An unmanned aerial vehicle anomaly diagnosis method and system based on multi-modal data fusion

Abstract

The application relates to the field of unmanned aerial vehicle (UAV) anomaly diagnosis, and in particular to a multi-modal data fusion UAV anomaly diagnosis method and system. The method comprises the following steps: acquiring multi-modal sensor data of the UAV, analyzing dust characteristics in a current environment in real time according to the multi-modal sensor data of the UAV, and generating a basic dust risk parameter set; acquiring UAV flight state parameters, taking the UAV flight state parameters as a dynamic correction factor, dynamically fusing the dynamic correction factor with the basic dust risk parameter set, and generating a dynamic dust risk influence parameter set; based on this, analyzing the influence degree of the dynamic dust on the flight safety of the UAV, and generating a UAV comprehensive dynamic risk level; based on this, diagnosing a current dust risk composition situation, and generating a UAV anomaly diagnosis report. In the UAV anomaly diagnosis process, the application can effectively support the flight system to make a forward-looking adaptive decision, and avoid serious safety hazards.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) anomaly diagnosis, and in particular to a UAV anomaly diagnosis method and system based on multimodal data fusion. Background Technology

[0002] In the field of drone safety control, drones have been widely used in key scenarios such as disaster relief, logistics transportation and complex environment monitoring. Their flight safety and reliability are directly related to the success or failure of missions, equipment maintenance and the safety of ground personnel. They are the core equipment of the modern low-altitude economy and intelligent emergency rescue system.

[0003] However, existing methods for diagnosing drone anomalies lack a mechanism for real-time, systematic fusion and collaborative perception of multi-dimensional environmental interference factors when facing complex, variable, and multi-factor coupled unstructured environments. This not only fails to effectively support the flight system in making forward-looking adaptive decisions, but may also trigger a chain of system failures under extreme conditions, posing serious safety hazards. Summary of the Invention

[0004] This application provides a method and system for anomaly diagnosis of unmanned aerial vehicles (UAVs) based on multimodal data fusion, in order to solve the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for anomaly diagnosis of unmanned aerial vehicles (UAVs) based on multimodal data fusion. The method includes: acquiring multimodal sensor data of the UAV; analyzing the dust characteristics in the current environment in real time based on the multimodal sensor data of the UAV to generate a basic dust risk parameter set; acquiring flight state parameters of the UAV; using the flight state parameters of the UAV as dynamic correction factors and dynamically fusing them with the basic dust risk parameter set to generate a dynamic dust risk impact parameter set; analyzing the degree of impact of dynamic dust on the flight safety of the UAV based on the dynamic dust risk impact parameter set to generate a comprehensive dynamic risk level of the UAV; and diagnosing the current dust risk composition based on the comprehensive dynamic risk level of the UAV and generating an anomaly diagnosis report of the UAV.

[0006] The above technical solution involves the UAV first collecting environmental data in real time using its onboard multimodal sensors. After fusion analysis by the edge computing unit, dust concentration, distribution, and other characteristics are extracted to generate a basic dust risk parameter set. Subsequently, the system simultaneously acquires the UAV's real-time flight status parameters and uses these dynamic parameters as correction factors to adaptively weight and fuse them with the basic risk parameter set, generating a more accurate dynamic dust risk impact parameter set. Next, based on this dynamic parameter set, the system uses a built-in lightweight rule engine or fuzzy inference model to comprehensively assess the threat level of dust from multiple aspects, ultimately mapping and outputting the UAV's comprehensive dynamic risk level. Finally, based on this risk level, the system diagnoses the risk type and proportion, executes corresponding adjustment measures, and generates a UAV anomaly diagnosis report containing all diagnostic decision details in real time, which is then output to the control personnel. By integrating multimodal data and dynamically correcting it, the comprehensiveness and accuracy of flight risk perception in complex dusty environments have been significantly improved, overcoming the limitations of traditional single-sensor perception. By closely combining environmental threats with the real-time status of UAVs, risk assessment has been transformed from static to dynamic, making early warnings more timely and decisions more accurate. The resulting closed-loop autonomous system of perception-assessment-decision-execution greatly enhances the autonomous adaptability and survivability of UAVs in harsh environments, effectively preventing potential failures and accidents caused by dust, ensuring the continuity and safety of critical tasks such as industrial inspection and disaster relief, and providing operators with clear and reliable decision support.

[0007] Optionally, the step of analyzing the dust characteristics in the current environment in real time based on the UAV multimodal sensor data to generate a basic dust risk parameter set includes: the UAV multimodal sensor data includes image optical characteristic data, active detection waveform data, and medium electromagnetic property data; based on the image optical characteristic data, analyzing the visual occlusion characteristics of dust to identify perceptible occlusion type dust risk and quantifying it into a perceptible occlusion type dust risk value; based on the active detection waveform data, analyzing the spatial concentration distribution and deposition characteristics of dust to identify deposition erosion type dust risk and quantifying it into a deposition erosion type dust risk value; based on the medium electromagnetic property data, analyzing the interference characteristics of dust on electromagnetic signals to identify conductive interference type dust risk and quantifying it into a conductive interference type dust risk value; and generating the basic dust risk parameter set based on the perceptible occlusion type dust risk value, the deposition erosion type dust risk value, and the conductive interference type dust risk value.

[0008] Optionally, the step of analyzing the visual occlusion characteristics of dust based on the image optical characteristic data to identify perceived occlusion-type dust risk and quantify it into a perceived occlusion-type dust risk value includes: extracting the global contour sharpness attenuation gradient and edge texture degradation rate of the current image frame relative to a historical reference frame based on the image optical characteristic data; determining a visual occlusion dynamic index characterizing the overall visibility deterioration trend of the environment based on the global contour sharpness attenuation gradient and the edge texture degradation rate; identifying and tracking key navigation markers on the preset flight path of the UAV based on the image optical characteristic data; analyzing the decay process of the morphological stability and feature recognizability of the key navigation markers in consecutive image frames to generate a navigation feature degradation coefficient; adding the visual occlusion dynamic index to the navigation feature degradation coefficient, and using the result as the perceived occlusion-type dust risk value.

[0009] Optionally, the step of analyzing the spatial concentration distribution and settling characteristics of dust based on the actively detected waveform data to identify the risk of depositional erosion dust and quantify it into a depositional erosion dust risk value includes: separating the main echo signal and the secondary echo signal clusters generated by the scattering of dust particle groups based on the actively detected waveform data; inverting the concentration field eddy current intensity characterizing the non-uniformity of dust spatial distribution based on the energy attenuation gradient and spatiotemporal distribution density of the secondary echo signal clusters; simultaneously extracting the waveform broadening characteristics and spectral distortion characteristics of the main echo signal after passing through the dust environment, and analyzing the average particle size trend and adsorption-agglomeration trend of dust particles to generate particle settling potential energy parameters; multiplying the concentration field eddy current intensity and the particle settling potential energy parameters to characterize their synergistic enhancement effect, and using the product result as the depositional erosion dust risk value.

[0010] Optionally, the step of analyzing the interference characteristics of dust on electromagnetic signals based on the electromagnetic property data of the medium to identify the risk of conductive interference dust and quantify it into a conductive interference dust risk value includes: analyzing the real part drift and imaginary part loss oscillation amplitude of the equivalent dielectric constant, which characterize the comprehensive influence of the medium on the electromagnetic wave propagation path, based on the electromagnetic property data of the medium; deduce the cumulative trend of carrier phase synchronization deviation caused by sudden changes in the propagation environment of the UAV communication and navigation signal based on the real part drift of the equivalent dielectric constant; track and analyze the irregular fading characteristics of the signal strength due to the absorption and attenuation of the dust medium based on the imaginary part loss oscillation amplitude; dynamically predict the bit error rate transition risk and instantaneous interruption probability of the wireless link of the UAV in each frequency band based on the cumulative trend of carrier phase synchronization deviation and the irregular fading period and depth; quantify the cumulative trend of carrier phase synchronization deviation and the irregular fading characteristics into phase instability parameters and signal attenuation parameters, respectively; square the phase instability parameters and the signal attenuation parameters respectively and add them together, and use the sum as the conductive interference dust risk value.

[0011] Optionally, the step of dynamically fusing the UAV flight state parameters as dynamic correction factors with the basic dust risk parameter set to generate a dynamic dust risk impact parameter set includes: the UAV flight state parameters including real-time flight attitude angle data and flight velocity vector data of the UAV; based on the real-time flight attitude angle data, calculating the ratio of the projected area of ​​each surface of the UAV relative to the dust deposition direction to generate a dynamic exposure factor characterizing the impact of flight attitude on deposition risk; based on the flight velocity vector data, calculating the relative kinetic energy of the UAV and the spatial dust particle group to generate a dynamic kinetic energy factor characterizing the impact of flight velocity on erosion and interference risk; multiplying the dynamic exposure factor by the deposition erosion type dust risk value to obtain a dynamic deposition erosion risk value modulated by flight attitude; multiplying the dynamic kinetic energy factor by the conductive interference type dust risk value to obtain a dynamic conductive interference risk value modulated by flight velocity; and adding the perception shielding type dust risk value, the dynamic deposition erosion risk value, and the dynamic conductive interference risk value to generate the dynamic dust risk impact parameter set.

[0012] Optionally, the step of analyzing the impact of dynamic dust on UAV flight safety based on the dynamic dust risk impact parameter set and generating a comprehensive dynamic risk level for the UAV includes: extracting the risk values ​​of perception-masking dust, deposition-erosion dust, and conductive interference dust based on the dynamic dust risk impact parameter set; calculating the product of the perception-masking dust risk value and the deposition-erosion dust risk value to obtain a first synergistic risk factor; calculating the product of the perception-masking dust risk value and the conductive interference risk value to obtain a second synergistic risk factor; calculating the product of the deposition-erosion risk value and the conductive interference risk value to obtain a third synergistic risk factor; adding the first synergistic risk factor, the second synergistic risk factor, and the third synergistic risk factor to obtain a total synergistic effect value; comparing the total synergistic effect value with a preset risk level threshold table, and outputting the comprehensive dynamic risk level of the UAV based on the comparison result, wherein the risk level threshold table defines risk levels corresponding to different numerical ranges.

[0013] Optionally, the step of diagnosing the current dust risk composition based on the comprehensive dynamic risk level of the UAV and generating an anomaly diagnosis report for the UAV includes: analyzing the proportions of the perception-masking dust risk value, the deposition-erosion dust risk value, and the conductive interference dust risk value based on the comprehensive dynamic risk level of the UAV to diagnose the dominant type and composition of the current dust risk; if the proportion of perception-masking dust risk exceeds the limit, the auxiliary navigation weight of the visual sensor is increased and a switch to a non-optical navigation mode is prompted; if the proportion of deposition-erosion dust risk exceeds the limit, the power system output power redundancy is increased and the surface anti-deposition program is activated; if the proportion of conductive interference dust risk exceeds the limit, the switch to the anti-interference communication frequency band is made and the signal transmission power is increased; and generating an anomaly diagnosis report for the UAV that includes the risk level, the values ​​of each risk type, and their countermeasures based on the dominant type and composition of the dust risk.

[0014] Optionally, the method further includes: establishing a dynamic dust correlation model among the perceived shielding dust risk value, the depositional erosion dust risk value, and the conductive interference dust risk value; tracing the induction and enhancement mechanism of high-value risk types on other risk types; tracking the spatial dust concentration distribution corresponding to high-value perceived shielding dust risk values, thereby inferring the potential area and rate of uneven dust accumulation on the surface of the UAV, and thus quantifying its depositional enhancement magnitude on the depositional erosion dust risk value; analyzing the dust composition and thickness distribution characterized by high-value depositional erosion dust risk values, thereby assessing the degree of change of the equivalent dielectric constant of the UAV radome by dust accumulation, and thus quantifying its dielectric induction intensity on the conductive interference dust risk value; and based on the depositional enhancement magnitude and the dielectric induction intensity, coupling and correcting the basic dust risk parameter set to generate a comprehensive dust risk parameter set with coupled chain effects.

[0015] Secondly, this application provides a multimodal data fusion-based unmanned aerial vehicle (UAV) anomaly diagnosis system, the system comprising:

[0016] The environmental perception module is used to acquire multimodal sensor data from the UAV, and based on the UAV multimodal sensor data, analyze the dust characteristics in the current environment in real time to generate a basic dust risk parameter set.

[0017] The data fusion module is used to acquire the flight status parameters of the UAV, use the UAV flight status parameters as dynamic correction factors, and dynamically fuse them with the basic dust risk parameter set to generate a dynamic dust risk impact parameter set.

[0018] The risk assessment module is used to analyze the impact of dynamic dust on the flight safety of UAVs based on the dynamic dust risk impact parameter set, and generate a comprehensive dynamic risk level for UAVs.

[0019] The decision control module is used to execute corresponding flight control commands and system configuration adjustment strategies based on the comprehensive dynamic risk level of the UAV, and generate an anomaly diagnosis report for the UAV. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram illustrating an application scenario provided in one embodiment of this application;

[0022] Figure 2 A flowchart illustrating a method for diagnosing unmanned aerial vehicle (UAV) anomalies based on multimodal data fusion, provided as an embodiment of this application;

[0023] Figure 3 This is a schematic diagram of the structure of a UAV anomaly diagnosis system based on multimodal data fusion, provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0025] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0026] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0027] Existing methods for diagnosing drone anomalies lack a mechanism for real-time, systematic fusion and collaborative perception of multi-dimensional environmental interference factors when facing complex, variable, and multi-factor coupled unstructured environments. This not only fails to effectively support the flight system in making forward-looking adaptive decisions, but may also trigger a chain of system failures under extreme conditions, posing serious safety hazards.

[0028] Based on this, this application provides a method and system for anomaly diagnosis of unmanned aerial vehicles (UAVs) using multimodal data fusion. First, the UAV collects environmental data in real time through its onboard multimodal sensors. After fusion analysis by an edge computing unit, characteristics such as dust concentration and distribution are extracted to generate a basic dust risk parameter set. Then, the system simultaneously acquires the UAV's real-time flight status parameters and uses these dynamic parameters as correction factors, adaptively weighting and fusing them with the basic risk parameter set to generate a more accurate dynamic dust risk impact parameter set. Next, based on this dynamic parameter set, a built-in lightweight rule engine or fuzzy inference model is used to comprehensively assess the threat level of dust from multiple aspects, ultimately mapping and outputting the UAV's comprehensive dynamic risk level. Finally, based on this risk level, the system diagnoses the risk type and proportion, executes corresponding adjustment measures, and generates a UAV anomaly diagnosis report containing all diagnostic decision details in real time, which is then output to the control personnel. By integrating multimodal data and dynamically correcting it, the comprehensiveness and accuracy of flight risk perception in complex dusty environments have been significantly improved, overcoming the limitations of traditional single-sensor perception. By closely combining environmental threats with the real-time status of UAVs, risk assessment has been transformed from static to dynamic, making early warnings more timely and decisions more accurate. The resulting closed-loop autonomous system of perception-assessment-decision-execution greatly enhances the autonomous adaptability and survivability of UAVs in harsh environments, effectively preventing potential failures and accidents caused by dust, ensuring the continuity and safety of critical tasks such as industrial inspection and disaster relief, and providing operators with clear and reliable decision support.

[0029] Figure 1 This diagram illustrates an application scenario provided by this application. In the process of diagnosing anomalies in unmanned aerial vehicles (UAVs), the method provided in this application can effectively support the flight system in making proactive and adaptive decisions, thus avoiding serious safety hazards.

[0030] Specifically, the method of this application is applied to any server that communicates with a multimodal sensor array. Through this server, the system acquires multimodal sensor data and flight status parameters of the UAV provided by the array. First, the UAV collects environmental data in real time using its onboard multimodal sensors. After fusion analysis by an edge computing unit, dust concentration, distribution, and other characteristics are extracted to generate a basic dust risk parameter set. Then, the system synchronously acquires the UAV's own real-time flight status parameters and uses these dynamic parameters as correction factors, adaptively weighting and fusing them with the basic risk parameter set to generate a more accurate dynamic dust risk impact parameter set. Next, based on this dynamic parameter set, a built-in lightweight rule engine or fuzzy inference model is used to comprehensively assess the threat level of dust from multiple aspects, ultimately mapping and outputting the UAV's comprehensive dynamic risk level. Finally, based on this risk level, the system diagnoses the risk type and proportion, executes corresponding adjustment measures, and generates a UAV anomaly diagnosis report containing all diagnostic decision details in real time, which is then output to the control personnel.

[0031] For specific implementation details, please refer to the following examples.

[0032] Figure 2 This is a flowchart illustrating a method for anomaly diagnosis of unmanned aerial vehicles (UAVs) based on multimodal data fusion, provided in one embodiment of this application. The method of this embodiment can be applied to servers in the above-described scenarios. Figure 2 As shown, the method includes:

[0033] S201. Acquire multimodal sensor data from the UAV, and based on the multimodal sensor data from the UAV, analyze the dust characteristics in the current environment in real time to generate a basic dust risk parameter set.

[0034] UAV multimodal sensor data can be a heterogeneous data set collected by various types of sensors (such as image sensors, lidar, infrared sensors, particulate matter sensors, etc.) carried on the UAV. It is used to comprehensively perceive the flight environment, including image optical characteristic data, active detection waveform data and medium electromagnetic property data. The data comes from the multimodal sensor array.

[0035] The basic dust risk parameter set can be a set of basic parameters used to characterize the risks that dust in the current environment may pose to drones, including perception-masking dust risk values, deposition-erosion dust risk values, and conductive interference dust risk values.

[0036] Specifically, data from a single sensor is insufficient to comprehensively capture the impact of dust on drones. Traditional drone anomaly diagnosis often relies on a single dust sensor to obtain data, but the hazards of dust to drones are multi-dimensional: excessive dust concentration may clog the engine air intake, affecting power output; different particle sizes cause varying degrees of wear on the fuselage structure, with fine particles potentially intruding into electronic components and causing short circuits, while coarse particles may impact the propeller, affecting aerodynamic performance. Relying solely on a single sensor cannot simultaneously acquire key information such as dust concentration, particle size, and distribution density, leading to a one-sided assessment of risk. For example, in mining operations, monitoring only dust concentration may overlook the presence of high-hardness particles, whose impact force during high-speed drone flight is sufficient to damage propeller blades and cause crashes. This step addresses this by using multiple onboard sensors to collect real-time environmental data from the drone; utilizing the built-in edge computing unit, this multimodal data is fused and analyzed to quickly extract key information such as dust concentration, distribution, and particulate matter characteristics in the environment, and based on this, a basic dust risk parameter set is generated. By generating a basic dust risk parameter set from multimodal sensor data, key information such as dust concentration and particle size can be comprehensively captured, overcoming the limitations of a single sensor and providing accurate basic data for subsequent risk assessment, thereby improving the accuracy of the initial judgment of the dust environment.

[0037] S202. Obtain the UAV flight status parameters, use the UAV flight status parameters as dynamic correction factors, and dynamically fuse them with the basic dust risk parameter set to generate a dynamic dust risk impact parameter set.

[0038] The drone flight status parameters can be a set of data reflecting the drone's operational status during flight, including real-time flight attitude angle data and flight velocity vector data, sourced from a multimodal sensor array. The dynamic dust risk impact parameter set can refer to a set of dust risk parameters, corrected for flight status, that are more practically instructive.

[0039] Specifically, there is a dynamic correlation between the flight status of drones and dust risk. Existing methods lack effective consideration of this correlation. The flight speed, altitude, attitude, and other state parameters of drones directly affect the intensity and manner in which dust affects them: when drones fly at high speeds, the airflow speed increases, the kinetic energy of dust particles increases, and the impact force on the fuselage surface and key components is significantly increased; when flying at low altitudes, the dust concentration near the ground is usually higher and may be accompanied by turbulence, resulting in uneven dust distribution and increasing the uncertainty of risk; when hovering, the airflow generated by the propellers may pick up dust from the ground, forming local high-concentration dust areas, directly affecting the normal operation of sensors and engines; traditional methods often treat dust risk as a static factor, assessing it only based on environmental parameters and ignoring the dynamic influence of flight status, resulting in a large deviation between risk assessment results and actual conditions. For example, in the same dust environment, the wear risk faced by drones flying at speeds of 10 m / s and 30 m / s can differ by several times. If the speed factor is not considered, the formulated protection strategy will lose its specificity. This step simultaneously acquires the real-time flight status of the UAV, such as speed and attitude parameters. These dynamic parameters are used as correction factors and weighted and adaptively fused with basic dust risk parameters in real time to generate a dynamic dust risk impact parameter set that better reflects actual flight conditions. By fusing flight status parameters as dynamic correction factors to generate a dynamic parameter set, the relationship between flight status and dust risk can be linked in real time, eliminating static assessment biases and making risk assessments more closely aligned with actual flight scenarios, thus enhancing the targeted judgment of dust risk under different flight conditions.

[0040] S203. Based on the dynamic dust risk impact parameter set, analyze the degree of impact of dynamic dust on the flight safety of UAVs and generate the comprehensive dynamic risk level of UAVs.

[0041] The comprehensive dynamic risk level of drones can be a grading index used to comprehensively evaluate the overall risk level faced by drones in the current dusty environment.

[0042] Specifically, existing diagnostic methods lack timeliness and struggle to address rapid changes in dust risks. Dust environments are highly dynamic; for example, in desert areas, a sandstorm can cause a rapid increase in dust concentration within a short period. Traditional diagnostic methods have long data processing cycles and cannot update risk assessment results in real time. When risks suddenly escalate, timely countermeasures cannot be triggered, leading to accidents. Furthermore, during mission execution, drones may constantly adjust their flight paths and mission objectives, resulting in changes in the dust environment. If risk parameters are not analyzed and updated in real time, drones will remain at potential risk for extended periods. For instance, in disaster relief scenarios, drones need to traverse areas with varying dust concentrations. If risk assessment is delayed, flight strategies may not be adjusted in time when entering high-risk areas, leading to equipment damage and hindering rescue operations. This step, based on a dynamic risk parameter set, employs a lightweight rule engine or fuzzy inference model to comprehensively assess and map the multi-dimensional risks posed by dust, ultimately outputting a clear and actionable comprehensive dynamic risk level for drones to guide subsequent decision-making. A comprehensive dynamic risk level is generated based on a dynamic parameter set, transforming complex parameters into intuitive levels. This provides a clear basis for flight control, facilitates rapid identification of risk levels, and improves the efficiency and accuracy of assessing the impact of dust on drone safety.

[0043] S204. Based on the comprehensive dynamic risk level of the drone, diagnose the current dust risk composition and generate a drone anomaly diagnosis report.

[0044] A drone anomaly diagnosis report can include information on the impact of dust on the drone during flight, the risk level, the numerical values ​​of each risk type, and corresponding countermeasures.

[0045] Specifically, existing anomaly diagnosis lacks a systematic response mechanism, making it difficult to form closed-loop management. Traditional methods, after assessing risks, often only issue simple alarm signals, failing to formulate specific flight control commands and system configuration adjustment strategies based on the risk level. This results in operators lacking clear action guidelines when facing risks. For example, when an increased dust risk is detected, operators may not know whether to reduce flight speed, change flight path, or adjust sensor monitoring frequency, delaying the optimal response time. Furthermore, the lack of detailed anomaly diagnosis reports fails to provide a basis for subsequent maintenance and improvement, leading to the potential recurrence of similar risks. This step diagnoses the dominant risk type and composition based on the comprehensive dynamic risk level of the UAV; simultaneously, it dynamically generates a UAV anomaly diagnosis report containing the risk level, numerical values ​​for each risk type, and corresponding countermeasures, and outputs it to control personnel. Executing control commands based on the risk level and generating reports enables timely implementation of corresponding measures to reduce risks, forming closed-loop management, ensuring flight safety, and providing a reference for maintenance and optimization, thereby improving the UAV's operational management level and adaptability to complex environments.

[0046] The method provided in this embodiment first involves the UAV collecting environmental data in real time using its onboard multimodal sensors. After fusion analysis by the edge computing unit, dust concentration, distribution, and other characteristics are extracted to generate a basic dust risk parameter set. Subsequently, the system synchronously acquires the UAV's own real-time flight status parameters and uses these dynamic parameters as correction factors to adaptively weight and fuse them with the basic risk parameter set to generate a more accurate dynamic dust risk impact parameter set. Next, based on this dynamic parameter set, the system uses a built-in lightweight rule engine or fuzzy inference model to comprehensively assess the threat level of dust in various aspects, and finally maps and outputs the comprehensive dynamic risk level of the UAV. Finally, based on this risk level, the system diagnoses the risk type and proportion, executes corresponding adjustment measures, and generates a UAV anomaly diagnosis report containing all diagnostic decision details in real time, which is then output to the control personnel. By integrating multimodal data and dynamically correcting it, the comprehensiveness and accuracy of flight risk perception in complex dusty environments have been significantly improved, overcoming the limitations of traditional single-sensor perception. By closely combining environmental threats with the real-time status of UAVs, risk assessment has been transformed from static to dynamic, making early warnings more timely and decisions more accurate. The resulting closed-loop autonomous system of perception-assessment-decision-execution greatly enhances the autonomous adaptability and survivability of UAVs in harsh environments, effectively preventing potential failures and accidents caused by dust, ensuring the continuity and safety of critical tasks such as industrial inspection and disaster relief, and providing operators with clear and reliable decision support.

[0047] In some embodiments, the UAV multimodal sensor data includes image optical characteristic data, active detection waveform data, and medium electromagnetic property data. Based on the image optical characteristic data, the visual occlusion characteristics of dust are analyzed to identify the risk of perceived occlusion-type dust and quantify it into a perceived occlusion-type dust risk value. Based on the active detection waveform data, the spatial concentration distribution and deposition characteristics of dust are analyzed to identify the risk of depositional erosion-type dust and quantify it into a depositional erosion-type dust risk value. Based on the medium electromagnetic property data, the interference characteristics of dust on electromagnetic signals are analyzed to identify the risk of conductive interference-type dust and quantify it into a conductive interference-type dust risk value. A basic dust risk parameter set is generated based on the perceived occlusion-type dust risk value, the depositional erosion-type dust risk value, and the conductive interference-type dust risk value.

[0048] Image optical property data can reflect the optical characteristics of dust, including the degree of absorption and scattering of light of different wavelengths by dust, and the resulting changes in image clarity and contrast. Active detection waveform data can be data formed by receiving waveform signals reflected back by dust after an active detection sensor emits a signal; it includes information such as dust concentration, distribution range, and settling velocity. Medium electromagnetic property data can reflect the characteristics of dust as a medium in the conduction and interference of electromagnetic signals, such as the degree of electromagnetic signal attenuation caused by dust and changes in signal transmission delay. Perception-obstructing dust risk refers to the risk that dust obstructs light, preventing the UAV's optical sensors from clearly perceiving the surrounding environment, thus affecting the UAV's identification of obstacles and targets, and ultimately causing flight safety hazards. The perception-obstructing dust risk value is a quantitative representation of this risk, with the numerical value reflecting the severity of the risk. Depositional erosion dust risk refers to the risk of dust depositing on critical components of a drone (such as propellers, sensor surfaces, and engine air intakes) or corroding these components, affecting the drone's normal operation and leading to malfunctions or performance degradation. The depositional erosion dust risk value is a quantitative indicator of this risk, reflecting the likelihood and extent of damage to drone components. Conductive interference dust risk refers to the risk of conductive dust entering the drone's circuitry, causing short circuits, poor contact, and other electromagnetic interference, affecting the normal operation of the drone's electronic equipment. The conductive interference dust risk value is a quantitative description of this risk, reflecting the severity of the interference to the drone's circuitry and electronic equipment.

[0049] Specifically, during drone flight, environmental dust is a significant factor affecting flight safety. Dust hazards to drones manifest in various forms, and single-type sensor data often only reflects one aspect of dust characteristics, failing to comprehensively and accurately assess the risks posed. For example, relying solely on image optical characteristic data only reveals the impact of dust on optical perception, but not the deposition of dust on components or its interference with electromagnetic signals. This leads to an incomplete assessment of dust risks faced by drones, potentially overlooking hidden safety hazards and impacting safe flight. Therefore, it is crucial to subdivide drone multimodal sensor data into image optical characteristic data, active detection waveform data, and medium electromagnetic property data, and to identify different types of dust risks based on these data. Different types of dust risks have different impact mechanisms and consequences on drones. Perception-obstructing dust risks primarily affect the drone's environmental perception capabilities, potentially causing collisions with obstacles; deposition-erosion dust risks mainly damage the drone's mechanical and sensing components, affecting its normal operation; and conductive interference dust risks target the drone's electronic circuitry, potentially causing equipment malfunctions or even loss of control. To address the above issues, this step simultaneously utilizes the UAV's optical sensors to acquire environmental images and extract image optical characteristic data (e.g., dust area accounting for 30% of the field of view), activates the active detection sensor to emit scanning waveforms and analyzes the active detection waveform data (e.g., particle size distribution in the concentration peak area is 10-50μm), and enables the electromagnetic sensor to monitor signal transmission quality and acquire medium electromagnetic property data (e.g., signal-to-noise ratio decrease of 20dB in the 2.4GHz band). Based on the image optical characteristic data, visual occlusion characteristics are analyzed (e.g., visibility decreases from 1000m to 300m), quantifying and generating a perception-occlusion type dust risk value (e.g., 75). Based on the active detection waveform data, spatial concentration and deposition characteristics are analyzed (e.g., deposition rate reaches 5mg / min), quantifying and generating a deposition-erosion type dust risk value (e.g., 80). Based on the medium electromagnetic property data, electromagnetic interference characteristics are analyzed (e.g., dielectric constant change rate is 0.5), quantifying and generating a conductive interference type dust risk value (e.g., 90). Finally, the three types of risk values ​​are encapsulated into a basic dust risk parameter set containing timestamps and geographic location tags (e.g., "Time 12:00 | Location X, Y | ..."). Shading value 75 | Deposition value 80 | Interference value 90”).

[0050] The method provided in this embodiment enables a detailed analysis of multimodal sensor data from drones, allowing for a comprehensive and accurate identification and quantification of various risks posed by dust in the environment. The generated set of basic dust risk parameters provides reliable foundational data for subsequent comprehensive dynamic risk level assessments. By assessing different types of dust risks separately, a clearer and more specific understanding of the dust risks faced by drones is achieved, which helps to develop targeted countermeasures and improve the adaptability of drones to dusty environments.

[0051] In some embodiments, based on image optical property data, the global contour sharpness attenuation gradient and edge texture degradation rate of the current image frame relative to the historical reference frame are extracted; based on the global contour sharpness attenuation gradient and edge texture degradation rate, a visual occlusion dynamic index characterizing the overall visibility deterioration trend of the environment is determined; based on image optical property data, key navigation markers on the UAV's preset flight path are identified and tracked; the decay process of the morphological stability and feature recognizability of key navigation markers in consecutive image frames is analyzed to generate a navigation feature degradation coefficient; the visual occlusion dynamic index is added to the navigation feature degradation coefficient, and the result is used as a perceived occlusion-type dust risk value.

[0052] The global contour sharpness attenuation gradient can be the rate and extent of the decrease in sharpness of the global contours of all objects in the current image frame compared to the corresponding contours in the historical reference frame, reflecting the overall deterioration of visual sharpness in the environment. The edge texture degradation rate can be the rate at which the richness and recognizability of the texture details of object edges in the current image frame decreases compared to the corresponding edge textures in the historical reference frame, characterizing the degree of loss of image detail information. The visual occlusion dynamic index, determined based on the global contour sharpness attenuation gradient and the edge texture degradation rate, is a quantitative indicator that characterizes the trend of overall environmental visibility deterioration over time, comprehensively reflecting the occlusion impact of dust on the overall visual environment. Key navigation markers can be objects with significant features on a preset flight path that can be used for UAV positioning and navigation (such as specific buildings, landmarks, markers, etc.). The navigation feature degradation coefficient can be generated based on the decay process of the morphological stability and feature recognizability of key navigation markers, used to quantify the degree of influence of dust on the recognition of navigation markers; the larger the coefficient, the more severely the navigation markers are affected by dust and the more difficult they are to identify.

[0053] Specifically, during drone flight, the risk of obscuring dust is a significant factor affecting flight safety. Accurately identifying and quantifying this risk is crucial. Single image feature analysis is insufficient to fully reflect the visual obscuring effect of dust. For example, focusing only on global outline sharpness may overlook detailed changes in key navigational markers, while focusing only on individual markers fails to grasp the overall trend of visibility deterioration. Combining the global outline sharpness attenuation gradient and edge texture degradation rate to generate a visual obscuring dynamic index can reflect the overall deterioration of environmental visibility. Drone flight depends on the perception of the overall environment. If the overall visual environment deteriorates, even if individual markers are identifiable, the inability to perceive surrounding obstacles may lead to collision risks. Therefore, the introduction of this index provides a comprehensive perspective for assessing the obscuring effect of dust on overall visual perception. To address the above issues, this step acquires the current image frame and a historical baseline frame (e.g., a clear frame from 10 seconds ago). Global contour features are extracted using image processing algorithms (e.g., Canny edge detection calculates an edge pixel loss rate of up to 15% / second), and the contour sharpness decay gradient is calculated (e.g., a gradient value of -0.5 / second). A texture analysis algorithm (e.g., LBP operator calculates a texture contrast decrease rate of up to 2% / second) is used to calculate the edge texture degradation rate. The contour sharpness decay gradient and texture degradation rate are then weighted and fused (e.g., gradient weight 0.6, rate weight 0.4) to generate the image. Visual occlusion dynamic index (e.g., index value 70); based on preset flight path data, identify key navigation markers (e.g., ground AR markers), track and calculate their morphological stability (e.g., shape distortion rate 5%) and feature discernibility (e.g., SIFT feature matching success rate drops from 90% to 60%) in continuous image frames; analyze the decay process of morphological stability and feature discernibility to generate navigation feature degradation coefficient (e.g., coefficient value 30); finally, add the visual occlusion dynamic index and the navigation feature degradation coefficient (e.g., 70+30=100) to obtain the perceived occlusion dust risk value.

[0054] By comprehensively analyzing the overall deterioration trend of the visual environment and the degree of impact on key navigation markers through the method provided in this embodiment, the risk of perceived dust obstruction can be fully and accurately quantified. The risk value of perceived dust obstruction provides a reliable quantitative basis for the diagnosis of UAV anomalies, enabling UAVs to more accurately identify visual perception risks caused by dust. At the same time, this multi-dimensional analysis method takes into account both the overall environment and key navigation information, avoiding the limitations of single-indicator evaluation and improving the comprehensiveness and accuracy of risk assessment.

[0055] In some embodiments, based on actively detected waveform data, the main echo signal and the secondary echo signal clusters generated by scattering from dust particle groups are separated; according to the energy attenuation gradient and spatiotemporal distribution density of the secondary echo signal clusters, the concentration field eddy current intensity characterizing the non-uniformity of dust spatial distribution is obtained by inversion; simultaneously, the waveform broadening characteristics and spectral distortion characteristics of the main echo signal after traversing the dust environment are extracted, and the average particle size trend and adsorption-agglomeration trend of dust particles are analyzed to generate particle settling potential energy parameters; the concentration field eddy current intensity and particle settling potential energy parameters are multiplied to characterize their synergistic enhancement effect, and the product result is used as the risk value of depositional erosion dust.

[0056] Secondary echo signal clusters can be a set of weak echo signals formed after an active detection signal is scattered by a group of dust particles in the air. Each secondary echo corresponds to the scattering of one or a group of dust particles, and the characteristics of the signal cluster reflect the distribution state of the dust particles. The concentration field eddy current intensity can be obtained based on the energy attenuation gradient and spatiotemporal distribution density inversion of the secondary echo signal cluster. It is a quantitative indicator used to characterize the non-uniformity of dust spatial distribution; higher intensity indicates a more disordered distribution of dust in space and a higher local concentration. Waveform broadening characteristics refer to the increase in the width of the waveform on the time axis after the main echo signal passes through a dusty environment. This is mainly caused by the difference in scattering delay of the signal by dust particles of different sizes, reflecting the particle size distribution characteristics of the dust particles. Spectral distortion characteristics refer to the deformation, shift, and other changes in the spectrum of the main echo signal after passing through a dusty environment, which are related to the physical properties (such as dielectric constant and shape) and motion state of the dust particles. The particle settling potential energy parameter can be generated by combining the average particle size trend and adsorption and agglomeration trend of dust particles. It is a quantitative indicator used to characterize the possibility of dust particles settling and the degree of accumulation after settling. The higher the parameter value, the easier it is for the particles to settle and accumulate.

[0057] Specifically, the risk of depositional erosion dust is one of the key factors affecting the safe operation of UAVs. Its severity is not only related to the spatial concentration of dust, but also closely related to the dust settling characteristics. A single indicator is insufficient to comprehensively assess this risk. If only dust concentration is considered and the settling trend of particles is ignored, it may be impossible to predict the rate of dust accumulation on components. If only settling characteristics are considered and concentration distribution is ignored, the instantaneous deposition risk in high-concentration areas may be underestimated. Secondary echo signal clusters are directly derived from the scattering of dust particles. Their energy attenuation gradient and spatiotemporal distribution density can accurately reflect the spatial distribution of dust. The concentration field eddy intensity obtained through inversion can effectively quantify the non-uniformity of dust distribution. In areas with high local eddy intensity, the dust concentration rises sharply, and a large amount of deposition is likely to occur when the UAV passes by. This is the spatial basis for assessing deposition risk. Without the analysis of the concentration field eddy intensity, high-risk concentration areas cannot be identified, resulting in insufficient spatial resolution of risk assessment. To address the above issues, this step, based on actively detected waveform data, first separates the main echo signal (e.g., a peak signal with an intensity exceeding a threshold of 200 dB) from the secondary echo signal clusters generated by scattering from dust particle groups (e.g., a high-frequency oscillating signal cluster with an intensity of 50-80 dB). Based on the energy attenuation gradient (e.g., a 15% decrease in signal intensity per meter) and spatiotemporal distribution density of the secondary echo signal clusters (e.g., 80 signal points per cubic meter per second), the concentration field eddy current intensity (e.g., eddy current coefficient 0.7), characterizing the spatial non-uniformity of dust distribution, is inverted. Simultaneously, the main echo signal is extracted... The waveform broadening characteristics (e.g., pulse width expansion rate of 18%) and spectral distortion characteristics (e.g., fundamental frequency attenuation rate of 12%) of the echo signal after passing through the dust environment are analyzed. Based on this, the average particle size trend (e.g., median particle size biased towards 30 μm) and adsorption and agglomeration trend (e.g., agglomeration growth rate of 5% per hour) of dust particles are analyzed to generate particle settling potential energy parameters (e.g., settling kinetic energy index of 110). The concentration field eddy current intensity is multiplied by the particle settling potential energy parameters (e.g., 0.7 × 110 = 77) to characterize their synergistic enhancement effect, and the product result is used as the risk value of depositional erosion dust.

[0058] By comprehensively analyzing the spatial concentration distribution and settling characteristics of dust through the method provided in this embodiment, the risk of depositional erosion dust is accurately quantified. The concentration field eddy current intensity accurately captures the non-uniformity of dust spatial distribution, and the particle settling potential energy parameter effectively reflects the settling trend of dust. The product of the two fully reflects the risk amplification effect under the synergistic effect of "concentration-settling", so that the risk value of depositional erosion dust can truly reflect the actual degree of harm.

[0059] In some embodiments, based on the electromagnetic property data of the medium, the drift of the real part of the equivalent dielectric constant and the oscillation amplitude of the imaginary loss of the medium, which characterize the comprehensive influence of the medium on the propagation path of the electromagnetic wave, are analyzed; based on the drift of the real part of the equivalent dielectric constant, the cumulative trend of carrier phase synchronization deviation caused by sudden changes in the propagation environment of the UAV communication and navigation signal is deduced; based on the oscillation amplitude of the imaginary loss, the irregular fading characteristics of the signal strength due to absorption and attenuation by the dust medium are tracked and analyzed; based on the cumulative trend of carrier phase synchronization deviation and the irregular fading period and depth, the bit error rate transition risk and instantaneous interruption probability of the wireless link of the UAV in each frequency band are dynamically predicted; the cumulative trend of carrier phase synchronization deviation and the irregular fading characteristics are quantified into phase instability parameters and signal attenuation parameters, respectively; the phase instability parameters and signal attenuation parameters are squared and then added together, and the sum is used as the conductive interference type dust risk value.

[0060] The real part drift of the equivalent dielectric constant can be the change in the real part of the equivalent dielectric constant from its reference value, reflecting the fluctuation degree of the influence of dust medium on the propagation speed of electromagnetic waves. The cumulative trend of carrier phase synchronization deviation can be the cumulative change trend of the synchronization deviation of the UAV communication and navigation signal over time due to changes in the propagation environment (dust interference), derived from the real part drift of the equivalent dielectric constant. Irregular fading characteristics can be the phenomenon of weakening of electromagnetic signal intensity without a fixed period and amplitude due to absorption and attenuation by dust medium, including characteristics such as the period and depth of fading. Bit error rate jump risk can refer to the possibility of a sudden and significant increase in the bit error rate of the UAV wireless link in a short period of time, related to the accumulation of carrier phase synchronization deviation and signal fading. The probability of instantaneous interruption can be the likelihood of a brief interruption of the electromagnetic signal under dust interference, determined by both irregular fading characteristics and phase synchronization deviation. The phase instability parameter can be a quantitative indicator of the cumulative trend of carrier phase synchronization deviation; the larger the parameter value, the more severe the accumulation of phase synchronization deviation and the more unstable the signal phase. Signal attenuation parameter can be a quantitative indicator of irregular fading characteristics. It reflects the degree to which signal strength is attenuated due to dust interference. The larger the value, the more obvious the attenuation.

[0061] Specifically, conductive dust interference poses a significant safety hazard to drones flying in dusty environments. Its main harm lies in disrupting communication and navigation signals, potentially leading to drone loss of control, mission failure, or even crashes. Single-dimensional signal interference analysis cannot comprehensively assess this risk; for example, focusing solely on signal attenuation may overlook synchronization failures caused by phase deviations. Monitoring only phase changes fails to reflect communication interruptions caused by sudden drops in signal strength. The real part shift of the equivalent dielectric constant directly reflects the impact of dust on the propagation speed of electromagnetic waves, and its cumulative effect leads to a continuous increase in carrier phase synchronization deviation. In drone communication… Phase synchronization is crucial for ensuring correct signal demodulation. Accumulated deviations can cause a spike in the bit error rate and even communication interruption. The oscillation amplitude of the imaginary part loss of the equivalent dielectric constant reflects the absorption characteristics of electromagnetic energy by dust. The irregular fading caused by this can lead to drastic fluctuations in signal strength. These fluctuations can cause the navigation signal received by the UAV to be inconsistent in strength, affecting positioning accuracy and easily leading to collision risks in complex environments. Quantifying the cumulative trend of carrier phase synchronization deviation into a phase instability parameter and the irregular fading characteristics into a signal attenuation parameter can accurately characterize the degree of signal interference from both phase and amplitude dimensions. To address the above issues, this step, based on the electromagnetic property data of the medium, analyzes the drift of the real part of the equivalent dielectric constant (e.g., a shift of 0.3 relative to a reference value) and the oscillation amplitude of the imaginary loss (e.g., an amplitude of 0.6) characterizing the combined influence of the medium along the electromagnetic wave propagation path. Based on the drift of the real part of the equivalent dielectric constant, it extrapolates the cumulative trend of carrier phase synchronization deviation caused by abrupt changes in the propagation environment (e.g., an increase in error of 0.15 radians per second). Based on the oscillation amplitude of the imaginary loss, it tracks and analyzes the irregular fading characteristics of signal strength due to absorption and attenuation by dust media (e.g., a sudden deep fading of 20 dB lasting 3 ms). Based on the cumulative trend of carrier phase synchronization deviation and the irregular fading period and depth, it dynamically predicts the bit error rate transition risk of the UAV's wireless links in each frequency band (e.g., from 10⁻⁻⁴). 6 The probability of transitioning to 10⁻² is 12% and the instantaneous interruption probability (e.g., 0.8% interruption per second); the cumulative trend of carrier phase synchronization deviation and irregular fading characteristics are quantified into phase instability parameters (e.g., exponent 75) and signal attenuation parameters (e.g., exponent 65), respectively; the phase instability parameters and signal attenuation parameters are squared (e.g., 75²=5625, 65²=4225) and then added together, and the sum (e.g., 5625+4225=9850) is used as the risk value of conductive interference dust.

[0062] By comprehensively analyzing the phase and amplitude changes of electromagnetic signals in a dusty environment, this embodiment achieves accurate quantification of the risk of conductive interference-type dust. The phase instability parameter and the signal attenuation parameter capture the characteristics of signal interference from different dimensions. The superposition after squaring strengthens the impact of severe interference, enabling the risk value to truly reflect the actual harm under the "synergistic effect of phase instability and signal attenuation".

[0063] In some embodiments, the UAV flight state parameters include real-time flight attitude angle data and flight velocity vector data of the UAV; based on the real-time flight attitude angle data, the ratio of the projected area of ​​each surface of the UAV relative to the dust deposition direction is calculated to generate a dynamic exposure factor characterizing the impact of flight attitude on deposition risk; based on the flight velocity vector data, the relative kinetic energy of the UAV and the spatial dust particle group is calculated to generate a dynamic kinetic energy factor characterizing the impact of flight velocity on erosion and interference risk; the dynamic exposure factor is multiplied by the deposition erosion type dust risk value to obtain the dynamic deposition erosion risk value modulated by flight attitude; the dynamic kinetic energy factor is multiplied by the conductive interference type dust risk value to obtain the dynamic conductive interference risk value modulated by flight velocity; the perceived shielding type dust risk value, the dynamic deposition erosion risk value, and the dynamic conductive interference risk value are added to generate a dynamic dust risk impact parameter set.

[0064] Real-time flight attitude angle data describes the UAV's attitude in space, including pitch, roll, and yaw angles, and is collected in real time by the UAV's attitude sensors. Flight velocity vector data describes the magnitude and direction of the UAV's flight speed, including horizontal and vertical velocity components, and is collected by the UAV's velocity sensors (such as GPS or pitot tubes). The dynamic exposure factor, generated based on the projected area ratio, is a quantitative indicator characterizing the impact of flight attitude on the risk of depositional erosion dust. A higher factor value indicates a higher risk of dust deposition on the airframe due to flight attitude. The dynamic kinetic energy factor, generated based on relative kinetic energy, is a quantitative indicator characterizing the impact of flight speed on the erosion risk of depositional erosion dust and the interference risk of conductive interference dust. A higher factor value indicates a greater degree of risk amplification by flight speed. The dynamic depositional erosion risk value is the risk value obtained by correcting the depositional erosion dust risk value with the dynamic exposure factor, reflecting the actual depositional erosion risk faced by the UAV under the current flight attitude. The dynamic conductive interference risk value can be obtained by correcting the conductive interference type dust risk value with a dynamic kinetic energy factor, reflecting the actual conductive interference risk faced by the UAV at the current flight speed.

[0065] Specifically, the basic dust risk parameter set only reflects the risk characteristics of dust itself in a specific environment. However, the flight state of the drone can significantly change the actual impact of dust risk. For example, in a dust environment with the same concentration, the area of ​​the drone's surface exposed to the dust settling path is different when the drone is flying in pitch versus in level flight, resulting in a significant difference in deposition risk. When flying at high speed, the impact of dust particles on the drone is stronger, which not only exacerbates the risk of erosion but may also increase interference with electromagnetic signals due to changes in particle motion. Therefore, it is impossible to accurately assess the real risk of a drone in actual flight based solely on the basic risk parameters. To address the above issues, this step uses real-time flight attitude angle data of the UAV (e.g., roll angle -5°, pitch angle 10°) to calculate the projected area ratio of each surface of the UAV relative to the dust deposition direction (e.g., a projection ratio of 1.2 at a pitch angle of 10°), generating a dynamic exposure factor characterizing the impact of flight attitude on deposition risk. Simultaneously, based on flight velocity vector data (e.g., horizontal velocity 8 m / s, vertical velocity 2 m / s), it calculates the relative kinetic energy of the UAV and the airborne dust particle group (e.g., a kinetic energy ratio of 1.4 at a composite velocity of 8.5 m / s), generating a dynamic kinetic energy ratio characterizing the impact of flight velocity on erosion and disturbance risk. The dynamic exposure factor (e.g., 1.2) is multiplied by the depositional erosion risk value (e.g., 60) to obtain the dynamic depositional erosion risk value modulated by flight attitude (e.g., 72); the dynamic kinetic energy factor (e.g., 1.4) is multiplied by the conductive interference risk value (e.g., 80) to obtain the dynamic conductive interference risk value modulated by flight speed (e.g., 112); finally, the perceived shielding dust risk value (e.g., 50), the dynamic depositional erosion risk value (e.g., 72), and the dynamic conductive interference risk value (e.g., 112) are added together to generate a set of dynamic dust risk impact parameters (e.g., a total of 234).

[0066] By introducing UAV flight state parameters as dynamic correction factors through the method provided in this embodiment, the dynamic fusion of basic dust risk and flight state is realized. The generated dynamic dust risk impact parameter set can accurately reflect the real dust risk faced by the UAV during actual flight. The dynamic exposure factor and dynamic kinetic energy factor respectively capture the modulation effect of flight attitude and flight speed on dust risk, making the assessment of deposition erosion risk and conductive interference risk more in line with the motion state of the UAV and avoiding the limitations of static assessment.

[0067] In some embodiments, based on a dynamic dust risk impact parameter set, risk values ​​for perceived shielding dust, deposition erosion dust, and conductive interference dust are extracted; the product of the perceived shielding dust risk value and the deposition erosion risk value is calculated to obtain a first synergistic risk factor; the product of the perceived shielding dust risk value and the conductive interference risk value is calculated to obtain a second synergistic risk factor; the product of the deposition erosion risk value and the conductive interference risk value is calculated to obtain a third synergistic risk factor; the first synergistic risk factor, the second synergistic risk factor, and the third synergistic risk factor are added to obtain a total synergistic effect value; the total synergistic effect value is compared with a preset risk level threshold table, and the comprehensive dynamic risk level of the UAV is output based on the comparison result. The risk level threshold table defines the risk level corresponding to different numerical ranges.

[0068] The first synergistic risk factor can be the product of the risk values ​​of perceived shielding dust and depositional erosion dust, used to characterize the superimposed effect of the interaction between these two types of risks. The larger the value, the more significant the synergistic effect. The second synergistic risk factor can be the product of the risk values ​​of perceived shielding dust and conductive interference dust, used to characterize the synergistic effect between optical perception risk and electromagnetic interference risk. The third synergistic risk factor can be the product of the risk values ​​of depositional erosion dust and conductive interference dust, used to characterize the synergistic effect between mechanical erosion risk and electromagnetic interference risk. The total synergistic effect value can be the sum of the first, second, and third synergistic risk factors, comprehensively reflecting the overall effect of the interaction and mutual aggravation among various dust risks. The preset risk level threshold table can be a pre-defined numerical range standard for classifying risk levels, which defines the risk level corresponding to different total synergistic effect value intervals (e.g., a total synergistic effect value of 0-5 corresponds to low risk, 5-10 corresponds to medium risk, etc.).

[0069] Specifically, the risks faced by drones in dusty environments are not simply the sum of individual risks, but rather involve significant synergistic effects between different risks. This synergy can lead to an overall risk far exceeding the sum of individual risks, and the actual risk level cannot be accurately reflected by a single risk value or its simple summation. The first synergistic risk factor captures the synergistic effect of "optical shielding-mechanical deposition." When dust simultaneously obscures vision and deposits on sensor surfaces, it accelerates the failure rate of the sensing system. The deposited dust reduces the transmittance of optical sensors, forming a "double shield" with the dust in the air, leading to a rapid decline in sensing capabilities. Ignoring this synergy may underestimate the urgency of sensing system failure. The second synergistic risk factor reflects the linked effect of "optical shielding-electromagnetic interference." In complex dusty environments, electromagnetic interference may cause communication delays between the drone and the ground. At the same time, optical shielding prevents the drone from autonomously identifying its path. The combination of these two factors significantly increases the risk of collision. For example, when a drone operating in a mine flies through dust, the combination of communication interruption and obstructed vision makes it highly likely to crash into the mine tunnel wall. The third synergistic risk factor reflects the cumulative hazards of "mechanical deposition-electromagnetic interference." To address the above issues, this step first extracts the risk values ​​for perceived shielding dust (e.g., 50), depositional erosion dust (e.g., 72), and conductive interference dust (e.g., 112) from the dynamic dust risk impact parameter set. Then, it calculates the product of the perceived shielding dust risk value and the depositional erosion dust risk value to obtain the first synergistic risk factor (e.g., 50 × 72 = 3600). It then calculates the product of the perceived shielding dust risk value and the conductive interference dust risk value to obtain the second synergistic risk factor (e.g., 50 × 112 = 5600). Finally, it calculates the product of the depositional erosion dust risk value and the conductive interference dust risk value to obtain the third synergistic risk factor. Factors (e.g., 72 × 112 = 8064); then, the first collaborative risk factor (e.g., 3600), the second collaborative risk factor (e.g., 5600), and the third collaborative risk factor (e.g., 8064) are added together to obtain the total collaborative effect value (e.g., 3600 + 5600 + 8064 = 17264); then, the total collaborative effect value (e.g., 17264) is compared with the preset risk level threshold table. If the threshold table defines 0-10000 as low risk, 10000-20000 as medium risk, and >20000 as high risk, then 17264 matches the medium risk level; finally, the comprehensive dynamic risk level of the UAV is output as "medium risk".

[0070] By analyzing the synergistic effects among various dust risks using the method provided in this embodiment, the generated total synergistic effect value can more realistically reflect the overall risk level faced by the drone, avoiding the limitations of single risk assessment. The comprehensive dynamic risk level obtained by comparing the total synergistic effect value with a preset threshold table provides drone operators with an intuitive and clear basis for risk judgment, which helps to quickly formulate response strategies.

[0071] In some embodiments, based on the comprehensive dynamic risk level of the UAV, the proportions of perceived shielding dust risk values, deposition erosion dust risk values, and conductive interference dust risk values ​​are analyzed to diagnose the dominant type and composition of the current dust risk: if the proportion of perceived shielding dust risk exceeds the limit, the auxiliary navigation weight of the visual sensor is enhanced and a switch to non-optical navigation mode is prompted; if the proportion of deposition erosion dust risk exceeds the limit, the power system output power redundancy is increased and the surface anti-deposition program is activated; if the proportion of conductive interference dust risk exceeds the limit, the switch to the anti-interference communication frequency band is made and the signal transmission power is increased; based on the dominant type and composition of dust risk, a UAV anomaly diagnosis report containing the risk level, the value of each risk type, and the corresponding countermeasures is generated.

[0072] The composition of dust risks can refer to the proportion and hierarchy of three types of dust risks—sensory shielding, depositional erosion, and conductive interference—in the overall risk. This reflects which type(s) of dust risks pose the primary threat in the current environment. Specifically, UAVs in complex dusty environments often face multiple types of risks coexisting, but the hazard mechanisms and coping methods for different types of risks differ significantly. For example, sensory shielding risks mainly affect navigation vision, depositional erosion risks mainly damage mechanical components, and conductive interference risks mainly disrupt communication signals. If only the overall risk level is known without clarifying the risk composition, operators or systems will be unable to take targeted measures, potentially leading to inefficient or even ineffective responses. Clarifying the risk proportion and dominant type is a prerequisite for accurate responses. When the proportion of sensory shielding risks exceeds the limit, continuing to rely on visual navigation may lead to collisions, at which point a switch to non-optical navigation mode must be made. Conversely, if depositional erosion risks are dominant, over-enhancing the navigation system is meaningless, and the power system should be prioritized for protection. To address the above issues, this step first receives the overall dynamic risk level of the UAV (e.g., "high risk") from the preceding process and extracts the risk values ​​for perception-masking dust (e.g., value 50), deposition-erosion dust (e.g., value 60), and conductive interference dust (e.g., value 80) from the dynamic dust risk impact parameter set. Then, the total risk value is calculated (e.g., 50+60+80=190), and the proportion of each risk type is calculated: perception-masking risk proportion (e.g., 50 / 190=26.3%), deposition-erosion risk proportion (e.g., 60 / 190=31.6%), and conductive interference risk proportion (e.g., 80 / 190=42.1%). Next, each proportion is compared with preset exceedance thresholds (e.g., perception-masking proportion threshold 35%, deposition-erosion proportion threshold 30%, and conductive interference proportion threshold 40%) to diagnose the dominant risk category. The system is categorized into several types (e.g., deposition erosion accounting for 31.6% > 30% and conductive interference accounting for 42.1% > 40%, which is classified as a dual-dominant risk). Then, targeted countermeasures are triggered based on the dominant type: for deposition erosion risk exceeding limits, the power system output power redundancy is increased (e.g., from 10% to 25%) and the surface anti-deposition program is activated (e.g., ultrasonic vibration is initiated); for conductive interference risk exceeding limits, the system switches to an anti-interference communication frequency band (e.g., from 2.4GHz to 900MHz) and the signal transmission power is increased (e.g., from 20dBm to 30dBm). Finally, the diagnostic results and measures are integrated to generate a structured UAV anomaly diagnostic report containing the risk level (e.g., "high risk"), various risk values ​​(e.g., perception shielding 50, deposition erosion 60, conductive interference 80), dominant type (e.g., "dual-dominant deposition erosion and conductive interference"), and specific countermeasures.

[0073] The method provided in this embodiment diagnoses the composition of dust risks, accurately identifying the dominant risk types currently affecting the flight safety of UAVs. This provides a clear basis for taking targeted countermeasures, avoiding the blindness and inefficiency of a "one-size-fits-all" approach. Differentiated measures taken for different dominant risks, such as switching navigation modes, increasing power redundancy, and switching communication frequency bands, can effectively mitigate the harm of the corresponding risks and improve the survivability and mission stability of UAVs in complex dusty environments.

[0074] In some embodiments, a dynamic correlation model of dust among perceived shielding dust risk values, depositional erosion dust risk values, and conductive interference dust risk values ​​is established to trace the induction and enhancement mechanism of high-value risk types on other risk types: tracking the spatial dust concentration distribution corresponding to high-value perceived shielding dust risk values, thereby inferring the potential area and rate of uneven dust accumulation on the surface of the UAV, thus quantifying its depositional enhancement magnitude on depositional erosion dust risk values; analyzing the dust composition and thickness distribution characterized by high-value depositional erosion dust risk values, thereby assessing the degree of change of the equivalent dielectric constant of the UAV radome by dust accumulation, thus quantifying its dielectric induction intensity on conductive interference dust risk values; based on the depositional enhancement magnitude and dielectric induction intensity, the basic dust risk parameter set is coupled and corrected to generate a comprehensive dust risk parameter set with coupled chain effects.

[0075] The dust dynamic correlation model can be used to describe the mutually induced and mutually reinforcing relationships among three types of dust risks: sensing shielding, depositional erosion, and conductive interference. By analyzing the dynamic changes and influence paths of the three types of risk values, the coupling effect between risks can be quantified. The depositional enhancement magnitude can be the degree to which a high numerical sensing shielding dust risk value leads to an increase in the depositional erosion dust risk value, calculated by quantifying the impact of potential area dust accumulation on depositional erosion risk. The dielectric induced intensity can be the degree to which a high numerical depositional erosion dust risk value leads to an increase in the conductive interference dust risk value, calculated by analyzing the change in the equivalent dielectric constant of the radome caused by dust accumulation.

[0076] Specifically, the three types of dust risks do not exist in isolation, but rather have significant coupled and chain effects. For example, high concentrations of perception-obstructing dust (such as dust clouds after mine blasting) not only obstruct vision, but their dense particles also rapidly deposit on the surface of the device, leading to a sharp increase in depositional erosion risks. Dust deposited on the radome (especially conductive components) alters the dielectric properties of the radome, thereby enhancing conductive interference risks. If the interaction between these risks is ignored and assessments are made based solely on a single risk value or a simple set of superimposed risk parameters, the actual risk level will be severely underestimated, resulting in inadequate countermeasures. To address the above issues, this step first receives a basic dust risk parameter set from the preceding process (e.g., including a risk value of 80 for perception shielding, 60 for deposition erosion, and 80 for conductive interference), and acquires multimodal sensor data (such as image optical property data, active detection waveform data, and medium electromagnetic property data). Simultaneously, a correlation database for the three risk values ​​is established. Then, the current high-value risk type is identified (e.g., perception shielding risk value 80 exceeds the threshold 70), and its triggering mechanism is traced: based on image optical property data, the spatial dust concentration distribution corresponding to the high-shielding area is tracked (e.g., concentration in front of the fuselage reaches 200 μg / m³), and the deposition pattern of high-concentration dust on the fuselage surface is analyzed (e.g., deposition at the rotor root). The deposition rate is 0.2 mm / min, and the deposition enhancement magnitude (e.g., 15%) is calculated accordingly. Simultaneously, the dust composition (e.g., 70% silicate) and thickness distribution (e.g., 0.6 mm at the radome) are retrieved based on active detection waveform data. The change in the equivalent dielectric constant of the antenna after dust accumulation (e.g., from 3.0 to 4.8) is measured using dielectric electromagnetic property data, and the dielectric induced intensity (e.g., 20%) is calculated accordingly. Next, the basic parameter set is coupled and corrected: the deposition erosion risk value is corrected to 60 × (1 + 15%) = 69, the conductive interference risk value is corrected to 80 × (1 + 20%) = 96, and the sensing shielding risk value remains unchanged at 80, generating a comprehensive dust risk parameter set (e.g., {sensing shielding: 80, deposition erosion: 69, conductive interference: 96}). Finally, the corrected parameter set is output to the risk level calculation module, providing data support for the anomaly diagnosis report.

[0077] The method provided in this embodiment establishes a dynamic correlation model of dust, revealing the chain reaction mechanism between the three types of dust risks. This upgrades risk assessment from single-type analysis to system coupling analysis, which is more in line with the generation and development laws of risks in actual dust environments. The quantification of deposition enhancement amplitude and dielectric induced intensity enables an accurate description of the interaction between risks, providing a scientific basis for the coupling correction of the basic dust risk parameter set. This allows the corrected comprehensive dust risk parameter set to truly reflect the actual level of risk.

[0078] Figure 3This is a schematic diagram of the structure of a UAV anomaly diagnosis system based on multimodal data fusion, provided in an embodiment of this application. Figure 3 As shown, the UAV anomaly diagnosis system 300 based on multimodal data fusion in this embodiment includes: an environment perception module 301, a data fusion module 302, a risk assessment module 303, a decision control module 304, and a dynamic communication module 305.

[0079] The environmental perception module 301 is used to acquire multimodal sensor data of the UAV, analyze the dust characteristics in the current environment in real time based on the UAV multimodal sensor data, and generate a basic dust risk parameter set; the data fusion module 302 is used to acquire UAV flight state parameters, use the UAV flight state parameters as dynamic correction factors, and dynamically fuse them with the basic dust risk parameter set to generate a dynamic dust risk impact parameter set; the risk assessment module 303 is used to analyze the degree of impact of dynamic dust on UAV flight safety based on the dynamic dust risk impact parameter set and generate a comprehensive dynamic risk level of the UAV; the decision control module 304 is used to execute corresponding flight control commands and system configuration adjustment strategies based on the comprehensive dynamic risk level of the UAV, and generate a UAV anomaly diagnosis report.

[0080] Optionally, when the environmental perception module 301 generates a basic dust risk parameter set by analyzing the dust characteristics in the current environment in real time based on the UAV multimodal sensor data, it specifically performs the following: the UAV multimodal sensor data includes image optical characteristic data, active detection waveform data, and medium electromagnetic property data; based on the image optical characteristic data, it analyzes the visual occlusion characteristics of dust to identify perceived occlusion type dust risk and quantifies it into a perceived occlusion type dust risk value; based on the active detection waveform data, it analyzes the spatial concentration distribution and deposition characteristics of dust to identify deposition erosion type dust risk and quantifies it into a deposition erosion type dust risk value; based on the medium electromagnetic property data, it analyzes the interference characteristics of dust on electromagnetic signals to identify conductive interference type dust risk and quantifies it into a conductive interference type dust risk value; and generates the basic dust risk parameter set based on the perceived occlusion type dust risk value, the deposition erosion type dust risk value, and the conductive interference type dust risk value.

[0081] Optionally, when the environmental perception module 301 analyzes the visual occlusion characteristics of dust based on the image optical characteristic data to identify and quantify the perceived occlusion dust risk as a perceived occlusion dust risk value, it specifically performs the following steps: based on the image optical characteristic data, extracting the global contour sharpness attenuation gradient and edge texture degradation rate of the current image frame relative to the historical reference frame; determining the visual occlusion dynamic index characterizing the overall visibility deterioration trend of the environment based on the global contour sharpness attenuation gradient and the edge texture degradation rate; identifying and tracking key navigation markers on the preset flight path of the UAV based on the image optical characteristic data; analyzing the decay process of the morphological stability and feature recognizability of the key navigation markers in continuous image frames to generate a navigation feature degradation coefficient; adding the visual occlusion dynamic index to the navigation feature degradation coefficient, and using the result as the perceived occlusion dust risk value.

[0082] Optionally, when the environmental sensing module 301 analyzes the spatial concentration distribution and sedimentation characteristics of dust based on the actively detected waveform data to identify the risk of depositional erosion dust and quantify it into a depositional erosion dust risk value, it specifically performs the following steps: based on the actively detected waveform data, it separates the main echo signal and the secondary echo signal cluster generated by the scattering of dust particle groups; according to the energy attenuation gradient and spatiotemporal distribution density of the secondary echo signal cluster, it inverts the concentration field eddy current intensity characterizing the non-uniformity of dust spatial distribution; simultaneously extracts the waveform broadening characteristics and spectral distortion characteristics of the main echo signal after passing through the dust environment, and analyzes the average particle size trend and adsorption-agglomeration trend of dust particles accordingly to generate particle sedimentation potential energy parameters; multiplies the concentration field eddy current intensity with the particle sedimentation potential energy parameters to characterize their synergistic enhancement effect, and uses the product result as the depositional erosion dust risk value.

[0083] Optionally, when the environmental perception module 301 analyzes the interference characteristics of dust on electromagnetic signals based on the electromagnetic property data of the medium to determine the risk of conductive interference dust and quantify it into a conductive interference dust risk value, it is specifically used to: analyze the real part drift and imaginary part loss oscillation amplitude of the equivalent dielectric constant characterizing the comprehensive influence of the medium on the electromagnetic wave propagation path based on the electromagnetic property data; deduce the cumulative trend of carrier phase synchronization deviation caused by sudden changes in the propagation environment of the UAV communication and navigation signal based on the real part drift of the equivalent dielectric constant; and deduce the cumulative trend of carrier phase synchronization deviation caused by sudden changes in the propagation environment based on the imaginary part loss oscillation amplitude. The method tracks and analyzes the irregular fading characteristics of signal strength due to absorption and attenuation by dust media; based on the cumulative trend of carrier phase synchronization deviation and the irregular fading period and depth, it dynamically predicts the bit error rate transition risk and instantaneous interruption probability of the UAV's wireless link in each frequency band; it quantifies the cumulative trend of carrier phase synchronization deviation and the irregular fading characteristics into phase instability parameters and signal attenuation parameters, respectively; it squares the phase instability parameters and the signal attenuation parameters and adds them together, using the sum as the conductive interference type dust risk value.

[0084] Optionally, the data fusion module 302 is specifically used for: the UAV flight state parameters including real-time flight attitude angle data and flight speed vector data of the UAV; based on the real-time flight attitude angle data, calculating the ratio of the projected area of ​​each surface of the UAV relative to the dust deposition direction, generating a dynamic exposure factor characterizing the impact of flight attitude on deposition risk; based on the flight speed vector data, calculating the relative kinetic energy of the UAV and the spatial dust particle group, generating a dynamic kinetic energy factor characterizing the impact of flight speed on erosion and interference risk; multiplying the dynamic exposure factor by the deposition erosion type dust risk value to obtain a dynamic deposition erosion risk value modulated by flight attitude; multiplying the dynamic kinetic energy factor by the conductive interference type dust risk value to obtain a dynamic conductive interference risk value modulated by flight speed; and adding the perceived shielding type dust risk value, the dynamic deposition erosion risk value, and the dynamic conductive interference risk value to generate the dynamic dust risk influence parameter set.

[0085] Optionally, the risk assessment module 303 is specifically used for: extracting the risk values ​​of the perceived shielding dust, the depositional erosion dust, and the conductive interference dust based on the dynamic dust risk impact parameter set; calculating the product of the perceived shielding dust risk value and the depositional erosion risk value to obtain a first synergistic risk factor; calculating the product of the perceived shielding dust risk value and the conductive interference risk value to obtain a second synergistic risk factor; calculating the product of the depositional erosion risk value and the conductive interference risk value to obtain a third synergistic risk factor; adding the first synergistic risk factor, the second synergistic risk factor, and the third synergistic risk factor to obtain a total synergistic effect value; comparing the total synergistic effect value with a preset risk level threshold table, and outputting the comprehensive dynamic risk level of the UAV based on the comparison result, wherein the risk level threshold table defines the risk level corresponding to different numerical ranges.

[0086] Feasible, the decision control module 304 is specifically used to: analyze the proportions of the perception-masking dust risk value, the deposition-erosion dust risk value, and the conductive interference dust risk value according to the comprehensive dynamic risk level of the UAV, in order to diagnose the dominant type and composition of the current dust risk; if the proportion of perception-masking dust risk exceeds the limit, the auxiliary navigation weight of the visual sensor is enhanced and a switch to non-optical navigation mode is prompted; if the proportion of deposition-erosion dust risk exceeds the limit, the power system output power redundancy is increased and the surface anti-deposition program is activated; if the proportion of conductive interference dust risk exceeds the limit, the switch to the anti-interference communication frequency band is made and the signal transmission power is increased; and generate an abnormal diagnosis report of the UAV containing the risk level, the value of each risk type, and the corresponding countermeasures according to the dominant type and composition of the dust risk.

[0087] Optionally, the system further includes a dynamic connection module 305, specifically used for: establishing a dynamic dust correlation model among the perceived shielding dust risk value, the depositional erosion dust risk value, and the conductive interference dust risk value; tracing the induction and enhancement mechanism of high-value risk types on other risk types; tracking the spatial dust concentration distribution corresponding to high-value perceived shielding dust risk values, thereby inferring the potential area and rate of uneven dust accumulation on the surface of the UAV, and thus quantifying its depositional enhancement magnitude on the depositional erosion dust risk value; analyzing the dust composition and thickness distribution represented by high-value depositional erosion dust risk values, thereby assessing the degree of change of the equivalent dielectric constant of the UAV radome by dust accumulation, and thus quantifying its dielectric induction intensity on the conductive interference dust risk value; and based on the depositional enhancement magnitude and the dielectric induction intensity, coupling and correcting the basic dust risk parameter set to generate a comprehensive dust risk parameter set with coupled chain effects.

[0088] The system in this embodiment can be used to execute the methods of any of the above embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

Claims

1. A method for unmanned aerial vehicle anomaly diagnosis based on multi-modal data fusion, characterized in that, The method comprises the following steps: acquiring unmanned aerial vehicle multi-modal sensor data, analyzing dust characteristics in the current environment in real time according to the unmanned aerial vehicle multi-modal sensor data, and generating a basic dust risk parameter set; acquiring unmanned aerial vehicle flight state parameters, taking the unmanned aerial vehicle flight state parameters as dynamic correction factors, dynamically fusing the unmanned aerial vehicle flight state parameters with the basic dust risk parameter set, and generating a dynamic dust risk impact parameter set; analyzing the influence degree of dynamic dust on the flight safety of the unmanned aerial vehicle according to the dynamic dust risk impact parameter set, and generating an unmanned aerial vehicle comprehensive dynamic risk level; diagnosing the current dust risk composition according to the unmanned aerial vehicle comprehensive dynamic risk level, and generating an unmanned aerial vehicle abnormal diagnosis report; the step of analyzing dust characteristics in the current environment in real time according to the unmanned aerial vehicle multi-modal sensor data and generating a basic dust risk parameter set comprises the following steps: the unmanned aerial vehicle multi-modal sensor data comprises image optical characteristic data, active detection waveform data and medium electromagnetic property data; based on the image optical characteristic data, the visual shielding characteristics of dust are analyzed to identify perceptual shielding type dust risks and quantify them into perceptual shielding type dust risk values; based on the active detection waveform data, the spatial concentration distribution and deposition characteristics of dust are analyzed to identify deposition erosion type dust risks and quantify them into deposition erosion type dust risk values; based on the medium electromagnetic property data, the interference characteristics of dust on electromagnetic signals are analyzed to identify conductive interference type dust risks and quantify them into conductive interference type dust risk values; the basic dust risk parameter set is generated according to the perceptual shielding type dust risk values, the deposition erosion type dust risk values and the conductive interference type dust risk values; the step of taking the unmanned aerial vehicle flight state parameters as dynamic correction factors, dynamically fusing the unmanned aerial vehicle flight state parameters with the basic dust risk parameter set, and generating a dynamic dust risk impact parameter set comprises the following steps: the unmanned aerial vehicle flight state parameters comprise real-time flight attitude angle data and flight speed vector data of the unmanned aerial vehicle; based on the real-time flight attitude angle data, the projection area ratio of each surface of the machine body relative to the dust deposition direction is calculated to generate a dynamic exposure factor representing the influence of the flight attitude on the deposition risk; based on the flight speed vector data, the relative motion kinetic energy of the machine body and the space dust particle group is calculated to generate a dynamic kinetic energy factor representing the influence of the flight speed on the erosion and interference risks; the dynamic deposition erosion risk value after the flight attitude modulation is obtained by multiplying the dynamic exposure factor and the deposition erosion type dust risk value; the dynamic conductive interference risk value after the flight speed modulation is obtained by multiplying the dynamic kinetic energy factor and the conductive interference type dust risk value; the dynamic dust risk impact parameter set is generated by adding the perceptual shielding type dust risk value, the dynamic deposition erosion risk value and the dynamic conductive interference risk value.

2. The method of claim 1, wherein, the step of analyzing the visual shielding characteristics of dust based on the image optical characteristic data to identify perceptual shielding type dust risks and quantify them into perceptual shielding type dust risk values comprises the following steps: based on the image optical characteristic data, the global outline clarity attenuation gradient and the edge texture degradation rate of the current image frame relative to the historical reference frame are extracted; According to the global contour clarity attenuation gradient and the edge texture degradation rate, a visual obscuration dynamic index is determined to represent the overall visibility deterioration trend of the environment; Based on the image optical characteristic data, key navigation markers on the preset flight path of the UAV are identified and tracked; The morphological stability and feature recognizability decay process of the key navigation markers in consecutive image frames are analyzed to generate a navigation feature degradation coefficient; The visual obscuration dynamic index and the navigation feature degradation coefficient are added together, and the result is taken as the perception obscuration type dust risk value.

3. The method of claim 2, wherein, Based on the active probe waveform data, the spatial concentration distribution and the deposition characteristics of the dust are analyzed to determine the deposition erosion type dust risk and quantify it as a deposition erosion type dust risk value, including: Based on the active probe waveform data, the primary echo signal and the secondary echo signal cluster generated by dust particle group scattering are separated; According to the energy attenuation gradient and the spatiotemporal distribution density of the secondary echo signal cluster, the concentration field vortex intensity representing the spatial distribution non-uniformity of the dust is inversely calculated; The waveform broadening characteristics and the frequency spectrum distortion characteristics of the primary echo signal after penetrating the dust environment are synchronously extracted, from which the average particle size trend and the adsorption and coagulation trend of the dust particles are analyzed to generate a particle deposition potential energy parameter; The concentration field vortex intensity and the particle deposition potential energy parameter are multiplied to represent the synergistic enhancement effect, and the product result is taken as the deposition erosion type dust risk value.

4. The method of claim 3, wherein, Based on the medium electromagnetic property data, the interference characteristics of the dust on the electromagnetic signal are analyzed to determine the conductive interference type dust risk and quantify it as a conductive interference type dust risk value, including: Based on the medium electromagnetic property data, the real part drift amount and the imaginary part loss oscillation amplitude of the equivalent permittivity representing the comprehensive influence of the medium on the electromagnetic wave propagation path are analyzed; According to the real part drift amount of the equivalent permittivity, the carrier phase synchronization deviation accumulation trend of the UAV communication navigation signal caused by the sudden change of the propagation environment is deduced; According to the imaginary part loss oscillation amplitude, the irregular fading characteristics of the signal strength caused by the absorption attenuation of the dust medium are tracked and analyzed; Based on the carrier phase synchronization deviation accumulation trend and the irregular fading period and depth, the bit error rate transition risk and the instantaneous interruption probability of the wireless link of each frequency band of the UAV are dynamically predicted; The carrier phase synchronization deviation accumulation trend and the irregular fading characteristics are quantified as a phase instability parameter and a signal attenuation parameter, respectively; The phase instability parameter and the signal attenuation parameter are squared and added together, and the sum value is taken as the conductive interference type dust risk value.

5. The method of claim 4, wherein, According to the dynamic dust risk influence parameter set, the influence degree of the dynamic dust on the flight safety of the UAV is analyzed to generate a comprehensive dynamic risk level of the UAV, including: Based on the dynamic dust risk influence parameter set, the perception obscuration type dust risk value, the deposition erosion type dust risk value, and the conductive interference type dust risk value are extracted; The product of the perception obscuration type dust risk value and the deposition erosion risk value is calculated to obtain a first synergistic risk factor; calculating a product of the perceptual obscuration type dust risk value and the conductive interference risk value to obtain a second synergistic risk factor; calculating a product of the deposition erosion risk value and the conductive interference risk value to obtain a third synergistic risk factor; adding the first synergistic risk factor, the second synergistic risk factor and the third synergistic risk factor to obtain a total synergistic effect value; comparing the total synergistic effect value with a preset risk level threshold table, and outputting the unmanned aerial vehicle comprehensive dynamic risk level according to the comparison result, wherein the risk level threshold table defines the risk level corresponding to different numerical ranges.

6. The method of claim 5, wherein, According to the unmanned aerial vehicle comprehensive dynamic risk level, diagnose the current dust risk composition situation, and generate an unmanned aerial vehicle abnormal diagnosis report, including: According to the unmanned aerial vehicle comprehensive dynamic risk level, analyze the proportion of the perceptual obscuration type dust risk value, the deposition erosion type dust risk value and the conductive interference type dust risk value to diagnose the dominant type and composition of the current dust risk: If the proportion of perceptual obscuration type dust risk exceeds the limit, the auxiliary navigation weight of the visual sensor is enhanced and the switch to the non-optical navigation mode is prompted; If the proportion of deposition erosion type dust risk exceeds the limit, the output power redundancy of the power system is improved and the surface deposition prevention program is activated; If the proportion of conductive interference type dust risk exceeds the limit, switch to the anti-interference communication frequency band and improve the signal transmission power. According to the dominant type and composition of the dust risk, generate the unmanned aerial vehicle abnormal diagnosis report containing the risk level, the numerical value of each risk type and the corresponding measures.

7. The method of claim 6, wherein, The method further comprises: Establishing a dust dynamic correlation model among the perceptual obscuration type dust risk value, the deposition erosion type dust risk value and the conductive interference type dust risk value to trace the induction and enhancement mechanism of high numerical value risk type to other risk types: Tracking the spatial dust concentration distribution corresponding to the high numerical value of the perceptual obscuration type dust risk value, and accordingly deducing the potential area and rate of uneven dust accumulation on the unmanned aerial vehicle body surface, so as to quantify the deposition enhancement amplitude of the deposition erosion type dust risk value; Analyzing the accumulation composition and thickness distribution represented by the high numerical value of the deposition erosion type dust risk value, and accordingly evaluating the change degree of the equivalent dielectric constant of the unmanned aerial vehicle radome caused by the accumulation, so as to quantify the dielectric induction intensity of the conductive interference type dust risk value; Based on the deposition enhancement amplitude and the dielectric induction intensity, coupling correction is performed on the basic dust risk parameter set to generate a comprehensive dust risk parameter set with coupling chain effect.

8. A multi-modal data fusion based unmanned aerial vehicle anomaly diagnosis system, characterized in that, Applied to the method of any one of claims 1-7, comprising: An environment perception module for acquiring unmanned aerial vehicle multi-modal sensor data, and analyzing the dust characteristics in the current environment in real time based on the unmanned aerial vehicle multi-modal sensor data to generate a basic dust risk parameter set; A data fusion module for acquiring unmanned aerial vehicle flight state parameters, taking the unmanned aerial vehicle flight state parameters as dynamic correction factors, and dynamically fusing the basic dust risk parameter set to generate a dynamic dust risk influence parameter set; A risk assessment module is configured to analyze the influence degree of the dynamic dust on the flight safety of the UAV according to the dynamic dust risk influence parameter set, and generate a comprehensive dynamic risk level of the UAV. A decision control module is configured to execute corresponding flight control instructions and system configuration adjustment strategies according to the comprehensive dynamic risk level of the UAV, and generate an abnormal diagnosis report of the UAV.

Images