45 results about "Vector distribution" patented technology

The invention relates to the technical field of image processing, and discloses a house quality detection method and system based on a live-action three-dimensional model. The method comprises the following steps: converting an OSGB format live-action three-dimensional building model into a body coordinate system and carrying out quality evaluation, carrying out semantic segmentation on a standardized model to obtain a component geographic entity set, establishing an independent local coordinate system of each component to form a single three-dimensional model set, generating a multi-view projection feature image set based on a normal vector distribution planning view angle, and carrying out multi-view projection on the single three-dimensional model set. And detecting defects by adopting a 3D-YOLO-BuildingDefect algorithm, and carrying out inverse transformation to obtain three-dimensional space defect distribution and a quality grade evaluation result. The technical problems that a traditional two-dimensional detection method is insufficient in spatial positioning precision, incomplete in detection coverage rate and insufficient in geometric information utilization are solved, and the spatial positioning accuracy of house quality detection and the comprehensiveness of defect recognition are improved.

The invention provides a multi-source data fusion analysis system for electronic commerce digitization insight, and relates to the technical field of data processing, and the system comprises a reference analysis unit used for selecting a reference analysis unit in a spatial relation structure, calculating fractal dimension features of feature vectors, carrying out multi-level recursive division on the analysis unit based on fractal dimension values, and carrying out multi-level data fusion analysis on the fractal dimension values; each level of division keeps a self-similarity characteristic, and feature partitions are obtained; the partition space distribution module is used for mapping the feature vector distribution to corresponding feature partitions and calculating partition space distribution weights based on vector density distribution and space position relations of the feature partitions; the value evaluation module is used for performing parameter adjustment on the deep learning network by using the partition space distribution weight to obtain a value evaluation result; and the value evaluation module is used for processing a value evaluation result through a multi-modal fusion algorithm, generating a personalized recommendation strategy based on a reinforcement learning framework, and calculating resources and an intelligent routing algorithm so as to complete a digital insight analysis process. The e-commerce data utilization efficiency and the decision-making effect can be improved.

The invention relates to a vector retrieval method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: receiving a query vector; searching according to the query vector and the global vector index of the vector index based on the search parameter to obtain an initial cluster; wherein the global vector index is determined based on a single-layer hierarchical navigable small-world graph; the search parameters are used for controlling the size of the initial cluster; under the condition that the vector distribution in the initial cluster set does not meet a preset condition, executing expansion search in the global vector index by taking the initial cluster set as a starting point to obtain each adjacent cluster; determining a target cluster set based on the initial cluster set and each adjacent cluster; the target cluster set is searched in a local vector index of the vector index, a target vector is obtained, and the local vector index is determined based on the inverted file plane vector index. By adopting the method, the vector recall precision can be improved.

The application relates to the field of engineering surveying and mapping technology, in particular to an engineering surveying and mapping marking method and system based on an AI robot, which comprises the following steps: controlling a surveying and mapping robot to move to a to-be-marked engineering area to collect environment point clouds and environment images, constructing a three-dimensional environment map based on the environment point clouds and the environment images, identifying the geometric features and spatial distribution of key surveying and mapping control points based on the three-dimensional environment map, planning a candidate position of a marking point, and generating a preliminary coordinate sequence of the marking point. The application introduces normal vector distribution analysis and a depth compensation mechanism, before surveying and mapping marking spraying, the robot calculates the local flatness and roughness of a base surface through line laser scanning, automatically fits an inclined plane and calculates a compensation depth, simultaneously adjusts the pitch angle of a mechanical arm, so that an inkjet head is always perpendicular to the base surface and is within a depth of field range, which ensures that no matter how the ground fluctuates, the marking point can keep a standard circle and clear boundaries, and solves the problem of difficult marking on uneven ground.

This invention relates to the field of large-model latent space mapping technology and discloses a cross-scene hazard migration identification method based on large-model latent space mapping. The method collects historical hazard data and normal operation data from multiple scenarios, and obtains a standardized scene dataset after format unification and noise filtering. Then, a large-model latent space mapping module is constructed to map hazard features and operation features from the standardized data to the latent space of a pre-trained large model, generating scene feature latent vectors. Based on these latent vectors, a cross-scene feature alignment model is constructed, obtaining a unified feature representation across scenarios by aligning the differences in latent vector distributions across different scenarios. Next, the hazard migration identification model is trained, using the unified cross-scene feature representation and real-time data from the target scenario to predict the probability of hazard presence. Finally, the scenario-level hazard identification threshold is calculated by combining the probability distribution of historical hazard data from the target scenario, and the feature alignment model parameters are updated periodically.

The application provides a water level regulation method and system for a water conservancy hub. The method includes the following steps: collecting data by underwater robots and satellite remote sensing data simultaneously, constructing a sediment deposition space-time atlas containing a deposition gradient, and periodically correcting the space-time atlas. A radar array is deployed in the curved section of the spillway, and a multi-frequency pulse signal is used to analyze the three-dimensional sediment flow velocity field vector distribution. Through the correlation analysis of the dynamic coupling of the deposition gradient and the flow velocity field, the correlation area of the flow velocity mutation and the sediment deposition is accurately identified, and a constraint condition set containing coordinate positioning and flow-silt matching rules is established. Combined with real-time water level monitoring data, the constraint condition set is used to calculate and generate a dynamic adjustment parameter sequence of the spillway gate opening, and the water level is accurately regulated. The technical scheme provided by the application can accurately regulate the water level of the water conservancy hub by intelligently optimizing the spillway parameters, effectively alleviate the sediment deposition problem in the high-sediment water area, and ensure the safe operation and flood control efficiency of the hub.

The present application relates to the technical field of data quality evaluation, in particular to a kind of automatic quality evaluation method of data set, comprising the following steps: extracting semantic fragment and vector information, clustering analysis covering structure, screening redundant sample, marking text difference, output consistency evaluation result.In the present application, the structured extraction and vector distribution measurement of semantic information enable high-precision identification of semantic coverage and structure density, vector overlap frequency and similarity threshold linkage enhance the discriminability of redundant detection, text position difference and semantic span joint labeling mechanism improves the systematicness of text consistency labeling, the synchronous comparison mechanism between text screening content and text difference labeling result ensures that the intersection sample of redundancy and deviation can be accurately identified, the dense clustering of cultural semantic content in vector space and the quantitative analysis ability of structure deviation point are synchronously improved, and the comprehensive dimension and logical closure of evaluation result are realized optimization.

The invention discloses a 3D printing small round hole part forming method which comprises the following steps: a) constructing a three-dimensional digital model of a part, and repairing a part model; b) according to part materials and small round hole structure characteristics, the width of a molten pool of a used metal powder material is included; the number of small circular hole contour laser vectors, the length of a single-section laser vector and the distance between an entity vector and a contour vector are set, a laser vector planning path at the small circular hole is obtained, and contour vector distribution is vector distribution with a staggered starting point and an end point; vector adjustment is carried out on the theoretical contour of the small round hole, a polygonal approximate contour formed by N sections of linear vectors is generated, at least one point of the starting point and the end point of each section of vector deviates from the boundary of the theoretical hole to the interior of a part entity, and a laser opening point and / or a laser closing point are / is located within the edge of the hole; and c) guiding the slices into equipment for printing. Energy accumulation is effectively restrained, hole blockage is avoided, and the precision of fine-structure parts such as small holes formed through 3D printing is improved.

Embodiments of the present application propose an imaging system based on super-Raleigh limit focusing lens and its application, which can be widely applied in the technical fields of photolithography, biomedical, optical imaging and nanometer structure precision measurement; the system can produce transverse wave vector distribution and excitation and detection vector beams with wave vector range meeting the requirements by wave vector ring modulation on incident linearly polarized light, focus the vector beams to make the focusing field meet the scale requirements, so as to obtain super-Raleigh limit excitation PSF and detection PSF, and obtain a confocal PSF capable of exciting high-order terms of spot intensity by superimposing the excitation and detection PSFs, thereby improving the imaging resolution of the system; on the other hand, the system can use solid and hollow focusing spots formed by focusing vector beams to form an image, and calculate according to the imaging result to reduce interference, so as to obtain a high-resolution detection result far exceeding the Raleigh diffraction limit, and further improve the resolution of the imaging system.

The application discloses a composition automatic scoring method based on multi-stage learning, and the method comprises the following steps: S1, extracting the shallow language features, emotional features and theme relevance features of the composition; S2, theme relevance feature extraction; S3, construction of a beautiful sentence identification model and extraction of composition style features; S4, training of a base learning machine; and S5, composition vector distributed representation and feature fusion model training and prediction. The application is applied to the field of automatic composition scoring, and a comprehensive and multi-dimensional composition scoring feature is designed for Chinese composition scoring, the detection and discovery of beautiful sentences in the composition are realized, and the beauty degree of language expression in the composition is better considered; meanwhile, the composition automatic scoring based on multi-stage learning is proposed, and multi-angle composition features are effectively combined for composition scoring.

The invention discloses a berthing and departing attitude optimization method and system based on model predictive control, and the method comprises the following steps: constructing a ship three-degree-of-freedom nonlinear kinematics and dynamics model, and taking the model as a prediction model of MPC; establishing a target function containing state tracking errors and control energy consumption, and controlling output through real-time rolling optimization calculation under the physical constraints of the thruster thrust and the rudder angle; in the berthing stage, weighted collaborative optimization is conducted on the ship position deviation and the attitude angle deviation through an MPC, and high-precision closed-loop control and thrust vector distribution of the pose in the low-speed environment are achieved; in the off-berthing stage, real-time prediction and momentum attenuation compensation are carried out on the inertial sliding track of the ship by combining MPC prediction step length based on preset off-berthing direction guidance, and output of a propeller is dynamically adjusted to ensure the safe envelope of an off-berthing path. According to the invention, high-precision attitude control and dynamic obstacle avoidance of the ship in a complex limited water area can be realized.

The invention discloses an electric power economic risk and customer value comprehensive evaluation system, which relates to the technical field of electric power system operation and comprises a multi-source data acquisition and preprocessing module, a multi-source data fusion module, a two-dimensional coupling evaluation modeling module, a dynamic weight adaptive adjustment module and an intelligent decision-making and visual output module. According to the method, the risk and the value are regarded as probability measurement, coupling is carried out by adopting optimal transmission methods such as Wasserstein gravity center and the like, and then quantiles, tail features and high-order moments are extracted from coupling distribution to serve as coupling evaluation vectors. The evaluation result of the distribution level fusion not only provides'single-point 'risk / value judgment, but also provides complete distribution form information. In combination with approximate Shapley-based interaction attribution, the system can reveal nonlinear interaction and marginal contribution between customer values and market / risk factors, thereby providing quantifiable and explainable decision basis for regionalized pricing, differentiated services, target customer retention strategies and risk slow release measures.

The application belongs to the technical field of three-dimensional modeling, and particularly relates to a three-dimensional scanning modeling method based on kinematics cooperation and sub-group fuzzy ant colony, which comprises the following steps: based on the kinematics theory of multi-body system, a global coordinate system of a scanning system and a homogeneous transformation matrix of each motion axis are established; a depth-of-field constraint is constructed; for a workpiece to be measured, a dynamic edge profile of the workpiece is acquired through the scanning system, a rough visual shell model of the workpiece is constructed, the normal vector distribution of each point on the surface of the workpiece is calculated, and a probe task set is distributed; the probe task set is split into discrete scanning task strips and solved; according to the optimal scanning task strip sequence obtained by solving, the scanning system is used to collect data of the workpiece; an adaptive sampling strategy based on a dynamic chord height error is used for point cloud gridding processing, a bidirectional curve fitting method is used to generate a NURBS surface, and a three-dimensional model of the workpiece is output. The application can realize full-automatic and high-precision digital reconstruction of a complex surface.

This invention relates to a photometric stereoscopic 2.5D mold monitoring imaging method, comprising a pre-calibration step and a real-time monitoring step. The real-time monitoring step includes: S1: sequentially triggering multiple light sources from different directions to illuminate individually via pulse control signals, and simultaneously triggering a camera to acquire multiple corresponding mold surface images; S2: performing pixel-level synthesis of the multiple mold surface images to generate a stable image with uniform illumination; S3: calculating the pixel-level surface normal vector distribution based on the photometric stereoscopic method; S4: performing integral reconstruction of the normal vector distribution to generate a 2.5D height map; S5: detecting mold surface anomalies based on the stable image and / or the 2.5D height map. This invention, through multi-directional light source collaborative illumination and image synthesis strategy, accurately solves the problem of uneven illumination in unilateral bright and dark areas, obtaining a bright and clear detection image of the entire surface, while also possessing high stability and real-time performance, meeting the needs of industrial mold monitoring.

This invention relates to the field of speech synthesis technology, specifically to a speech synthesis method based on an implicit continuous consistency model. The method involves constructing a dataset containing audio and its text annotations; building a residual vector quantization variational autoencoder, training it using a joint loss until convergence, and extracting the mean and variance of all audio data mapped to a latent vector distribution; sampling Gaussian noise latent variables based on the mean and variance, sampling Gaussian noise and time steps, and adding noise to obtain speaker features; calculating the continuous consistency loss using the time steps, speaker features, text, audio data, and the consistency model; and optimizing the continuous consistency loss until the consistency model converges. This method utilizes residual vector quantization technology to achieve high-rate audio feature compression and decoupling, and combines the single-step sampling characteristics of the consistency model in the latent space to improve the inference efficiency and training stability of speech synthesis, enabling the model to generate high-fidelity speech in a very small number of iterations.

The invention discloses an open winding motor control precision improvement method based on ABC three-dimensional space modulation, and relates to the technical field of motor control. Vector distribution in an ABC three-dimensional space has more obvious regularity and symmetry, and geometric relationship analysis is simpler and more visual. Firstly, candidate vectors are simplified according to vector distribution under an ABC coordinate system, so that the strategy complexity is reduced as much as possible under the condition that the torque plane and zero-sequence dimension control precision are not reduced. Then obtaining four optimal vectors meeting conditions through online traversal and rolling optimization processes, then obtaining a vector action sequence according to a distribution relation of action vectors and action time of each vector obtained in the traversal process, and obtaining voltage level signals of each phase in a switching period according to the voltage vectors and the action sequence; and finally, a unified modulation signal generation strategy is formulated.

The invention discloses a laser point cloud map precision evaluation system and method based on multi-modal data fusion, and the system comprises an MME evaluation module which constructs matrix elements through the coding and hierarchical processing principle, and is used for removing dynamic points of a point cloud map, retaining static points of the point cloud map, and calculating the proportion of the dynamic points; calculating a normal vector distribution entropy value of a local area on the basis of the static points and quantitatively evaluating the microscopic quality; introducing intensity reflection consistency detection identification and calculating an intensity anomaly proportion of the special material area; comprehensively calculating the comprehensive score of the MME; the APE optimization module is used for correcting an absolute pose error by fusing multi-sensor data and calculating an APE error and an APE comprehensive score; then, the weight balance module automatically adjusts the weight ratio of the MME to the APE under the condition of defining an environment adaptability weighting function, and calculates the precision evaluation indexes of the MME and the APE; therefore, a high-precision map evaluation result can still be provided in a complex environment.

The application relates to the technical field of distributed photovoltaic power prediction, in particular to a distributed photovoltaic power intelligent prediction method and system. N The two-dimensional plane coordinate points corresponding to the installation geographical positions of the distributed photovoltaics are recorded as the first position to the nth position. N The wind vector distribution function and the current irradiance distribution function are fitted; the actual active power of the first distributed photovoltaic to the nth distributed photovoltaic is obtained. N The spatial distribution of the actual active power is time-space deduced, and the predicted output active power of the first distributed photovoltaic to the nth distributed photovoltaic at the predicted time point is calculated. N The application comprehensively considers the wind speed, wind direction and other multi-dimensional data and the self-voltage regulation characteristics of the photovoltaic and other factors to realize the output power prediction of the distributed photovoltaic in the high-altitude and light-cloud area with rich sunshine.

The invention discloses a permanent magnet synchronous motor phase current reconstruction method and device for reducing harmonic waves, and belongs to the technical field of motor control. According to the method, a sampling topological structure is optimized, a sampling resistor is shifted from a direct-current bus side to a position between upper bridge arms of two phases A and B of an inverter, and low-harmonic and high-precision phase current reconstruction is realized by combining a novel switch state phase shift method (N switch state phase shift strategy) and a dynamic zero vector distribution strategy. The method specifically comprises the steps that sectors are divided according to a space vector diagram, and sampling vectors are selected; phase shift operation is carried out in a low modulation degree area through minimization of PWM waveform modification, and the average voltage conservation principle is followed; zero vector time is dynamically allocated in a high modulation degree region to extend a linear modulation range. The phase current total harmonic distortion (THD) is remarkably reduced, the modulation range is expanded, meanwhile, the cost effectiveness and the real-time performance are considered, and the method is suitable for low-cost permanent magnet synchronous motor driving systems such as electric automobile fans and water pumps.

The application discloses a kind of based on edge computing's tap water real-time purification control method, equipment and medium, it is related to automatic control technical field, including, obtain offset fluorescence wavelength and attenuation intensity characteristic, generate optical characteristic data stream;In fixed time window, the attenuation gradient and displacement vector of optical characteristic data stream are extracted, and attenuation gradient and displacement vector are convolved in space and construct three-dimensional pollutant atlas;According to the vector distribution mode diagram of pollutant distribution in three-dimensional pollutant atlas, output magnetic field control parameter;Magnetic field control parameter is input to magnetic field generator, and generates branched catalytic cluster;Through branched catalytic cluster, catalytic degradation is carried out to unpurified water source, and generates catalytic process curve;The coincidence degree of comparison catalytic process curve and preset target curve is completed tap water real-time purification control.The application realizes the adaptive real-time purification of heavy metal pollution by the synergy of three-dimensional pollutant atlas construction and catalytic process closed-loop control.

The invention relates to the technical field of industrial visual inspection, and particularly discloses a wood board surface defect detection method and system.The method comprises the steps that firstly, a multi-exposure image sequence of the surface of a wood board is collected from multiple polarization angles, an unsaturation intensity graph is reconstructed through high-dynamic-range fusion, and a Stokes parameter field representing the whole surface point polarization state is solved; performing physical inversion on the parameter field, resolving a micro-plane normal vector distribution field for describing the surface microscopic geometrical morphology, performing multi-scale decomposition and adaptive fusion on the intensity graph and the normal vector field, generating an enhanced feature graph for simultaneously coding appearance texture and bottom layer geometrical physical information, inputting the feature graph into a pre-trained identification process, and identifying the appearance texture and the bottom layer geometrical physical information. A normal vector field is introduced as an internal geometric consistency constraint for feature correction and guidance, so that accurate positioning, classification and confidence evaluation of defects are realized; according to the method, the accuracy, the robustness and the practicability in surface defect detection of the high-reflection wood board are improved.

This invention discloses a quality control method and system for the production of soft capsules, relating to the field of soft capsule production and testing technology. This invention achieves a closed-loop quality control through pre-processing state detection, graded rejection, and batch balance verification. The method includes: acquiring point cloud data of soft capsules on the conveyor belt using 3D vision technology; accurately identifying and rejecting soft capsules in an adhered / stacked state to a first recycling container based on the height and normal vector distribution characteristics of the point cloud; then performing appearance defect detection on single-capsule soft capsules and rejecting defective capsules to a second recycling container; finally, using a batch balance verification module, dynamically comparing the balance relationship between the upstream discharge benchmark and the rejection counts and good product entry counts at each level, achieving anomaly diagnosis and quality traceability in the production process. This invention significantly improves the accuracy of soft capsule detection and the reliability of the production process through decoupling of state and defect processing and closed-loop quantity verification.

The invention discloses an orchestrable highway tunnel lining surrounding rock multi-medium seepage analysis system and method.The method comprises the steps that firstly, contour lines are automatically generated based on section parameters, a section contour is obtained through sweeping and truncation, and quadrilateral grids are generated; quadrilateral grids of the tunnel lining surrounding rock section are automatically grouped, permeability coefficients and boundary water heads are mapped, and after a water head field is solved, seepage pressure scalar quantity and specific flow vector distribution are output; a sample library is constructed, machine learning regression is adopted to predict the Tunnel 3D key unit seepage pressure, the confidence coefficient is output, and in combination with boundary clamping, interface gradient constraint, adjacency graph smoothing and residual error driven section encryption recharge updating, three-dimensional coloring visualization and standardized output of a seepage analysis result are achieved. According to the method, the modeling automation and traceability are improved, the cross-mileage result consistency and visualization readability are improved, the three-dimensional global numerical solution cost is remarkably reduced, and the reliability and engineering applicability of seepage analysis are improved through physical constraints and self-adaptive closed loops.

The invention relates to the technical field of image pattern recognition, in particular to a diopter parameter analysis model optimization method and system, and the method comprises the following steps: collecting image sequence recognition region features, constructing edge segments, judging the structural continuity, executing concurrent recognition and trajectory analysis, and outputting a diopter parameter analysis optimization graph. According to the method, through constructing curvature section and gray vector distribution in an image sequence, extracting continuous features of a boundary structure, screening stable graph frames in combination with inter-frame direction consistency, recording corresponding boundaries by using a dual-channel recognition rhythm alignment mode, establishing a structure difference value and generating an error trajectory graph; the stable nodes are identified based on the offset trend, and the key graph segment is backtracked, so that the restoration coherence and response consistency of the image structure are enhanced, and the expression ability of the graph result to the change characteristics of the crystalline lens and the accuracy of parameter extraction are improved.

This invention relates to the field of visual AI technology, specifically to a high-precision assembly and quality traceability management system for KD (knock-down) components based on visual recognition and AI. In this invention, a dynamic point cloud acquisition and generation module generates point cloud flow data, and a normal field construction module establishes a surface normal vector distribution field. A dual-mode feature extraction module performs low-pass smoothing filtering to extract manifold surfaces to construct a flexible deformation compensation model, and performs high-pass residual filtering to extract random high-frequency normal disturbance data to generate microscopic quality fingerprint features. An adaptive assembly and traceability module generates an adaptive assembly trajectory based on the flexible deformation compensation model to control the assembly mechanism, achieving precise assembly that corrects physical deformation. The microscopic quality fingerprint features serve as a unique index to identify associated data and generate quality traceability files, thus realizing high-reliability assembly and label-free traceability of KD assembled parts.

The invention discloses a vector retrieval method and device for a balanced graph topological structure, and relates to the field of vector retrieve.The method comprises the steps that firstly, a graph is initialized into a minimum directed complete graph with any vector meeting the target in-out degree, a gradual vector adding mode is adopted, a forward optimization multi-level edge selection strategy and an edge refining mode are combined, and the minimum directed complete graph with the target in-out degree is obtained; and finally, constructing a neighbor graph of a balanced graph topological structure, and then performing greedy approximate nearest neighbor retrieval on the neighbor graph for the query vector to obtain a query result. According to the method, the balance of the topological structure of the neighbor graph is realized, the problem of deflection of the topological structure of the neighbor graph caused by unbalanced vector distribution is eliminated, and the accuracy and efficiency of vector retrieval are effectively improved. Compared with a current mainstream neighbor graph composition method, the composition method does not depend on reverse edge adding operation to update neighbors for existing vectors in the graph, and composition complexity and time expenditure are reduced.

This invention discloses a method and system for life assessment of composite bushing equipment based on multi-factor coupling, relating to the field of power equipment condition assessment. The method includes: collecting multi-source operating status data of the composite bushing equipment during operation; constructing a three-dimensional structural model of the composite bushing based on the three-dimensional structural design parameters of the equipment; performing stress field and environmental field coupling analysis on the model based on the multi-source operating status data to identify multiple structurally sensitive regions; extracting local operating status characteristic parameters for each sensitive region to construct regional degradation feature vectors; identifying the coupling influence relationships between vectors to construct a composite bushing degradation propagation topology network; obtaining the distribution of regional degradation trend prediction feature vectors based on the topology network; and performing life assessment according to the vector distribution to output the remaining service life. This application solves the technical problem of the difficulty in accurately predicting the life trend in existing life assessment methods, achieving the technical effect of reliably predicting the equipment life trend.

The invention relates to the technical field of image classification and computer vision crossing, and discloses an image classification-based full-surface bare concrete crack intelligent identification method. The method comprises the following steps: acquiring a three-dimensional point cloud and a surface normal vector distribution diagram through multi-view structured light projection and polarization imaging, and constructing a crack three-dimensional enhanced representation fusing geometric height and microstructure direction; and inputting the data to a multi-scale attention feature fusion network to generate a crack semantic segmentation mask, and extracting parameters of a center line track, a crack width, a depth change rate and a trend curvature to realize high-precision classification of four types of cracks. The system integrates structured light, polarization imaging and an embedded heterogeneous computing platform, and has ambient light suppression and adaptive exposure functions. According to the invention, the identification sensitivity and accuracy of tiny, closed and complex cracks are improved.

The utility model discloses a height difference torque vector distribution device which comprises a differential mechanism, a left half shaft is connected to one end of the differential mechanism, and a right half shaft is connected to the other end of the differential mechanism; a motor is arranged on one side of the left half shaft, a first-stage speed reduction assembly is arranged on the motor, a third-stage speed reduction assembly is arranged on the left half shaft, and the third-stage speed reduction assembly is in transmission connection with the first-stage speed reduction assembly through a second-stage speed reduction assembly; the three-stage speed reduction assembly transmits received power to the planetary gear assembly, and a first gear ring and a second gear ring on the planetary gear assembly transmit the power to a shell and a left half shaft of the differential mechanism respectively. The beneficial effects of the utility model are mainly reflected in that the arrangement has a torque distribution function, and gives consideration to the energy consumption problem and driving experience in yaw dynamics control; and the energy consumption problem and the driving experience in yaw dynamics control are considered.