47 results about "Geometric distance" patented technology

The invention discloses a method for identifying boundary features of a three-dimensional model of a closed cavity of an airplane, and provides a method for automatically and intelligently detecting the boundary features of the closed cavity in order to solve the problems that the extraction precision of the boundary features of the closed cavity is poor and the efficiency is low. Providing a minimum allowable gap recommendation mechanism based on a decision tree, and automatically extracting boundary part information; analyzing and screening an accurate boundary surface through a geometric distance and a normal vector included angle; and constructing a part attribute adjacency graph, and identifying the gap type of the closed cavity based on a GNN enhanced VF3 sub-graph isomorphism algorithm to realize automatic identification and marking of boundary parts and gaps of the closed cavity in the three-dimensional model. The method solves the problems of low manual detection efficiency and large error in a complex assembly environment, remarkably improves the precision and automation level of closed cavity boundary feature recognition, and is suitable for assembly area modeling scenes with few samples and high precision requirements in the field of aeronautical manufacturing.

The invention provides a trajectory clustering method based on space-time weighting and a density peak value, and relates to the technical field of trajectory clustering. The method provided by the invention comprises the following steps: based on a minimum description length criterion, fusing an adaptive space-time weight and a space-time geometric distance to segment trajectory data to obtain sub-trajectory segments; calculating a space-time local density, a relative distance and a decision value of each sub-track segment based on a density peak clustering algorithm; performing noise identification on the sub-track segments based on the average space-time geometric distance and the average local space-time density, iteratively selecting a class cluster center based on the decision values of the non-noise sub-track segments and the time-space factor decision values, and generating a class cluster set; and determining a candidate representative trajectory set from the class cluster set based on the spatio-temporal local density and the maximum density, extracting continuous sub-trajectory segments from the candidate representative trajectory set based on spatio-temporal continuity constraints, and sequentially connecting end points of the continuous sub-trajectory segments to generate a representative trajectory. According to the method, the trajectory clustering effect is improved through space-time weighted segmentation and density peak trajectory clustering.

The present application relates to the technical field of fusion design, in particular to a three-period minimal surface spatial positioning fusion structure design method based on the inverse distance weighting method, in the present application, by setting the implicit function equation of the three-period minimal surface and screening the control points according to the spatial distribution and geometric constraints, the accurate selection and positioning of the structure unit in the design target area are realized, the accuracy of geometric mapping is improved, the inverse distance weighting calculation is carried out through the convolutional neural network, the difference of the geometric region in the space can be effectively identified, the response ability of structure fusion is improved, the weight is calculated combined with the spatial distribution of the control points, and the weight distribution mechanism is formed by using the actual geometric distance between the control points, the continuity and smoothness of the geometric response in the interpolation region are guaranteed, the graph neural network is introduced to process the surface continuity problem, the local geometric mutation area is identified and corrected through the edge weight and topological structure between nodes, and the geometric smoothness and boundary consistency of the whole structure are improved.

The invention discloses a joint correction method and system for GNSS (Global Navigation Satellite System) non-structural deformation and a storage medium. According to the method, GNSS pseudo-range and carrier phase observation data and precise products such as a precise orbit and clock correction are obtained, a three-component displacement time sequence caused by a non-tidal atmospheric load, a non-tidal ocean load and a land hydrological load and a temperature-driven thermoelastic displacement time sequence are constructed, and the three-component displacement time sequence and the temperature-driven thermoelastic displacement time sequence are unified to a reference frame and a time system consistent with GNSS calculation. After interpolating the non-structural displacement vector to an observation epoch according to time, converting the non-structural displacement vector to a geocentric rectangular coordinate system from a site local coordinate system ENU, and projecting the non-structural displacement vector to a sight line direction from a satellite to an observation station reference point to form an equivalent geometric distance correction item; the correction terms are respectively applied to pseudo-range observation and carrier phase equivalent meter domain observation in the form of the same domain as observation, and then GNSS daily arc precise calculation is performed to output a coordinate time sequence, and preferably, the method can be used for a precise single-point positioning integer ambiguity fixing technology.

The invention discloses an underground structure scanning method for a controlled active source. According to the method, a known signal generated by an active source is equivalent to a record of a virtual station, the virtual station and an actual six-component seismograph station form a station pair, and the phase velocity of a wave under the distance is estimated through a phase difference / travel time difference between the virtual station and the actual six-component seismograph station and a geometric distance relationship between the source and the station; the obtained phase velocity is actually equivalent to a frequency dispersion curve at the midpoint of the connecting line of the source and the station, so that the phase velocity-depth structure from the source to the midpoint of the station can be obtained by performing inversion by using the obtained frequency dispersion curve; as the active source moves in the measurement area, a group of signals can be generated at different positions respectively, and a series of virtual midpoints are formed for the single-station six-component seismograph, so that the density of measurement points is remarkably increased in space. On the basis, through combination with an inversion result of a single station, three-dimensional underground velocity structure modeling with high credibility and high spatial resolution can be realized.

The present invention discloses a pairing selection method based on machine learning, comprising: first, calculating the compatibility scores between pairs using two geometric distance constraints, resulting in two compatibility scores; then, for each pairing, calculating its compatibility score with all other pairs to form two feature vectors; combining the two feature vectors into a single feature vector using a maximum pooling method; sorting the elements in the combined feature vector from largest to smallest, and taking the top k elements to form the final feature vector. Finally, the feature vectors and labels of the training data are input into a support vector machine classifier to obtain a trained support vector machine classifier, which is then used to predict the labels of the test data pairs to identify the correct pairing. The method of the present invention has high precision and recall, can accurately identify the correct pairing, and improve the success rate of registration.

A method of determining device locations includes receiving a set of candidate locations in an environment and generating training data representing synthetic distance measurements between pairs of the candidate locations. The synthetic distance measurements are generated by applying random variations to the geometric distances between the candidate locations. The training data and candidate locations are used to train a machine learning model. A set of measured distances between devices are input to the trained learning model, which is used to determine a respective location in the environment of each of the devices.

The invention discloses a ground-air collaborative mapping method based on semantic feature and ground plane multi-constraint fusion, which comprises the following steps: generating a sub-map set through local point cloud data collected by a fire-fighting unmanned vehicle and an unmanned aerial vehicle, and generating a rotation-invariant semantic descriptor by utilizing a maximum height value rasterization strategy and rotation group enhancement; by calculating a geometric distance and a semantic distance and combining isomorphic and heterogeneous mode weighted fusion, loopback candidate sub-map pairs are screened, geometric registration is carried out, and loopback constraints are generated. Meanwhile, ground point cloud is extracted through an RANSAC algorithm, a sliding window smoothly updates a ground plane model, and height constraint and attitude alignment constraint are constructed. And the loopback constraint, the ground plane constraint and the pose information are input into a Ceres optimization framework for joint optimization, a globally consistent total map is generated, the precision and robustness of ground-air collaborative mapping are effectively improved, and the method is suitable for emergency scenes such as complex fire rescue and the like.

This invention discloses a linear base station array positioning method based on geometric modeling, aiming to solve the technical challenges of existing methods in terms of positioning accuracy, computational efficiency, and convergence under linear base station conditions. The method includes: receiving signals from a linear base station array with known location coordinates and real-time acquisition of the time difference of arrival (TDOA) between the target and each base station; selecting base station combinations and, considering the geometric characteristics of the linear base station array, constraining the rationality of geometric distance measurements based on the trigonometric inequality to construct a dynamic correction model for ranging errors, reducing errors caused by equipment deviations and signal interference; generating preliminary positioning results for each base station combination based on the environmental characteristics of the linear base station array using the trigonometric cosine theorem; and optimizing the fusion method of multiple combination positioning results by designing an improved Cramer-Rao lower bound (CRLB) weighting strategy, ultimately obtaining a more robust positioning result. The algorithm proposed in this invention has strong error resistance and broad application prospects in target positioning of linear base station arrays.

The application discloses a multi-constellation multi-fault detection and elimination method and device and electronic equipment, and belongs to the technical field of satellite navigation and positioning. The method comprises the following steps: acquiring GNSS observation data and a fisheye image, performing semantic segmentation on the image to identify a sky area, and judging satellite visibility; constructing an Euclidean distance matrix, a geometric angle matrix and an inter-satellite pseudo-range difference matrix for each constellation; based on a triangular geometric relationship and a cosine theorem, a geometric distance from a reference star to a terminal is obtained by using median robust estimation, and a residual error matrix is constructed; a fault voting matrix is generated through K-means clustering, a fault satellite set is output, and positioning calculation is performed by using non-fault satellites. The application fuses visual semantic information and inter-satellite geometric constraints, does not need iterative search and a large amount of training data, realizes one-time multi-fault real-time detection and elimination in a complex urban environment, and has good generalization ability and robustness.

A distance-based EKF positioning method, system, device and medium, comprising: collecting inertial measurement unit information, three-dimensional laser scanning points and three-dimensional point cloud map information; finding corresponding points of three-dimensional laser scanning points in three-dimensional point cloud map to realize data association; after data association, solving the distance between corresponding points to obtain geometric distance measurement and robust distance measurement; predicting the pose state at the current time according to the inertial measurement unit information; updating the pose state at the current time by using the geometric distance measurement and the robust distance measurement, and then outputting the result. The application discloses a distance-based EKF multi-sensor fusion positioning scheme, which takes the distance between three-dimensional laser scanning points and corresponding points of a three-dimensional point cloud map as the observation of a filter, realizes corresponding point finding through a k-NN fast nearest neighbor search algorithm, does not need to perform feature extraction, and does not need to perform an iterative calculation step, so that the method has great advantages in calculation efficiency compared with existing methods.

The invention provides a voice interactive CT robot sickbed cooperation method, which is based on an environment grid map, adopts an improved algorithm to plan an initial global path, fuses a geometric distance heuristic term to a target point, a risk potential field value inversely proportional to an obstacle distance and a path turning penalty term in a cost function, and performs local trajectory optimization by utilizing a TEB algorithm, so as to improve the robustness of the system. The method comprises the following steps: representing a robot as a plurality of safety coverage circles, minimizing a cost function comprising a trajectory jerk smoothness penalty term and an obstacle safety gap term under the condition of satisfying space-time and kinematics constraints, reconstructing the weight of the cost function, enhancing the pose precision and tail end zero speed requirement, weakening a time optimization target and solving the problem that the robot is close to a target when the target is close to the target. And if it is detected that the topological homotopy categories of the local trajectory and the global path around the obstacle are inconsistent and the cost value exceeds a blocking threshold value, a risk potential field value is temporarily increased around the obstacle, re-planning is triggered from path points before topological deviation, and the path reliability and the environmental adaptability are ensured.

The invention relates to the technical field of global navigation satellite system data processing, and discloses a GARCH-based GNSS cycle slip detection threshold real-time determination method, which comprises the following steps: acquiring GNSS double-frequency observation data, constructing a Morse-Wiberner and geometric distance-free combination observation value, and carrying out differential processing to obtain a noise time sequence. Secondly, traversal calculation is carried out based on an akaike information criterion, and the optimal order and the sliding window length of the GARCH model are determined; then, a current epoch condition standard deviation is predicted using the model, and a multiple factor is adjusted in combination with an environmental state to construct a dynamic detection threshold. And finally, comparing an absolute value of the noise sequence with a threshold value to judge a cycle slip, and calculating an integer solution by constructing an observation equation to complete repair. According to the method, the heterovariance characteristic of the observation noise is modeled by using the GARCH model, so that the adaptive adjustment of the detection threshold along with the environmental noise level is realized, the misjudgment rate in a complex environment is effectively reduced, and the detection accuracy is improved.

The application discloses a joint correction method and system of GNSS non-tectonic deformation and a storage medium. The method obtains GNSS pseudo-range and carrier phase observation data and precise orbit, clock error and other precise products, constructs three-component displacement time series caused by non-tidal atmospheric load, non-tidal ocean load and land hydrology load and temperature-driven thermal elastic displacement time series, and unifies to a reference frame and a time system consistent with GNSS solution. The non-tectonic displacement vector is interpolated to the observation epoch by time, converted from the local coordinate system ENU to the geocentric rectangular coordinate system, and projected to the line-of-sight direction of the satellite to the reference point of the station to form an equivalent geometric distance correction term; the correction term is applied to the pseudo-range observation and the carrier phase equivalent meter domain observation in the same domain as the observation, and then GNSS daily arc segment precise solution is performed to output the coordinate time sequence, which is preferably used for precise point positioning integer ambiguity fixing technology.

This application discloses an automatic optimization method and apparatus for 3D bounding box labels in 4D dynamic object interactive annotation, comprising: acquiring all key 3D bounding boxes and their corresponding dynamic object point sets, and all interpolated 3D bounding boxes and their corresponding dynamic object point sets in a dynamic object sequence under a global coordinate system; calculating the principal directions of the two dynamic object point sets and calculating their differences, adjusting the direction of the interpolated 3D bounding boxes according to the differences to obtain the interpolated 3D bounding boxes with optimized directions; calculating the density distribution characteristics and geometric distance distribution characteristics of the interpolated 3D bounding boxes and key 3D bounding boxes and their corresponding dynamic object point sets, and constructing a composite loss function; iteratively optimizing the center point position of the interpolated 3D bounding boxes with optimized directions, fusing the composite loss function in each iteration to obtain the optimal center point position of the interpolated 3D bounding boxes, and then updating the center point position. Compared with previous solutions, this invention significantly improves the accuracy of interpolated 3D bounding boxes and further reduces manual workload.

The invention discloses a contact type measurement mechanical arm calibration method based on distance constraint, and relates to the technical field of mechanical arm calibration, and the method comprises the following steps: a plurality of non-collinear space reference points are fixedly installed in a mechanical arm working space, a contact type measuring head is installed on a flange at the tail end of a mechanical arm, and a measurement basis is established; the mechanical arm is controlled to move to different poses covering the working space, the measuring head is driven to make contact with the reference points in sequence, and the actual space distance is measured; according to the initial kinematics parameters of the mechanical arm, the theoretical position of the tail end is calculated through forward kinematics, and the theoretical geometric distance between the measuring head and the reference point is deduced; constructing a nonlinear optimization objective function based on a deviation value between an actual measurement distance and a theoretical calculation distance; the target function is converted into a matrix equation form, and a regularization iterative algorithm is adopted to calculate the correction amount of the kinematics parameters of the mechanical arm; and the obtained parameter correction amount is written into a mechanical arm controller, and parameter updating and calibration compensation are completed.

The invention discloses a cloth simulation method, device and equipment and a computer readable storage medium. The cloth simulation method comprises the steps that collision detection is conducted on finite element units through a linear hierarchical bounding box collision detection algorithm; calculating the minimum geometric distance corresponding to each detected collision candidate pair; determining total collision potential energy according to each minimum geometric distance, a preset global rigidity coefficient and a preset minimum safety gap distance by utilizing a local obstacle function; determining local friction potential energy respectively corresponding to each collision candidate pair according to the relative position vector and the normal contact force scalar quantity respectively corresponding to each collision candidate pair, and determining total friction potential energy; according to the boundary stiffness coefficient, the augmented Lagrangian multiplier and the target positions corresponding to the vertexes of the finite element units, local boundary potential energy corresponding to the vertexes of the finite element units is determined, and total boundary potential energy is determined; and performing cloth simulation according to the total collision potential energy, the total friction potential energy and the total boundary potential energy. According to the invention, penetration between the cloth and the cloth or an external object is avoided.

The invention discloses a three-dimensional constellation shaping method based on regional probability folding and weighted neighborhood mapping, which belongs to the technical field of communication, and comprises the following steps of: calculating a radial distance of a constellation point coordinate, performing hierarchical division to obtain a radial energy hierarchical set, and calculating an effective connectivity degree of the constellation point by identifying a primary neighbor and a secondary neighbor to obtain an effective connectivity degree of the constellation point; carrying out folding frequency distribution based on a hierarchical scale factor to obtain folding frequency and non-uniform probability distribution of each constellation point, and constructing a probability-geometry joint weighted edge weight based on a geometric distance between the constellation points to obtain a weighted neighborhood matrix; according to the weighted neighborhood matrix and the bit tag Hamming distance, a global mapping cost function is established, and an optimal bit mapping relation is obtained for three-dimensional constellation shaping. The method solves the problems that in the prior art, probability distribution only depends on the energy criterion, but the local topology difference of constellation points on the same layer is ignored, so that the error code risk is unbalanced; and system performance limitation caused by mismatching of bit mapping and probability distribution under a non-uniform probability condition is solved.

The invention discloses a method for constructing a high temporal-spatial resolution ionosphere model for a dense GNSS deformation observation station. The method comprises the following steps: collecting double-frequency observation data and broadcast ephemeris of a dense GNSS deformation monitoring station in a modeling area; calculating a multi-path effect, a cycle slip ratio and a data effective rate, and screening out data meeting quality requirements; respectively acquiring a satellite end DCB and a receiver end DCB; using the screened double-frequency observation data to calculate a non-geometric distance combination, and combining a satellite end DCB and a receiver end DCB to obtain an oblique path TEC; based on thin layer hypothesis, calculating a puncture point position of a satellite signal in a thin layer and a satellite elevation angle and an azimuth angle at the puncture point, and converting an oblique path TEC into a vertical TEC by using a projection function; and constructing a TEC grid model, and averaging the vertical TECs of all puncture points falling on each grid to obtain a TEC value and a system. According to the invention, calculation is carried out by using the data of the existing large-scale dense GNSS deformation network monitoring stations, and the value-added effect of the data is realized on the premise that the construction cost is not increased.

The application provides a reasonable path set determination method for combined transport based on fuzzy conversion, which comprises the following steps: an ideal point solution set satisfying a multi-objective function is obtained in a Q-dimensional target space, and a limited geometric distance of a combined transport node network is calculated according to a fuzzy membership function; compared with a traditional hard constraint contradiction without solution, the method can make the calculation and iteration process easier, improve the solving efficiency, save the operation time, greatly improve the convergence speed, save the iteration time, and is very suitable for solving the multi-objective reasonable path set problem of the combined transport.

Techniques and systems are provided for generating depth information for an image. For instance, a process can include obtaining one or more images of an environment. The process can further include generating a set of features for the one or more images. The process can also include combining the set of features with one or more distance maps to generate combined feature distance information, wherein the one or more distance maps indicate distances based on relative height above a ground level. The process can further include generating depth information of the environment based on the combined feature distance information, and outputting the depth information of the environment.

This application discloses a real-time decision-making method, system, and medium for intelligent agents based on Riemannian manifolds, relating to the field of artificial intelligence technology. The method, executed by an intelligent decision-making system, includes: constructing a behavior transition graph based on a historical trajectory dataset; mapping the attention matrix generated by a large language model in the decision-making task to a symmetric positive definite matrix manifold for representation; embedding the behavior transition graph into a hyperbolic Riemannian space for representation; learning the geometric alignment mapping from the symmetric positive definite matrix manifold and hyperbolic Riemannian space to a common metric space by optimizing the objective loss function; in the inference phase, acquiring current environmental observation data and inputting it into the large language model to obtain the corresponding attention representation, which is then mapped to the common metric space to obtain a query vector; and using actions corresponding to action nodes with a geometric distance below a preset threshold as control commands. This significantly reduces the computational overhead of decision-making and improves cross-scenario generalization ability and decision interpretability.

The invention discloses a satellite-ground distance determination method and device, a computer storage medium and a terminal, and the method comprises the steps: dividing a first region needing the calculation of the satellite-ground distance into a plurality of second regions under the condition that the geometric distances between a large number of grid points and a satellite need to be calculated, and selecting more than one first grid point for each second region, determining a first satellite-ground distance of the first grid point through an iterative calculation mode; by utilizing the characteristic that the signal receiving moments of the grid points in the second area are all aligned, according to the first satellite-ground distance of the first grid point, the satellite position corresponding to the first grid point and the first displacement related information, the calculation of the second satellite-ground distance is realized by combining the position of the second grid point; in conclusion, for the grid points in the second area, only the first grid point needs to be subjected to iterative calculation to determine the satellite-to-ground distance, and the other grid points do not need to be subjected to iterative calculation processing, so that the calculation of the satellite-to-ground distance of the second grid point is simplified, the calculation amount of the satellite-to-ground distance is reduced, and the calculation processing efficiency of the satellite-to-ground distance is improved.

The invention relates to path planning, in particular to an autonomous search path planning method and system for a rescue robot, and specifically, the method comprises the following steps: obtaining a grid map of a rescue environment, and calculating a weighted extension distance fusing a geometric distance, terrain trafficability and environment risk cost for adjacent nodes; a forward open list starting from a starting point and a reverse open list starting from a target are established, and in bidirectional search, environmental risks are included in a forward evaluation function, and the environmental risks are not included in a reverse evaluation function; when a node accessed in the other direction is searched in a certain direction, recording the node as an initial intersection, and continuing bidirectional expansion by taking the total expansion distance of the current path as an upper bound to search a better intersection; in all the intersection points, selecting a point which enables the total extension distance to be minimum after the forward path and the reverse path are spliced as an optimal intersection point, and generating an initial path; and smoothing the initial path to ensure that the terrain trafficability and the environmental risk value of each point on the smoothed path are within a safety threshold, and obtaining a feasible path.

The invention provides a space intelligent path generation method and device based on space function topology and a medium, belongs to the technical field of space intelligence, and can enable the space intelligent navigation and path planning technology not only to depend on geometric distance, but also to model space function semantics and topology logic constraints. The space function topology-based space intelligent path generation method comprises the following steps: obtaining structure information of a space environment to construct a topological graph, the structure information comprising space structure information and function region information; performing information propagation and aggregation on nodes of the topological graph by using a graph neural network to generate a global topological embedding matrix; and based on the global topology embedding matrix, selecting an optimal path between the nodes, so that the optimal path can be reached in a physical space and logic constraints are met. The invention discloses an intelligent spatial path generation device based on spatial functional topology and a computer storage medium.

The present invention provides a DOA estimation method based on geometric distance metric under low signal-to-noise ratio. The method constructs a signal model of a sensor array according to the direction vector of the signal source, introduces information geometric metric, gives a definition of geometric distance metric, selects a reference matrix for implementing DOA estimation through geometric distance metric, performs regularization processing on the selected reference matrix, solves the distance of geometric distance metric, and obtains the estimated target direction. The present invention transforms the direction estimation problem into a study of the geometric structure difference problem between two points on a matrix manifold, and can well implement DOA estimation through geometric structure difference. After selecting the reference matrix and performing reversible processing on the reference matrix, it is substituted into the information geometric distance calculation formula of LEM, KLD, SKLD and JBLD to obtain the geometric distance values corresponding to different angles. The angle corresponding to the minimum geometric distance value is the target direction. Under the conditions of low signal-to-noise ratio and small number of snapshots, the target direction estimation has a high estimation accuracy.

The invention is applicable to the technical field of computer aided design, and provides a parameter adaptive adjustment method for keeping a mechanical structure connection relationship, comprising the following operation steps: S1, structural connectivity identification: modeling an assembly structure with a directed acyclic graph; s2, displacement decomposition and parameter priority ranking: ranking parameter sets in each direction according to geometric distances between the parameter sets and tail end parts; s3, response direction judgment based on the relative position: calculating direction vectors of the central point of the part where the control parameters are located and the central point of the adjacent part; and S4, allocation and execution: respectively processing in each direction, and allocating parameters to bear displacement according to priorities. According to the parameter adaptive adjustment method for keeping the connection relation of the mechanical structure, tail end displacement is decomposed into three main direction components, priority ranking is carried out according to the spatial distance between each adjustable parameter and a tail end part, and modeling or simulation is not needed by utilizing the relative geometric position between the parts.

The present invention discloses a method for robust collaborative positioning of multiple robots, which relates to the technical field of intelligent robots and can achieve correct collaborative positioning of multiple robots in the presence of a large number of erroneous data associations. The method comprises the following steps: for a system composed of multiple robots, wherein the multiple robots each obtain environmental information, each robot constructs a local map, each local map is composed of multiple sub-maps, and constraints between sub-maps are constructed; the constraints between sub-maps are composed of a number of matching point pairs. A vector filtering model is constructed to filter out erroneous constraints and point pairs. A hybrid probability model is constructed to calculate the correct probabilities of constraints and points in data associations from the perspectives of geometric distance, semantic association and vector consistency. A double-layer expectation maximization algorithm is used to determine the position and posture of the sub-maps and remove erroneous data associations to achieve collaborative positioning of multiple robots.

This invention proposes a remote sensing rotating target detection algorithm based on dynamic orientation-aware attention. First, addressing the problem of insufficient multi-scale feature fusion, a cross-scale dense fusion feature pyramid network is proposed. Through full-scale dense connections and feature distillation mechanisms, the deep fusion of semantic and detailed information is enhanced. Second, addressing the problem of insufficient target orientation awareness, a dynamic orientation-aware attention module is proposed. This module dynamically generates rotating convolutional kernels guided by angle information, achieving alignment between the receptive field and the target orientation. Finally, to address the training instability caused by angle periodicity, a Gaussian weighted chamfered distance loss is constructed, transforming angle error into a geometric distance penalty. An angle-aware Gaussian weight is introduced to alleviate angle discontinuities. This invention effectively solves the problems of insufficient multi-scale feature fusion, insufficient orientation awareness, and training instability caused by angle periodicity in remote sensing rotating target detection.

The invention relates to the field of generative artificial intelligence. The invention provides a user interface code generation method and system and a medium. The user interface code generation method comprises the steps of obtaining user interface screenshot data to extract an atomic component set and a text element set; constructing and solving a weighted bipartite graph based on the atomic component set and the text element set so as to merge the successfully matched atomic components and the text into a composite atomic component, and reserving the unsuccessfully matched atomic components and the text as independent atomic components, the weighted bipartite graph realizes semantic matching of the atomic component and the text according to the geometric distance, the alignment direction and the semantic correlation; performing structural analysis on the atomic component, wherein the structural analysis comprises the steps of generating a strong constraint container by adopting a differential aggregation strategy and generating a pre-mask area based on the strong constraint container; constructing a full-page component tree by using an improved recursive projection segmentation algorithm under the protection mechanism of the pre-mask area; and traversing the full-page component tree by adopting a post-order traversal strategy to generate codes in a divide-and-conquer manner.