A method and system for precise landing of a UAV based on adaptive switching of dual cameras

Abstract

The present application relates to the technical field of unmanned aerial vehicle control, in particular to a kind of unmanned aerial vehicle precision landing method and system based on double camera adaptive switching, method includes: using main camera high-definition image to solve first pose, guide unmanned aerial vehicle to approach target;Real-time monitoring height and mark imaging ratio, judge whether to meet switching condition;When switching vice camera, inherit main camera last frame pose, combine inertial navigation data to carry out prediction fusion, ensure that control closed loop is continuous;Using wide-angle image of vice camera to solve second pose, fine-tune pose until meet ground contact condition.The present application solves the contradiction between long-distance recognition accuracy and near distance field of view limitation by double camera switching and multi-source data fusion, eliminates the pose step in the process of hardware switching, realizes the whole continuous, high-precision autonomous precision landing.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control technology, and in particular to a method and system for precise UAV landing based on adaptive switching of dual cameras. Background Technology

[0002] With the increasing application of drones in scenarios such as inspection, delivery, automatic return and recovery, and mobile platform takeoff and landing, how to continuously obtain stable and accurate relative attitude information during the landing phase has become a key issue affecting the safety and accuracy of autonomous landing. Existing visual guidance landing schemes mostly rely on a single imaging unit combined with ground visual markers to complete target recognition and attitude estimation. However, in actual descent, it is often difficult to balance imaging field of view, recognition clarity, close-range complete framing capability, and attitude disturbance adaptability: strong long-range recognition capability is required at higher altitudes, while ensuring the entire target remains within the effective field of view as it approaches the ground; otherwise, unstable recognition, fluctuations in attitude calculation, discontinuous control adjustments, and even decreased terminal landing accuracy can easily occur. Especially in complex environments or dynamic carrier helipad scenarios, the existing solutions still fall short in maintaining continuous and stable visual guidance across the entire altitude range. Therefore, it is necessary to propose a visual guidance technology solution more suitable for the precise landing of drones throughout the entire process. Summary of the Invention

[0003] In view of the above technical problems, the present invention provides a method and system for precise landing of drones based on dual-camera adaptive switching, which aims to solve the problem that existing drone visual landing technology is difficult to simultaneously take into account high-altitude recognition capability, low-altitude complete framing capability and positioning and control continuity during landing, thereby achieving more stable, accurate and reliable autonomous landing of drones throughout the entire landing phase.

[0004] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part from practice of this disclosure.

[0005] According to one aspect of the present invention, a method for precise landing of a drone based on adaptive switching of dual cameras is proposed, applicable to a drone equipped with a main camera and a secondary camera, the method comprising the following stages: High-altitude guidance phase: When the UAV is in the first altitude range, the main camera is activated to capture high-definition images containing ground-identifiable visually coded markers, and the first pose information of the UAV relative to the identifiable visually coded markers is calculated based on the high-definition images. The UAV is then controlled to fly towards the location of the identifiable visually coded markers according to the first pose information. Adaptive switching decision phase: During the descent of the UAV, the current relative altitude of the UAV and the imaging proportion of the identifiable visual coded mark in the current image are acquired in real time, and it is determined whether the current relative altitude and the imaging proportion simultaneously meet the preset switching conditions. Smooth transition and pose fusion stage: When the current relative height and the imaging ratio simultaneously meet the preset switching conditions, the image acquisition source is switched from the main camera to the secondary camera. During the switching process, the last valid pose output by the main camera is inherited as the prior pose, and pose prediction is performed in combination with the inertial navigation data of the UAV. The predicted pose is fused with the pose initially calculated by the secondary camera to output smooth and continuous transition pose information, so as to keep the control closed loop of the UAV uninterrupted. Low-altitude precision landing phase: When the drone is in a second altitude range lower than the first altitude range, the secondary camera is activated to capture a wide-angle image containing the complete identifiable visual coding mark. Based on the wide-angle image, the second pose information of the drone relative to the identifiable visual coding mark is continuously calculated. The horizontal position and yaw attitude of the drone are finely adjusted according to the second pose information until the drone meets the preset ground contact conditions and completes the landing.

[0006] Furthermore, determining whether the current relative height and the imaging percentage simultaneously meet preset switching conditions includes: Set the main camera to sub-camera switching trigger condition and the sub-camera to main camera switching trigger condition. A hysteresis interval is formed between the main camera to sub-camera switching trigger condition and the sub-camera to main camera switching trigger condition to avoid frequent switching of image acquisition source within the critical height range. The trigger condition for switching from the main camera to the secondary camera is set as follows: the current relative height is less than or equal to a preset descent height threshold, and the imaging ratio is greater than or equal to a preset magnification ratio threshold. The trigger condition for switching from the secondary camera to the primary camera is set as follows: the current relative height is greater than or equal to a preset rise height threshold, and the imaging ratio is less than or equal to a preset shrinkage ratio threshold. To form the hysteresis interval, the descent height threshold is set to be less than the ascent height threshold, while the magnification ratio threshold is set to be greater than the shrinkage ratio threshold.

[0007] Furthermore, the algorithm implementation process for calculating the first pose information of the UAV relative to the identifiable visual coded marker based on the high-definition image, and for continuously calculating the second pose information of the UAV relative to the identifiable visual coded marker based on the wide-angle image, specifically includes: Extract the two-dimensional pixel coordinates of multiple edge corner points of the identifiable visually encoded marker in the currently captured image; Obtain the prior information of the three-dimensional physical dimensions of the identifiable visually encoded mark pre-calibrated during the initialization phase; Using a perspective point projection pose estimation algorithm, combined with the two-dimensional pixel coordinates and the three-dimensional physical size prior information, a spatial projection mapping relationship is constructed; By iteratively optimizing and solving the spatial projection mapping relationship, the six-degree-of-freedom pose data of the UAV relative to the identifiable visual coded mark is obtained. The six-degree-of-freedom pose data includes horizontal lateral offset, horizontal longitudinal offset, vertical height, roll angle, pitch angle, and yaw angle. The horizontal lateral offset, the horizontal longitudinal offset, and the yaw angle are extracted from the six-degree-of-freedom pose data for UAV horizontal closed-loop position control. The vertical height is extracted for UAV vertical closed-loop altitude control. The roll angle and the pitch angle are used for system health monitoring and stability compensation.

[0008] Furthermore, prior to the high-altitude guidance phase, the method also includes a system coordinate unification and intrinsic parameter compensation initialization step: The perspective distortion parameters of the main camera are calibrated, the wide-angle fisheye distortion parameters of the secondary camera are calibrated, and the distortion correction parameters are injected into the airborne image processing front end for real-time distortion elimination to ensure the calculation accuracy under low-altitude distortion field of view. Spatial joint calibration is used to obtain the external parameter spatial transformation relationship matrix between the main camera, the secondary camera and the coordinate system of the UAV's inertial measurement unit. At the instant of switching into the smooth transition and pose fusion stage, the intrinsic parameters of the secondary camera and the distortion correction parameters are immediately invoked to process the image. The pose calculation module is not re-initialized, and all the calculated pose data are uniformly converted to the coordinate system of the UAV body inertial measurement unit according to the extrinsic parameter spatial transformation relationship matrix for global expression.

[0009] Furthermore, the step of fusing the predicted pose with the pose initially calculated by the secondary camera to output smooth and continuous transition pose information includes the following specific calculation steps: Using real-time high-frequency motion acceleration and angular velocity data provided by the UAV inertial measurement unit, motion state integral calculation is performed on the inherited prior pose to obtain a short-time predicted pose for compensating for the short-time delay caused by camera switching and data processing. When the secondary camera completes the acquisition of a new frame of image and calculates the initial observation pose, a filtering estimation model is constructed. The short-term predicted pose is used as the predicted value for state update, and the initial observed pose is used as the observed value for measurement update. These are synchronously input into the filter estimation model for weighted smooth fusion calculation, and the continuous pose sequence that eliminates step jumps is output to the position loop and attitude loop of the UAV flight controller to maintain the smoothness of the landing command throughout the entire process.

[0010] Furthermore, the determination logic for the drone to complete landing under preset ground contact conditions is based on the simultaneous fulfillment of multiple sub-conditions to prevent accidental locking: The current absolute altitude value above the ground measured by the UAV's onboard ranging sensor is consistently lower than the preset ground contact safety height limit; The descent speed of the UAV in the vertical downward direction, calculated by integrating airborne vision and inertial data, is consistently lower than the preset descent speed limit. Received a mechanical physical trigger signal from the UAV landing gear contacting the ground; When the above sub-conditions are continuously met within the preset time window, the ground contact is determined to be successful, triggering the UAV landing lock-on procedure, stopping the attitude calculation and horizontal power output, maintaining only attitude stability and gradually locking the rotor motors.

[0011] Furthermore, when the identifiable visually encoded marker is deployed on a mobile carrier helipad in a dynamically swaying state, the method further includes a dynamic adaptive compensation step: The motion compensation module based on inertial navigation prediction is activated to estimate in real time the undulation and tilt attitude changes of the mobile carrier's helipad caused by environmental disturbances; Based on the estimated fluctuations and tilt changes, the target tracking trajectory of the UAV control system is dynamically adjusted and compensated to counteract external interference from the carrier motion on the visual relative pose measurement. The system synchronously and adaptively fine-tunes the height threshold and percentage threshold parameters in the preset switching conditions based on the dynamic swaying amplitude of the mobile carrier's helipad.

[0012] According to a second aspect of this disclosure, a drone precision landing system based on dual-camera adaptive switching is provided, applied to perform the drone precision landing method based on dual-camera adaptive switching as described above, the system comprising: The main and secondary dual-camera visual acquisition hardware module includes a high-resolution main camera with a narrow field of view and a wide-angle secondary camera with a wide field of view. It can also be integrated into a dual-focal-length integrated optoelectronic pod that achieves optical zoom by switching between internal prisms and sensors, and is used to capture images of ground-recognizable visually coded markers across the entire altitude range. The pose calculation and dual decision processing module is communicatively connected to the main and secondary dual-camera visual acquisition hardware module, and is used to perform image decoding, corner feature extraction, continuous six-degree-of-freedom pose calculation, and adaptive acquisition source switching decision based on parameter threshold and hysteresis logic. The smooth fusion and inertial compensation module is used to inherit the historical prior pose during the switching of image acquisition sources, and combine it with the UAV inertial navigation data for filtering and fusion output to eliminate control command jumps; The flight control closed-loop execution module receives the continuous attitude signals output by the smooth fusion and inertial compensation module, and drives the rotor system to perform high-altitude approach, smooth transition, low-altitude fine-tuning, vertical descent and safe ground touchdown locking actions without interrupting the control loop.

[0013] The technical solution disclosed herein has the following beneficial effects: This invention optimizes the visual perception requirements at different altitude stages during UAV descent, improving the continuity of target marker recognition, the stability of pose calculation, and the smoothness of descent control as the UAV approaches from high altitude to low altitude and touches down. This balances long-range guidance capabilities with close-range fine-tuning capabilities. Compared to traditional single-vision solutions, this invention helps reduce recognition instability and control fluctuations caused by changes in imaging conditions, enhancing robustness, accuracy, and environmental adaptability during descent. Furthermore, this solution primarily relies on the collaboration between the airborne vision unit and conventional flight control perception information, without depending on high-cost external positioning infrastructure. It possesses good engineering feasibility, deployment convenience, and applicability for widespread adoption, making it particularly suitable for UAV point-to-point recovery and precise landing scenarios requiring a high degree of autonomy. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a precise landing method for a drone based on adaptive switching of dual cameras, as described in the embodiments of this specification. Figure 2 This is a structural block diagram of a drone precision landing system based on dual-camera adaptive switching, as described in an embodiment of this specification. Detailed Implementation

[0015] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of these specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0016] Furthermore, the accompanying drawings are merely illustrative of this disclosure. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0017] This invention provides a method for precise drone landing based on adaptive switching of dual cameras. (Refer to...) Figure 1 The diagram shown is a flowchart illustrating a precise landing method for a drone based on adaptive switching of dual cameras, according to an embodiment of the present invention. This method can be applied to drones equipped with a main camera and a secondary camera. The method can be executed by a device, which can be implemented in software and / or hardware. Specifically, the method may include the following steps S101-S104: In step S101, during the high-altitude guidance phase: when the UAV is in the first altitude range, the main camera is activated to acquire a high-definition image containing a ground-identifiable visually coded marker, and the first pose information of the UAV relative to the identifiable visually coded marker is calculated based on the high-definition image. The UAV is then controlled to fly towards the location of the identifiable visually coded marker according to the first pose information.

[0018] In practical applications, the main camera typically has a small field of view (e.g., 60 to 80 degrees) and high resolution. Within the higher first altitude range, if a wide-angle camera is used directly, the visually encoded markers will occupy very few pixels in the image, making them difficult to detect and recognize effectively. Therefore, enabling the main camera to shoot vertically downwards ensures that sufficiently clear details of the visually encoded markers are obtained even at long distances, providing a foundation for subsequent accurate positioning.

[0019] The algorithm implementation process for the first pose information and the second pose information specifically includes: Extract the two-dimensional pixel coordinates of multiple edge corner points of the identifiable visually encoded markers in the currently captured image; and typically it is necessary to detect and extract four or more key corner points of the visually encoded markers to obtain enough data points to support geometric calculations in the spatial dimension; The prior information of the three-dimensional physical dimensions of the identifiable visually encoded mark is obtained in the pre-calibrated input during the initialization phase; this prior information is mainly defined by the known physical dimensions and spatial layout of the visually encoded mark, and is the absolute basis for establishing real-world 3D coordinates and calculating the scale. Using a perspective point projection pose estimation algorithm, combined with the prior information of the two-dimensional pixel coordinates and the three-dimensional physical dimensions, a spatial projection mapping relationship is constructed. Specifically, this process typically calls the Perspective-n-Point (PnP) algorithm, such as the SolvePnP algorithm combined with the random sample consensus algorithm, to build the model. The spatial projection mapping relationship can be expressed by the classic pinhole camera imaging model formula: ; in, As a scale factor, This is a homogeneous representation of the two-dimensional pixel coordinates of the edge corner points. This is the homogeneous representation of the corresponding three-dimensional physical coordinates. The intrinsic parameter matrix of the camera, matrix It includes the rotation matrix R and the translation vector t, which are the pose parameters that need to be solved in the end.

[0020] By iteratively optimizing the spatial projection mapping relationship, the six-degree-of-freedom (6DOF) pose data of the UAV relative to the identifiable visual coded marker is obtained. The six-DOF pose data includes horizontal lateral offset, horizontal longitudinal offset, vertical height, roll angle, pitch angle, and yaw angle. That is, the iterative optimization based on the random sampling consensus algorithm can effectively eliminate mismatched corner points, thereby outputting a complete six-DOF pose that accurately reflects the spatial relative state of the UAV.

[0021] The horizontal lateral offset, horizontal longitudinal offset, and yaw angle are extracted from the six-degree-of-freedom pose data for UAV horizontal closed-loop position control. The vertical altitude is extracted for UAV vertical closed-loop altitude control. The roll angle and pitch angle are used for system health monitoring and stability compensation. In the control loop, the offsets in the X and Y directions and the yaw angle directly determine the UAV's horizontal approach maneuver. The roll angle and pitch angle are typically not directly used as inputs for horizontal position control, but are primarily used for UAV attitude stability compensation to prevent excessive tilting caused by external wind disturbances, while also serving as indicators for monitoring system health.

[0022] In step S102, during the adaptive switching decision stage: during the descent of the UAV, the current relative altitude of the UAV and the imaging proportion of the identifiable visual coded marker in the current image are acquired in real time, and it is determined whether the current relative altitude and the imaging proportion simultaneously meet the preset switching conditions.

[0023] The core of this decision-making mechanism lies in employing dual threshold conditions of relative height and image proportion. Since the specific physical size of visually encoded markers may vary in different scenarios, relying solely on a single height data or a single image proportion to trigger switching can easily lead to switching too early or too late.

[0024] The step of determining whether the current relative height and the imaging proportion simultaneously meet the preset switching conditions includes: Set the main camera to sub-camera switching trigger condition and the sub-camera to main camera switching trigger condition. A hysteresis interval is formed between the main camera to sub-camera switching trigger condition and the sub-camera to main camera switching trigger condition to avoid frequent switching of image acquisition source within the critical height range. The purpose of setting a hysteresis interval is to cope with fluctuations near the critical altitude. During descent, drones may experience slight fluctuations in altitude due to light wind disturbances. Without hysteresis logic, these slight fluctuations would cause the system to perform high-frequency, ineffective switching between the main and secondary cameras, severely consuming processing resources and causing abrupt changes in control commands. The trigger condition for switching from the main camera to the secondary camera is set as follows: the current relative height is less than or equal to a preset descent height threshold, and the imaging ratio is greater than or equal to a preset magnification ratio threshold. For example, when the drone descends to a height of 3 to 5 meters and the imaging ratio of the visual coding mark in the image reaches 70% to 85%, it indicates that the current main camera image has enough pixels for stable calculation and is about to exceed the narrow field of view of the main camera. At this time, the condition for switching downwards is met.

[0025] The trigger condition for switching from the secondary camera to the primary camera is set as follows: the current relative height is greater than or equal to a preset rise height threshold, and the imaging ratio is less than or equal to a preset shrinkage ratio threshold. To form the hysteresis interval, the descent height threshold is set to be less than the ascent height threshold, while the magnification ratio threshold is set to be greater than the shrinkage ratio threshold.

[0026] By setting the above parameter combination, we can not only ensure that the visually encoded marker always has enough pixels in the picture for positioning, but also reserve sufficient field of view margin for the drone's attitude swaying.

[0027] In step S103, during the smooth transition and pose fusion stage: when the current relative height and the imaging ratio simultaneously meet the preset switching conditions, the image acquisition source is switched from the main camera to the secondary camera. During the switching process, the last valid pose output by the main camera is inherited as the prior pose, and pose prediction is performed in combination with the inertial navigation data of the UAV. The predicted pose is fused with the pose initially calculated by the secondary camera to output smooth and continuous transition pose information, so as to keep the control closed loop of the UAV uninterrupted.

[0028] Because the main camera has a narrow field of view, the target will inevitably be lost as the altitude continues to decrease. At this point, a smooth switch to the secondary camera with a larger field of view (e.g., no less than 120 degrees) ensures that the target remains within the frame. Introducing inertial navigation data for pose prediction and fusion effectively fills the brief data gaps caused by hardware switching and image processing, truly achieving continuous positioning without frame loss.

[0029] The specific calculation steps for the transition pose information include: By utilizing the real-time high-frequency motion acceleration and angular velocity data provided by the UAV's inertial measurement unit, motion state integration is performed on the inherited prior pose to obtain a short-term predicted pose used to compensate for the short-term delay caused by camera switching and data processing. The inertial measurement unit can provide high-frequency and low-latency motion data, and by integrating acceleration and angular velocity within a very short time window, it can very accurately predict the instantaneous relative displacement and attitude changes of the UAV.

[0030] When the secondary camera completes the acquisition of a new frame of image and calculates the initial observation pose, a filtering estimation model is constructed. The short-term predicted pose is used as the predicted value for state update, and the initial observed pose is used as the observed value for measurement update. These are synchronously input into the filtered estimation model for weighted smooth fusion calculation, outputting a continuous pose sequence that eliminates step jumps to the UAV flight controller's position and attitude loops, maintaining the smoothness of the landing command throughout the entire process. The fused continuous pose sequence can be smoothly received by the flight controller, which integrates visual positioning results, inertial data, and altimeter data to stably output uninterrupted landing commands.

[0031] In step S104, during the low-altitude precision landing phase: when the drone is in a second altitude range lower than the first altitude range, the secondary camera is activated to capture a wide-angle image containing the complete identifiable visual coding mark. Based on the wide-angle image, the second pose information of the drone relative to the identifiable visual coding mark is continuously calculated. The horizontal position and yaw attitude of the drone are finely adjusted according to the second pose information until the drone meets the preset ground contact conditions and completes the landing.

[0032] At this stage, for example when the altitude decreases to about 0.5 to 1 meter, even if the drone's position shifts due to crosswinds at extremely close range, the wide-angle imaging of the secondary camera can still ensure that the visually encoded markers remain completely within the frame. The system continues to perform high-precision calculations and fine-tuning, executing the vertical descent maneuver.

[0033] The determination logic for the drone to complete landing under preset ground contact conditions is based on the simultaneous fulfillment of multiple sub-conditions to prevent accidental locking: The current absolute altitude value above the ground measured by the UAV's onboard ranging sensor is consistently lower than the preset ground contact safety height limit; The descent speed of the UAV in the vertical downward direction, calculated by integrating airborne vision and inertial data, is consistently lower than the preset descent speed limit. Received a mechanical physical trigger signal from the UAV landing gear contacting the ground; When the above sub-conditions are continuously met within the preset time window, the ground contact is determined to be successful, triggering the UAV landing lock-on procedure, stopping the attitude calculation and horizontal power output, maintaining only attitude stability and gradually locking the rotor motors.

[0034] This multi-dimensional cross-validation logic fundamentally prevents the risk of misjudgment and ground contact caused by relying solely on visual calculations, ensuring the absolute safety of the final ground contact and locking action.

[0035] In one embodiment, prior to the high-altitude guidance phase, the method further includes a system coordinate unification and intrinsic parameter compensation initialization step: The perspective distortion parameters of the main camera are calibrated, the wide-angle fisheye distortion parameters of the secondary camera are calibrated, and the distortion correction parameters are injected into the airborne image processing front end for real-time distortion elimination to ensure the calculation accuracy under low-altitude distortion field of view. Spatial joint calibration is used to obtain the extrinsic spatial transformation relationship matrices between the main camera, the secondary camera and the UAV inertial measurement unit coordinate system, respectively.

[0036] At the instant of switching into the smooth transition and pose fusion phase, the intrinsic parameters of the secondary camera and the distortion correction parameters are immediately invoked to process the image. The pose calculation module is not reinitialized, and all calculated pose data are uniformly transformed into the coordinate system of the UAV's inertial measurement unit according to the extrinsic parameter spatial transformation relationship matrix for global representation. Whether the original relative coordinates extracted by the main camera or the secondary camera, after transformation by the extrinsic parameter spatial transformation relationship matrix, they can be unified to the same physical reference system, ensuring that the spatial coordinate reference altitude received by the flight control algorithm before and after the visual switch is consistent.

[0037] In one embodiment, when the identifiable visually encoded marker is deployed on a mobile carrier helipad in a dynamically swaying state, the method further includes a dynamic adaptive compensation step: The motion compensation module based on inertial navigation prediction is activated to estimate in real time the undulation and tilt attitude changes of the mobile carrier's helipad caused by environmental disturbances; Based on the estimated fluctuations and tilt changes, the target tracking trajectory of the UAV control system is dynamically adjusted and compensated to counteract external interference from the carrier motion on the visual relative pose measurement. The system synchronously and adaptively fine-tunes the height threshold and percentage threshold parameters in the preset switching conditions based on the dynamic swaying amplitude of the mobile carrier's helipad.

[0038] The uncertainty introduced by the moving carrier makes static thresholds prone to failure. Real-time adaptive fine-tuning can effectively avoid erroneous switching caused by sudden carrier turbulence leading to markers instantly moving out of the field of view or abrupt changes in proportion, significantly enhancing the system's adaptability in complex dynamic scenarios.

[0039] Based on the same line of thought, such as Figure 2 The diagram shown is a structural block diagram of a drone precision landing system based on dual-camera adaptive switching, according to an embodiment of the present invention. The system includes: The main and secondary dual-camera visual acquisition hardware module 201 includes a high-resolution main camera with a narrow field of view and a wide-angle secondary camera with a wide field of view. It can also be integrated into a dual-focal-length integrated optoelectronic pod that achieves optical zoom by switching between internal prisms and sensors, and is used to capture images of ground-recognizable visually coded markers covering the entire altitude range. The pose calculation and dual decision processing module 202 is communicatively connected to the main and secondary dual-camera visual acquisition hardware module, and is used to perform image decoding, corner feature extraction, continuous six degrees of freedom pose calculation, and adaptive acquisition source switching decision based on parameter threshold and hysteresis logic. The smooth fusion and inertial compensation module 203 is used to inherit the historical prior pose during the switching of image acquisition sources, and combine it with the UAV inertial navigation data for filtering and fusion output to eliminate control command jumps; The flight control closed-loop execution module 204 receives the continuous attitude signal output by the smooth fusion and inertial compensation module, and drives the rotor system to perform high-altitude approach, smooth transition, low-altitude fine adjustment, vertical descent and safe ground touchdown locking actions without interrupting the control loop.

[0040] The specific details of the above system have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0041] Compared with existing technologies, this system improves the control method for open-circuit fault conditions, enabling the three-phase induction motor to maintain continuous operation even when current imbalance and related disturbances occur due to stator phase open circuits. This reduces speed fluctuations and electromagnetic torque pulsations, and improves the dynamic response during the fault recovery phase, thereby enhancing the system's operational stability and reliability. Furthermore, this invention eliminates the need for independent and complex fault detection mechanisms and additional fault-tolerant mode switching procedures, reducing implementation complexity and application costs. It is suitable for drive scenarios requiring high reliability and continuous operation.

[0042] The accompanying drawings are merely illustrative of the processes included in the methods according to exemplary embodiments of this disclosure and are not intended to be limiting. It is readily understood that the processes shown in the drawings do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0043] It should be noted that although several modules or units of the system have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0044] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

[0045] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for precise landing of a drone based on adaptive switching of dual cameras, characterized in that, Applied to drones equipped with a main camera and a secondary camera, the method includes the following stages: High-altitude guidance phase: When the UAV is in the first altitude range, the main camera is activated to capture high-definition images containing ground-identifiable visually coded markers, and the first pose information of the UAV relative to the identifiable visually coded markers is calculated based on the high-definition images. The UAV is then controlled to fly towards the location of the identifiable visually coded markers according to the first pose information. Adaptive switching decision phase: During the descent of the UAV, the current relative altitude of the UAV and the imaging proportion of the identifiable visual coded mark in the current image are acquired in real time, and it is determined whether the current relative altitude and the imaging proportion simultaneously meet the preset switching conditions. Smooth transition and pose fusion stage: When the current relative height and the imaging ratio simultaneously meet the preset switching conditions, the image acquisition source is switched from the main camera to the secondary camera. During the switching process, the last valid pose output by the main camera is inherited as the prior pose, and pose prediction is performed in combination with the inertial navigation data of the UAV. The predicted pose is fused with the pose initially calculated by the secondary camera to output smooth and continuous transition pose information, so as to keep the control closed loop of the UAV uninterrupted. Low-altitude precision landing phase: When the drone is in a second altitude range lower than the first altitude range, the secondary camera is activated to capture a wide-angle image containing the complete identifiable visual coding mark. Based on the wide-angle image, the second pose information of the drone relative to the identifiable visual coding mark is continuously calculated. The horizontal position and yaw attitude of the drone are finely adjusted according to the second pose information until the drone meets the preset ground contact conditions and completes the landing.

2. The method for precise drone landing based on dual-camera adaptive switching according to claim 1, characterized in that, The step of determining whether the current relative height and the imaging percentage simultaneously meet preset switching conditions includes: Set the main camera to sub-camera switching trigger condition and the sub-camera to main camera switching trigger condition. A hysteresis interval is formed between the main camera to sub-camera switching trigger condition and the sub-camera to main camera switching trigger condition to avoid frequent switching of image acquisition source within the critical height range. The trigger condition for switching from the main camera to the secondary camera is set as follows: the current relative height is less than or equal to a preset descent height threshold, and the imaging ratio is greater than or equal to a preset magnification ratio threshold. The trigger condition for switching from the secondary camera to the primary camera is set as follows: the current relative height is greater than or equal to a preset rise height threshold, and the imaging ratio is less than or equal to a preset shrinkage ratio threshold. To form the hysteresis interval, the descent height threshold is set to be less than the ascent height threshold, while the magnification ratio threshold is set to be greater than the shrinkage ratio threshold.

3. The method for precise drone landing based on dual-camera adaptive switching according to claim 1, characterized in that, The algorithm implementation process for calculating the first pose information of the UAV relative to the identifiable visual coded marker based on the high-definition image and the second pose information of the UAV relative to the identifiable visual coded marker based on the wide-angle image specifically includes: Extract the two-dimensional pixel coordinates of multiple edge corner points of the identifiable visually encoded marker in the currently captured image; Obtain the prior information of the three-dimensional physical dimensions of the identifiable visually encoded mark pre-calibrated during the initialization phase; Using a perspective point projection pose estimation algorithm, combined with the two-dimensional pixel coordinates and the three-dimensional physical size prior information, a spatial projection mapping relationship is constructed; By iteratively optimizing and solving the spatial projection mapping relationship, the six-degree-of-freedom pose data of the UAV relative to the identifiable visual coded mark is obtained. The six-degree-of-freedom pose data includes horizontal lateral offset, horizontal longitudinal offset, vertical height, roll angle, pitch angle, and yaw angle. The horizontal lateral offset, the horizontal longitudinal offset, and the yaw angle are extracted from the six-degree-of-freedom pose data for UAV horizontal closed-loop position control. The vertical height is extracted for UAV vertical closed-loop altitude control. The roll angle and the pitch angle are used for system health monitoring and stability compensation.

4. The method for precise drone landing based on dual-camera adaptive switching according to claim 1, characterized in that, Prior to the high-altitude guidance phase, the method further includes a system coordinate unification and intrinsic parameter compensation initialization step: The perspective distortion parameters of the main camera are calibrated, the wide-angle fisheye distortion parameters of the secondary camera are calibrated, and the distortion correction parameters are injected into the airborne image processing front end for real-time distortion elimination to ensure the calculation accuracy under low-altitude distortion field of view. Spatial joint calibration is used to obtain the external parameter spatial transformation relationship matrix between the main camera, the secondary camera and the coordinate system of the UAV's inertial measurement unit. At the instant of switching into the smooth transition and pose fusion stage, the intrinsic parameters of the secondary camera and the distortion correction parameters are immediately invoked to process the image. The pose calculation module is not re-initialized, and all the calculated pose data are uniformly converted to the coordinate system of the UAV body inertial measurement unit according to the extrinsic parameter spatial transformation relationship matrix for global expression.

5. The method for precise drone landing based on dual-camera adaptive switching according to claim 1, characterized in that, The step of fusing the predicted pose with the pose initially calculated by the secondary camera to output smooth and continuous transition pose information includes the following specific calculation steps: Using real-time high-frequency motion acceleration and angular velocity data provided by the UAV inertial measurement unit, motion state integral calculation is performed on the inherited prior pose to obtain a short-time predicted pose for compensating for the short-time delay caused by camera switching and data processing. When the secondary camera completes the acquisition of a new frame of image and calculates the initial observation pose, a filtering estimation model is constructed. The short-term predicted pose is used as the predicted value for state update, and the initial observed pose is used as the observed value for measurement update. These are synchronously input into the filter estimation model for weighted smooth fusion calculation, and the continuous pose sequence that eliminates step jumps is output to the position loop and attitude loop of the UAV flight controller to maintain the smoothness of the landing command throughout the entire process.

6. The method for precise drone landing based on dual-camera adaptive switching according to claim 1, characterized in that, The decision-making logic for the drone to complete landing under preset ground contact conditions is based on the simultaneous fulfillment of multiple sub-conditions to prevent accidental locking: The current absolute altitude value above the ground measured by the UAV's onboard ranging sensor is consistently lower than the preset ground contact safety height limit; The descent speed of the UAV in the vertical downward direction, calculated by integrating airborne vision and inertial data, is consistently lower than the preset descent speed limit. Received a mechanical physical trigger signal from the UAV landing gear contacting the ground; When the above sub-conditions are continuously met within the preset time window, the ground contact is determined to be successful, triggering the UAV landing lock-on procedure, stopping the attitude calculation and horizontal power output, maintaining only attitude stability and gradually locking the rotor motors.

7. The method for precise drone landing based on dual-camera adaptive switching according to claim 1, characterized in that, When the identifiable visually encoded marker is deployed on a mobile vehicle helipad in a dynamically swaying state, the method further includes a dynamic adaptive compensation step: The motion compensation module based on inertial navigation prediction is activated to estimate in real time the undulation and tilt attitude changes of the mobile carrier's helipad caused by environmental disturbances; Based on the estimated fluctuations and tilt changes, the target tracking trajectory of the UAV control system is dynamically adjusted and compensated to counteract external interference from the carrier motion on the visual relative pose measurement. The system synchronously and adaptively fine-tunes the height threshold and percentage threshold parameters in the preset switching conditions based on the dynamic swaying amplitude of the mobile carrier's helipad.

8. A drone precision landing system based on dual-camera adaptive switching, applied to perform the drone precision landing method based on dual-camera adaptive switching as described in any one of claims 1 to 7, the system comprising: The main and secondary dual-camera visual acquisition hardware module includes a high-resolution main camera with a narrow field of view and a wide-angle secondary camera with a wide field of view. It can also be integrated into a dual-focal-length integrated optoelectronic pod that achieves optical zoom by switching between internal prisms and sensors, and is used to capture images of ground-recognizable visually coded markers across the entire altitude range. The pose calculation and dual decision processing module is communicatively connected to the main and secondary dual-camera visual acquisition hardware module, and is used to perform image decoding, corner feature extraction, continuous six-degree-of-freedom pose calculation, and adaptive acquisition source switching decision based on parameter threshold and hysteresis logic. The smooth fusion and inertial compensation module is used to inherit the historical prior pose during the switching of image acquisition sources, and combine it with the UAV inertial navigation data for filtering and fusion output to eliminate control command jumps; The flight control closed-loop execution module receives the continuous attitude signals output by the smooth fusion and inertial compensation module, and drives the rotor system to perform high-altitude approach, smooth transition, low-altitude fine-tuning, vertical descent and safe ground touchdown locking actions without interrupting the control loop.

Images