45 results about "Continuous field" patented technology

The invention relates to the technical field of vehicle trajectory analysis, in particular to a vehicle trajectory anomaly detection method and system based on deep learning, and the method comprises the following steps: obtaining trajectory coordinates and boundary distance to generate symbol offset, constructing a road offset continuous field through interpolation and spatial embedding, calculating the ratio of instantaneous speed to reference speed, and obtaining a vehicle trajectory anomaly detection result. And generating an unbalance degree parameter, executing vector dimension stretching, extracting a trajectory embedding vector and an abnormal score by using a long short-term memory network, executing square summation and difference operation, and generating a trajectory energy gradient value. According to the method, migration characteristics are mapped to a continuous field to achieve track and road geometric constraint association, unbalance degree parameters are used for remarking directional quantity scales to enhance speed dynamic perception, energy gradient modulation scoring is cooperated, a confidence probability curved surface is constructed by using trilinear interpolation, and a curved surface peak value is positioned to output a detection result. Positioning noise is suppressed; and abnormal precision under sparse sampling is improved.

The invention relates to the field of conveyor state monitoring, and discloses a conveyor running state monitoring system based on edge calculation, which is used for running state monitoring and fault early warning of an industrial conveyor system. The conveyor operation state monitoring system based on edge calculation comprises a sensing module, a prediction module, an identification module, a decision module and an optimization module. The sensing module constructs a virtual sensing field by deploying a multi-physical-quantity sensor, and continuous field description of the running state of the conveyor is established; the prediction module performs multi-state parallel deduction based on a quantum superposition principle to generate a prediction result; the identification module uses a biological field resonance principle to realize abnormity identification and fault positioning; the decision module generates a control strategy according to the abnormity severity and the resource state; and the optimization module forms a closed-loop system by adjusting resources. Accurate monitoring, early warning and intelligent decision making of the running state of the conveyor are achieved, and the reliability and efficiency of equipment monitoring are improved.

The invention discloses a shape memory alloy multi-physics field analysis method based on an expansion depth operator network, belongs to the technical field of material science and engineering calculation, solves the problem of phase field dynamics analysis of multi-physics field coupling under various initial conditions, and comprises the following steps: S1, constructing the expansion depth operator network, the expansion depth operator network comprises a branch network and a backbone network; and S2, determining the total error of the neural network according to the data information error and the physical information error, optimizing the parameters of the neural network by adopting an AdamW optimizer and a Minibatch technology, training the neural network, and completing the shape memory alloy multi-physical field analysis based on the extended depth operator network. According to the method, the expansion depth operator network is used for replacing a finite element method, the solving process of the multi-physical field phase field equation is accelerated, field quantity distribution continuous in time and space is obtained, the solving process is accelerated, and continuous field distribution is obtained.

The invention provides a laser communication device for rapid link establishment and networking and a satellite internet constellation. The device utilizes the wavelength division multiplexing unit and the pre-aiming unit to realize beam combination of receiving and transmitting optical path signals so as to improve the coaxiality of receiving and transmitting branches; an optical fiber displacement device arranged on the transmitting-receiving lens is used for adjusting the relative position of the beam combining optical fiber and the focal point of the transmitting-receiving lens, and continuous beam changing of a transmitted light path signal within the divergence angle range of 55-1000 microrad is achieved; and a position sensing unit is matched to focus a receiving light path signal, so that a 0.2-1-degree receiving light path continuous field of view is realized. Through 10000 times of offset error estimation simulation, the coverage rate of an uncertain area to a target light spot reaches 99.32%, and about 1mrad beam change can be realized, so that minute-level capture efficiency is realized under the condition of the continuous field of view.

This invention provides a method for processing streaming data throughout the entire lifecycle of an engineering project, relating to the field of engineering data processing technology. The method includes: constructing permanent spatial anchoring expression units for target structural units during the construction phase of deeply buried structures, generating a set of permanent spatial expressions for concealed components; performing lifecycle flow mapping modeling of the spatial anchoring expression units to generate a set of traceable spatial evolution paths for the concealed phase; performing spatial expression reconstruction and closed-state association processing for the concealed structural phase to generate a continuous field of spatial expression for the concealed structural phase; and performing continuous generation and flow expression modeling of the spatial semantics of streaming data throughout the entire lifecycle of the engineering project to generate a streaming data flow expression model for the entire lifecycle of the engineering project, completing a traceable flow expression of the spatial state of the concealed structural phase. This achieves continuous streaming data flow throughout the entire lifecycle, improves the digital management and safety assurance capabilities of engineering projects, and solves the problem of non-retrospective spatial flow after the closure of deeply buried structures.

The invention discloses an X-ray and NeRF algorithm-based luneberg lens internal structure analysis method, and relates to the technical field of three-dimensional imaging of optical lenses, and the method comprises the following implementation steps: S1, configuring an RGB-X-ray combined imaging system, and carrying out data acquisition, and S2, carrying out the calibration of external parameters of a camera through a calibration plate: carrying out the imaging through a specially-made calibration plate, and carrying out the calibration of the external parameters of the camera through the calibration plate. The method comprises the following steps: acquiring a plurality of frames of RGB (Red, Green, Blue) images to obtain external parameters of an RGB camera under a calibration plate coordinate system, acquiring a plurality of frames of X-ray images to obtain an external parameter matrix of an X-ray camera about the calibration plate coordinate system, performing implicit representation on a continuous field in a lens by utilizing NeRF, mapping three-dimensional coordinates to differentiable density and radiation values, and calculating the calibration plate coordinate system. According to the method, high-fidelity reconstruction from multi-angle X-ray projection to a three-dimensional structure is achieved, detail information in the lens can be effectively reserved, meanwhile, the continuity and symmetry of the structure are guaranteed, and compared with a traditional voxelization or CT reconstruction method, the problems that noise artifacts, fractures and unstable numerical values are likely to be generated in the traditional voxelization method are solved.

The embodiment of the invention provides an automatic driving scene mining method and device, equipment, a medium and a program, and the method comprises the steps: obtaining a mining demand text of an automatic driving scene; target scene features in the demand text are identified and mined, a database is inquired based on the target scene features, the database stores time sequence features related to the automatic driving scene, and the time sequence features are composed of feature frames at different moments; querying a target feature frame and a single-frame moment corresponding to the target scene feature from the time sequence feature, and querying at least one time sequence segment from the time sequence feature based on the single-frame moment of the target feature frame; and generating scene mining data of the automatic driving scene according to all the queried time sequence fragments. Therefore, the problems of low working efficiency, poor flexibility and expandability and the like caused by the fact that continuous scene fragments conforming to complex spatial-temporal logic cannot be efficiently, accurately and automatically mined from massive time series data due to dependence on manual rules and static retrieval in related technologies are solved.

The invention provides a broken shaft protection method for a circuitous wheel of an aerial passenger device, and relates to the field of coal mine safety monitoring. The problems that in the prior art, space-time alignment errors of multi-source data are large, a physical field is distorted, and real-time performance is insufficient are solved. Through dynamic graph structure modeling, heterogeneous sensors such as a vibration sensor, a temperature sensor, a displacement sensor and the like are mapped into space-time association graph nodes, edge connection is constructed based on spatial proximity and time synchronism, and continuity characteristics of an equipment structure are coded. And generating a three-dimensional continuous field and dynamically visualizing the three-dimensional continuous field through a ray casting volume rendering technology. According to the scheme, high-precision and high-real-time technical support is provided for broken shaft fault early warning.

The application discloses an F2 layer critical frequency continuous field reconstruction method and device, equipment and medium, and relates to the ionosphere space parameter prediction field, and comprises the following steps: adding a to-be-predicted space position as a node to an initial heterogeneous graph to obtain a target heterogeneous graph; processing the target heterogeneous graph through a preset graph attention network to obtain an F2 layer critical frequency prediction value corresponding to the to-be-predicted space position; the F2 layer critical frequency prediction value is a prediction value generated based on a scaling relationship between the F2 layer critical frequency and electron density; and performing a continuous field reconstruction operation based on a plurality of F2 layer critical frequency prediction values. Thus, the correlation of different discrete ionosphere sonde stations is utilized by the graph attention network, the F2 layer critical frequency is predicted in combination with the scaling relationship between the F2 layer critical frequency and electron density, the F2 layer critical frequency accuracy and the robustness under disturbance conditions can be improved, and the F2 layer critical frequency of any space position can be predicted, so that the reconstruction of the continuous space field is realized.

This application relates to the technical field of smart blackboards and discloses an interactive resource scheduling and optimization system for a blackboard, comprising a field mapping unit, a field solving unit, a topology unit, a control unit, and an instruction unit connected in sequence. The field mapping unit converts discrete interactive events into a continuous demand density distribution field and a resource capability potential field on the blackboard plane. The field solving unit dynamically obtains the current state of the demand density distribution field and the resource capability potential field by simultaneously solving a system of partial differential equations containing nonlinear coupling terms. The topology unit simultaneously optimizes the resource density distribution and the topology of the resource region by minimizing a preset energy functional. The control unit generates continuous control quantities for resource scheduling accordingly. The instruction unit finally converts these into a real-time sequence of underlying hardware control instructions. By introducing a spatially continuous field description, coupled dynamic system modeling, and topology optimization, the system improves the adaptive capability of resource scheduling, global optimization efficiency, and real-time response performance.

This application relates to a method, system, equipment, and medium for construction site environmental monitoring based on multi-source fusion. The method improves data reliability and consistency at the source by spatiotemporally aligning multi-source sensor data and innovatively calculating a real-time data quality factor to guide the weighted fusion of neighboring data. Furthermore, it uses physical constraints to repair missing values ​​and combines spatial interpolation to generate a refined environmental data cube covering the entire area, achieving data enhancement from discrete points to a continuous field. Finally, it extracts multi-dimensional features based on this cube and constructs a dynamic benchmark model combining construction stages and weather conditions for intelligent comparison and early warning. This fundamentally solves the problems of low data quality, information silos, and delayed early warning, achieving a comprehensive improvement in the reliability, panoramic perception capability, and forward-looking early warning of construction site environmental monitoring.

The application provides an AIGC card point editing method and system, solves the problems of audio and picture splitting, hard transition, lack of global planning and inability to complete when the material is insufficient caused by the dependence of the prior art on discrete beat point editing. The method comprises the following steps: constructing an audio energy field and a visual energy field based on background audio and video material respectively; establishing a two-field coupled partial differential equation, solving a globally optimal editing path, obtaining an editing point sequence and a segment allocation scheme; for the interval with a matching error exceeding a threshold, calling an AIGC to generate a supplementary video segment, and generating a transition frame sequence according to the visual energy field jump amplitude at the editing point; dynamically adjusting the playing speed according to the ratio of the energy field time gradient; finally, splicing and fusing the variable speed segment, the transition frame and the supplementary segment to output the card point video. The application realizes global optimization editing of continuous field coupling, and improves the card point precision and visual fluency.

The application provides a mine heavy metal pollution simulation prediction method and system based on digital twinning, relates to the technical field of digital twinning, and first acquires a multi-source environmental monitoring data set containing heavy metal concentration sampling values and spatial coordinate information of a target mine, injects the multi-source environmental monitoring data set into a pre-constructed mine digital twinning body, generates a discrete distribution of heavy metal concentration spatial scatter point sets in the digital twinning body according to spatial coordinate information mapping, then performs three-dimensional space continuous field reconstruction processing, generates a heavy metal three-dimensional distribution field, further calls a heavy metal migration and diffusion simulation model based on a physical process drive, performs multi-medium coupling migration and diffusion simulation on the three-dimensional distribution field in the digital twinning body, and generates a space-time evolution prediction result containing heavy metal space distribution prediction fields and pollution diffusion front advancing trajectories of future time periods. The application realizes deep fusion of monitoring data and physical simulation in the digital twinning body, and improves the accuracy of mine heavy metal pollution prediction.

The present application relates to the technical field of agricultural intelligent irrigation and precise pesticide application, and discloses a quantitative pesticide spraying system and method for standardized planting of traditional Chinese medicinal materials. A coordinate system and an equidistant sampling grid are established on the surface of a field, a capacitive soil humidity sensor is laid out and calibrated, real-time voltage data is accurately converted into the soil water content of the grid nodes, and four-point bilinear interpolation is used to realize the acquisition of the soil water distribution of the continuous field in the field. Based on the node difference calculation, the water content spatial gradient and the average gradient of the whole field are calculated, the area is divided according to the number of nozzles, the regional boundary is dynamically adjusted to ensure the data validity, and the average gradient of each area is extracted. Combined with the unit area pesticide spraying demand and the coverage area, the basic pesticide spraying volume is obtained, the adjustment factor is generated by linear mapping of the regional gradient and the global extreme value, the pesticide spraying volume is differentiated and controlled, and the boundary clamping is implemented, and finally the opening time is calculated according to the spraying speed, and the controller executes the operation in sequence.

The application provides a macro-micro conversion individual generation and trajectory fusion method of ghost car fleet vehicles, and relates to the technical field of intelligent transportation. The method comprises the following steps: in a target trajectory generation period in multiple trajectory generation periods, obtaining a traffic state label of a target traffic flow according to a continuous field quantity buffer of the target traffic flow; generating a micro vehicle according to the traffic state label, and obtaining a multidimensional state of the micro vehicle; fusing a target trajectory with the highest candidate weight in multiple candidate trajectories generated in a first area and a pre-acquired measured trajectory to obtain a completed trajectory; taking the completed trajectory and a micro vehicle trajectory as a vehicle trajectory generation result of the target trajectory generation period, and storing the vehicle trajectory generation result in a discrete individual quantity buffer; wherein the micro vehicle trajectory is generated according to the multidimensional state. Compared with a macro-micro conversion method of randomly scattering points, the method can reduce false speed waves and false queues, and improve the consistency of macro-micro conversion, data fusion accuracy and trajectory completion quality.

This invention relates to a time-domain electromagnetic simulation method based on Tree-Cotree decomposition and a hybrid DG-FE / SE domain decomposition. The method includes: dividing the computational domain into a tetrahedral finite element subdomain and a hexahedral spectral element subdomain according to geometric dimensions and electromagnetic properties; constructing a minimum spanning tree and cotree based on mesh edges to apply Gaussian divergence constraints; discretizing both the finite element and spectral element regions using scalar and vector basis functions to obtain Galerkin weak solutions; uniformly processing the interface between the finite element and spectral element subdomains using a topology identification-polygon decomposition-Gaussian integration method, and combining this with upwind flux to achieve continuous field transfer; assembling the spatial discretization matrix of the Galerkin weak solutions, and completing time propagation using an explicit fourth-order Runge-Kutta scheme. Compared with existing technologies, this invention systematically eliminates the low-frequency collapse problem, ensures the non-singularity of the system matrix, and guarantees solution accuracy and efficiency.

The invention provides a method and a device for rapidly generating crop growth vigor information, which are used for monitoring the regular growth vigor condition of crops by utilizing satellite remote sensing inversion and supporting the continuous monitoring of the field growth vigor of crops such as rice, corn, rape, citrus and tea and the monitoring of the growth state of the growth stage. The method comprises the following steps: firstly, performing time sequence optical image reconstruction and data space-time fusion on an obtained multi-temporal optical image to generate high space-time remote sensing data; then, remote sensing parameters related to crop growth are extracted from the high space-time remote sensing image, a plurality of parameters such as a normalized vegetation index and a normalized yellow difference index are calculated, a high space-time crop growth grid result is obtained, and under the spatial constraint of a crop vector plot, elements are converted into grids to serve as mask data; and performing mask extraction on a resampled high-space-time crop growth grid result to obtain spectral characteristics of a corresponding unit of crops, performing wave band set statistical calculation to obtain a growth index mean value of a certain crop, and obtaining a comprehensive evaluation result of the plot-level crop growth.

The invention discloses an aero-engine rolling bearing remaining service life prediction method and system based on a coordinate neural field, and belongs to the technical field of fault prediction and health management. Comprising the following steps: constructing and training a conditionally modulated coordinate neural field network, representing time-frequency observation of a vibration signal as a continuous field, and applying a random mask during training to learn missing detection repair; features extracted by the network are aggregated into latent representation, and a conditional diffusion model is trained to learn evolution priori of the conditional diffusion model; during prediction, a current latent representation is obtained based on latest observation, and a gradient based on observation likelihood is introduced to carry out posterior guidance in the process of executing reverse sampling of a diffusion model to generate a future trajectory; and finally, decoding the track and calculating a health index to obtain probability distribution of the remaining service life. According to the method, data missing is uniformly processed through continuous representation, and the generative model is combined with posterior guidance, so that high-precision, calibratable and highly-adaptive residual service life probabilistic prediction is realized.

The application discloses a kind of reference calibration methods for the size drift of mobile phone shell positioning hole, it is related to mobile phone shell production technical field, comprising: obtaining the contour boundary data of multiple positioning holes on mobile phone shell, contour action energy field is established in the internal region of positioning hole, and according to the distance relationship of each position to contour boundary, distance potential field is constructed, and center probability field is generated;According to the energy variation continuity and propagation direction consistency of contour action energy field analysis positioning hole contour boundary each area, identify abnormal action area and carry out energy attenuation processing to abnormal action area.The application constructs contour action energy field and distance potential field in the internal positioning hole, and further generates center probability field, so that the effect of positioning hole contour structure on internal space can be expressed in the form of continuous field, so that the stable center position of hole region can be stably determined when the positioning hole contour appears ovalization or local anomaly.

The invention discloses an under-forest multi-parameter collaborative two-dimensional environment field modeling method and related equipment, and relates to the field of environment field modelling. According to the method, related sensing data, environment temperature, environment humidity and harmful gas concentration data of a wind field are synchronously collected through all detection nodes, and real-time geographic positions are obtained; the remote monitoring platform performs time synchronization and spatial position alignment on the multi-parameter environment data based on the sampling timestamp and the geographic position to form a multi-parameter environment data set with a unified space-time reference, and further constructs a two-dimensional environment field covering a monitoring area based on the data set and the geographic position of each detection node; therefore, the conversion of the multi-parameter environmental information under the forest from discrete perception to the spatial continuous field model is realized, and the problems that the monitoring data is isolated and the global environment situation cannot be formed in the prior art are effectively solved.

This invention belongs to the field of welding monitoring technology, and particularly relates to a multi-field coupled dynamic monitoring and quantitative prediction method for the initiation and propagation of hydrogen-induced cracks in welding. This method simultaneously collects multi-source data including process parameters, temperature, strain, and hydrogen concentration during the welding process, performs time synchronization alignment and finite element mesh spatial matching, and reconstructs discrete measurement points into continuous temperature, strain-related, and hydrogen concentration fields under observation consistency and physical constraints. The continuous field is used as boundary conditions and correction quantities to drive the thermo-mechanical-hydrogen physical field coupling model in real time for transient calculations. Furthermore, a risk index is calculated based on hydrogen-induced crack initiation criteria, and risk regions are identified. Combined with a hydrogen-assisted cracking model, the crack propagation path and rate are predicted, a comprehensive early warning index is constructed, and a threshold determination is performed to achieve early warning output. This invention can achieve quantitative assessment and early warning of crack risk during welding, reducing the risk of missed defects and rework after welding.

The invention provides a mainboard configuration method and electronic equipment, which can be applied to the technical field of computers. The hardware configuration method comprises the steps of obtaining a global fingerprint of a target mainboard through a basic input and output system in a boot loader stage when the target mainboard is started; in response to a target continuous field existing in a firmware program of the basic input and output system preloaded in the preset memory, performing data verification on the target continuous field according to a preset verification field corresponding to the global fingerprint; and under the condition that the verification result of the target continuous field indicates that the target continuous field is the same as the preset verification field, pre-compiling a configuration file of the target mainboard into a mirror image of mainboard firmware, so that the target mainboard executes configuration information in the configuration file.

This disclosure provides systems, methods, and devices for vehicle driving assistance systems that support image processing. In a first aspect, a method of image processing includes receiving point cloud data comprising a sequence of a plurality of point clouds with respect to time; determining, for a first point cloud, a first set of patches; determining, for the first point cloud, a second set of patches having patches of a different size than patches of the first set; determining a latent representation of the point cloud data and of motion across the sequence of the plurality of point clouds based on the first and second sets of patches and a second point cloud; and determining a scene flow representing the motion depicted across the sequence of the plurality of point clouds based on the latent representation. Other aspects and features are also claimed and described.

The application provides a deep earth target monitoring method and device for spatiotemporal continuous modeling. The method embeds sensor discrete sampling data into a four-dimensional spatiotemporal coordinate domain and establishes a global topological graph structure through a dynamic graph attention mechanism; an implicit neural spatiotemporal continuous field based on coordinate mapping is constructed, deep earth signal features are extracted by using a multi-scale position coding and a periodic activation architecture; a geomechanics partial differential equation is introduced in parameter optimization to construct a physical constraint operator, realizing physical consistency deduction of non-observed coordinate points; a Fourier neural operator is used to perform operator mapping in the frequency domain, reconstructing a cross-scale global continuous feature field; and a companion gradient inversion algorithm is used to realize reverse source location and microcrack trajectory tracking. The application realizes conversion from discrete observation to high-fidelity continuous physical field, significantly improving monitoring accuracy and early warning reliability in extreme working conditions such as blind area monitoring, noise interference and local sensor failure.

The invention discloses an optical flow estimation method and device based on layered geometric injection, a medium and a product, and relates to the field of optical flow estimation.The method comprises the steps that a feature pyramid is extracted from two continuous frames of images through an encoder, and the feature pyramid comprises multiple layers of motion features from deep to shallow; except the motion features of the shallowest layer, according to the sequence from deep to shallow, the motion features of the deeper layer are injected into the adjacent motion features of the shallower layer in sequence through a geometric injection module, and finally target features are obtained; constructing a full-pair cost volume according to the target features; according to the full-pair cost volume, the motion feature of the shallowest layer and the decoder, the target optical flow is estimated, high optical flow estimation accuracy is still achieved in the scene with discontinuous motion, and the problem that an existing optical flow estimation method based on deep learning is low in optical flow estimation accuracy in the scene with discontinuous motion is solved.

The embodiment of the invention discloses a method for constructing a topology description function of a three-dimensional spline component. The method comprises the following steps: constructing a first function and a second function; the first function and the second function take the parameter coordinates as variables, and are respectively used for determining the roundness degree of cross section edges corresponding to the parameter coordinates and the number of polygon-like edges; the first function and the second function are continuous differentiable mathematical operation functions; constructing a third function according to the first function and the second function; and constructing a topology description function of the three-dimensional spline component according to the third function. Thus, the first function and the second function are both continuous differentiable functions with parameter coordinates as variables, a continuous field along a path (parameter coordinates) can be obtained, geometric edges and corners and additional design variables caused by segment jump can be thoroughly eliminated, and the third function can enable the cross section size to continuously change along with the high order of the parameter coordinates by using a single analytic expression. The complexity of a topology description function expression can be reduced.

The invention belongs to the technical field of wireless communication, and particularly relates to a robust communication resource allocation method for multi-IRS auxiliary USV security data unloading. The method comprises the following steps: constructing a multi-IRS-assisted USV-UAV network data unloading system model; calculating a USV average information age based on a multi-IRS-assisted USV-UAV network data unloading system model; according to the USV average information age, constructing a minimum average information age optimization problem by taking a CUAV calculation time delay constraint, an edge server resource constraint, a service scheduling constraint, a UIRS trajectory constraint, a safety rate constraint, a phase parameter constraint and an error range constraint as constraint conditions; solving a minimum average information age optimization problem to obtain a communication resource allocation scheme; according to the invention, communication security can be guaranteed, system information age can be effectively reduced, and data timeliness and system robustness in a multi-task continuous scene can be improved.

The present application relates to a kind of based on sparse sounding data phased array radar reflectivity factor simulation method, belong to radio field, based on the sparse sounding data of multiple point multiple height layer collected, sparse reconstruction simulation scheme is designed, the high-quality restoration of original sparse observation data and three-dimensional continuous field construction are realized.Sparse reconstruction simulation method based on sparse sounding data is provided, and the method comprises the following steps: constructing multivariate dictionary base matrix;By sparse representation theory and bayesian optimization model, the sparse coefficient reflecting the essential characteristics of precipitation is obtained, and the high spatial resolution reconstruction of temperature and humidity field and precipitation microphysical parameter field is realized.Afterwards, the phase state determination and probability field calculation model driven by meteorological parameters is constructed, the three-dimensional simulation of liquid, solid and mixed state precipitation echo is realized.Finally, the unit direction vector and azimuth guide code matrix of phased array radar beam ray are calculated, the radar volume scanning data implicit sampling simulator is introduced, and the high-precision reflectivity factor simulation under the stereoscopic scanning mode of phased array radar is realized.

The application discloses a kind of based on knowledge graph's rubber surface defect real-time detection system and method.The system includes defect fingerprint extraction module, defect dynamics modeling module, hierarchical decision graph module, adaptive strategy optimization module, defect field analysis module and intelligent compensation control module.Through multiscale feature extraction, construct defect fingerprint and pedigree network, based on physical and chemical principles, predict defect evolution, use four-layer knowledge structure to realize multi-hop graph reasoning, use Bayesian optimization and meta-learning to dynamically adjust detection strategy, model rubber surface as continuous field for anomaly positioning, and generate hierarchical compensation strategy according to impact score.The application realizes the integration of detection, prediction, reasoning and control, and improves detection accuracy and efficiency.

The present application relates to the technical field of vehicle trajectory analysis, in particular to a vehicle trajectory anomaly detection method and system based on deep learning, comprising the following steps: obtaining trajectory coordinates and boundary distance to generate symbolic offset, constructing road offset continuous field through interpolation and space embedding, calculating instantaneous speed and reference speed ratio, generating imbalance degree parameter and performing vector dimension stretching, extracting trajectory embedding vector and anomaly score using long short-term memory network, performing square sum and difference operation, and generating trajectory energy gradient value. In the present application, the offset feature is mapped to the continuous field to realize the correlation of trajectory and road geometry constraint, the imbalance degree parameter is used to recalibrate the vector scale to enhance the speed dynamic perception, the energy gradient is modulated to score and the trilinear interpolation is used to construct the confidence probability surface, the peak value of the surface is located to output the detection result, and the positioning noise is suppressed to improve the anomaly accuracy under sparse sampling.