757 results about "Global Map" patented technology

The invention discloses an ultra-wideband laser radar inertial navigation cooperative SLAM (Simultaneous Localization and Mapping) method and system, which are suitable for robot positioning and mapping tasks in a GPS (Global Positioning System)-free environment. According to the method, high-frequency motion priori of an IMU, relative pose constraint of LiDAR and absolute ranging information of UWB are fused, a unified factor graph optimization model is constructed, and multi-sensor cooperative positioning is realized; in the system initialization stage, IMU bias calibration, UWB base station coordinate configuration and LiDAR initial attitude estimation are completed; in the operation process, IMU pre-integration is utilized to predict the pose, LiDAR point cloud registration is utilized to obtain the relative motion between key frames, UWB ranging is combined to construct a residual item, a self-adaptive weight mechanism is introduced to suppress ranging abnormity, and the fusion robustness is improved; the system supports geometric feature loopback detection, cross-frame constraints are constructed in combination with UWB to carry out closed-loop optimization, and an optimization result is used for real-time incremental updating of a global map; the method has the advantages of high positioning precision, strong anti-interference capability, wide application scene and the like.

The invention discloses a polling path dynamic planning method, medium and equipment, and the method comprises the steps: obtaining environment information in real time, dynamically switching an RTK-GPS mode, a visual inertial navigation mode and a laser SLAM positioning mode, generating the current positioning information of a robot in combination with a global map, and constructing a local environment map; and planning an adaptive second preset inspection path based on the first preset inspection path. Environmental feature information is actively collected in an outdoor environment, a real-time obstacle avoidance path is generated by dynamically updating a local environment map, and the real-time obstacle avoidance path is fused to a second preset inspection path, so that smooth switching and dynamic obstacle avoidance cooperation of a positioning mode in an indoor and outdoor scene are realized; the problem of planning failure caused by multiple positioning source conflicts, local path oscillation and environment sudden change in a complex scene is solved, and the continuity of the inspection process and the path planning robustness are guaranteed.

The invention provides a robot semantic navigation method and system, a terminal and a storage medium, and the method comprises the steps: inputting a received natural language command and a pre-constructed semantic map into a pre-trained large language model, so that the large language model extracts a target navigation object in the natural language command, matching the target navigation object with a semantic map to determine a target position point of the target navigation object on the semantic map; performing path planning based on the current position of the robot and the target position point to obtain a corresponding target path so that the robot navigates to the target position point based on the target path; the semantic map is a global map which corresponds to the current space environment and contains semantic information. The robot can execute accurate positioning and navigation tasks in a complex environment through the semantic map containing semantic information and the large language model, the requirement for high spatial precision is met, and the robot can understand natural languages of human beings.

The invention relates to the technical field of automatic driving, and particularly discloses an automatic driving interpretation text determination method based on a large visual language model, and the method comprises the steps: obtaining a multi-view camera image, a BEV map image and a text query template; video features in the multi-view camera image are extracted, and image features in the BEV map image are extracted; aligning the image features with the video features based on a modal encoder; encoding the text query template, generating a text mark, mapping the aligned image features and video features to a text embedding space based on a shared projector, generating a visual mark, and inputting the visual mark and the text mark into an LLM backbone model; comprehensive three-dimensional dynamic / static situation perception and visual representation unification of a local scene and a global map are realized, the 3D LVLM architecture has excellent three-dimensional space understanding and dynamic time reasoning capabilities, and the accuracy and safety of end-to-end automatic driving are effectively improved.

The invention discloses a robot vision-inertia SLAM method and device and a medium, and belongs to the technical field of computer vision and robot navigation. The method comprises the following steps: synchronously acquiring images and inertial data through a robot binocular camera and an IMU, performing feature enhancement on an original image in combination with a pre-trained deep learning model, introducing an adaptive brightness compensation mechanism and designing an image light supplementing module based on a generative adversarial network (GAN), recovering low-illumination image details, and improving the image quality of a low-illumination area; an entropy-based adaptive dynamic interference rejection algorithm is provided, and dynamic interference feature points are rejected in combination with IMU (Inertial Measurement Unit) data; a low-rank approximate improved graph optimization algorithm is adopted, and global map construction and pose optimization are accelerated; and through entropy-based nonlinear dynamic smoothing coefficient adjustment, the track stability is improved. According to the method, the positioning precision and robustness in a low-light environment are improved, the calculation efficiency is improved, and dynamic interference is effectively resisted.

The invention discloses a vehicle-road collaborative sensing and parking method and system for a shaft mine draw shaft environment, and relates to the field of unmanned driving, and the method comprises the steps: collecting roadside laser radar data, roadside camera data, roadside laser range finder data, vehicle-mounted laser radar data and vehicle-mounted camera data; respectively calculating roadside obstacle information and vehicle-mounted obstacle information; vehicle-road obstacles are matched through a Hungary algorithm, and a vehicle-road collaborative perception result is generated; performing global path planning by calculating an optimal driving path from the starting point to the unloading point by using the global map to generate a global path; performing local path planning through a path obstacle avoidance algorithm to generate a local path; and according to the global path and the local path, controlling the driving path, the vehicle speed and the parking of the unmanned vehicle. In order to solve the problems that in the prior art, an unmanned vehicle cannot comprehensively sense obstacles and cannot stop in place under the mine shaft environment, comprehensive sensing of the obstacles in the mine shaft environment and accurate parking of the vehicle are achieved.

The invention relates to the technical field of laser SLAM, and particularly discloses a laser SLAM method based on ground segmentation, and the method comprises the steps: obtaining laser radar point cloud data information, carrying out the point cloud preprocessing of the laser radar point cloud data, and obtaining ground point cloud features and non-ground point cloud features; respectively carrying out feature extraction, feature matching and pose estimation on the ground point cloud features and the non-ground point cloud features to obtain a pose nonlinear optimization constraint result; selecting a key frame according to a pose nonlinear optimization constraint result, constructing a local map according to pose information of the key frame, and obtaining a local map construction result; loopback detection is carried out according to the local map construction result, and a loopback detection constraint result is obtained; and performing global pose optimization processing according to a loopback detection constraint result and a local map construction result to obtain a global pose and global map construction result. The laser SLAM method based on ground segmentation provided by the invention can improve the positioning precision and the operation stability.

The invention provides a fruit stem clamping and shearing agricultural robot and a picking control method thereof.The robot comprises a control cabinet, a mechanical arm and a visual navigation system which are installed on a crawler-type chassis supporting platform, and an upper computer control system and a lower computer control system are installed in the control cabinet; and the tail end of the mechanical arm is provided with a movable end effector with an identifying and positioning system and a light supplementing device. The movable end effector can flexibly adapt to picking requirements and complex working conditions of different crops; the upper computer control system is combined with a depth camera and a laser radar to collect data, and a YOLOv8-MC model is utilized to perform identification and key point positioning on a target crop in an image, so that the crop identification precision in a complex environment is remarkably improved, the probability of missing collection and mistaken collection is reduced, and the damage to fruits is reduced; when the upper computer control system constructs a global map, depth correction and map optimization are carried out on illumination changes, and the path planning and operation accuracy of the robot is further improved.

The invention discloses a cooperative positioning and trajectory planning method based on an air-ground robot system, and the method comprises the steps: enabling unmanned vehicles to construct a global environment map through a laser radar, achieving the positioning of the unmanned vehicles, obtaining the distances between the unmanned vehicles through the combination of an IMU (Inertial Measurement Unit) and a UWB (Ultra Wideband) distance measurement chip of the unmanned vehicles, and carrying out the real-time pose estimation. Based on the fused pose and global map information, a method of combining mixed A * search and nonlinear trajectory optimization is adopted, a collaborative trajectory planning problem is converted into a multi-constraint optimization model, the sight distance quality between air-ground robots, an unmanned vehicle team-shaped geometric structure and an unmanned aerial vehicle rapid navigation problem are comprehensively considered, and the space-ground robot cooperative trajectory planning method is established. And generating a collision-free optimized cooperative trajectory meeting multi-robot kinematics constraints in real time. According to the method, the positioning precision and the coordination degree of the air-ground coordination system in a complex dynamic environment are remarkably improved, and a safe, efficient and real-time heterogeneous multi-robot trajectory planning method is provided for various application scenes such as routing inspection, search and rescue, environment monitoring and the like.

The invention provides a mobile robot active positioning method and device for an indoor dynamic scene, and the method comprises the steps: collecting indoor environment data, and constructing an indoor global map which comprises the geometric structure and semantic information of an environment; sensing data collected by the mobile robot in real time are matched with an indoor global map, the pose of the mobile robot in the three-dimensional space is estimated in real time, and the uncertainty index of pose estimation is calculated; evaluating information gains of different potential observation points based on uncertainty indexes of pose estimation, and selecting the observation point with the maximum information gain as a candidate target point; planning a path to the candidate target point; the mobile robot collects sensing data at the candidate target points, the sensing data are matched with the indoor global map again, and the pose of the mobile robot is updated; the indoor environment change is monitored in real time, the indoor global map is dynamically updated, and the indoor global map is regularly corrected and optimized.

A laser SLAM system for slope compensation by using IMU and an implementation method thereof are provided. The system includes: a data preprocessing and slope detection module for acquiring and processing original point cloud data in a roadway to obtain slope information; acquiring and processing IMU data of a mobile vehicle to obtain initial values of a position and posture of the mobile vehicle and the IMU factors; a laser point cloud position and posture estimation module for classifying feature points based on the slope information and the initial values of the position and posture of the mobile vehicle to obtain a joint error; performing inter-frame matching position and posture estimation and obtaining key frame factors; and a high-precision mapping module for optimizing a factor map, and obtaining a global map of a ramp to complete the slope compensation.

The invention relates to an unmanned aerial vehicle autonomous obstacle avoidance method for an electric power inspection complex environment. The method comprises the following steps: S1, acquiring an image, LiDAR, an optical flow sensor and IMU data on an inspection path through an airborne camera and a sensor; s2, fusing the multi-source data through an EKF (Extended Kalman Filter) algorithm; s3, real-time obstacle detection and recognition are carried out by using a Tiny YOLO model, the Tiny YOLO model is introduced into a double-flow network, and the detection capability of the Tiny YOLO model in a resource limited scene is enhanced; s4, performing semantic segmentation on the image by using deep learning, obtaining positions and states of different types of objects, combining with a geometric map generated by LiDAR, updating a global map through an ICP algorithm, performing continuous pose map optimization by using g2o, reducing drift in path estimation, and obtaining a semantic map; and S5, calculating a new safety path by using a D * Lite algorithm, and dynamically adjusting the flight strategy of the unmanned aerial vehicle. According to the invention, the obstacle avoidance capability and the path planning efficiency of the unmanned aerial vehicle in a power inspection complex environment are effectively improved.

The invention discloses a multi-mode SLAM method, system and device, a medium and a program product, and the method comprises the steps: segmenting a current image frame through employing a lightweight semantic segmentation network, obtaining a semantic mask of the current image frame, carrying out the correction, and projecting the corrected semantic mask to a three-dimensional point cloud, and obtaining a point cloud image of the current image frame under a radar coordinate system; performing point cloud segmentation to obtain a plurality of point cloud sets with different semantic classification tags; predicting a saliency mask of the current image frame, extracting visual feature points of the current image frame, and obtaining depth values of the visual feature points; and finally, tracking the visual feature points and constructing a local map. And meanwhile, a Scan-Context descriptor and a similar distance are utilized to carry out loopback detection and global BA on the point cloud set, and a global map is constructed. According to the invention, various sensor data are comprehensively utilized, the perception capability of environment characteristics is improved, and the problems of insufficient perception and reduced positioning precision in a complex environment are effectively solved.

The invention belongs to a synchronous localization and mapping (SLAM) method in the field of robots, and discloses a binocular vision SLAM method based on a deep learning feature extraction and matching algorithm. The system comprises three modules: a front-end tracking module, a local mapping module and a loopback detection module. For each frame of input binocular image, firstly, feature points of the image are extracted, and then frame-to-frame matching is used to track a current image frame and determine whether the current frame is set as a key frame. In the local mapping module, matching from a key frame to a local map is used to obtain a more accurate pose and a global map, and in addition, the loopback detection module is used for inhibiting accumulative errors of a large scene. Experiments on a public data set and a data set collected by a robot platform show that robust, accurate and globally consistent pose estimation and mapping are realized by the method, and quick real-time operation can be realized on an embedded platform at a speed exceeding 10 FPS.

The invention relates to a robot positioning method and device and a robot. The method comprises the following steps: acquiring motion sensing data and local scene point cloud information, and performing local pose estimation based on the motion sensing data and the local scene point cloud information to obtain current local pose information; projecting the local scene point cloud information to obtain local scene contour data, and determining a target map corresponding to the local scene contour data in a preset global map; acquiring a corresponding pose correction parameter based on the matching degree of the target map and the local scene contour data; performing global pose conversion on the current local pose information to obtain initial global pose information, and correcting the initial global pose information by using pose correction parameters to obtain target global pose information; and fusing the current local pose information and the target global pose information to obtain target pose information. The method can improve the positioning accuracy of the robot.

The invention discloses an unmanned aerial vehicle obstacle avoidance method based on sensing fusion and reinforcement learning, and belongs to the technical field of aviation flight control, and the method comprises the following steps: S1, sensor fusion time-space synchronization; s2, sensing fusion obstacle recognition and positioning; s3, obstacle avoidance route planning based on reinforcement learning; the low-resolution global map and a local occupation grid map with the unmanned aerial vehicle as the center are superposed to serve as multi-channel input, and the current state and the historical action sequence of the unmanned aerial vehicle are combined to be input into an intelligent agent model fused by a trained convolutional neural network and a long-short-term memory network; and an obstacle avoidance decision is output through a reward function of the agent model, and the airborne computer executes the obstacle avoidance decision to control the unmanned aerial vehicle to complete an obstacle avoidance action. Through sensor fusion, the obstacle sensing capability under different weather conditions and different flight environments is enhanced, the sensing distance is also increased, and the flight safety under extreme conditions can be ensured.

The invention belongs to the field of computer vision, and particularly relates to a method, a device and a system for monitoring suction-scrabbling falling of a cleaning robot, and the method comprises the following steps: 1) a rear-view camera arranged on the cleaning robot collects suction-scrabbling images, and carries out data labeling; 2) building a deep learning labeling and training environment in a server or a workstation, selecting a YOLOv8 target detection network model, sending labeled suction and scrabbling image data into the model for training, and performing format conversion and quantization on the model after the training is completed; 3) inputting an image acquired in real time into the trained target detection model, and judging whether the suction rake falls off or not; (4) if the suction scrabbling device falls off, the upper computer sends falling abnormal information to a robot controller of the cleaning robot, and the robot controller controls a hub motor to brake emergently; and meanwhile, the upper computer transmits the abnormal information of the suction-scrabbling falling and the position coordinate information of the robot in the global map to the operation and maintenance personnel through the cloud platform and then transmits the abnormal result and the position information of the suction-scrabbling falling to the operation and maintenance personnel.

The invention discloses a multi-level hybrid obstacle avoidance control method adaptive to environment complexity, and relates to the technical field of robot control, and the method comprises the steps: obtaining the global map information of an inspection environment, dividing a map into grid units, marking static obstacles, and calculating a global path of an inspection robot from a starting position to a target position through an A * algorithm; position and speed information of a dynamic obstacle is obtained in real time through a laser radar, a gravitational field model and a repulsive force field model are established, a gravitational field is generated by a target point, and a repulsive force field is generated by the detected dynamic obstacle. According to the invention, by establishing an obstacle classification mechanism based on the distance threshold, quantitative evaluation of the threat degree is realized, and the problem of low control efficiency caused by adopting a unified processing strategy for all obstacles in a traditional method is avoided.

The invention discloses a construction robot positioning and mapping method in a long tunnel environment, and relates to the technical field of computer vision, and the method comprises the steps that a robot obtains data information in the long tunnel environment through a carried sensor, and the data information comprises IMU information, an RGB image and a depth image; the method comprises the following steps: calculating IMU pre-integration according to IMU information; carrying out adaptive ORB point feature extraction and EDlines line feature extraction and optimization on an RGB image; constructing a deep learning network to carry out feature matching on extracted point-line features to obtain a matching result; constructing a point-line feature re-projection error model based on the matching result; preliminarily estimating an inter-frame pose based on the depth image and a point-line feature re-projection error model; the method comprises the following steps: selecting a key frame and detecting an identification code based on a depth image and a preliminarily estimated inter-frame pose, constructing a factor graph model based on the key frame, an identification code detection result and IMU pre-integration, updating global positioning based on a factor graph optimization result, and constructing a global map.

A detection device, such as an unmanned vehicle, is adapted to traverse a search area and generate sensor data associated with objects that may be present in the search area. The generated sensor data is used by a system including object detection inference models configured to receive the sensor data and output object data, a local object tracker configured to track detected objects in a local map, and a global object tracker configured to track detected objects on a global map. The local object tracker is configured to fuse object detections from the object detection inference models to identify locally tracked objects, and a Kalman filter processes frames of fused object data to resolve duplicates and / or invalid object detections. The global object tracker includes a pose manager, configured to track global objects in the global map and update the pose based on a map optimization process. User-in-the-loop processing includes a user interface for displaying and manual editing of detected object data.

The invention relates to an unmanned aerial vehicle dynamic environment sensing method based on a visual large model, electronic equipment and a medium, and the method comprises the steps: carrying out the instance segmentation and optical flow estimation of multiple frames of original images based on the visual large model and an optical flow estimation network, and obtaining a dynamic mask and inter-frame optical flow information; performing feature point classification on the depth information, the dynamic masks and the inter-frame optical flow information corresponding to the multiple frames of images to obtain a classification result; constructing a re-projection error model based on the classification result to obtain an estimated pose; evaluating the quality of the feature points based on the estimated pose to obtain an optimized pose; and constructing a feature point global map based on the world coordinates of the continuous frame road sign points and the optimized poses. Through a multi-information fusion and step-by-step optimization mechanism, the method has relatively high robustness, can adapt to various dynamic environments, and improves environment perception performance and operation reliability in different scenes.

The invention provides a land unmanned equipment semantic preserving large model deployment method and system, and relates to the technical field of navigation control. According to the deployment method, a preliminary control suggestion is directly output through a lightweight fusion model, optimization fusion is carried out through an MPC framework, underlying dynamics and environmental constraints, and seamless connection from semantics to control is achieved; in combination with a VILO odometer and instance segmentation, a global map containing dynamic obstacle semantic information is constructed and updated in real time, and an accurate context is provided for planning and decision making; model pruning, knowledge distillation and edge calculation scheduling are adopted, so that a complex multi-modal semantic model can run in real time on an embedded platform; the decision basis from instruction analysis to control execution is recorded and visualized in the whole process, and the transparency and credibility of the system are improved; on-line re-planning, multi-stage fault detection and switching strategies are integrated, and the robustness of the system in a dynamic environment and an abnormal condition is remarkably improved.

The invention discloses a multi-unmanned-aerial-vehicle cooperative SLAM system based on common-view area detection. According to the system, firstly, an adaptive threshold value and a feature homogenization method are introduced to improve a feature extraction algorithm to extract image features; secondly, on the basis of a feature point screening algorithm fusing YOLO and geometric constraints, the influence of motion exterior points is reduced; then, an improved feature point screening mechanism is provided for the problem of feature mismatching existing in the unmanned aerial vehicle image; thirdly, on the basis of local positioning and mapping of the unmanned aerial vehicle, a common-view area detection algorithm based on a visual word bag and relative pose calculation are fused into a local map; and finally, performing global optimization by adopting an improved L-M solution algorithm to improve pose estimation, and generating a high-precision global map. According to the method, the multi-unmanned aerial vehicle cooperative SLAM system is improved, so that the interference of movement obstacles in an unknown environment is reduced, the mapping and positioning precision is improved, and better real-time performance and robustness are achieved.

The invention discloses a tight coupling laser inertial vision fusion method based on radiation reconstruction, and relates to the technical field of vision fusion. Comprising the following steps: removing laser radar point cloud motion distortion based on IMU measurement, dividing the laser radar point cloud into multi-scale voxel grids, and obtaining a plane normal vector of the laser radar point cloud contained in each voxel grid to calculate a point-surface residual error; real radiation information is obtained through photometric calibration, meanwhile, a global map is projected to a current frame to obtain a tracking point, the obtained tracking point is tracked through an LK optical flow method, and a frame-to-frame re-projection error and a radiation error are obtained; a multi-constraint factor graph containing IMU pre-integration, laser radar constraint and visual constraint is constructed based on the factor graph, and robustness enhancement of the system in a degradation environment is realized in combination with a key frame sliding window optimization mechanism. The method provided by the invention has high positioning precision and robustness in a highly challenged degradation scene.

A method for calculating an optimum route for a first vehicle to travel from a current location to a selected parking space includes receiving map data of a parking zone, the map data including local map data and global map data of the parking zone, sensor data, parking space data indicating availability of parking spaces within the parking zone, and vehicle dynamics and location data, selecting, as the selected parking space, at least one available parking space based on the parking space data, determining a plurality of target positions between the current location and the selected parking space, calculating, in response to the selecting of the selected parking space and based on the plurality of target positions, the optimum route, including calculating trajectories of the first vehicle along an entire route between the current location and the selected parking space, and controlling the first vehicle to travel from the current location to the selected parking space.

The invention discloses a three-dimensional point cloud map construction method, system and device and a medium. The method comprises the following steps: acquiring a color image, a depth image and inertial measurement unit data; constructing an image pyramid according to the color image, and extracting feature points; performing dynamic target detection and feature point screening on the image pyramid through target detection model detection; according to the feature points and inertial measurement unit data, performing preliminary estimation and refined optimization on the pose; performing thread tracking and key frame judgment according to the distribution and poses of the feature points to obtain a key frame, and updating a local map; carrying out loopback detection on the current frame, and optimizing a global map; and performing point cloud generation, point cloud splicing and point cloud optimization according to the color image and the depth image corresponding to the key frame to obtain a three-dimensional point cloud map. By screening the dynamic feature points and fusing the IMU data, high-precision indoor positioning and map modeling are realized, and the method can be widely applied to the technical field of intelligent positioning and mapping.

The invention discloses a map updating method, a robot and a storage medium, and the method comprises the steps: carrying out the probability rapid increase and probability slow attenuation of a neighbor key frame and a global map according to a current frame, and then updating a target global map according to the neighbor key frame and the observation probability of a position point in the global map. In the operation of rapid probability increase, the newly-added content in the map has a larger probability increase gradient, so that the newly-added position point in the map is more rapidly kept consistent with the global map, and the problems of positioning deviation and loss caused by inconsistency of the actual environment and the global map are reduced. In the operation of probability slow attenuation, a smaller probability attenuation gradient is set for a map position point, the retention period of historical information by the map is prolonged, the recall rate of a map matching algorithm during subsequent robot positioning is improved, and the influence of positioning deviation caused by inconsistency of an actual environment map and a global map is further reduced.

The invention discloses an unmanned loading vehicle autonomous loading and unloading method suitable for mining area bulk cargo transfer, and relates to the technical field of mining unmanned loading and unloading and transportation control, and the method comprises the following steps: S1, carrying out function division on an open-air mining area to obtain a corresponding bulk cargo stacking area, an unmanned loading vehicle waiting area and a truck waiting area; and S2, an autonomous loading and unloading system is integrated on the unmanned loading vehicle, mining area environment information is obtained through the autonomous loading and unloading system, a global map of an operation area is constructed, an operation path is planned, loading and unloading operation is completed, and in the autonomous loading and unloading system, a loading and unloading operation logic module serves as a main process. According to the unmanned loading vehicle autonomous loading and unloading method suitable for mining area bulk cargo transfer, all scenes in the mining area bulk cargo loading and unloading process are independently achieved through the unmanned loading vehicle, that is, the unmanned loading vehicle can accurately construct an operation environment map through an advanced sensor and an intelligent algorithm, and efficient autonomous loading and unloading are achieved.

The invention relates to a multi-robot unknown environment collaborative exploration method based on a dynamic layering and zoning strategy, which comprises the steps that a plurality of robots are distributed in different areas of an unknown environment for collaborative exploration, each robot constructs a local map in real time through an SLAM algorithm, and a global map is constructed based on information sharing among the robots; based on the exploration area in the global map, segmenting the two-dimensional exploration area corresponding to the global map into a plurality of sub-areas through a dynamic layering and partitioning strategy to obtain a layering and partitioning result; wherein the layering and partitioning result is a sub-region set when the two-dimensional exploration region reaches the optimal hierarchy; based on the starting point position of each robot, distributing the layering and partitioning results among the robots, and balancing the task load between any two adjacent robots in the multiple robots to obtain a task unit set of each robot; and a minimum coverage path is planned on the optimal level based on the task unit set of each robot, and collaborative exploration is carried out based on the minimum coverage path.

The invention discloses an unmanned aerial vehicle image seamless rapid splicing method based on multi-view geometry, and particularly relates to the technical field of unmanned aerial vehicle image splicing, and the method comprises the steps: extracting feature points of video frames through a deep learning model, and carrying out the inter-frame matching, and obtaining a matching point set; based on the matching point set, solving the three-dimensional coordinates of the 3D map points, determining the initial pose of the key frame and initializing the global map, matching the feature points of the current frame with the 3D map points in the global map, and solving the camera pose of the current frame; a new key frame is selected according to the feature coverage rate, a local optimization window is constructed for 3D map points jointly observed by the newly added key frame and the associated key frame, error correction is carried out on key frame poses and 3D map point coordinates in the window through local light beam adjustment optimization, and a global map is updated; and based on the optimized global 3D map point fitting reference surface, projecting the key frame image to the fitting reference surface, generating a panoramic image by adopting a multi-scale pyramid fusion method, and outputting a final image spliced image.