125 results about "Elevation map" patented technology

The invention provides an elevation map path planning method and system based on a terrain adaptive potential energy field, and relates to the technical field of path planning. Comprising the steps of calculating the gradient and roughness of the position of an intelligent agent, setting a terrain adaptive gain, dynamically adjusting the gradient and roughness, and obtaining a terrain adaptive repulsive force field and a composite potential energy field in combination with a basic repulsive force field; determining the motion direction of the agent based on the composite potential energy field, and determining the next position of the agent; in the process that the intelligent agent moves to the next position, whether local re-planning needs to be carried out on the next position of the intelligent agent or not is judged based on a preset environment change judgment rule; and when re-planning is needed, recalculating the terrain adaptive repulsive force field and the dynamic obstacle repulsive force field, updating the composite potential energy field, and re-determining the next position of the intelligent body. According to the invention, the terrain can be effectively sensed, the planning strategy is adaptively adjusted, and high precision, high efficiency and strong robustness are considered.

The invention discloses an image recognition system for industrial part defect detection, and particularly relates to the field of part defect detection, which comprises a multi-axis controllable light source array module, a high-speed polarization camera module, an edge computing node module, a double-branch semantic segmentation network module and a physical constraint post-processing module, according to the invention, through combination of time-sharing stroboscopic illumination and polarization image sequence acquisition, multi-dimensional perception of surface topography and material differences is realized; generating an elevation map and a normal map by using photometric stereo solution, constructing a differential rendering layer reverse matching CAD model, and extracting flash sensitive features; a double-branch U-Net network is adopted to fuse geometric and polarization characteristics, the characterization capability is enhanced through a trans-attention mechanism, and a pixel-level mask is output; and finally, mapping a two-dimensional result to a three-dimensional coordinate system by means of calibration parameters, carrying out geometric verification in combination with a tolerance zone and a height threshold value, and automatically generating a structured defect report containing position, size, grade and visual information.

The invention provides a mechanical arm real-time stacking model generation method and device based on a 2.5 D elevation map. The method comprises the steps that depth point cloud data of a stacking area is obtained; preprocessing the depth point cloud data to obtain de-noised point cloud data; performing external parameter calibration by adopting Eye-to-hand to obtain an accurate conversion relation between a camera coordinate system and a mechanical arm flange coordinate system, and performing coordinate system conversion to obtain de-noised point cloud data under a unified world coordinate system; performing space clipping on the de-noised point cloud data to obtain effective point cloud data in the stacking area range; a mechanical arm directional bounding box is generated under the unified world coordinate system, and whether the mechanical arm directional bounding box intersects with the stacking area or not is detected through the separation axis theorem; if not, a 2.5 D elevation map is generated according to the effective point cloud data in the stacking area range, and a real-time stacking model of the mechanical arm is obtained according to the 2.5 D elevation map; therefore, quick response and flexible adjustment are realized in a dynamic scene.

The invention discloses a quadruped robot landing planning method based on scene decoupling and risk avoidance. The quadruped robot landing planning method aims at solving the problems that in the prior art, perception is not precise in a complex environment, and the obstacle avoidance capacity is insufficient. The method comprises the following steps: acquiring and fusing multi-source sensor data, and segmenting an original point cloud; generating a scene decoupling elevation map of a multi-layer structure by using the segmentation point cloud; constructing a multi-objective optimization problem taking risk avoidance as a core based on the elevation map, wherein the multi-objective optimization problem comprises a comprehensive cost function and a strict obstacle avoidance constraint; utilizing a reaction formula adjusting module of a capturable region theory to cope with a dynamic instability risk; and solving the multi-target and multi-constraint optimization problem in real time by adopting a hierarchical solving strategy, and finally generating an optimal foot end drop point considering safety and stability. According to the method, the terrain adaptability, the motion stability and the decision intelligence of the quadruped robot in an unstructured environment are remarkably improved through fine decoupling of a scene and quantitative avoidance of multi-source risks.

The invention discloses a method for automatically generating an ancient building elevation by using a multi-modal neural network, and the method comprises the steps: obtaining and preprocessing a three-dimensional inclined point cloud model and a three-dimensional laser point cloud model of a target ancient building, and obtaining the indoor and outdoor point cloud data of the ancient building; performing feature point matching and rigid body transformation fusion on the ancient building indoor and outdoor point cloud data to obtain a three-dimensional ancient building point cloud model; performing semantic segmentation on the three-dimensional ancient building point cloud model through projection and a multi-modal neural network to obtain each structure area of the ancient building; and according to each structural region of the ancient building, extracting each component contour from the three-dimensional point cloud model to generate an ancient building elevation.

The invention discloses a non-flat terrain autonomous navigation method based on path pruning and trajectory smoothing. The method comprises the following steps: constructing an elevation map with three-dimensional map information as a navigation map according to robot navigation space information; based on the navigation map, performing global path search by adopting a path pruning strategy to obtain an initial path from the starting point to the target point; based on the initial path, track optimization is carried out by using a B spline curve, and a smooth track meeting robot dynamics constraints is generated; in response to an obstacle which is detected on the smooth trajectory and cannot be avoided, performing three-dimensional obstacle crossing trajectory planning based on terrain elevation information of the elevation map, and generating an obstacle crossing trajectory capable of crossing the obstacle; wherein the robot performs autonomous navigation based on the smooth track or the obstacle crossing track. According to the invention, efficient global search, smooth trajectory optimization and intelligent obstacle crossing decision of the robot in unknown and non-flat terrains can be realized.

The invention provides a cross-layer path planning method and system based on a multilayer elevation map, and the method comprises the steps: obtaining three-dimensional point cloud data of a multilayer environment, carrying out the adaptive layering of the three-dimensional point cloud data, and obtaining the number of layers of a point cloud group correspondingly covering all elevation information; constructing an elevation map of each layer of point cloud group; calculating the height difference between the current grid and the adjacent grid to obtain the trafficability of the current grid, analyzing the terrain, and optimizing the elevation map; identifying an effective point at the junction of each layer as a passing gateway point; and performing cross-layer path search based on the constructed multi-layer elevation map and the passage gateway point. According to the method, seamless path planning of cross-layer scenes such as stairs and slopes is realized based on the multi-layer elevation map and the layering strategy, and the problem of path breakage during interlayer transition of a traditional method is solved.

According to the hybrid elevation map generation method based on multi-modal data fusion and reinforcement learning, through vision-LiDAR deep fusion, direction perception feature coding, space-time consistency optimization and reinforcement learning dynamic parameter adjustment, elevation map generation with 15-meter long distance and 0.05-meter / pixel high resolution in a medium-speed unmanned driving scene is realized. According to the method, the elevation error is reduced to be lower than 15 cm, the path planning success rate is increased to 90%, the system cost is reduced by 70%, and robust navigation support is provided for a complex unstructured environment.

The invention discloses an unmanned aerial vehicle route global planning method based on key feature points, and relates to the field of unmanned aerial vehicle route planning, and the method comprises the steps: carrying out the hierarchical division of a three-dimensional elevation map according to the performance of an unmanned aerial vehicle; according to obstacle pixel proportions of different layers, an optimal layer is screened out; on the basis of the segmentation maps of the optimal layer and the adjacent layer and in combination with the three-dimensional geographic features, obtaining corrected key feature points; utilizing a heuristic weight calculation formula to calculate heuristic weight values of the corrected key feature points; according to a sorting result of the heuristic weight values, a plurality of corrected key feature points of the unmanned aerial vehicle route are obtained according to a connection barrier-free principle and in combination with a three-dimensional path algorithm until a target point is selected; and based on the initial point, the target point and the plurality of corrected key feature points of the unmanned aerial vehicle route, generating an optimized unmanned aerial vehicle flight path after corner optimization. According to the method, the three-dimensional feasible flight route can be efficiently generated, and the calculation time is short.

The invention discloses a wheat harvesting navigation path generation method and an automatic harvester, and relates to the technical field of automatic driving of agricultural machinery, and the wheat harvesting navigation path generation method comprises the following steps: S1, collecting three-dimensional point cloud data of a wheat field through a laser radar, and enabling a collection region to at least cover the region boundary of a harvested region and a non-harvested region; s2, performing noise reduction and voxel grid downsampling processing on the three-dimensional point cloud data to generate a two-dimensional elevation map; s3, setting grid characteristics and threshold values thereof according to the wheat variety height characteristics; and S4, carrying out binarization processing on the two-dimensional elevation map pixel by pixel based on the grid features and the threshold values thereof. According to the wheat harvesting navigation path generation method, boundary extraction and path conversion are carried out based on radar point cloud, the method has the characteristics of robustness and real-time performance, an automatic harvester based on the method can reduce the recutting rate and land rolling compaction during operation, and resources are effectively saved.

The invention provides an underground pipe network detection positioning method and device, and the method comprises the steps: obtaining the multi-source data of an underground pipe network of a to-be-detected point and the surrounding environment, including ground image data, inspection well internal image data, adjacent inspection well spacing data, ground elevation data, pipe bottom elevation data, and position coordinates; inputting the ground image data into a semantic segmentation network, and outputting a ground semantic segmentation image; respectively extracting contours of the ground semantic segmentation image and the image data in the inspection well to obtain a first edge image and a second edge image; constructing a pavement information measurement model; pipeline diameter information, pipeline gradient and rain and sewage pipeline mixed connection conditions are determined; and integrating the pavement information measurement model, the first edge image, the second edge image, the pipeline diameter information, the pipeline gradient and the like, and drawing a planar graph and an elevation graph of the underground drainage pipe network. According to the method, high-precision detection and positioning of the underground pipe network are remarkably improved, and the method is suitable for city updating and city informatization construction.

The invention discloses a single-frame unmanned aerial vehicle image pixel positioning method and system based on an elevation map, and relates to the technical field of unmanned aerial vehicle image processing and geographic positioning, and the method comprises the steps: obtaining a single-frame original image shot by an unmanned aerial vehicle, unmanned aerial vehicle attitude information during shooting, camera internal and external parameters, and a digital elevation map of a shooting region; according to the internal reference of the camera, converting the pixel coordinate of the target pixel in the original image into a normalized plane coordinate system of the camera, and carrying out coordinate distortion correction to obtain the back projection point coordinate of the target pixel; converting a camera-unmanned aerial vehicle body coordinate system and a body-geographic inertial reference coordinate system according to camera external parameters and unmanned aerial vehicle attitude information to obtain normalized ray direction vectors corresponding to target pixels; an earth surface height constraint is constructed based on a digital elevation map, and an intersection point of a ray and the ground is determined by performing iterative approximation on a space point along a ray direction. According to the invention, rapid and accurate conversion from the single-frame unmanned aerial vehicle image pixel to the geographic coordinate can be realized.

The invention discloses a digital surface model (DSM)-guided hyperspectral and LiDAR combined unmixing method, relates to the field of multi-modal image processing, and aims to solve the problems of end member confusion and insufficient space structure maintenance caused by spectrum similarity in hyperspectral unmixing. According to the method, a hyperspectral image and LiDAR data of the same area are obtained, a LiDAR elevation map is expanded into a multiband profile through an attribute configuration file method, and a digital surface model (DSM) is generated. Constructing a spectrum and space double-branch auto-encoder, respectively extracting spectrum and space features and carrying out fusion mapping, obtaining an abundance matrix by using a normalized exponential function, and reconstructing a hyperspectral image; a double-branch adaptive mixed channel attention mechanism is designed in a spectrum branch, and a space attention mechanism is introduced in a space branch. In the training process, a self-defined loss function is formed by combining DSM-guided structure entropy regularization, spectral angular distance and root-mean-square error, and the space continuity and boundary retention of the abundance graph are improved. The method is suitable for hyperspectral unmixing and multi-modal remote sensing fine identification under complex terrains.

The invention relates to a training method of a perception and control model, and a robot control method and device. The method comprises the following steps: acquiring a plurality of sample image sets; wherein each sample image set comprises a depth image and an elevation image which represent the same terrain; performing multi-stage training on the initial model according to the plurality of sample image sets to obtain a sensing and control model for controlling the action of the robot; wherein the multi-stage training comprises the following steps of: training an elevation image codec by utilizing an elevation image to obtain a trained elevation image codec; training a depth image codec by using the depth image and the trained elevation image codec to obtain a trained depth image codec; and performing reinforcement learning training on the action strategy neural network by using the trained depth image codec, thereby facilitating the improvement of the model training efficiency through the interaction of perception and control, helping the robot to accurately perceive the working environment, and realizing accurate action control.

The application discloses a kind of single frame unmanned plane image pixel positioning method and system based on elevation map, it is related to unmanned plane image processing and geographic positioning technical field, obtain the single frame original image of unmanned plane shooting and the digital elevation map of the shooting area when unmanned plane attitude information, camera internal and external parameters, shooting;According to camera internal parameter, the pixel coordinates of target pixel in original image are converted to camera normalization plane coordinate system, and target pixel back projection point coordinates are obtained after coordinate distortion correction;According to camera external parameter and unmanned plane attitude information, camera-unmanned plane body coordinate system, body-geographic inertial reference coordinate system conversion is carried out, and the normalized ray direction vector corresponding to target pixel is obtained;Based on digital elevation map, construct ground height constraint, by iterative approximation to space point along ray direction, determine the intersection of ray and ground.The application can realize the fast and accurate conversion of single frame unmanned plane image pixel to geographic coordinate.

This invention relates to a method for constructing an elevation map, belonging to the field of autonomous navigation technology, and solves the problem of driving safety of existing unmanned tracked vehicles in complex environments. The method includes: determining the vehicle's first pose information using point cloud data; determining the vehicle's second pose information using inertial navigation data; extracting a global descriptor from the point cloud data and performing loop closure detection to obtain the vehicle's third pose information; constructing and solving a factor graph with the first, second, and third pose information as observation factors to obtain the vehicle's pose information; stitching the point cloud data based on the pose information to obtain a point cloud map; clustering the point clouds in the point cloud map and constructing an elevation map based on the clustering results.

Embodiments of the present disclosure disclose an airport abnormal area identification method and device, electronic equipment and readable medium. A specific embodiment of the method comprises: selecting reference area data and contrast area data; performing registration correction on the reference area remote sensing image and the contrast area remote sensing image to generate a corrected reference area image and a corrected contrast area image; performing area shadow detection on the corrected reference area image and the corrected contrast area image; performing constraint feature fusion on the corrected reference area image and the corrected contrast area image respectively to generate reference area main features and contrast area main features; performing abnormal area identification on the reference area main features and the contrast area main features; and performing verification grading on the candidate abnormal area identification information to obtain airport target area abnormal information. The embodiment can perform shadow detection on the airport area to identify abnormal shadow areas, and perform verification on the abnormal areas through a grid elevation map, thereby improving the safety of the airport area clearance.

The application discloses a building height detection method and device, a storage medium and an electronic device, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring a target remote sensing image corresponding to a target building, wherein the target building is a building to be detected in height; inputting the target remote sensing image into a target detection model for detection processing, and outputting height information of the target building, wherein the target detection model is a model generated by learning and training a target neural network model by using M remote sensing images, M shadow maps and M target elevation maps, the target neural network model is a neural network model with residual connection, the remote sensing image is a remote sensing image corresponding to a building, the shadow map is a shadow map of the building, the target elevation map is used for representing height information of the building, and M is a positive integer. Through the application, the problem that the effect of detecting the height of a building is poor in the related art is solved.

The embodiment of the invention provides a movement control method for a foot type robot. The movement control method is used for controlling the foot type robot to stably and accurately move in diversified terrains. The foot-type robot is loaded with a depth camera and an inertial measurement unit IMU, and the movement control method comprises the steps that in the walking process of the foot-type robot, point cloud data, IMU attitude data, remote control instruction data and motor state data collected by the depth camera are obtained; generating elevation map data based on the point cloud data and the IMU attitude data; inputting the elevation map data, the remote control instruction data, the IMU attitude data and the motor state data into a pre-trained reinforcement learning model to generate a motor control instruction; the reinforcement learning model is obtained by configuring various terrain scenes in a simulation training environment and training by adopting a priority sampling strategy; according to the priority sampling strategy, the sampling frequencies of different learning action sequences are determined according to the difficulty priorities of the different learning action sequences; and the motor control instruction is issued to a motor control system of the foot type robot, and the foot type robot is driven to execute corresponding actions.

The invention relates to the technical field of robot positioning, discloses a four-foot robot three-dimensional positioning method and system, a storage medium and a robot, and aims to solve the technical problems of height drift and foot end error accumulation of a four-foot robot in a visual degradation environment without a GPS (Global Positioning System) by a traditional positioning method. The method comprises the following steps: firstly, constructing an initial state vector containing motion states of a fuselage and four foot ends, and establishing a nonlinear continuous time state equation considering foot end dynamics; relying on an unscented Kalman filtering framework, a prior state is obtained through Sigma point sampling and time updating; a foot end observation value is corrected through foot bottom rolling compensation, a height absolute reference is provided by combining a local invariant elevation map, and multi-leg slippage detection and observation weight self-adaptive adjustment are synchronously completed; and finally, the optimal state is updated and output through observation. According to the invention, the visual sensor is used for topographic mapping instead of direct positioning, the height drift is effectively inhibited, the robustness and precision of positioning in a complex environment are improved, and the method is suitable for unstructured scenes such as rescue exploration and the like.

The application discloses an extraterrestrial celestial body task area search path planning method based on an elevation map and belongs to the technical field of space robots. The method is as follows: the slope data in a task area is solved through the point set data of the elevation map, and the alternative passing position points of the rover are determined; the planning grid size is determined, and whether the grid is passable is determined according to the proportion of the number of passable points in the grid to the total number as a judgment standard; the exploration value of each grid is determined, and a weight value is assigned to the grid. The current grid is taken as a center point, the detectable nodes in a predetermined range around the current grid are marked as detected nodes, and the weight values of the detected nodes are reduced; the value of each grid in the passable range around the current grid is judged with the current grid as a center point, if there is a grid with a value not less than an end threshold, the grid with the maximum value is selected as a next path node, the above steps are repeated, and the extraterrestrial celestial body task area search path planning based on the elevation map is realized.

The present invention provides a method and system for collaborative motion control of a quadruped single-arm robot, relating to the field of quadruped single-arm robot planning and control. The method comprises: establishing a multi-rigid body dynamics model and a single rigid body reduced-order model as model constraints; screening feasible landing areas; generating a safe landing area; formulating a landing point cost function and determining the optimal landing point; discretizing the state transition equation of the quadruped single-arm robot; introducing the traversable steps as an inequality constraint and the optimal landing point as an elevation map equality constraint into a model predictive controller; optimizing the multi-task planning method based on the multi-rigid body dynamics model, rationally allocating control strategies, dynamically coordinating conflicting tasks between quadruped motion and manipulator operation, and collaboratively controlling multiple tasks; and integrating target data from a depth camera at the end of the manipulator arm to autonomously plan the posture adjustment path of the robot body. The present invention significantly improves the stability and accuracy of collaborative grasping by the robot.

The invention provides a base station coverage rate prediction method and device, a storage medium and a program product. The method comprises the following steps: determining a line-of-sight / non-line-of-sight state of a first user test point and a base station according to first elevation map data and first base station deployment information so as to determine a wireless propagation model of the first user test point; calculating the base station coverage rate of the first target area according to the wireless propagation model of the first user test point; according to the first elevation map data, the first base station deployment information and the base station coverage rate of the first target area, training an initial coverage rate prediction model to obtain a trained coverage rate prediction model; wherein the initial coverage rate prediction model comprises a generator network and a discriminator network; wherein the trained coverage rate prediction model is used for outputting the base station coverage rate of the second target area according to the input second elevation map data of the second target area and the second base station deployment information. The problem of insufficient prediction precision in a complex terrain environment in the prior art is solved.

The invention provides a map data generation method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring point cloud data and motion state data, and fusing the point cloud data and the motion state data to obtain pose estimation data; performing ground elevation modeling based on the pose estimation data and the point cloud data to obtain an elevation map corresponding to the point cloud data; performing incremental mapping processing based on the elevation map corresponding to the multi-frame point cloud data to obtain a global elevation map; performing feature extraction on the global elevation map to obtain topographic feature data of each grid in the global elevation map, and determining a traffic risk score of each grid based on the topographic feature data of each grid; and based on the traffic risk score of each grid, updating the global elevation map to obtain a target map. According to the invention, the integrity of terrain modeling and the accuracy of traffic risk detection can be improved.

Methods and systems for improved generation and georeferencing of floor plans are presented. In one embodiment, a method is presented that includes receiving images that depict sheets of a blueprint of a structure. Subsets of the images depicting floor sheets and elevation sheets may be identified. Exterior contours may be extracted from the images depicting floor sheets and elevation contours may be extracted from the images depicting elevation sheets. A corresponding structure within a three-dimensional map may be identified based on the exterior contours and the elevation contours. A three-dimensional contour of the exterior of the structure may be extracted from the three-dimensional map.

The invention discloses a handwritten character recognition method and device based on multi-modal image fusion and a medium, and relates to the technical field of visual identification, and the method comprises the steps: carrying out the feature point matching and geometric transformation of a standardized multi-modal image set, obtaining registration multi-modal data, and generating a separation feature image set through physical-spectrum joint inversion; extracting stroke candidate segments from the ink mark region on the basis of the separation feature image set and the micro-elevation map, constructing stroke nodes, judging connectivity, and generating a stroke topological information set; constructing a multi-scale feature pyramid according to the separated feature image set, the registered multi-modal data and the stroke topological information set, generating a unified fusion feature image through cross-modal feature interaction, and performing text detection and line segmentation to generate a text region boundary and segmentation information; according to the method, accurate detection and row segmentation of the character region are realized, and the stroke integrity and the structure consistency are ensured, so that the character recognition accuracy and the system robustness in a complex handwritten manuscript are improved.

The invention relates to the technical field of robot environment perception, in particular to a real-time ground segmentation method and system. The method comprises the following steps: receiving and preprocessing laser radar point cloud data, and dividing the data into a plurality of concentric annular areas; carrying out adaptive ground fitting on each area and processing the multilayer structure through vertical plane fitting to obtain an initial ground point set; correcting an under-segmentation region of the current frame by using a time sequence backspacing mechanism and updating parameters; constructing an elevation map based on the initial point set, and obtaining a fine ground point set through adaptive resolution CSF simulation; and carrying out confidence evaluation on the two types of point sets, and generating a final result through multi-criterion fusion. Accurate segmentation of the complex terrain is realized, and robustness and real-time performance of a dynamic environment are improved.

The invention relates to the technical field of bulk material flow loading monitoring, in particular to a train compartment loading material flow height online monitoring method and device. The method comprises the following steps: continuously collecting point cloud data of carriages and internal material flows passing through a monitoring area, and performing preprocessing; determining a front boundary surface position parameter and a rear boundary surface position parameter of the carriage in the advancing direction based on the preprocessed point cloud data, and establishing a carriage model in combination with the prior parameters of the carriage; performing space judgment on the preprocessed point cloud data based on a carriage model to obtain a material flow point cloud in the carriage; based on the material flow point cloud, constructing a carriage material flow three-dimensional elevation model and a two-dimensional carriage material flow elevation map; and extracting a height statistical sequence from the two-dimensional compartment material flow elevation map, obtaining a height-time curve in combination with train running speed information, and outputting the three-dimensional compartment material flow elevation model and the height-time curve. The problems that online dynamic monitoring requirements are difficult to meet, and local loading abnormity cannot be found in time can be solved.

A method of estimating a radio map from environmental context comprises designating a training area and for each transmitter location in the training area, generating a channel parameter radio map that is centered on the transmitter and is based on a ray-tracing twin model and generating a two-dimensional elevation map, centered on the transmitter, corresponding to the training area. The method may further comprise training a U-Net-based or any other neural network-based model using the channel parameter radio maps and the two-dimensional building elevation maps corresponding to the transmitters in the area, to produce a trained model, and designating an area to be radio-mapped. The method may further comprise providing the trained with a two-dimensional input elevation map derived from a three-dimensional model of the area to be radio mapped, and producing, by the model, an estimate of the radio map based on the two-dimensional elevation map.

This application relates to the field of image processing technology and discloses an online correction method for the installation error of a UAV's downward-looking camera. The method includes: acquiring a first image frame and a second image frame; obtaining a third set of matching points and a fourth set of matching points based on the first image frame, the second image frame, and a satellite map; obtaining the intersection of the third set of matching points and the fourth set of matching points, and performing uniformization processing on the intersection to obtain a fifth set of matching points; generating an elevation map; obtaining the three-dimensional coordinates of each feature point in the fifth set of matching points based on the pixel coordinates matched in the satellite map, and the corresponding satellite map and elevation map; and obtaining the attitude error by solving a least-squares problem, and compensating for the current attitude of the UAV using the attitude error. This application can calculate the installation error of the downward-looking camera in real time, thereby ensuring the positioning accuracy of the UAV during visual navigation.