Unmanned aerial vehicle inspection method based on unmanned aerial vehicle nest

Abstract

The invention discloses an unmanned aerial vehicle routing inspection method based on an unmanned aerial vehicle nest, and particularly relates to the technical field of unmanned aerial vehicle management, and the method comprises the steps of unmanned aerial vehicle communication safety, comprehensive health data collection, unmanned aerial vehicle power system depth diagnosis, in-nest precise vibration diagnosis, and multi-vehicle cooperative routing inspection scheduling optimization. When entering a communication range, a reinforced unmanned aerial vehicle nest network security system is constructed, sensitive data and communication security are guaranteed, full-aircraft health data and vibration signal data are acquired through a built-in sensor, and all-directional data coverage and deep diagnosis of the data are realized through multi-device cooperative acquisition. The unmanned aerial vehicle body fault early warning is realized, task load analysis guarantees the effectiveness of inspection image data through definition calculation and infrared temperature calibration, vibration spectrum analysis is extremely sensitive to early degradation of mechanical parts, early warning can be given out when the fault is still in the bud stage, and a longer early warning window is provided.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) management technology, and more specifically, to a UAV inspection method based on UAV nests. Background Technology

[0002] Unmanned aerial vehicle (UAV) nesting systems have been widely applied in various fields due to their advantages of automation and intelligence. For example, in the field of power line inspection, UAV nesting systems can perform regular inspections of transmission lines, promptly detecting line faults and safety hazards. In the field of security monitoring, the system can conduct real-time monitoring of key areas. In today's rapidly developing technological landscape, UAV nesting systems have gained widespread application in the field of inspection due to their unique advantages.

[0003] With continuous technological advancements and growing application demands, the market size of drone storage facilities is expanding rapidly, and its development prospects are extremely broad. As a "smart base" for drones, drone storage facilities not only provide safe and reliable parking spaces but also enable a series of key functions such as automatic charging, task scheduling, and data processing, greatly improving the efficiency and automation level of inspection operations.

[0004] However, it still has some shortcomings in actual use. For example, as the networking and intelligence of the drone nest system continue to improve, the nest involves a large amount of sensitive information, which poses a risk of cybersecurity threats. Meanwhile, during long-term operation, the condition of the drone itself becomes the key factor affecting the reliability of the next mission. The existing health diagnosis function of the drone nest relies on basic parameter detection and lacks in-depth diagnostic capabilities. The in-nest diagnostic technology cannot fully detect the health status of the drone. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a drone inspection method based on drone nests to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a drone inspection method based on drone nests, comprising the following steps: Step S01: Drone Communication Security: After completing the inspection mission, the drone flies to the drone nest. When entering the communication range, a reinforced drone nest network security system is constructed.

[0007] Step S02: Comprehensive health data collection: After the drone arrives at the preset position of the nest, the main controller of the nest automatically sends a health diagnosis start command to the drone and collects the whole machine's health data and vibration signal data through its built-in sensors.

[0008] Step S03: In-depth diagnosis of UAV power system: including acoustic print analysis sub-step, blade deformation analysis sub-step, fuselage structural integrity analysis sub-step and mission load analysis sub-step, to identify whether there are faults in each component of the UAV.

[0009] Step S04: Precise vibration diagnosis within the nest: This step is used to analyze the vibration signals of the drone within the nest and provide early warning of drone malfunctions.

[0010] Step S05: Multi-drone collaborative inspection scheduling optimization: Based on the preset mapping relationship between health diagnosis results and health levels, the nest main controller obtains the health level of each drone based on the health diagnosis results of each drone, and allocates tasks using a genetic algorithm in combination with the requirements of the next inspection task.

[0011] Preferably, step S01: Unmanned Aerial Vehicle (UAV) communication security specifically includes: After a drone enters the communication range of the nest, the nest sends a wake-up command containing a timestamp and a random number via radio frequency signal. After receiving the command, the drone concatenates the preset hardware unique identifier and dynamic factor in a fixed format to generate a dynamic identity token, which is then fed back to the nest through encrypted transmission technology. The nest verifies the validity of the token. If the verification is successful, a temporary communication link is established. Devices that fail to pass the verification are blacklisted and the nest alarm is triggered. The identity token contains the drone device ID, timestamp signature, and instruction random number hash value; The encrypted transmission technology encrypts the communication data between the data center and the drone, and between the data center and the remote control center; After successful identity authentication, the nest and the drone use elliptic curve cryptography to negotiate a key: the nest generates a random private key and calculates a public key, which is then sent to the drone using asymmetric encryption; the drone generates its own random private key and public key, calculates a shared session key using the nest's public key, and sends back the shared key verification information after encrypting it with the nest's public key; after both parties verify the consistency of the shared key, all subsequent communication data is symmetrically encrypted.

[0012] Preferably, step S02: comprehensive health data collection specifically includes: Once the drone reaches the preset location in the nest, the main controller of the nest automatically sends a health diagnosis start command to the drone. The drone responds to the command, starts a self-test program, and collects full-machine health data through its built-in sensors. The self-test procedure is as follows: An industrial camera array scans the drone and uses visual recognition algorithms to obtain the precise three-dimensional coordinates of the drone's fuselage, arms, propellers, and camera gimbal. The drone motors are controlled to run unloaded according to a preset speed sequence, and the full-frequency audio signals of the motors during operation are acquired through acoustic sensors. Based on the obtained precise three-dimensional coordinates of the propellers, the robotic arm scans each propeller along a predetermined trajectory using a 3D line laser scanner to obtain its high-density three-dimensional point cloud data and generate a three-dimensional point cloud model of the propellers; at the same time, it scans the surface of the propellers using an industrial camera to obtain two-dimensional digital images of the propellers; the robotic arm switches its end effector to an ultrasonic flaw detector and contacts the probe to the root of the propeller to obtain the ultrasonic echo signal of the propeller. The distance data of the detection points on the fuselage is obtained by emitting lasers at key detection points on the main frame of the drone using a laser displacement sensor and receiving the reflected light. Based on the obtained precise three-dimensional coordinates of the camera gimbal, the robotic arm autonomously adjusts the attitude of the UAV gimbal so that its visible light camera is facing the standard optical calibration target set on the inner wall of the nest, and controls the visible light camera to automatically focus on and shoot the target to obtain a digital image of the calibration target; the robotic arm autonomously adjusts the attitude of the UAV gimbal so that its infrared thermal imager is facing the blackbody radiation source target set on the inner wall of the nest, and controls the infrared thermal imager to shoot the blackbody target to obtain an infrared thermal image.

[0013] The vibration signal data specifically refers to: controlling the UAV motor to run unloaded according to a preset speed sequence, and synchronously acquiring the vibration signal of the motor during operation through a vibration sensor.

[0014] Preferably, the voiceprint analysis sub-step specifically includes: After preprocessing the acquired full-band audio signal, Fourier transform is performed on each frame of the signal to convert the time-domain audio signal into a frequency-domain power spectrum. Feature vectors of the acoustic characteristics of the UAV motor are extracted by frequency domain power spectrum. The feature vectors include energy values, spectral centroids, and spectral entropy of a specific frequency band. The feature vector is input into a pre-trained deep learning classification model. The model extracts deep features through multi-layer convolution and pooling operations. After mapping through a fully connected layer, it outputs the probability distribution of each fault category. The probability distribution of each fault category includes the specific fault type, severity level, and diagnostic confidence. The diagnostic conclusion is transmitted to the remote control center through the nest security communication link.

[0015] Preferably, the blade deformation analysis sub-step includes three-dimensional blade deformation analysis, blade surface defect analysis, and internal damage analysis, used to detect the health status of the propeller; 3D deformation analysis of propeller blades: The 3D point cloud model of the propeller is registered with the standard CAD model of the propeller stored in the database. Then, point cloud difference calculation is performed to generate a deformation cloud map and obtain the bending and twisting deformation of the propeller blade in the deformation cloud map area. The blade surface defect analysis uses computer vision algorithms to filter out candidate regions that meet the defect characteristics from the acquired two-dimensional digital images of the propeller. Within the candidate region, the local binary pattern variance is obtained. If the local binary pattern variance is greater than the local binary pattern variance threshold, it indicates that the local texture features are significantly different from the surrounding intact regions, and it is identified as a defect on the surface of the propeller blade. Internal damage analysis: The ultrasonic echo signal of the propeller was obtained and filtered using a digital bandpass filter to obtain the filtered signal; The Hilbert transform is applied to the filtered signal to obtain the analytic signal, and the signal envelope is obtained from the magnitude of the analytic signal. In the signal envelope, the first time point in time where the amplitude exceeds the initial wave threshold is extracted and marked as the initial pulse. In the signal envelope after the initial pulse, the last time point in time where the amplitude exceeds the bottom wave threshold is extracted and marked as the bottom wave. If there is an echo peak in the signal envelope between the initial pulse and the bottom wave, and the amplitude exceeds the abnormal wave threshold, record its time position. Based on the propagation speed of ultrasound in the blade material, the depth corresponding to the abnormal echo peak is calculated. If 0 < depth < maximum blade thickness, then it is determined that there is an internal reflector. In the signal envelope corresponding to the initial pulse moment, the peak amplitude is extracted as the initial wave amplitude, and in the signal envelope corresponding to the bottom wave moment, the peak amplitude is extracted as the bottom wave amplitude. The amplitude ratio is calculated by comparing the amplitude of the bottom wave with the amplitude of the starting wave, and then compared with a preset threshold. If the amplitude ratio is less than the preset threshold, it is determined that there is internal damage that causes sound energy attenuation. Extract the original ultrasonic signal segments 5 μs before and after the abnormal echo peak, and perform Fourier transform on the signal segments to obtain the frequency domain spectrum; The center frequency of the abnormal echo signal is calculated using the spectral centroid algorithm and compared with the preset center frequency of the echo signal at that location. If the absolute value of the difference between the center frequency and the preset center frequency of the echo signal at that location is greater than the allowable center frequency difference, then it is determined that there is micro-crack-like damage.

[0016] Preferably, the sub-step of fuselage structural integrity analysis specifically includes: Extract the distance of the drone's fuselage detection point in its initial state from the database, calculate the difference between the distance of the fuselage detection point and the distance of the corresponding fuselage detection point in the initial state, and compare it with a preset safety threshold. If the absolute value of the difference is greater than the preset safety threshold, it is determined that the structure has undergone deformation beyond the allowable range.

[0017] Preferably, the task load analysis sub-step specifically includes: The clarity of the digital image of the calibration target is extracted and compared with a preset clarity threshold. If the clarity is less than the preset clarity threshold, it is determined to be a stain, and the control system sends a cleaning command to the robot arm. Extract the grayscale values ​​of the pixels in the infrared thermal image, calculate the average pixel grayscale value of the image by dividing the sum of the grayscale values ​​by the total number of pixels, call the current calibration curve parameters built into the infrared thermal imager, substitute the average pixel grayscale value into the curve formula, and calculate the measured temperature value. For each calibration temperature point, the temperature deviation is calculated by measuring the temperature value and comparing it with the actual temperature of the blackbody. If the absolute value of the temperature deviation at any temperature point is greater than the allowable temperature deviation, a temperature calibration curve is automatically generated and transmitted to the infrared thermal imager through the secure communication link between the machine and the instrument to complete the online calibration of the temperature measurement accuracy.

[0018] Preferably, step S04: precise vibration diagnosis within the nest specifically includes: For each new drone model, standard testing procedures are performed inside the facility to collect baseline health vibration data of the drone. Compare the current vibration signal with the baseline healthy vibration data point by point: extract the current power spectral density at each frequency point, and obtain the power spectral density difference at each frequency point by the difference between the current power spectral density and the baseline power spectral density, and generate the differential spectrum; Output fault anomaly judgment: If the energy integral of the differential spectrum in the 1000-2000Hz band is greater than the preset value, and is accompanied by discrete spur peaks in the 500-800Hz range, it is judged as early pitting wear of the UAV bearing, and the fault location is sent to the remote control center through a secure link.

[0019] Preferably, step S05: multi-machine collaborative scheduling optimization specifically includes: Based on the preset mapping relationship between health diagnosis results and health levels, the main controller of the drone will classify each drone into level A, level B, and level C according to the health diagnosis results of each drone. The main controller of the drone nest receives the next inspection task instruction issued by the remote control center, parses the task parameters including the inspection area, key area inspection requirements, and auxiliary area inspection requirements, and uses a genetic algorithm to allocate tasks based on the requirements of the next inspection task: Class A drones are responsible for key area inspections, Class B drones are responsible for auxiliary area inspections, and Class C drones are marked as needing maintenance and are prohibited from participating in the task. Based on the location of the drone nest, the mission area, and the health status of the drone, the shortest path is planned, and the drone with the best health status in Class A is set as the lead drone. The task allocation results and flight paths are sent to each UAV through the nest's secure communication link. After the lead UAV confirms receipt, it sends a signal to the nest indicating that the cluster task is ready to be completed and awaits inspection instructions.

[0020] The technical effects and advantages of this invention are as follows: 1. This invention provides a drone inspection method based on a drone nest. After the drone enters the nest's communication range, the nest sends a wake-up command containing a timestamp and a random number via radio frequency signal. After receiving the command, the drone concatenates a preset hardware unique identifier and a dynamic factor in a fixed format to generate a dynamic identity token, which is then transmitted back to the nest via encrypted transmission technology. The nest verifies the validity of the token. If successful, a temporary communication link is established. Devices that fail verification are blacklisted and a nest alarm is triggered. After successful identity authentication, the nest generates a random private key and calculates a public key, which is then sent to the drone via asymmetric encryption. The drone generates its own random private key and public key, calculates a shared session key using the nest's public key, and encrypts it with the nest's public key before sending back the shared key verification information. After both parties verify the consistency of the shared key, all subsequent communication data is symmetrically encrypted and transmitted through the drone-nest secure communication link, thus realizing the construction of a highly secure network protection system and ensuring the security of sensitive data and communication. 2. This invention provides a drone inspection method based on a drone nest. When the drone arrives at a preset position in the nest, the nest's main controller automatically sends a health diagnosis start command to the drone. The drone's built-in sensors collect overall health data and vibration signal data. The acquired full-band audio signals are analyzed using acoustic signatures to output the probability distribution of each fault category. The acquired 3D point cloud model, 2D digital image, and ultrasonic echo signal of the propeller are analyzed using blade deformation to output the propeller's health status. The distances to the fuselage detection points are analyzed using fuselage structural integrity to output whether structural deformation exceeds the allowable range. Clarity is calculated using calibrated target digital images. The control system sends a cleaning command to the robotic arm via infrared... The average pixel grayscale value of the external thermal image is calculated. The current calibration curve parameters built into the infrared thermal imager are called, and the average pixel grayscale value is substituted into the curve formula to calculate the measured temperature value. The temperature deviation is calculated to complete the online calibration of the temperature measurement accuracy. By comparing the current vibration signal with the benchmark healthy vibration data point by point, the fault anomaly judgment is output. By collecting UAV health data through multi-device collaboration, comprehensive data coverage is achieved, and further in-depth diagnosis of the data is carried out to realize early warning of UAV fuselage faults. The mission load analysis ensures the effectiveness of the inspection image data through sharpness calculation and infrared temperature calibration. Vibration spectrum analysis is extremely sensitive to the early degradation of mechanical components and can issue warnings when the fault is still in its infancy, providing a longer warning window. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a drone inspection method based on a drone nest according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 As shown, this invention provides a drone inspection method based on drone nests, comprising the following steps: Step S01: UAV communication security: After completing the inspection task, the UAV flies to the UAV nest and constructs a reinforced UAV nest network security system when entering the communication range.

[0024] In one possible design, step S01: UAV communication security specifically includes: After a drone enters the communication range of the nest, the nest sends a wake-up command containing a timestamp and a random number via radio frequency signal. After receiving the command, the drone concatenates the preset hardware unique identifier and dynamic factor in a fixed format to generate a dynamic identity token, which is then fed back to the nest through encrypted transmission technology. The nest verifies the validity of the token. If the verification is successful, a temporary communication link is established. Devices that fail to pass the verification are blacklisted and the nest alarm is triggered. The identity token contains the drone device ID, timestamp signature, and instruction random number hash value; The encrypted transmission technology encrypts the communication data between the data center and the drone, and between the data center and the remote control center; After successful identity authentication, the nest and the drone use elliptic curve cryptography to negotiate a key: the nest generates a random private key and calculates a public key, which is then sent to the drone using asymmetric encryption; the drone generates its own random private key and public key, calculates a shared session key using the nest's public key, and sends back the shared key verification information after encrypting it with the nest's public key; after both parties verify the consistency of the shared key, all subsequent communication data is symmetrically encrypted.

[0025] As an explanation: Dynamic factor extraction: After the drone receives the wake-up command, it extracts the timestamp and random number in the wake-up command as dynamic factors to ensure that the token generated for each authentication is different.

[0026] Step S02: Comprehensive health data collection: When the drone arrives at the preset position of the nest, the main controller of the nest automatically sends a health diagnosis start command to the drone and collects the whole machine's health data and vibration signal data through its built-in sensors.

[0027] In one possible design, step S02: comprehensive health data collection specifically includes: Once the drone reaches the preset location in the nest, the main controller of the nest automatically sends a health diagnosis start command to the drone. The drone responds to the command, starts a self-test program, and collects full-machine health data through its built-in sensors. The self-test procedure is as follows: An industrial camera array scans the drone and uses visual recognition algorithms to obtain the precise three-dimensional coordinates of the drone's fuselage, arms, propellers, and camera gimbal. The drone motors are controlled to run unloaded according to a preset speed sequence, and the full-frequency audio signals of the motors during operation are acquired through acoustic sensors. Based on the obtained precise three-dimensional coordinates of the propellers, the robotic arm scans each propeller along a predetermined trajectory using a 3D line laser scanner to obtain its high-density three-dimensional point cloud data and generate a three-dimensional point cloud model of the propellers; at the same time, it scans the surface of the propellers using an industrial camera to obtain two-dimensional digital images of the propellers; the robotic arm switches its end effector to an ultrasonic flaw detector and contacts the probe to the root of the propeller to obtain the ultrasonic echo signal of the propeller. The distance data of the detection points on the fuselage is obtained by emitting lasers at key detection points on the main frame of the drone using a laser displacement sensor and receiving the reflected light. Based on the obtained precise three-dimensional coordinates of the camera gimbal, the robotic arm autonomously adjusts the attitude of the UAV gimbal so that its visible light camera is facing the standard optical calibration target set on the inner wall of the nest, and controls the visible light camera to automatically focus on and shoot the target to obtain a digital image of the calibration target; the robotic arm autonomously adjusts the attitude of the UAV gimbal so that its infrared thermal imager is facing the blackbody radiation source target set on the inner wall of the nest, and controls the infrared thermal imager to shoot the blackbody target to obtain an infrared thermal image.

[0028] The vibration signal data specifically refers to: controlling the UAV motor to run unloaded according to a preset speed sequence, and synchronously acquiring the vibration signal of the motor during operation through a vibration sensor.

[0029] In this embodiment, it should be specifically explained that the industrial camera array scans the drone and obtains the precise three-dimensional coordinates of the drone's fuselage, arms, propellers, and camera gimbal through a visual recognition algorithm. Specifically: An industrial camera array captures images of the drone at the parking spot from at least three different angles, obtaining a multi-view two-dimensional digital image set that includes the drone as a whole and key components (drone fuselage, arms, propellers, camera gimbal); The acquired two-dimensional digital image is input into a visual recognition algorithm, and based on a pre-trained component recognition model, the outline and feature points of the key component are identified from the image. By matching and triangulating the features of two-dimensional digital images of the same component from different cameras, the precise three-dimensional coordinates of the key component in the nest world coordinate system are obtained. The generated precise three-dimensional coordinate mapping table of the component is output to the robot arm control center to execute the positioning reference.

[0030] Step S03: In-depth diagnosis of the UAV power system: including acoustic print analysis sub-step, blade deformation analysis sub-step, fuselage structural integrity analysis sub-step and mission load analysis sub-step, to identify whether there are faults in each component of the UAV.

[0031] In one possible design, step S03: in-depth diagnosis of the UAV power system specifically includes: Voiceprint analysis sub-steps: After preprocessing the acquired full-band audio signal, Fourier transform is performed on each frame of the signal to convert the time-domain audio signal into a frequency-domain power spectrum. Feature vectors of the acoustic characteristics of the UAV motor are extracted by frequency domain power spectrum. The feature vectors include energy values, spectral centroids, and spectral entropy of a specific frequency band. The feature vector is input into a pre-trained deep learning classification model. The model extracts deep features through multi-layer convolution and pooling operations. After mapping through a fully connected layer, it outputs the probability distribution of each fault category. The probability distribution of each fault category includes the specific fault type, severity level, and diagnostic confidence. The diagnostic conclusion is transmitted to the remote control center through the nest security communication link. The blade deformation analysis sub-step includes three-dimensional blade deformation analysis, blade surface defect analysis, and internal damage analysis, which are used to detect the health status of the propeller. 3D deformation analysis of propeller blades: The 3D point cloud model of the propeller is registered with the standard CAD model of the propeller stored in the database. Then, point cloud difference calculation is performed to generate a deformation cloud map and obtain the bending and twisting deformation of the propeller blade in the deformation cloud map area. The blade surface defect analysis uses computer vision algorithms to filter out candidate regions that meet the defect characteristics from the acquired two-dimensional digital images of the propeller. Within the candidate region, the local binary pattern variance is obtained. If the local binary pattern variance is greater than the local binary pattern variance threshold, it indicates that the local texture features are significantly different from the surrounding intact regions, and it is identified as a defect on the surface of the propeller blade. Internal damage analysis: The ultrasonic echo signal of the propeller was obtained and filtered using a digital bandpass filter to obtain the filtered signal; The Hilbert transform is applied to the filtered signal to obtain the analytic signal, and the signal envelope is obtained from the magnitude of the analytic signal. In the signal envelope, the first time point in time where the amplitude exceeds the initial wave threshold is extracted and marked as the initial pulse. In the signal envelope after the initial pulse, the last time point in time where the amplitude exceeds the bottom wave threshold is extracted and marked as the bottom wave. If there is an echo peak in the signal envelope between the initial pulse and the bottom wave, and the amplitude exceeds the abnormal wave threshold, record its time position. Based on the propagation speed of ultrasound in the blade material, the depth corresponding to the abnormal echo peak is calculated. If 0 < depth < maximum blade thickness, it is determined that there is an internal reflector (such as delamination or crack interface). As an explanation: the initial wave threshold is set at 50% of the peak value of the initial wave in a healthy sample, the bottom wave threshold is set at 30% of the peak value of the bottom wave in a healthy sample, and the values ​​are obtained by statistical analysis of historical health data in the machine nest database. The abnormal wave threshold is set at twice the peak value of the normal clutter in a healthy sample. In the signal envelope corresponding to the initial pulse moment, the peak amplitude is extracted as the initial wave amplitude, and in the signal envelope corresponding to the bottom wave moment, the peak amplitude is extracted as the bottom wave amplitude. The amplitude ratio is calculated by comparing the amplitude of the bottom wave with the amplitude of the starting wave, and then compared with a preset threshold. If the amplitude ratio is less than the preset threshold, it is determined that there is internal damage (such as delamination or pores) that causes sound energy attenuation. Extract the original ultrasonic signal segments 5 μs before and after the abnormal echo peak, and perform Fourier transform on the signal segments to obtain the frequency domain spectrum; The center frequency of the abnormal echo signal is calculated using the spectral centroid algorithm and compared with the preset center frequency of the echo signal at that location. If the absolute value of the difference between the center frequency and the preset center frequency of the echo signal at that location is greater than the allowable center frequency difference, it is determined that there is micro-crack-like damage. As an explanation: Cracks cause ultrasonic wave scattering, resulting in a significant downward shift in the center frequency; Sub-steps for fuselage structural integrity analysis: Extract the distance of the drone's fuselage detection point in its initial state from the database, calculate the difference between the distance of the fuselage detection point and the distance of the corresponding fuselage detection point in the initial state, and compare it with a preset safety threshold. If the absolute value of the difference is greater than the preset safety threshold, it is determined that the structure has undergone deformation beyond the allowable range. Task payload analysis sub-steps: The clarity of the digital image of the calibration target is extracted and compared with a preset clarity threshold. If the clarity is less than the preset clarity threshold, it is determined to be a stain, and the control system sends a cleaning command to the robot arm. Extract the grayscale values ​​of the pixels in the infrared thermal image, calculate the average pixel grayscale value of the image by dividing the sum of the grayscale values ​​by the total number of pixels, call the current calibration curve parameters built into the infrared thermal imager, substitute the average pixel grayscale value into the curve formula, and calculate the measured temperature value. For each calibration temperature point, the temperature deviation is calculated by measuring the temperature value and comparing it with the actual temperature of the blackbody. If the absolute value of the temperature deviation at any temperature point is greater than the allowable temperature deviation, a temperature calibration curve is automatically generated and transmitted to the infrared thermal imager through the secure communication link between the machine and the instrument to complete the online calibration of the temperature measurement accuracy.

[0032] In this embodiment, it should be specifically noted that the formula for calculating the depth corresponding to the abnormal echo peak is as follows: in, This is represented by the depth corresponding to the abnormal echo peak. Expressed as propagation speed, This is represented as the abnormal echo time. This is represented as the initial pulse time.

[0033] Step S04: Precise vibration diagnosis within the nest: This step is used to analyze the vibration signals of the drone within the nest and provide early warning of drone malfunctions.

[0034] In one possible design, step S04: precise vibration diagnosis within the nest specifically includes: For each new drone model, standard testing procedures are performed inside the facility to collect baseline health vibration data of the drone. Compare the current vibration signal with the baseline healthy vibration data point by point: extract the current power spectral density at each frequency point, and obtain the power spectral density difference at each frequency point by the difference between the current power spectral density and the baseline power spectral density, and generate the differential spectrum; Output fault anomaly judgment: If the energy integral of the differential spectrum in the 1000-2000Hz band is greater than the preset value, and is accompanied by discrete spur peaks in the 500-800Hz range, it is judged as early pitting wear of the UAV bearing, and the fault location is sent to the remote control center through a secure link.

[0035] In this embodiment, it should be specifically noted that the formula for calculating the energy integral of the differential spectrum in the 1000-2000Hz frequency band is as follows: in, Expressed as the power spectral density difference in the 1000-2000Hz frequency band. This is expressed as the integral operation of the power spectral density difference in the frequency band of 1000-2000Hz, where S represents the energy integral and f represents the frequency variable.

[0036] Step S05: Multi-drone collaborative inspection scheduling optimization: Based on the preset mapping relationship between health diagnosis results and health levels, the nest main controller obtains the health level of each drone based on the health diagnosis results of each drone, and allocates tasks using a genetic algorithm in combination with the requirements of the next inspection task.

[0037] In one possible design, step S05: multi-machine collaborative scheduling optimization specifically includes: Based on the preset mapping relationship between health diagnosis results and health levels, the main controller of the drone will classify each drone into level A, level B, and level C according to the health diagnosis results of each drone. The main controller of the drone nest receives the next inspection task instruction issued by the remote control center, parses the task parameters including the inspection area, key area inspection requirements, and auxiliary area inspection requirements, and uses a genetic algorithm to allocate tasks based on the requirements of the next inspection task: Class A drones are responsible for key area inspections, Class B drones are responsible for auxiliary area inspections, and Class C drones are marked as needing maintenance and are prohibited from participating in the task. Based on the location of the drone nest, the mission area, and the health status of the drone, the shortest path is planned, and the drone with the best health status in Class A is set as the lead drone. The task allocation results and flight paths are sent to each UAV through the nest's secure communication link. After the lead UAV confirms receipt, it sends a signal to the nest indicating that the cluster task is ready to be completed and awaits inspection instructions.

[0038] In this embodiment, it should be specifically explained that after the UAV enters the nest communication range, the nest sends a wake-up command containing a timestamp and a random number via radio frequency signal. After receiving the command, the UAV concatenates a preset hardware unique identifier and a dynamic factor in a fixed format to generate a dynamic identity token, which is then fed back to the nest via encrypted transmission technology. The nest verifies the validity of the token. If successful, a temporary communication link is established. Devices that fail verification are blacklisted and a nest alarm is triggered. After successful identity authentication, the nest generates a random private key and calculates a public key, which is then sent to the UAV via asymmetric encryption. The UAV generates its own random private key and public key, calculates a shared session key using the nest's public key, and encrypts it with the nest's public key before feeding back the shared key verification information. After both parties verify the consistency of the shared key, all subsequent communication data is symmetrically encrypted and transmitted through the UAV-nest secure communication link, thus realizing the construction of a highly secure network protection system and ensuring the security of sensitive data and communication. When the UAV reaches the preset position in the nest, the main controller of the nest automatically sends a health diagnosis start command to the UAV. The UAV's built-in sensors collect overall health data and vibration signal data. The acquired full-band audio signals are analyzed using acoustic signatures to output the probability distribution of each fault category. The acquired 3D point cloud model, 2D digital image, and ultrasonic echo signal of the propeller are analyzed using blade deformation to output the propeller's health status. The distances to the fuselage detection points are analyzed using fuselage structural integrity to output whether structural deformation exceeds the allowable range. The clarity is calculated using the calibrated target digital image. The control system sends a cleaning command to the robotic arm and calculates the average image value using infrared thermal images. The system calculates the measured temperature value by calling the current calibration curve parameters built into the infrared thermal imager, substituting the average pixel gray value into the curve formula, calculating the temperature deviation, and completing the online calibration of temperature measurement accuracy. By comparing the current vibration signal with the benchmark healthy vibration data point by point, it outputs the fault anomaly judgment. Through the collaborative collection of UAV health data by multiple devices, it achieves comprehensive data coverage and further performs in-depth diagnosis of the data, realizing early warning of UAV fuselage faults. The mission load analysis ensures the effectiveness of inspection image data through sharpness calculation and infrared temperature calibration. Vibration spectrum analysis is extremely sensitive to the early degradation of mechanical components and can issue warnings when the fault is still in its infancy, providing a longer warning window.

[0039] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A drone inspection method based on drone nests, characterized in that, Includes the following steps: Step S01: Drone Communication Security: After completing the inspection mission, the drone flies to the drone nest. When entering the communication range, a reinforced drone nest network security system is constructed. Step S02: Comprehensive health data collection: After the drone arrives at the preset position of the nest, the main controller of the nest automatically sends a health diagnosis start command to the drone and collects the whole machine's health data and vibration signal data through its built-in sensors; Step S03: In-depth diagnosis of UAV power system: including acoustic print analysis sub-step, blade deformation analysis sub-step, fuselage structural integrity analysis sub-step and mission load analysis sub-step, to identify whether there are faults in each component of the UAV; Step S04: Precise vibration diagnosis within the nest: This step is used to analyze the vibration signals of the drone within the nest and provide early warning of drone malfunctions. Step S05: Multi-drone collaborative inspection scheduling optimization: Based on the preset mapping relationship between health diagnosis results and health levels, the nest main controller obtains the health level of each drone based on the health diagnosis results of each drone, and allocates tasks using a genetic algorithm in combination with the requirements of the next inspection task.

2. The UAV inspection method based on UAV nests according to claim 1, characterized in that: Step S01: Unmanned aerial vehicle (UAV) communication security specifically includes: After a drone enters the communication range of the nest, the nest sends a wake-up command containing a timestamp and a random number via radio frequency signal. After receiving the command, the drone concatenates the preset hardware unique identifier and dynamic factor in a fixed format to generate a dynamic identity token, which is then fed back to the nest through encrypted transmission technology. The nest verifies the validity of the token. If the verification is successful, a temporary communication link is established. Devices that fail to pass the verification are blacklisted and the nest alarm is triggered. The identity token contains the drone device ID, timestamp signature, and instruction random number hash value; The encrypted transmission technology encrypts the communication data between the data center and the drone, and between the data center and the remote control center; After successful identity authentication, the nest and the drone use elliptic curve cryptography to negotiate a key: the nest generates a random private key and calculates a public key, which is then sent to the drone using asymmetric encryption; the drone generates its own random private key and public key, calculates a shared session key using the nest's public key, and sends back the shared key verification information after encrypting it with the nest's public key; after both parties verify the consistency of the shared key, all subsequent communication data is symmetrically encrypted.

3. The UAV inspection method based on UAV nests according to claim 1, characterized in that: Step S02: Comprehensive health data collection specifically includes: Once the drone reaches the preset location in the nest, the main controller of the nest automatically sends a health diagnosis start command to the drone. The drone responds to the command, starts a self-test program, and collects full-machine health data through its built-in sensors. The self-test procedure is as follows: An industrial camera array scans the drone and uses visual recognition algorithms to obtain the precise three-dimensional coordinates of the drone's fuselage, arms, propellers, and camera gimbal. The drone motors are controlled to run unloaded according to a preset speed sequence, and the full-frequency audio signals of the motors during operation are acquired through acoustic sensors. Based on the obtained precise three-dimensional coordinates of the propellers, the robotic arm scans each propeller along a predetermined trajectory using a 3D line laser scanner to obtain its high-density three-dimensional point cloud data and generate a three-dimensional point cloud model of the propellers; at the same time, it scans the surface of the propellers using an industrial camera to obtain two-dimensional digital images of the propellers; the robotic arm switches its end effector to an ultrasonic flaw detector and contacts the probe to the root of the propeller to obtain the ultrasonic echo signal of the propeller. The distance data of the detection points on the fuselage is obtained by emitting lasers at key detection points on the main frame of the drone using a laser displacement sensor and receiving the reflected light. Based on the obtained precise three-dimensional coordinates of the camera gimbal, the robotic arm autonomously adjusts the attitude of the UAV gimbal so that its visible light camera is facing the standard optical calibration target set on the inner wall of the nest, and controls the visible light camera to automatically focus on and shoot the target to obtain a digital image of the calibration target; the robotic arm autonomously adjusts the attitude of the UAV gimbal so that its infrared thermal imager is facing the blackbody radiation source target set on the inner wall of the nest, and controls the infrared thermal imager to shoot the blackbody target to obtain an infrared thermal image. The vibration signal data specifically refers to: controlling the UAV motor to run unloaded according to a preset speed sequence, and synchronously acquiring the vibration signal of the motor during operation through a vibration sensor.

4. The UAV inspection method based on UAV nests according to claim 1, characterized in that: The specific steps of the voiceprint analysis sub-step are as follows: After preprocessing the acquired full-band audio signal, Fourier transform is performed on each frame of the signal to convert the time-domain audio signal into a frequency-domain power spectrum. Feature vectors of the acoustic characteristics of the UAV motor are extracted by frequency domain power spectrum. The feature vectors include energy values, spectral centroids, and spectral entropy of a specific frequency band. The feature vector is input into a pre-trained deep learning classification model. The model extracts deep features through multi-layer convolution and pooling operations. After mapping through a fully connected layer, it outputs the probability distribution of each fault category. The probability distribution of each fault category includes the specific fault type, severity level, and diagnostic confidence. The diagnostic conclusion is transmitted to the remote control center through the nest security communication link.

5. The UAV inspection method based on UAV nests according to claim 1, characterized in that: The blade deformation analysis sub-step includes three-dimensional blade deformation analysis, blade surface defect analysis, and internal damage analysis, which are used to detect the health status of the propeller. 3D deformation analysis of propeller blades: The 3D point cloud model of the propeller is registered with the standard CAD model of the propeller stored in the database. Then, point cloud difference calculation is performed to generate a deformation cloud map and obtain the bending and twisting deformation of the propeller blade in the deformation cloud map area. The blade surface defect analysis uses computer vision algorithms to filter out candidate regions that meet the defect characteristics from the acquired two-dimensional digital images of the propeller. Within the candidate region, the local binary pattern variance is obtained. If the local binary pattern variance is greater than the local binary pattern variance threshold, it indicates that the local texture features are significantly different from the surrounding intact regions, and it is identified as a defect on the surface of the propeller blade. Internal damage analysis: The ultrasonic echo signal of the propeller was obtained and filtered using a digital bandpass filter to obtain the filtered signal; The Hilbert transform is applied to the filtered signal to obtain the analytic signal, and the signal envelope is obtained from the magnitude of the analytic signal. In the signal envelope, the first time point in time where the amplitude exceeds the initial wave threshold is extracted and marked as the initial pulse. In the signal envelope after the initial pulse, the last time point in time where the amplitude exceeds the bottom wave threshold is extracted and marked as the bottom wave. If there is an echo peak in the signal envelope between the initial pulse and the bottom wave, and the amplitude exceeds the abnormal wave threshold, record its time position. Based on the propagation speed of ultrasound in the blade material, the depth corresponding to the abnormal echo peak is calculated. If 0 < depth < maximum blade thickness, then it is determined that there is an internal reflector. In the signal envelope corresponding to the initial pulse moment, the peak amplitude is extracted as the initial wave amplitude, and in the signal envelope corresponding to the bottom wave moment, the peak amplitude is extracted as the bottom wave amplitude. The amplitude ratio is calculated by comparing the amplitude of the bottom wave with the amplitude of the starting wave, and then compared with a preset threshold. If the amplitude ratio is less than the preset threshold, it is determined that there is internal damage that causes sound energy attenuation. Extract the original ultrasonic signal segments 5 μs before and after the abnormal echo peak, and perform Fourier transform on the signal segments to obtain the frequency domain spectrum; The center frequency of the abnormal echo signal is calculated using the spectral centroid algorithm and compared with the preset center frequency of the echo signal at that location. If the absolute value of the difference between the center frequency and the preset center frequency of the echo signal at that location is greater than the allowable center frequency difference, then it is determined that there is micro-crack-like damage.

6. The UAV inspection method based on UAV nests according to claim 1, characterized in that: The specific sub-step for analyzing the fuselage structural integrity is as follows: Extract the distance of the drone's fuselage detection point in its initial state from the database, calculate the difference between the distance of the fuselage detection point and the distance of the corresponding fuselage detection point in the initial state, and compare it with a preset safety threshold. If the absolute value of the difference is greater than the preset safety threshold, it is determined that the structure has undergone deformation beyond the allowable range.

7. The UAV inspection method based on UAV nests according to claim 1, characterized in that: The specific steps of the task load analysis sub-step are as follows: The clarity of the digital image of the calibration target is extracted and compared with a preset clarity threshold. If the clarity is less than the preset clarity threshold, it is determined to be a stain, and the control system sends a cleaning command to the robot arm. Extract the grayscale values ​​of the pixels in the infrared thermal image, calculate the average pixel grayscale value of the image by dividing the sum of the grayscale values ​​by the total number of pixels, call the current calibration curve parameters built into the infrared thermal imager, substitute the average pixel grayscale value into the curve formula, and calculate the measured temperature value. For each calibration temperature point, the temperature deviation is calculated by measuring the temperature value and comparing it with the actual temperature of the blackbody. If the absolute value of the temperature deviation at any temperature point is greater than the allowable temperature deviation, a temperature calibration curve is automatically generated and transmitted to the infrared thermal imager through the secure communication link between the machine and the instrument to complete the online calibration of the temperature measurement accuracy.

8. The UAV inspection method based on UAV nests according to claim 1, characterized in that: Step S04: Precise vibration diagnosis within the nest specifically involves: For each new drone model, standard testing procedures are performed inside the facility to collect baseline health vibration data of the drone. Compare the current vibration signal with the baseline healthy vibration data point by point: extract the current power spectral density at each frequency point, and obtain the power spectral density difference at each frequency point by the difference between the current power spectral density and the baseline power spectral density, and generate the differential spectrum; Output fault anomaly judgment: If the energy integral of the differential spectrum in the 1000-2000Hz band is greater than the preset value, and is accompanied by discrete spur peaks in the 500-800Hz range, it is judged as early pitting wear of the UAV bearing, and the fault location is sent to the remote control center through a secure link.

9. The UAV inspection method based on UAV nests according to claim 1, characterized in that: Step S05: Multi-machine collaborative scheduling optimization specifically includes: Based on the preset mapping relationship between health diagnosis results and health levels, the main controller of the drone will classify each drone into level A, level B, and level C according to the health diagnosis results of each drone. The main controller of the drone nest receives the next inspection task instruction issued by the remote control center, parses the task parameters including the inspection area, key area inspection requirements, and auxiliary area inspection requirements, and uses a genetic algorithm to allocate tasks based on the requirements of the next inspection task: Class A drones are responsible for key area inspections, Class B drones are responsible for auxiliary area inspections, and Class C drones are marked as needing maintenance and are prohibited from participating in the task. Based on the location of the drone nest, the mission area, and the health status of the drone, the shortest path is planned, and the drone with the best health status in Class A is set as the lead drone. The task allocation results and flight paths are sent to each UAV through the nest's secure communication link. After the lead UAV confirms receipt, it sends a signal to the nest indicating that the cluster task is ready to be completed and awaits inspection instructions.

Images