23 results about "Vector field" patented technology

The present application relates to the technical field of additive manufacturing online monitoring and closed-loop control, in particular to a 3D printing workpiece defect identification system based on image detection; containing coaxial acquisition, potential mapping, diffusion analysis, singular point extraction and bidirectional feedback module; the system converts the collected image stream into a virtual flow velocity vector field; the core is to simulate the evolution of the physical field by using the virtual viscosity diffusion algorithm, calculate the difference of the vector field before and after diffusion to extract the topological mutation, and generate a defect potential thermal map; based on the potential value of the thermal map, the system performs bidirectional feedback of generating G code intervention instructions and adjusting the virtual viscosity coefficient; the present application reduces the pattern recognition to local physical field simulation, without sample training, and realizes millisecond-level real-time response.

The application provides an industrial character image data enhancement method and related equipment, decodes an obtained industrial character normal line image to be processed to obtain a normal line vector field of each pixel, performs mirror inversion and / or an angle measurement-based morphological operation on the normal line vector field to obtain an updated normal line vector field, encodes the updated normal line vector field to obtain an RGB image, stores the RGB image in an image matrix, and obtains an enhanced industrial character normal line image; compared with the prior art, the application directly performs vector field operation on the normal line image to realize controllable deformation of the character surface geometry, thereby generating an industrial character sample with rich morphological diversity, and solving the problems of fixed character geometry morphology, difficulty in simulating changes in concave-convex depth caused by process differences, edge blunting and other deformations in the prior art industrial character data generation method.

The present application relates to the technical field of three-dimensional geological modeling, in particular to a strata-bound lead-zinc deposit resource reserve evaluation method and system based on a three-dimensional model, comprising the following steps: calculating a stratum grid normal and dip vector, constructing a unit set and giving a vector to generate an occurrence vector field, screening the same lithology sample to construct a list and establish a local coordinate system, analyzing the sample relative position vector mapping distance components to the local coordinate system and normalizing and synthesizing structural distance, solving weight coefficients based on the structural distance and weightedly estimating grade, and combining density and volume to calculate resource quality and accumulate and output. In the present application, the grade evaluation is constrained in the stratum extension direction by constructing the occurrence vector field and the local coordinate system, the spatial correlation is corrected by using anisotropic structural distance, the evaluation main shaft is adaptively rotated with the fold, the cross-layer diffusion phenomenon is eliminated, the natural occurrence state of the ore body is restored, and the reserve estimation accuracy and model fidelity are improved.

The application discloses a kind of earth magnetic field horizontal component representation methods of removing pole singularity, comprising: based on adolf schmidt semi-normalized spherical harmonic theory, the spherical harmonic representation of geomagnetic field magnetic potential function is obtained by spherical harmonic polynomial expansion calculation method;With south pole or north pole as center, mirror center symmetry extension is carried out to great circle on sphere, so that each point on sphere has a mirror point, so that the mirror extension function of geomagnetic field magnetic potential function is obtained;Two-dimensional Fourier expansion is carried out to mirror extension function using double Fourier series, and double coverage function expansion is obtained;Based on double coverage function expansion, the horizontal component of geomagnetic field is calculated, i.e. the spherical gradient of spherical magnetic potential function is calculated.The realization is based on the existing IGRF model magnetic potential expansion coefficient, the magnetic field tangent vector representation is obtained without being polluted by pole problem, and a new modeling idea and calculation method of spherical tangent vector field are provided.

The application provides a generator set running, leaking and dripping intelligent monitoring method and system based on AI recognition, relates to the field of generator set intelligent monitoring, and comprises the following steps: collecting an image sequence and performing multi-waveband spectral decomposition to generate a spectral feature map, so as to identify a leakage area and a material type; then, a leakage area centroid change is calculated to construct a diffusion vector field, on the basis of which, a diffusion boundary is predicted and a diffusion risk coefficient is calculated; the shortest distance between the predicted boundary and the unit boundary is combined to determine an early warning level; finally, the leakage source is located through reverse tracing of the diffusion vector field. The application realizes early and accurate identification of leakage, diffusion trend prediction and source positioning, and improves the automation and early warning timeliness of monitoring.

This invention belongs to the field of defect detection, specifically a method and system for identifying defects in pipeline welds. The method includes: acquiring an initial X-ray image of the pipeline weld; extracting a global feature map using a first deep convolutional network; decoding to generate a defect prediction confidence map; extracting key geometric structures of the weld and constructing a geometric prior weight map; applying Bayesian variational inference to the network; statistically analyzing the dispersion of multiple random forward propagations to represent cognitive uncertainty and generate an uncertainty map; pixel-level weighted fusion of the confidence map, uncertainty map, and geometric prior weight map to obtain a probability heatmap; constructing a composite sampling guidance vector field based on this heatmap; deploying sampling points along the vector field; controlling the imaging system to scan and acquire high-resolution local projection data from multiple angles; three-dimensional reconstruction of the projection data to obtain local features; fusing these features with global features through a cross-attention module to generate an enhanced defect representation; and outputting the category, three-dimensional spatial location, and size of the weld defect.

The invention relates to a method for transmitting at least one parameter (3,3.1,3.2,3.3,3.4,3.5) for at least one road segment (2,2.3,2.4,5) from a source map (1) to a target map (4) comprising the following process steps: - Determining a vector field (7) based on at least one parameter (3,3.1,3.2,3.3,3.4,3.5) present in the source map (1), - Transferring the vector field (7) to the target map (4), - Determining at least one parameter (3,3.1,3.2,3.3,3.4,3.5) for at least one road segment (2,2.3,2.4,5) in the target map (4) based on the vector field (7).

The application discloses a kind of automatic fiber placement reference line generation methods based on vector field analysis, first, draw several auxiliary lines on the surface along the vertical direction of 0 degree fiber direction, draw several reference line starting points on the surface, draw a geodesic line at each starting point as the basis reference line, calculate the nearest point of the basis reference line and each auxiliary line as the starting position of the reference line node optimization;Again, the laying surface is meshed, the fiber direction at each grid node is calculated, the geodesic curvature and angle deviation are calculated using the vector field analysis method, and the path planning quality comprehensive evaluation function is calculated;Finally, the optimal position of the reference line node is solved using the optimization algorithm, the objective function of the optimization algorithm is set as the reference line comprehensive evaluation function, the optimization variable is set as the position of the reference line node, and the optimal reference line is obtained by fitting node;The application can improve the design efficiency of reference line, thereby improving the efficiency and effect of automatic fiber placement path planning generation.

This invention provides a method and system for quantitatively detecting material crack depth based on thermal gradient vector field, belonging to the field of nondestructive testing technology. The invention proposes a quantitative detection scheme of unilateral heat flow excitation + orthogonal gradient scanning: First, a unilateral high-temperature boundary is constructed using a portable chemical self-heating pack and insulating material, forcing the heat flow to transversely cross the crack, forming a significant temperature step; second, infrared thermographic sequences of the cooling process are acquired to construct a two-dimensional thermal potential energy field and calculate the global gradient vector field; morphological algorithms are used to extract the crack topological skeleton, and virtual rays are emitted along the local orthogonal detection direction determined by the local thermal gradient direction at the skeleton nodes. Adaptive orthogonal integration is performed to extract multidimensional spatial features under multiple time series, which are then input into a supervised learning regression model to invert the physical depth of the crack. This invention effectively eliminates the geometric projection error caused by crack curvature, and features high computational accuracy and strong anti-interference ability, making it suitable for large-scale rapid detection in engineering sites.

The present disclosure provides a method and system for rendering 3D data fields with variations. The method includes receiving a 3D representation of a physical object or process, discretized into points defined by a surface mesh or point cloud, values of a main field at a set of points in the 3D representation, and values of a variation field associated with the main field for at least a subset of the set of points. A combined visualization of the main field values and the variation field values is generated and overlaid on the 3D representation, with the overlay configured to spatially correlate the variation field values with the main field values. The combined visualization is then displayed via an output interface. The method can generate dynamic pulsating visualizations or static overlays for both scalar and vector fields, and can be applied in various display environments, such as virtual reality and augmented reality interfaces.

This invention provides a method and system for monitoring the deformation of slender marine structures, belonging to the field of marine engineering structure monitoring technology based on computer data processing. The invention first acquires measured strain using sparsely arranged sensors, and then generates a continuous curvature vector field that allows for stiffness steps through cross-sectional equivalent mapping and physical piecewise interpolation. Subsequently, based on the piecewise constant curvature theory, the structure is discretized, and the local geometric features of the infinitesimal elements are extracted and analyzed using a three-dimensional special Euclidean group. SE (3) Construct a local homogeneous transformation matrix in space; finally, combine the chain rule to recursively deduce the global pose and reconstruct the three-dimensional large deformation curve. This invention avoids the cumbersome modal parameter identification, strictly decouples spatial bending deformation, eliminates the geometric nonlinearity accumulation error caused by plane integration from the bottom layer, and breaks through the high-fidelity monitoring limitation of complex and slender marine structures such as variable cross-section towers and irregular blades with extremely low sensing cost.

The application relates to the technical field of fault diagnosis, and discloses a water pump fault diagnosis method based on variational mode decomposition and a deep residual network, aiming at the mode aliasing problem caused by the overlapping of fluid noise and mechanical fault frequency bands, the application reconstructs a vibration signal into a phase space trajectory manifold, calculates a transient optical flow velocity field and a local optical flow divergence, establishes an augmented Lagrange objective functional containing a divergence penalty term to carry out iterative decoupling, outputs a one-dimensional time sequence mode component and a two-dimensional optical flow vector field matrix, constructs a double-flow residual network to extract impact waveform semantic features and manifold topological distortion space features respectively, carries out cross-modal feature alignment and fusion through a cross-attention mechanism, inputs a multilayer perception machine to output a fault prediction probability, and carries out end-to-end system joint optimization. The application can accurately separate fluid-solid coupling components under complex hydraulic working conditions, and improves the weak fault diagnosis accuracy.

The application provides a generator set running, leaking and dripping intelligent monitoring method and system based on AI recognition, relates to the field of generator set intelligent monitoring, and comprises the following steps: collecting an image sequence and performing multi-waveband spectral decomposition to generate a spectral feature map, so as to identify a leakage area and a material type; then, a leakage area centroid change is calculated to construct a diffusion vector field, on the basis of which, a diffusion boundary is predicted and a diffusion risk coefficient is calculated; the shortest distance between the predicted boundary and the unit boundary is combined to determine an early warning level; finally, the leakage source is located through reverse tracing of the diffusion vector field. The application realizes early and accurate identification of leakage, diffusion trend prediction and source positioning, and improves the automation and early warning timeliness of monitoring.

The application provides a somatic agent training method based on flow matching to generate 3D perception and thinking chain, comprising: generating a three-dimensional geometric structure perception representation based on multi-modal input; generating a plurality of semantic heterogeneous reasoning paths by a thinking chain generator adopting a high-temperature sampling strategy based on the three-dimensional geometric structure perception representation and analyzing the paths into structured semantic strategies; fusing the structured semantic strategies with the three-dimensional geometric structure perception representation to guide a flow matching model to generate a candidate trajectory; verifying the disturbance stability of the candidate trajectory, calculating a consistency score by using a deterministic vector field, and selecting an optimal trajectory; executing the optimal trajectory and introducing an after-the-fact semantic verification and dynamic backtracking mechanism based on the thinking chain to realize self-correction of task execution. The application significantly improves the three-dimensional perception ability of the somatic agent in an uncertain environment, the diversity of task planning, the physical feasibility of action and the completion robustness of long-range tasks, and realizes robust planning and execution of complex long-range tasks.

This application provides a neural network modeling method and apparatus for nonadiabatic time-dependent density functional dynamics, relating to the technical field of interdisciplinary artificial intelligence and computational physics. The method includes: constructing a phase space state vector based on the real and imaginary parts of the complex state coefficients of the occupied orbits to obtain a feature representation; inputting the feature representation into a Hamiltonian dynamics branch to obtain a Hamiltonian-induced vector field; inputting the feature representation into a direct vector field branch to obtain a direct vector field; and inputting the feature representation into a memory exchange-correlation potential branch to obtain an exchange-correlation potential. The method and apparatus provided in this application, by constructing a three-headed neural network including a Hamiltonian dynamics branch, a direct vector field branch, and a memory exchange-correlation potential branch, simultaneously model the structure, residuals, and memory. This achieves higher accuracy, better long-term stability, and stronger physical interpretability compared to existing technologies in simulating strongly nonequilibrium, long-term nonadiabatic dynamic processes.

This invention discloses a GNSS noise identification method and system based on Riemannian manifold evolution and neural operators, belonging to the field of high-precision data processing for satellite navigation. The method first reconstructs the phase space of the GNSS coordinate sequence, using a manifold encoder to map it to the latent space of the Riemannian manifold to obtain the metric tensor. Second, it constructs a continuously evolving vector field controlled by a neural constant differential operator, and generates the eigenphysical evolution ideal trajectory through spatiotemporal integration. Subsequently, it introduces the Lie derivative operator to calculate the geometric deviation rate of the observation field along the ideal field evolution to strip away the physical signal, outputting a high-purity random noise manifold. Finally, it uses a Fourier neural operator to extract frequency domain features, fits the power spectral density, inputs a fully parameterized characteristic equation to calculate the significance coefficients, and dynamically outputs the optimal combined noise model. This invention overcomes the shortcomings of traditional methods that are sensitive to discrete sampling and missing data, achieving high-precision geometric decoupling of physical signals and random noise, and adaptive identification of complex noise.

The application provides a kind of multi-layer non-parallel medium antenna radiation characteristic simulation method and device, wherein the method comprises: obtaining the model information of the grid model of multi-layer non-parallel medium and the initialization information of incident ray;Model information includes the geometric information and material information of triangular grid;The initialization information of ray includes: incident origin, direction vector and vector field;Multi-order ray tracing is carried out on the split rays generated after the incident ray from incident origin to the interface of multi-layer non-parallel medium;The tracking content includes the split point, direction vector and vector field of each order split ray;Based on the tracking content obtained after multi-order ray tracing, the vector field of all split rays with split point located on the outer surface of multi-layer non-parallel medium is coherently superimposed to obtain the outer surface vector field;According to the outer surface vector field, the far-field pattern for simulating antenna radiation characteristics is calculated.The scheme can quickly calculate the antenna radiation characteristics of multi-layer non-parallel medium.

The present application relates to the technical field of image recognition, and discloses a river channel floating object intelligent recognition method based on multi-device image fusion, comprising: acquiring the visible light polarization signal and the long-wave infrared absolute temperature of a target water area, mapping into a polarization radiation intensity field and a thermal radiation intensity field and extracting corresponding gradient vector fields; calculating the divergence of the polarization gradient vector field to obtain a polarization anomaly divergence field; combining the time partial derivative of the thermal radiation intensity field and the thermal force gradient vector field to deduce a thermal force phase flow velocity field; calculating the spatial first-order gradient vector of the polarization anomaly divergence field and performing an inner product operation with the thermal force phase flow velocity field to obtain a cross-damping tensor field; substituting into a Poisson equation to reconstruct a global closure topological potential energy field; solving an intrinsic real phase Gaussian curvature field of the potential energy field; calculating the global energy intrinsic expectation ground state, and performing a phase change collapse judgment on the curvature field according to the same to output a binary mask.

This invention provides a method and system for detecting abnormal diver behavior based on a spatiotemporal attention mechanism. The method includes: constructing an underwater local optical distortion vector field by tracking the displacement of suspended particles in a video. This vector field is then used in a feature extraction network to perform regularization correction on the image and guide the receptive field shift to extract limb spatial features. When a limb is obscured by a bubble, cross-frame cross-correlation calculation is performed by analyzing the deformation of the bubble boundary and the limb motion velocity vector field of the last frame before obscuration to infer the implicit motion trajectory of the obscured limb, and a compensating feature vector is constructed to generate a spatiotemporal feature matrix. Finally, the divergence value of the gradient field of this matrix is ​​calculated to quantify the degree of behavioral abruptness; when the divergence value breaks through the stable range, it is determined to be abnormal. This invention solves the technical problems of inaccurate behavioral feature extraction and interruption of temporal information caused by underwater environmental interference and bubble obscuration.

This invention provides a method for generating spline animations of surface flow in a 3D model based on Hodge decomposition. The method includes generating a vector field from an input 3D mesh model; performing Helmholtz-Hodge decomposition on the current vector field to generate an irrotational field, a divergence-free field, and a harmonic field; using the divergence-free field as the initial vector field and calculating the corresponding curl; integrating the current vector field over time using the curl as velocity to obtain the vector field at the next time step; calculating the corresponding curl based on the vector field at the current time step, and then performing calculations for the next time step until a set time step is reached; and finally generating a spline animation by visualizing the trajectory. This invention ensures that the vector fields are continuously correlated and do not intersect in each consecutive time step, guaranteeing the continuously correlated flow pattern of the generated visualized trajectory, and can automatically and efficiently simulate the animation effect of flowing lines.

This invention discloses a near-field microwave SAR imaging scanning trajectory planning method for non-destructive testing of aircraft coatings. By introducing Poisson surface reconstruction to perform global implicit surface fitting on noisy point clouds, a smooth and continuous triangular mesh model and a stable normal vector field are obtained, fundamentally eliminating the interference of original point cloud noise on normal vector calculation. An adaptive region partitioning strategy based on curvature and probe scanning range enables intelligent partitioning of arbitrarily complex surfaces. By constructing a local scanning coordinate system based on normal vectors within each sub-region and generating a zigzag trajectory, high-precision fitting of the scanning path to the actual surface is achieved. By simultaneously using the reconstructed 3D model for collision detection environment construction and implementing dual collision safety checks during planning and execution using a motion planner, absolute safety during the detection process is ensured.

This invention relates to the field of industrial machine vision inspection technology, specifically disclosing an intelligent detection system for surface defects in mechanical parts. The system acquires bright-field and dark-field image sequences of mechanical parts, and filters valid images based on image sharpness and integrity evaluation. It aligns and corrects the initial images with a standard template through multi-scale spatial registration. Based on the structural tensor, it constructs a local feature descriptor, calculates the geodesic distance of the Riemannian manifold to construct a global differential energy field, and solves for a dense motion vector field by functional inversion including data fidelity terms and vector field regularization terms. Geometric deformation normalization is achieved through vector field structure enhancement and divergence-curl decomposition. Finally, multi-directional contour wave transform is used to enhance gradient features, and adaptive threshold segmentation and morphological tomography are combined to generate a defect segmentation map. This invention effectively eliminates the interference of part posture differences and systematic deformation, achieving accurate identification and quantitative analysis of surface defects.