Shipborne unmanned aerial vehicle recovery method and system based on visual navigation and six-degree-of-freedom compensation

Abstract

The invention relates to a ship-borne unmanned aerial vehicle recycling method and system based on visual navigation and six-degree-of-freedom compensation. The method comprises the following steps: extracting infrared, polarization and fluorescence frequency domain features and fusing to obtain an anti-shielding composite descriptor; constructing a ship dynamics differential model, discretizing the ship dynamics differential model by adopting an explicit multi-step method, and predicting six-degree-of-freedom disturbance data of the ship; a vision and inertia combined cost function is constructed, a vision residual item is calculated, the continuity of the motion track of the unmanned aerial vehicle is restrained, and the optimized real-time pose of the unmanned aerial vehicle is obtained; decomposing the six-degree-of-freedom disturbance data into a body coordinate system compensation amount, setting a backstepping sliding mode control law containing feed-forward compensation, and generating an anti-disturbance trajectory instruction; and air data are collected in real time, the rotor rotating speed is corrected according to the air data, the electromagnetic adsorption force is calculated, and the ship-borne unmanned aerial vehicle is recycled. And the navigation stability under complex sea conditions is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of shipborne UAV recovery, and in particular to a shipborne UAV recovery method and system based on visual navigation and six-degree-of-freedom compensation. Background Art

[0002] Shipborne drone recovery refers to the technology used to safely recover a drone from the deck of a moving or swaying ship, with important applications in areas such as ocean monitoring and maritime rescue. Visual navigation utilizes visual sensors such as cameras to acquire images of the surrounding environment, and uses image processing and analysis to achieve drone positioning and navigation. This technology provides critical information for drone recovery, such as target identification and pose estimation. Six-degree-of-freedom compensation compensates for the six-degree-of-freedom motion (translation and rotation along three coordinate axes) caused by wind and waves at sea. This compensates for the impact of the ship's motion on recovery.

[0003] Under complex sea conditions, shipborne drone recovery faces the problem that the ship will produce six-degree-of-freedom motion due to factors such as wave impact, causing the deck to shake continuously, making it difficult for the drone to accurately aim at the recovery area and land stably; on the other hand, environmental interference such as wave spray and shadows caused by sunlight can easily cause partial occlusion or loss of deck marking features in visual navigation, resulting in the drone being unable to reliably obtain its own position information relative to the deck.

[0004] Therefore, traditional ship-borne drone recovery technology often results in defects in the collected images and unpredictable ship movements due to environmental interference such as wave impact, wave spray, and sunlight exposure, thus affecting the accuracy and reliability of drone recovery. Summary of the Invention

[0005] Based on this, in order to solve the above technical problems, a shipborne UAV recovery method and system based on visual navigation and six-degree-of-freedom compensation is provided, which can enhance the navigation stability under complex sea conditions and meet the recovery of shipborne UAVs under harsh conditions such as high sea conditions.

[0006] A shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation, the method comprising:

[0007] A multispectral camera mounted on a drone is used to capture a deck mark image containing concentric ring-shaped optical marks. Infrared, polarization, and fluorescence frequency domain features of the marked area in the deck mark image are extracted and fused to obtain an anti-occlusion composite descriptor.

[0008] Acquiring the six-degree-of-freedom motion parameters of the ship in real time, constructing a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretizing the ship dynamics differential model using an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship;

[0009] Determine the Lie group representation of the drone's pose, obtain the weight coefficients of the set visual term and inertial term, and construct a joint visual and inertial cost function; based on the cost function, calculate the visual residual term and constrain the continuity of the drone's motion trajectory based on the anti-occlusion composite descriptor and six-degree-of-freedom perturbation data, and obtain the optimized real-time pose of the drone;

[0010] Decomposing the six-degree-of-freedom disturbance data into compensation amounts for the body coordinate system, setting a backstepping sliding mode control law with feedforward compensation, and generating an anti-disturbance trajectory instruction;

[0011] Air data is collected in real time, the rotor speed is corrected according to the air data, the electromagnetic attraction force is calculated, and the ship-borne drone is recovered based on the corrected rotor speed, electromagnetic attraction force, real-time posture of the drone, and anti-disturbance trajectory instructions.

[0012] In one embodiment, a multispectral camera mounted on a drone is used to collect deck marking images, including:

[0013] determining an image acquisition frequency, wherein the multispectral camera carried by the drone acquires images of the deck markings at the image acquisition frequency;

[0014] The multispectral camera includes visible light, infrared, and polarization sensors; the deck mark image contains several groups of concentric ring-shaped optical marks, and each group of marks includes an infrared reflection layer, a polarization coding layer, and a fluorescent texture layer from the inside to the outside.

[0015] In one embodiment, infrared, polarization, and fluorescence frequency domain features of the marked area in the deck mark image are extracted and fused to obtain an anti-occlusion composite descriptor, including:

[0016] Extracting the spiral gradient features of the marked area in the deck mark image, performing radial gradient analysis on the infrared reflective layer, and calculating the polar coordinate distribution of the feature point set to obtain the infrared forward gradient features;

[0017] Based on the polarization coding layer, generating optical features through multi-frequency modulation, defining a polarization layer azimuth coding function according to the optical features, and obtaining polarization feature coding according to the polarization layer azimuth coding function;

[0018] Performing Log-Gabor filtering on the fluorescence frequency domain features to extract frequency domain feature vectors;

[0019] The infrared input gradient feature, polarization feature encoding, and frequency domain feature vector are cross-modally correlated to obtain an anti-occlusion composite descriptor.

[0020] In one embodiment, six-degree-of-freedom motion parameters of a ship are acquired in real time, a ship dynamics differential model is constructed based on the six-degree-of-freedom motion parameters, and the ship dynamics differential model is discretized using an explicit multi-step method to predict six-degree-of-freedom disturbance data of the ship, including:

[0021] The ship's six-degree-of-freedom motion parameters are acquired in real time by using the ship's onboard IMU, and a ship dynamics differential model in a non-inertial system is constructed based on the six-degree-of-freedom motion parameters; the ship dynamics differential model is used to describe the six-degree-of-freedom motion coupling relationship;

[0022] Determining a spatial coordinate system corresponding to the infrared, polarization, and fluorescence frequency domain features of the mark area in the extracted deck mark image, and inputting the six-degree-of-freedom motion parameters of the ship into the ship dynamics differential model using the spatial coordinate system as an initial reference;

[0023] The ship dynamics differential model is discretized by an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship.

[0024] In one embodiment, determining the Lie group representation of the drone's posture, obtaining the weight coefficients of the set visual term and inertial term, and constructing a joint visual and inertial cost function include:

[0025] Determine the Lie group representation of the drone's posture, set a visual term and an inertial term, and set corresponding weight coefficients for the visual term and the inertial term respectively;

[0026] Based on the Lie group representation, visual term, inertial term, and weight coefficient, a joint vision and inertial cost function is constructed.

[0027] In one embodiment, based on the cost function, according to the anti-occlusion composite descriptor and the six-degree-of-freedom disturbance data, the visual residual term is calculated and the continuity of the UAV motion trajectory is constrained to obtain the optimized UAV real-time pose, including:

[0028] Performing projection matching constraint pose estimation on the anti-occlusion composite descriptor, and calculating a visual residual term through the visual term based on the cost function;

[0029] The predicted compensation amount of the ship motion is determined according to the six-degree-of-freedom disturbance data, Lie group calculation is performed based on the cost function, and the continuity of the UAV motion trajectory is constrained by the inertia term to obtain the optimized real-time posture of the UAV.

[0030] In one embodiment, the six-degree-of-freedom disturbance data is decomposed into compensation amounts of the body coordinate system, a backstepping sliding mode control law with feedforward compensation is set, and an anti-disturbance trajectory instruction is generated, including:

[0031] Obtaining a rotation matrix from the ship coordinate system to the UAV coordinate system according to the six-degree-of-freedom disturbance data;

[0032] Obtaining the ship's linear velocity and ship's linear acceleration, and calculating the translation compensation amount and the rotation compensation amount according to the rotation matrix, the ship's linear velocity, and the ship's linear acceleration;

[0033] A feedforward and feedback composite controller is designed to generate a backstepping sliding mode control law, and an anti-disturbance trajectory instruction is generated according to the backstepping sliding mode control law.

[0034] In one embodiment, real-time collection of air data, correction of rotor speed based on the air data, and calculation of electromagnetic attraction force include:

[0035] Collect air density, standard air density, and deck wind speed vector in real time to obtain the reference rotor speed and rotor blade tip speed, correct the rotor speed, and obtain the corrected rotor speed;

[0036] The vacuum magnetic permeability, the magnetic moments of the UAV and the deck permanent magnets, the relative distance between the UAV and the deck, and the vertical component are obtained to calculate the electromagnetic adsorption force.

[0037] In one embodiment, the method further comprises:

[0038] When the UAV is controlled to enter the landing altitude range, local airflow data of the deck is collected, and the rotor parameters of the UAV are adjusted according to the local airflow data of the deck;

[0039] The electromagnetic adsorption force is adjusted according to the local airflow data of the deck to realize the recovery of the ship-borne UAV.

[0040] A shipborne UAV recovery system based on visual navigation and six-degree-of-freedom compensation, the system comprising:

[0041] An anti-occlusion feature encoding module is used to collect deck mark images containing concentric ring-shaped optical marks using a multispectral camera mounted on a drone, extract infrared, polarization, and fluorescence frequency domain features of the marked areas in the deck mark images, and fuse them to obtain an anti-occlusion composite descriptor;

[0042] A ship motion prediction module is used to obtain the ship's six-degree-of-freedom motion parameters in real time, construct a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretize the ship dynamics differential model using an explicit multi-step method to predict the ship's six-degree-of-freedom disturbance data;

[0043] The pose optimization module is used to determine the Lie group representation of the drone's pose, obtain the weight coefficients of the set visual and inertial terms, and construct a joint visual and inertial cost function. Based on the cost function, the visual residual term is calculated according to the anti-occlusion composite descriptor and the six-degree-of-freedom perturbation data, and the continuity of the drone's motion trajectory is constrained to obtain the optimized real-time pose of the drone.

[0044] A dynamic compensation module is used to decompose the six-degree-of-freedom disturbance data into compensation amounts in the body coordinate system, set a backstepping sliding mode control law with feedforward compensation, and generate an anti-disturbance trajectory instruction;

[0045] The drone recovery module is used to collect air data in real time, correct the rotor speed according to the air data, calculate the electromagnetic attraction force, and realize the ship-borne drone recovery based on the corrected rotor speed, electromagnetic attraction force, real-time posture of the drone, and anti-disturbance trajectory instructions.

[0046] The above-mentioned ship-borne UAV recovery method and system based on visual navigation and six-degree-of-freedom compensation can overcome the problem of feature loss of visual markings due to wave splashing and lighting changes by integrating infrared, polarization, and fluorescence frequency domain features; the ship dynamics differential model predicts the six-degree-of-freedom disturbance data of the ship through discretization processing, and optimizes the real-time position and posture of the UAV in combination with the cost function, which can compensate for the ship displacement at future moments in advance; by setting a backstepping sliding mode control law with feedforward compensation, the low-frequency shaking of the ship and the dynamic response of the UAV itself can be effectively separated; by correcting the rotor speed and calculating the electromagnetic adsorption force, soft landing cushioning for UAV recovery can be achieved, reducing the risk of structural damage caused by hard impact. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A diagram illustrating an application environment of a shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation in one embodiment;

[0048] Figure 2 1 is a flow chart of a method for recovering a shipborne UAV based on visual navigation and six-degree-of-freedom compensation in one embodiment;

[0049] Figure 3 1 is a block diagram of a shipborne UAV recovery system based on visual navigation and six-degree-of-freedom compensation in one embodiment;

[0050] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0052] The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Figure 1 As shown, the application environment includes a computer device 110, a drone 120 equipped with a multispectral camera, and a ship 130, wherein the computer device 110, the drone 120 equipped with a multispectral camera, and the ship 130 communicate via a network connection. The multispectral camera carried by the drone 120 collects the deck mark image, which contains concentric ring-shaped optical marks. The computer device 110 can extract the infrared, polarization, and fluorescence frequency domain features of the marked area in the deck mark image, and fuse them to obtain an anti-occlusion composite descriptor; the computer device 110 can obtain the six-degree-of-freedom motion parameters of the ship in real time, construct a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and use an explicit multi-step method to discretize the ship dynamics differential model to predict the six-degree-of-freedom disturbance data of the ship; the computer device 110 can determine the Lie group representation of the drone's posture, and obtain the set visual terms and inertial terms. The weight coefficient is used to construct a joint visual and inertial cost function. Based on the cost function, the visual residual term is calculated according to the anti-occlusion composite descriptor and the six-degree-of-freedom disturbance data, and the continuity of the UAV's motion trajectory is constrained to obtain the optimized real-time UAV pose. The computer device 110 can decompose the six-degree-of-freedom disturbance data into body coordinate system compensation, set a backstepping sliding mode control law with feedforward compensation, and generate anti-disturbance trajectory instructions. The computer device 110 can collect air data in real time, correct the rotor speed based on the air data, calculate the electromagnetic attraction force, and realize ship-borne UAV recovery based on the corrected rotor speed, electromagnetic attraction force, UAV real-time pose, and anti-disturbance trajectory instructions. The computer device 110 can be, but is not limited to, various personal computers, laptops, smartphones, robots, tablet computers, and other devices.

[0053] In one embodiment, Figure 2 As shown, a shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation is provided, comprising the following steps:

[0054] In step 202, a multispectral camera mounted on a drone is used to collect a deck mark image containing concentric ring-shaped optical marks. The infrared, polarization, and fluorescence frequency domain features of the marked area in the deck mark image are extracted and fused to obtain an anti-occlusion composite descriptor.

[0055] Computer equipment can use the multispectral labeling layer to extract composite descriptors that integrate infrared radial gradients, polarization azimuth coding, and fluorescence frequency domain features through the multispectral camera carried by drones to solve the problem of occlusion and interference.

[0056] In one embodiment, a provided shipborne drone recovery method based on visual navigation and six-degree-of-freedom compensation may further include a process of collecting deck mark images. The specific process includes: determining an image acquisition frequency, and using a multispectral camera mounted on the drone to collect the deck mark images at the image acquisition frequency; wherein the multispectral camera includes visible light, infrared, and polarization sensors; the deck mark images contain several groups of concentric annular optical markers, each group of markers including, from the inside out, an infrared reflection layer, a polarization coding layer, and a fluorescent texture layer.

[0057] The drone's landing deck can be equipped with three sets of concentric ring-shaped optical markers. Each set of markers, from the inside out, consists of an infrared reflective layer, a polarization coding layer, and a fluorescent texture layer. In this embodiment, the drone can be equipped with a multispectral camera that includes visible light, infrared, and polarization sensors to capture images of the deck markings at a frequency of 20 Hz.

[0058] Specifically, in one embodiment, a shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation is provided, which may also include a process of obtaining an anti-occlusion composite descriptor by feature fusion. The specific process includes: extracting the spiral gradient features of the marked area in the deck mark image, performing radial gradient analysis on the infrared reflection layer, and calculating the polar coordinate distribution of the feature point set to obtain the infrared forward gradient features; based on the polarization coding layer, generating optical features through multi-frequency modulation, defining the polarization layer azimuth coding function according to the optical features, and obtaining the polarization feature code according to the polarization layer azimuth coding function; performing Log-Gabor filtering on the fluorescence frequency domain features to extract the frequency domain feature vector; cross-modally correlating the infrared forward gradient features, polarization feature codes, and frequency domain feature vectors to obtain the anti-occlusion composite descriptor.

[0059] The computer equipment can extract the spiral gradient features of the marked area, perform radial gradient analysis on the infrared layer, and calculate the feature point set P i =(r i ,θ i ) polar coordinate distribution; To solve the problem of partial obstruction of the mark caused by sea spray and shadows, it is necessary to construct a polarization feature encoding with azimuth resolution capability. Azimuth-sensitive optical features are generated through multi-frequency modulation, and the polarization layer azimuth angle coding function is defined as: Among them, φ is the output value of the polarization layer azimuth encoding function, θ is the polar angle of the infrared feature point, α k is the transmittance modulation coefficient of the k-th order polarizer, 2 k is the frequency domain multiplication coefficient, β kis the preset phase offset of the k-th order polarizer; the computer device can perform Log-Gabor filtering on the fluorescent texture layer to extract the frequency domain feature vector G = (g1, g2, ..., g8).

[0060] Then, the computer equipment can fuse the multispectral features, cross-modally correlate the infrared geometric features, polarization coding and fluorescence frequency domain features to form an anti-occlusion composite descriptor: F i =[P i ,φ(θ i ),G i ] T Among them, F i is the composite descriptor of the i-th feature point, P i is the coordinate of the feature point in the polar coordinate system, G i is the frequency domain feature vector extracted from the fluorescent texture layer, and T is the vector transpose operator.

[0061] Step 204 , obtaining the six-degree-of-freedom motion parameters of the ship in real time, constructing a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretizing the ship dynamics differential model using an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship.

[0062] Computer equipment can construct a ship dynamics differential model by obtaining the ship's six-degree-of-freedom motion parameters. Based on the ship dynamics differential model and the Adams-Bashforth discretization method, it can predict the six-degree-of-freedom disturbance data of the ship at the future time Δt in real time.

[0063] In one embodiment, a provided shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation may also include a process of predicting six-degree-of-freedom disturbance data, the specific process including: obtaining the six-degree-of-freedom motion parameters of the ship in real time through the shipborne IMU, and constructing a ship dynamics differential model in a non-inertial system based on the six-degree-of-freedom motion parameters; the ship dynamics differential model is used to describe the six-degree-of-freedom motion coupling relationship; determining the spatial coordinate system corresponding to the infrared, polarization, and fluorescence frequency domain characteristics of the marked area in the extracted deck mark image, and using the spatial coordinate system as the initial reference, inputting the six-degree-of-freedom motion parameters of the ship into the ship dynamics differential model; the ship dynamics differential model is discretized through an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship.

[0064] Computer equipment can obtain the ship's six-degree-of-freedom motion parameters in real time through the ship's IMU, where the linear acceleration is a b =[a x ,a y ,a z ] T ; angular velocity is ω b =[ω x ,ωy ,ω z ] T ; The attitude angle is Θ=[φ,θ,ψ] T In this embodiment, to address the problem that the ship motion compensation lags behind the actual disturbance, a ship dynamics differential model in a non-inertial system is established to describe the six-degree-of-freedom motion coupling relationship: Among them, v b is the linear velocity in the ship coordinate system, J is the ship's moment of inertia matrix, and τ is the external torque vector.

[0065] In order to achieve real-time prediction, the explicit multi-step method can be used to discretize the differential equation and calculate the predicted value of the ship's posture at the future time Δt: Among them, X t+Δt is the ship state at the next Δt moment, Δt is the prediction time step, f t is the right-hand side of the differential equation at time t.

[0066] In this embodiment, after completing the spiral gradient feature extraction and composite descriptor construction of the deck optical markers, the six-degree-of-freedom motion parameters of the ship at the current moment are obtained in real time through the ship-borne inertial measurement unit, the visual feature space coordinate system established in step S202 is used as the initial reference, the ship motion parameters are input into the rigid body motion differential equation, and the position change trend of the ship in the future time window is deduced through the multi-step explicit integration method, providing time-advanced disturbance prediction data for subsequent compensation control.

[0067] Step 206: Determine the Lie group representation of the drone's pose, obtain the weight coefficients of the set visual term and inertial term, and construct a joint visual and inertial cost function; based on the cost function, according to the anti-occlusion composite descriptor and six-degree-of-freedom perturbation data, calculate the visual residual term and constrain the continuity of the drone's motion trajectory to obtain the optimized real-time pose of the drone.

[0068] Computer equipment can construct a joint visual and inertial cost function, and synchronously optimize the mathematical consistency of the drone's posture and the ship's motion compensation through Lie group operations.

[0069] Specifically, in one embodiment, a shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation is provided, which may also include a process of constructing a cost function. The specific process includes: determining the Lie group representation of the UAV posture, setting the visual term and the inertial term, and setting corresponding weight coefficients for the visual term and the inertial term respectively; based on the Lie group representation, the visual term, the inertial term, and the weight coefficient, a joint visual and inertial cost function is constructed.

[0070] In this embodiment, in order to eliminate the data conflict between visual and inertial measurement, a tightly coupled optimization model in Lie group space is constructed, and the cost function is defined as: E(T uav)=λ1E vision +λ2E inertial ; Among them, λ1,λ2 are the weight coefficients of the visual term and the inertial term respectively, T uav Lie group representation of the drone’s posture, E vision is the visual term, E inertial is the inertia term.

[0071] In one embodiment, a shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation is provided, which may also include a process for optimizing the real-time pose. The specific process includes: performing projection matching constraint pose estimation based on the anti-occlusion composite descriptor, and calculating the visual residual term through the visual term based on the cost function; determining the ship motion prediction compensation amount based on the six-degree-of-freedom disturbance data, performing Lie group calculation based on the cost function, and constraining the continuity of the UAV motion trajectory through the inertia term to obtain the optimized UAV real-time pose.

[0072] The computer device constrains the pose estimation by projecting matching of the composite descriptor, and the visual residual term is calculated as: Among them, F i cam is the i-th composite descriptor observed by the camera, F i model (T uav ) is based on the posture T uav The projected model predicts the descriptor, ρ(·) is the Huber robust kernel function, and ||·||2 is the Euclidean norm.

[0073] In this embodiment, the ship motion prediction results can also be integrated, and the inertia term can constrain the continuity of the UAV motion trajectory: Where Ξ is the Lie group difference operator, is the Lie group composition operator, T pred is the UAV pose predicted by inertial measurement, ΔT ship is the compensation from the ship motion prediction, is the Mahalanobis distance.

[0074] In this embodiment, based on the ship motion prediction result obtained in step S204, a dual constraint mechanism for UAV pose estimation is established. On the one hand, the composite visual descriptor generated in step S202 is used for feature matching to calculate the relative spatial relationship between the UAV and the deck mark; on the other hand, the predicted ship motion is converted into an equivalent disturbance offset in the inertial reference frame of the UAV, and the visual observation data and inertial prediction data are weighted and fused by a nonlinear optimization algorithm to output the interference-resistant optimal pose estimation value of the UAV.

[0075] Step 208: Decompose the six-degree-of-freedom disturbance data into compensation values ​​for the body coordinate system, set a backstepping sliding mode control law with feedforward compensation, and generate an anti-disturbance trajectory instruction.

[0076] The computer equipment can decompose the predicted disturbance into the translation / rotation compensation of the body coordinate system, and design the backstepping sliding mode control law to realize the feedforward and feedback composite tracking.

[0077] In one embodiment, a provided shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation can also include a process of generating anti-disturbance trajectory instructions, the specific process including: obtaining the rotation matrix from the ship coordinate system to the UAV coordinate system based on the six-degree-of-freedom disturbance data; obtaining the ship linear velocity and ship linear acceleration, and calculating the translation compensation amount and rotation compensation amount based on the rotation matrix, ship linear velocity and ship linear acceleration; designing a feedforward and feedback composite controller, generating a backstepping sliding mode control law, and generating the anti-disturbance trajectory instructions based on the backstepping sliding mode control law.

[0078] The computer equipment can convert the predicted ship translation disturbance into the compensation amount of the UAV body coordinate system. The translation compensation amount calculation formula is: Among them, ΔΡ c is the translation compensation, R bu is the rotation matrix from the hull coordinate system to the drone coordinate system, v ship is the ship's linear speed, a ship is the ship’s linear acceleration. Next, the rotation compensation is calculated based on the predicted value of the ship’s angular motion: Where Δθ c is the rotation compensation, ω ship is the angular velocity of the ship, α ship is the ship's angular acceleration, is the integral within the time interval [t, t+Δt].

[0079] Next, the computer device can have a feedforward and feedback composite controller to generate the backstepping sliding mode control law of the UAV power system: Where u is the backstepping sliding mode control law, M(q) is the generalized inertia matrix of the UAV, q is the generalized coordinate of the UAV, is the generalized velocity, is the expected generalized acceleration, is the expected generalized velocity, K p ,K d are the proportional and differential gain matrices respectively, is the Coriolis force matrix, and G(q) is the gravity compensation term.

[0080] In this embodiment, the optimized UAV pose estimate is used as the initial state of trajectory control. Combined with the provided ship motion differential prediction results, the translational and rotational disturbances of the ship at future moments are decomposed into the UAV body coordinate system. Based on the UAV dynamics model, a feedforward-feedback composite control instruction is constructed to generate a power system input signal containing a motion compensation amount, so that the UAV can actively offset the ship disturbance that has not actually occurred when performing trajectory tracking.

[0081] Step 210 , collect air data in real time, correct the rotor speed according to the air data, calculate the electromagnetic attraction force, and realize the recovery of the ship-borne UAV according to the corrected rotor speed, electromagnetic attraction force, real-time posture of the UAV, and anti-disturbance trajectory instructions.

[0082] Computer equipment can integrate airflow density compensation rotor power and electromagnetic adsorption force gradient control to complete anti-interference docking between drones and ships.

[0083] In one embodiment, a method for recovering a shipborne drone based on visual navigation and six-degree-of-freedom compensation is provided, which may include a process of correcting the rotor speed and calculating the electromagnetic adsorption force. The specific process includes: real-time collection of air density, standard air density, and deck wind speed vector, obtaining the reference rotor speed and rotor blade tip speed, correcting the rotor speed, and obtaining the corrected rotor speed; obtaining vacuum magnetic permeability, the magnetic moment of the drone and the deck permanent magnet, the relative distance between the drone and the deck, and the vertical direction component, and calculating the electromagnetic adsorption force.

[0084] When the UAV enters the final landing phase, the deck airflow compensation is activated and the rotor speed is corrected according to the bow anemometer data. The correction formula is: Among them, Ω corr is the rotor speed correction, Ω0 is the reference rotor speed, ρ is the current air density, ρ0 is the standard air density, v wind is the deck wind speed vector, v tip is the rotor blade tip speed.

[0085] Next, the electromagnetic adsorption device can be activated and powered on to generate a gradient magnetic field at a distance of 0.3m from the deck. The electromagnetic adsorption force calculation model is: Among them, F mag is the electromagnetic adsorption force, μ0 is the vacuum magnetic permeability, m uav ,m deck are the magnetic moments of the UAV and the deck permanent magnets, r is the relative distance between the UAV and the deck, and z is the vertical component.

[0086] In one embodiment, a method for recovering a shipborne UAV based on visual navigation and six-degree-of-freedom compensation is provided, which may also include a process for recovering the shipborne UAV. The specific process includes: when the UAV is controlled to enter the landing height range, collecting local airflow data on the deck, and adjusting the rotor parameters of the UAV according to the local airflow data on the deck; adjusting the electromagnetic adsorption force according to the local airflow data on the deck to achieve the recovery of the shipborne UAV.

[0087] When the UAV is controlled to enter the final landing altitude range, the environmental adaptation strategy is further superimposed based on the generated compensation control quantity: the rotor aerodynamic parameters are dynamically adjusted according to the changes in the local airflow on the deck, and the predictive magnetic field gradient control of the electromagnetic adsorption device is activated at the same time, so that the UAV forms a stable force constraint relationship at the moment of contact with the deck, completing the mode switch from dynamic compensation control to physical contact stability.

[0088] The present application provides a method for recovering shipborne UAVs based on visual navigation and six-degree-of-freedom compensation, which enhances navigation stability in complex sea conditions by fusing multispectral visual features with a ship motion prediction model. Specifically, spiral gradient feature encoding combines multi-layer information of infrared, polarization, and fluorescent textures to more effectively overcome the problem of feature loss of visual markers due to wave splashing and illumination changes. The differential prediction model decomposes the six-degree-of-freedom motion of the ship into an equivalent perturbation of the UAV body coordinate system, so that the pose estimation can compensate for the ship displacement at future moments in advance. This design avoids control lags caused by sensor delays, and can maintain continuous and stable relative pose solutions, especially in high sea conditions. At the same time, the tightly coupled optimization framework uniformly processes the spatiotemporal consistency of visual and inertial data through Lie group constraints, reducing positioning drift caused by failure of a single sensor.

[0089] A ship differential prediction model based on rigid-body dynamics uses a high-order numerical integration method to achieve a balance between computational efficiency and accuracy. The time-domain differential equation directly relates the ship's inertial parameters and motion state, avoiding the spectrum leakage problem introduced by Fourier transform. The prediction results are embedded in the UAV control law in the form of Lie algebras, forming a feedforward and feedback composite compensation architecture, which effectively separates the ship's low-frequency swaying from the UAV's own dynamic response. This design reduces the complexity of the controller's parameter adjustment while ensuring phase synchronization between the compensation instructions and the ship's actual disturbances.

[0090] By triggering different physical field auxiliary devices in a graded manner, it adapts to the terminal landing requirements under complex sea conditions; the deck airflow compensation module dynamically adjusts the rotor aerodynamic force distribution according to real-time wind field data to offset the asymmetric load caused by the ship's swaying; the gradient magnetic field design of the electromagnetic adsorption device produces a capture force field that varies with distance, forming a soft landing buffer in the final contact stage; through field-force coupling, progressive energy dissipation is achieved to reduce the risk of structural damage caused by hard impact; the entire recovery process does not require high-precision absolute positioning, but only needs to maintain the gradual convergence of relative posture, so that the system can still work reliably in a satellite denial environment.

[0091] In order to explain the shipborne drone recovery method based on visual navigation and six-degree-of-freedom compensation provided in this application, in one embodiment, a specific application scenario is provided:

[0092] The implementation scenarios are as follows:

[0093] The ship parameters are: displacement 12 tons, length 18 meters, and the diagonal element of the moment of inertia matrix is ​​j x =1500kg·m 2 、j y =3200kg·m 2 、j z =2800kg·m 2 ;

[0094] The deck markings are arranged as follows: three sets of concentric circles with diameters of 0.5m, 1.0m, and 1.5m, with the centers of the circles spaced 2m apart;

[0095] The drone parameters are: quad-rotor drone, mass 4.2 kg, rotor diameter 0.3 m, maximum thrust 120 N; the camera uses a multispectral camera with a resolution of 2048×2048 and a frame rate of 20 Hz.

[0096] The implementation steps are as follows:

[0097] 1. Anti-occlusion visual feature encoding:

[0098] Image acquisition: The drone activates its camera at an altitude of 15 meters, with an infrared exposure time of 5 milliseconds and a polarization filter switching cycle of 50 milliseconds. The captured image resolution is 1024×1024 ROI area, and single-frame processing takes 8 milliseconds.

[0099] Feature extraction: Infrared layer radial gradient: Determine the coordinates of the circle center from the acquired image as (512.3, 507.8) pixels, with an allowable radius error of plus or minus 0.5 pixels; use the standard two-dimensional distance formula to calculate the distance from each feature point to the circle center: The calculated radii of the three rings are 243.2 pixels, 486.5 pixels, and 729.7 pixels, respectively. These pixel distances correspond to actual physical distances of 0.5 meters, 1.0 meters, and 1.5 meters. Polarization layer coding: preset α k =[0.7,0.4,0.2,0.1],β k =[0,π / 3,2π / 5,π / 2]; calculate the encoding value for the direction of θ=45°: Log-Gabor filter center frequency f c =0.25 cycles / pixel, bandwidth 2.5 octaves; extract frequency domain vector G = [0.87, 0.12, -0.05, 0.33, 0.21, -0.18, 0.09, 0.04]; feature fusion: generate composite descriptor F i The dimension is 3+1+8=12, which is stored in the feature library after normalization;

[0100] 2. Differential prediction of ship motion:

[0101] Data acquisition: IMU outputs the current linear acceleration a b =[0.3,-0.2,1.1]m / s 2 Angular velocity ω b =[0.15,-0.08,0.05]rad / s;

[0102] Solve the differential equation: Substitute the ship parameters and calculate the right-hand side of the differential equation:

[0103]

[0104] Prediction calculation: Select the time window Δt = 0.2s and use the fourth-order Adams-Bashforth formula: Historical data t-1 =[0.298,-0.135,1.092,0.0007,-0.0016,0.0002] T ;

[0105] 3. Fusion of visual and inertial data:

[0106] Cost function construction: Set weights λ1 = 0.6, λ2 = 0.4, and Huber kernel threshold δ = 0.1; match the current frame to 32 feature points and calculate the visual residual: The maximum residual e max =0.087, total visual error E vision =0.024; Inertial constraint term: predicted pose transformation ΔT shipThe corresponding Lie algebra ξ=[0.062,-0.041,0.028,0.0003,-0.0007,0.0001] T ; Calculate Mahalanobis distance: E inertial =ξ T ∑ -1 ξ=0.017;covariance matrix ∑=diag(0.01 2 ,0.01 2 ,0.01 2 ,0.005 2 ,0.005 2 ,0.005 2 ); Optimization solution: initial pose estimation T init =[0.5,-0.3,12.1,0.02,-0.01,0.005] T After 7 Levenberg-Marquardt iterations, the damping factor μ = 1e-4, the final pose

[0107] 4. Six-degree-of-freedom dynamic compensation control:

[0108] Perturbation decomposition: rotation matrix R from the hull to the drone coordinate system bu Generated by Euler angles [0.019,-0.011,0.0048]:

[0109] Calculation of translation compensation:

[0110] Backstepping sliding mode control: UAV mass matrix M = diag(4.2,4.2,4.2,0.12,0.12,0.12); set gain K p =0.5I,K d =1.2I;

[0111] Generate control input:

[0112] 5. Multi-modal landing guidance:

[0113] Airflow compensation: measured deck wind speed v wind =[3.2,-1.5,0]m / s;

[0114] Corrected rotor speed:

[0115] Electromagnetic adsorption control: magnetic moment m of the drone's permanent magnet uav =1.2A·m 2 ; Magnetic moment of deck permanent magnet m deck =2.5A·m 2; Relative distance r = 0.28m, calculate the adsorption force:

[0116] At the moment of touching the ground (r=0.05m), the adsorption force increases to 215N, ensuring that there is no relative sliding between the drone and the deck.

[0117] The implementation results are as follows:

[0118] In the test under level 5 sea conditions (wave height 2.5 meters), the drone was recovered from a height of 15 meters:

[0119] Pose estimation error: position ±0.012m, attitude ±0.3°;

[0120] Trajectory tracking latency: only 8 milliseconds from prediction to compensation execution;

[0121] Final landing accuracy: lateral deviation ≤ 0.02 m, vertical contact speed ≤ 0.1 m / s;

[0122] Time required for the entire process: 3.2 seconds from identification to locking, meeting the ship-borne real-time requirements.

[0123] It should be understood that, although the various steps in the above flow chart are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the above flow chart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0124] In one embodiment, Figure 3 As shown, a shipborne UAV recovery system based on visual navigation and six-degree-of-freedom compensation is provided, including: an anti-occlusion feature encoding module 310, a ship motion prediction module 320, a posture optimization module 330, a dynamic compensation module 340 and a UAV recovery module 350, wherein:

[0125] The anti-occlusion feature encoding module 310 is used to collect deck mark images containing concentric ring-shaped optical marks using a multispectral camera mounted on a drone, extract infrared, polarization, and fluorescence frequency domain features of the marked areas in the deck mark images, and fuse them to obtain an anti-occlusion composite descriptor;

[0126] The ship motion prediction module 320 is used to obtain the ship's six-degree-of-freedom motion parameters in real time, construct a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretize the ship dynamics differential model using an explicit multi-step method to predict the ship's six-degree-of-freedom disturbance data;

[0127] The pose optimization module 330 is used to determine the Lie group representation of the UAV's pose, obtain the weight coefficients of the set visual and inertial terms, and construct a joint visual and inertial cost function. Based on the cost function, the visual residual term is calculated according to the anti-occlusion composite descriptor and the six-degree-of-freedom perturbation data, and the continuity of the UAV's motion trajectory is constrained to obtain the optimized real-time pose of the UAV.

[0128] The dynamic compensation module 340 is used to decompose the six-degree-of-freedom disturbance data into compensation quantities in the body coordinate system, set a backstepping sliding mode control law with feedforward compensation, and generate anti-disturbance trajectory instructions;

[0129] The drone recovery module 350 is used to collect air data in real time, correct the rotor speed according to the air data, calculate the electromagnetic attraction force, and realize the ship-borne drone recovery based on the corrected rotor speed, electromagnetic attraction force, real-time posture of the drone, and anti-disturbance trajectory instructions.

[0130] In one embodiment, the anti-occlusion feature encoding module 310 is also used to determine the image acquisition frequency, and the multispectral camera equipped on the drone acquires the deck mark image at the image acquisition frequency; wherein the multispectral camera includes visible light, infrared, and polarization sensors; the deck mark image contains several groups of concentric ring-shaped optical marks, and each group of marks includes an infrared reflection layer, a polarization encoding layer, and a fluorescent texture layer from the inside to the outside.

[0131] In one embodiment, the anti-occlusion feature coding module 310 is also used to extract the spiral gradient features of the marking area in the deck marking image, perform radial gradient analysis on the infrared reflection layer, and calculate the polar coordinate distribution of the feature point set to obtain the infrared forward gradient features; based on the polarization coding layer, generate optical features through multi-frequency modulation, define the polarization layer azimuth coding function according to the optical features, and obtain the polarization feature coding according to the polarization layer azimuth coding function; perform Log-Gabor filtering on the fluorescence frequency domain features to extract the frequency domain feature vector; perform cross-modal correlation on the infrared forward gradient features, polarization feature coding, and frequency domain feature vector to obtain an anti-occlusion composite descriptor.

[0132] In one embodiment, the ship motion prediction module 320 is further used to obtain the six-degree-of-freedom motion parameters of the ship in real time through the ship-borne IMU, and to construct a ship dynamics differential model in a non-inertial system based on the six-degree-of-freedom motion parameters; the ship dynamics differential model is used to describe the six-degree-of-freedom motion coupling relationship; determine the spatial coordinate system corresponding to the infrared, polarization, and fluorescence frequency domain characteristics of the marked area in the extracted deck mark image, and use the spatial coordinate system as the initial reference to input the six-degree-of-freedom motion parameters of the ship into the ship dynamics differential model; the ship dynamics differential model is discretized through an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship.

[0133] In one embodiment, the posture optimization module 330 is also used to determine the Lie group representation of the drone's posture, set visual terms and inertial terms, and set corresponding weight coefficients for the visual terms and inertial terms respectively; based on the Lie group representation, visual terms, inertial terms, and weight coefficients, a joint visual and inertial cost function is constructed.

[0134] In one embodiment, the pose optimization module 330 is also used to perform projection matching constrained pose estimation based on the anti-occlusion composite descriptor, and calculate the visual residual term through the visual term based on the cost function; determine the ship motion prediction compensation amount based on the six-degree-of-freedom disturbance data, perform Lie group calculation based on the cost function, constrain the continuity of the UAV motion trajectory through the inertia term, and obtain the optimized UAV real-time pose.

[0135] In one embodiment, the dynamic compensation module 340 is also used to obtain the rotation matrix from the ship coordinate system to the UAV coordinate system based on the six-degree-of-freedom disturbance data; obtain the ship linear velocity and ship linear acceleration, and calculate the translation compensation amount and rotation compensation amount based on the rotation matrix, ship linear velocity and ship linear acceleration; design a feedforward and feedback composite controller, generate a backstepping sliding mode control law, and generate an anti-disturbance trajectory instruction based on the backstepping sliding mode control law.

[0136] In one embodiment, the drone recovery module 350 is also used to collect air density, standard air density, and deck wind speed vector in real time, obtain the reference rotor speed and rotor blade tip speed, correct the rotor speed, and obtain the corrected rotor speed; obtain the vacuum magnetic permeability, the magnetic moment of the drone and the deck permanent magnet, the relative distance between the drone and the deck, and the vertical direction component, and calculate the electromagnetic adsorption force.

[0137] In one embodiment, the drone recovery module 350 is also used to collect local airflow data on the deck when the drone is controlled to enter the landing height range, and adjust the rotor parameters of the drone according to the local airflow data on the deck; adjust the electromagnetic adsorption force according to the local airflow data on the deck to realize the recovery of the ship-borne drone.

[0138] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface, display screen and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a ship-borne drone recovery method based on visual navigation and six-degree-of-freedom compensation is implemented. The display screen of the computer device can be a liquid crystal display or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0139] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0140] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of a shipborne drone recovery method based on visual navigation and six-degree-of-freedom compensation are implemented.

[0141] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of a ship-borne drone recovery method based on visual navigation and six-degree-of-freedom compensation are implemented.

[0142] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0143] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0144] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation, characterized in that: The method comprises: A multispectral camera mounted on a drone is used to collect deck mark images containing concentric ring-shaped optical marks. Infrared, polarization, and fluorescence frequency domain features of the marked areas in the deck mark images are extracted and fused to obtain an anti-occlusion composite descriptor. Acquiring the six-degree-of-freedom motion parameters of the ship in real time, constructing a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretizing the ship dynamics differential model using an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship; Determine the Lie group representation of the drone's pose, obtain the weight coefficients of the set visual term and inertial term, and construct a joint visual and inertial cost function; based on the cost function, calculate the visual residual term and constrain the continuity of the drone's motion trajectory based on the anti-occlusion composite descriptor and six-degree-of-freedom perturbation data, and obtain the optimized real-time pose of the drone; Decomposing the six-degree-of-freedom disturbance data into compensation amounts for the body coordinate system, setting a backstepping sliding mode control law with feedforward compensation, and generating an anti-disturbance trajectory instruction; Air data is collected in real time, the rotor speed is corrected according to the air data, the electromagnetic attraction force is calculated, and the ship-borne drone is recovered based on the corrected rotor speed, electromagnetic attraction force, real-time posture of the drone, and anti-disturbance trajectory instructions.

2. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 1 is characterized in that: Deck marking images were collected using a drone equipped with a multispectral camera, including: determining an image acquisition frequency, wherein the multispectral camera carried by the drone acquires images of the deck markings at the image acquisition frequency; The multispectral camera includes visible light, infrared, and polarization sensors; the deck mark image contains several groups of concentric ring-shaped optical marks, and each group of marks includes an infrared reflection layer, a polarization coding layer, and a fluorescent texture layer from the inside to the outside.

3. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 2 is characterized in that: Extract infrared, polarization, and fluorescence frequency domain features of the marked area in the deck mark image, and fuse them to obtain an anti-occlusion composite descriptor, including: Extracting the spiral gradient features of the marked area in the deck mark image, performing radial gradient analysis on the infrared reflective layer, and calculating the polar coordinate distribution of the feature point set to obtain the infrared forward gradient features; Based on the polarization coding layer, generating optical features through multi-frequency modulation, defining a polarization layer azimuth coding function according to the optical features, and obtaining polarization feature coding according to the polarization layer azimuth coding function; Performing Log-Gabor filtering on the fluorescence frequency domain features to extract frequency domain feature vectors; The infrared input gradient feature, polarization feature encoding, and frequency domain feature vector are cross-modally correlated to obtain an anti-occlusion composite descriptor.

4. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 1 is characterized in that: Acquire the six-degree-of-freedom motion parameters of the ship in real time, construct a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretize the ship dynamics differential model using an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship, including: The ship's six-degree-of-freedom motion parameters are acquired in real time by using the ship's onboard IMU, and a ship dynamics differential model in a non-inertial system is constructed based on the six-degree-of-freedom motion parameters; the ship dynamics differential model is used to describe the six-degree-of-freedom motion coupling relationship; Determining a spatial coordinate system corresponding to the infrared, polarization, and fluorescence frequency domain features of the mark area in the extracted deck mark image, and inputting the six-degree-of-freedom motion parameters of the ship into the ship dynamics differential model using the spatial coordinate system as an initial reference; The ship dynamics differential model is discretized by an explicit multi-step method to predict the six-degree-of-freedom disturbance data of the ship.

5. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 1 is characterized in that: Determine the Lie group representation of the drone's posture, obtain the weight coefficients of the set visual term and inertial term, and construct the joint visual and inertial cost function, including: Determine the Lie group representation of the drone's posture, set a visual term and an inertial term, and set corresponding weight coefficients for the visual term and the inertial term respectively; Based on the Lie group representation, visual term, inertial term, and weight coefficient, a joint vision and inertial cost function is constructed.

6. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 5 is characterized in that: Based on the cost function, according to the anti-occlusion composite descriptor and the six-degree-of-freedom disturbance data, the visual residual term is calculated and the continuity of the UAV motion trajectory is constrained to obtain the optimized UAV real-time pose, including: Performing projection matching constraint pose estimation on the anti-occlusion composite descriptor, and calculating a visual residual term through the visual term based on the cost function; The predicted compensation amount of the ship motion is determined according to the six-degree-of-freedom disturbance data, Lie group calculation is performed based on the cost function, and the continuity of the UAV motion trajectory is constrained by the inertia term to obtain the optimized real-time posture of the UAV.

7. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 1 is characterized in that: Decomposing the six-degree-of-freedom disturbance data into compensation amounts for the body coordinate system, setting a backstepping sliding mode control law with feedforward compensation, and generating anti-disturbance trajectory instructions, including: Obtaining a rotation matrix from the ship coordinate system to the UAV coordinate system according to the six-degree-of-freedom disturbance data; Obtaining the ship's linear velocity and ship's linear acceleration, and calculating the translation compensation amount and the rotation compensation amount according to the rotation matrix, the ship's linear velocity, and the ship's linear acceleration; A feedforward and feedback composite controller is designed to generate a backstepping sliding mode control law, and an anti-disturbance trajectory instruction is generated according to the backstepping sliding mode control law.

8. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 1 is characterized in that: Collect air data in real time, correct the rotor speed based on the air data, and calculate the electromagnetic attraction force, including: Collect air density, standard air density, and deck wind speed vector in real time to obtain the reference rotor speed and rotor blade tip speed, correct the rotor speed, and obtain the corrected rotor speed; The vacuum magnetic permeability, the magnetic moments of the UAV and the deck permanent magnets, the relative distance between the UAV and the deck, and the vertical component are obtained to calculate the electromagnetic adsorption force.

9. The shipborne UAV recovery method based on visual navigation and six-degree-of-freedom compensation according to claim 1, characterized in that: The method further comprises: When the UAV is controlled to enter the landing altitude range, local airflow data of the deck is collected, and the rotor parameters of the UAV are adjusted according to the local airflow data of the deck; The electromagnetic adsorption force is adjusted according to the local airflow data of the deck to realize the recovery of the ship-borne UAV.

10. A shipborne UAV recovery system based on visual navigation and six-degree-of-freedom compensation, characterized in that: The system comprises: An anti-occlusion feature encoding module is used to collect deck mark images containing concentric ring-shaped optical marks using a multispectral camera mounted on a drone, extract infrared, polarization, and fluorescence frequency domain features of the marked areas in the deck mark images, and fuse them to obtain an anti-occlusion composite descriptor; A ship motion prediction module is used to obtain the ship's six-degree-of-freedom motion parameters in real time, construct a ship dynamics differential model based on the six-degree-of-freedom motion parameters, and discretize the ship dynamics differential model using an explicit multi-step method to predict the ship's six-degree-of-freedom disturbance data; The pose optimization module is used to determine the Lie group representation of the drone's pose, obtain the weight coefficients of the set visual and inertial terms, and construct a joint visual and inertial cost function. Based on the cost function, the visual residual term is calculated according to the anti-occlusion composite descriptor and the six-degree-of-freedom perturbation data, and the continuity of the drone's motion trajectory is constrained to obtain the optimized real-time pose of the drone. A dynamic compensation module is used to decompose the six-degree-of-freedom disturbance data into compensation amounts in the body coordinate system, set a backstepping sliding mode control law with feedforward compensation, and generate an anti-disturbance trajectory instruction; The drone recovery module is used to collect air data in real time, correct the rotor speed according to the air data, calculate the electromagnetic attraction force, and realize the ship-borne drone recovery based on the corrected rotor speed, electromagnetic attraction force, real-time posture of the drone, and anti-disturbance trajectory instructions.

Images