimensional spatial position of the screened echo points is traced to form the inherent scattering signal related to the UAV configuration; A set of reflection points not covered by the wavelength reflection intensity distribution band and the dynamic interference filtering model is classified as the instantaneous noise signal.

4. The method according to claim 1, wherein Determining a dynamic structural response spectrum representing the external components of the UAV from the inherent scattering signal includes: Dividing a plurality of spatial reflection points in the inherent scattered signal into position areas corresponding to the external components of the UAV; Fitting the motion trajectory vector of each position area based on the displacement vector sequence of the spatial reflection point changing with time; Matching the motion trajectory vector of each position area with the standard component stiffness parameters in a preset UAV manufacturing material property library to obtain the matched standard component stiffness parameters; generating a local vibration wave propagation direction chain of the external component of the UAV based on the matched standard component stiffness parameters; On the local vibration wave propagation direction chain, mapping the vibration energy envelope of the spatial reflection point corresponding to each propagation direction node; generating a time-space response intensity mapping relationship of the position area according to an amplitude variation relationship of the vibration energy envelope in a scanning time sequence corresponding to the position area; The time-space response intensity mapping relationship of all position areas is integrated to form the dynamic structural response map.

5. The method according to claim 1, wherein The dynamic structural response spectrum is synchronously compared with a plurality of pre-stored known UAV model benchmark response spectra, and an optimal fit determination result is generated based on the comparison results, including: Extracting the characteristic change frequency of the time-space response intensity mapping relationship of each position area in the dynamic structural response map; Aligning the characteristic change frequency with the reference frequency band corresponding to the pre-stored UAV model reference response spectrum in time and space coverage area; After the spatiotemporal coverage areas are aligned, calculating the degree of continuous distribution overlap of the characteristic change frequencies within the reference frequency band; Based on the continuous distribution overlap, identifying the similarity of spectral energy transfer paths between the dynamic structural response spectrum and each benchmark response spectrum; generating a frequency offset phase difference set according to a sequence of overlapping positions of the characteristic change frequency and the reference frequency band; Suppressing the contribution of phase anomaly points associated with a meteorological interference intensity of a preset threshold value in the frequency offset phase difference set to obtain a phase anomaly point suppression result; The continuous distribution coincidence, the spectrum energy transfer path similarity and the phase abnormal point suppression result are combined to generate the optimal fit determination result.

6. The method according to claim 1, characterized in that Based on the optimal fit determination result, the drone model is output, including: Extracting the aggregate value of multiple source indicators in the optimal fit determination result, wherein the multiple source indicators include continuous distribution coincidence, spectrum energy transfer path similarity, and phase anomaly suppression result; Performing regional mapping of the local structural vibration source position on the aggregated value and the pre-stored model feature space to generate a regional mapping result; Based on the region mapping result, activating the response feature core corresponding to the drone model; Verifying the spatial compatibility of the response characteristic kernel with the current atmospheric turbulence correlation factor to obtain a spatial compatibility verification result; Locking the target model identifier according to the spatial compatibility verification result and the weight distribution of the aggregation value; Perform a feature backtracking consistency check between the target model identifier and the historical recognition result. When the deviation value of the feature backtracking consistency check is less than a preset dynamic fluctuation tolerance threshold, output the target model identifier as a drone model.

7. The method according to claim 4, characterized in that Based on the matched standard component stiffness parameters, a local vibration wave propagation direction chain of the external component of the UAV is generated, including: Determining the main direction of axial stress conduction in the position area according to the matched standard component stiffness parameters; Based on the boundary constraints of adjacent areas of the position area, extending the main direction of axial stress conduction to form a continuous conduction path; The connection nodes marking the boundaries of adjacent position areas on the continuous conduction path are vibration wave refraction nodes; Based on the direction change rate of the motion trajectory vector, correcting the conduction angle offset of the continuous conduction path at the vibration wave refraction node to obtain a corrected conduction path segment; All corrected conduction path segments are connected to generate a chain of local vibration wave propagation directions.

8. A real-time identification system for drone models based on laser radar point cloud features, characterized in that: include: An acquisition module is used to acquire an original point cloud sequence generated by the laser radar scanning the UAV in real time under the target flight area, wherein the original point cloud sequence includes disturbed three-dimensional space reflection points; a separation module, configured to separate the inherent scattered signal associated with the UAV configuration and the transient noise signal associated with complex meteorological interference based on the physical reflectance characteristics of the original point cloud sequence; An analysis module, configured to analyze the inherent scattering signal to obtain a dynamic structural response spectrum representing the external components of the UAV; a comparison module, configured to synchronously compare the dynamic structural response spectrum with a plurality of pre-stored reference response spectra of known UAV models, and generate an optimal fit determination result based on the comparison results; The output module is used to output the drone model based on the optimal fit determination result.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a real-time drone model recognition method based on lidar point cloud features as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a real-time identification method for drone models based on laser radar point cloud features as described in any one of claims 1 to 7 is implemented.

Images