220 results about "Dynamic motion" patented technology

The invention discloses a multi-view child motion coordination ability evaluation system and method based on deep learning, belongs to the technical field of motion evaluation, and solves the problems that child motion evaluation in the prior art mainly depends on manual observation and simple physical testing, and multiple angles and key details of child actions are difficult to synchronously track. The method comprises the following steps: training to obtain a skeleton point detection model and an athletic ability evaluation model, acquiring real-time video streams of personnel entering a field based on an acquisition camera, identifying skeleton key points in a preprocessing data set by the skeleton point detection model, performing multi-person detection on a personnel matching result based on an athletic area division method, and evaluating the athletic ability of the personnel. The exercise ability evaluation model carries out quantitative analysis on multi-person detection results; according to the invention, visual identification and motion state detection technologies are combined, a front view angle and side view angle dual-camera layout is adopted, and a deep learning algorithm is matched, so that automatic children dynamic motion evaluation is realized. The action process can be completely captured, and the detection accuracy and efficiency are improved.

The invention provides a method for predicting the vibration reduction effect of exercise compression pants, and relates to the technical field of functional costume design, and the method comprises the steps: collecting a static costume pressure value of a lower limb target muscle group and soft tissue amplitude data in exercise, constructing a pressure-amplitude mathematical model through data processing, and verifying the model; constructing a human body-clothing finite element model, inputting fabric mechanical parameters into the model, and verifying the effectiveness of the model after iterative optimization; inputting fabric mechanical parameters into the verified finite element model, inputting a pressure prediction value output by the model into the pressure-amplitude mathematical model, and predicting soft tissue amplitude in a dynamic motion scene; according to the predicted soft tissue amplitude in the dynamic motion scene, the vibration reduction effect of the motion compression trousers in the target motion scene is evaluated. The method provided by the invention solves the problem that the vibration reduction effect evaluation of the traditional sports compression pants depends on a large number of repeated physical experiments and is difficult to realize accurate quantitative prediction.

The invention relates to the technical field of motion capture and data processing, in particular to a method for realizing real-time dynamic capture of a three-dimensional space of a human spine by applying an IMU (Inertial Measurement Unit). According to the method, a personalized three-dimensional dynamic spine model is constructed based on CT image data of a user, and a mathematical model of the spine is established by fusing a physiological anatomical structure of the spine and a robot mechanism motion theory so as to accurately describe motion characteristics of the spine. An IMU (inertial measurement unit) sensor is adhered to a key part of the back of a user, an individualized spine model is driven by acquiring and processing IMU data in real time, real-time reconstruction of dynamic movement of the spine is realized, and space angles among vertebral sections of the spine are accurately calculated. The dynamic change of the spine in the movement process can be accurately captured, the movement posture of the spine can be deeply analyzed, and the movement effect of the spine can be accurately and visually tracked and quantified in real time.

The invention relates to a vehicle trajectory tracking control method considering dynamic motion requirements. Comprising the following steps: 1, establishing a nonlinear vehicle dynamics model by comprehensively considering longitudinal motion, lateral motion and yawing motion of a vehicle; 2, searching saddle node bifurcation points with qualitative change of the stability state of the vehicle through a particle swarm optimization algorithm and a Lyapunov criterion, and constructing a dynamic stability domain boundary of the vehicle under the change of the longitudinal vehicle speed and the road adhesion coefficient; and 3, designing a vehicle multi-target trajectory tracking controller based on a nonlinear model predictive control theory, quantitatively evaluating the stability state of the vehicle by adopting a stability index, and meanwhile, designing a controller built-in parameter adjustment strategy to match the dynamic motion requirements of the distributed driving electric vehicle in different stability states. According to the method, the driving safety and the movement comfort of the distributed driving electric automobile are effectively ensured, and a set of solution is provided for realizing the optimal cooperative control of the chassis subsystem of the automatic driving automobile.

A wearable system for pedestrians / users is adapted to detect threats posed by, for example, approaching vehicles, bikers, and issue an alert. The wearable system may include an inertial measurement unit to measure the gait of the pedestrian and actively adjust alert generation thresholds to account for the noisy and large movements of the gait using present and past IMU data. The thresholds can be raised during active phases of the gait, and lowered in the quieter (i.e., less active) phases of the gait, keeping the range, and therefore time to react, relatively steady with respect to the pedestrian / user's motion. The alert generation threshold may be combined with the processed receiver data to determine threats and alert the pedestrian / user using any combination of haptic, audio, vibrational, or visual outputs.

The invention discloses a five-dimensional motion vector real-time construction and calibration method based on multi-sensor fusion, and relates to the technical field of multi-sensor information fusion and dynamic state estimation, and the method comprises the steps: collecting and preprocessing motion carrier data, obtaining a preprocessing data set and a feature data set, inputting the feature data set into a long short-term memory network model, and obtaining a multi-sensor fusion model; and outputting a sensor error offset prediction vector to the extended Kalman filtering model, outputting a preliminary five-dimensional motion vector, and performing consistency verification and correction on the preliminary five-dimensional motion vector through a kinematics constraint equation to obtain a corrected five-dimensional motion vector so as to drive a virtual model corresponding to a motion carrier to perform synchronous position and attitude updating. According to the method, the sensor error is predicted through the long-short-term memory network model, and the dynamic motion model constraint module is additionally arranged to perform physical constraint correction, so that the problems of inaccurate error compensation and lack of physical authenticity of the calculation result under the dynamic working condition are solved, and the construction precision and reliability of the five-dimensional motion vector are improved.

The invention provides a virtual fitting method and device, equipment and a storage medium. The method comprises the following steps: acquiring a video containing a figure and a clothing image; determining a character potential feature map according to the video potential features obtained from the video without the clothing part and the depth posture in the video; determining a clothing feature map and clothing semantic latent features according to the clothing image; performing first self-attention calculation according to the character latent feature map and the clothing feature map to obtain a first feature map; performing first cross attention calculation according to the clothing semantic latent features and the first feature map to obtain a second feature map; performing space-time attention processing according to the second feature map to obtain a third feature map; and generating a virtual fitting image according to the third feature map and the clothing semantic latent features. According to the method provided by the invention, learning is carried out for clothing properties through space-time attention processing, and an efficient global space-time attention mechanism is realized, so that a more real and coherent try-on image is generated, and the visual fidelity and dynamic motion consistency of clothing are improved.

The invention discloses a moving object grabbing method based on deep learning, and the method comprises the steps: collecting a dynamic scene image through an industrial camera, and detecting the 2D image features of a target object in a dynamic motion process through an improved YOLOv11 algorithm; a depth sensor is used for collecting 3D point cloud data of a target object to be grabbed in advance for improved YOLOv11 model training, preprocessing and feature extraction are carried out on the 3D point cloud data, a 2D detection result detected by an improved YOLOv11 algorithm is fused with the 3D point cloud data, then registration of the 3D point cloud data and a 2D image is carried out, and the three-dimensional attitude of the dynamic target object is estimated. On the basis of the fused and registered data and the motion track of the dynamic target object, the optimal grabbing point and posture are predicted, and an improved S-shaped curve track planning algorithm is adopted to generate a grabbing path of the mechanical arm; the mechanical arm completes the grabbing action according to the planned path, and meanwhile, the grabbing force and posture are adjusted in real time through force sense feedback; and the grabbed target object is placed at a designated position, and the next grabbing task is prepared.

The invention provides a mechanical arm movement track generation method and device, a storage medium and a program product. The method comprises the steps that observation data of the current environment and state data of the mechanical arm at the current moment are obtained; respectively carrying out feature coding on the observation data and the state data to obtain an observation feature vector and a state feature vector; the observation feature vector, the state feature vector and the state data are input into a trained mechanical arm motion trail generation model, and the motion trail of the mechanical arm is obtained; wherein the mechanical arm motion trail generation model comprises a denoising network and a probabilistic dynamic motion primitive ProDMP decoder, the denoising network is used for generating ProDMP weight parameters based on the observation feature vector and the state feature vector, and the ProDMP decoder is used for generating a motion trail according to the ProDMP weight parameters and the state data. The continuity, smoothness and boundary controllability of the motion trail of the mechanical arm are improved.

The invention provides a target object 6D pose estimation method and system oriented to electronic equipment AR assembly guidance. The method comprises the steps of electronic part coarse positioning based on improved YOLOv5s-CBAM, multi-mode (RGB-point cloud) feature fusion fine registration and LSTM dynamic motion prediction compensation. The system is developed based on an ROS framework and a PyTorch platform, and depends on an Open3D point cloud library and a TensorRT acceleration engine. According to the method, the problems of small target detection, metal reflection and shielding in an electronic assembly scene are solved through a cascaded visual positioning technology, the problem of insufficient precision caused by dependence on mark points or general SLAM in a traditional AR guide system is avoided, and the pose estimation robustness in a complex working condition is improved. Compared with an existing assembling method depending on manual calibration or two-dimensional drawings, the assembling efficiency and the first-pass yield are remarkably improved.

The invention belongs to the technical field of ocean information engineering, and discloses a joint simulation method for a structure collision effect in a floating installation process of an offshore converter station. The joint simulation method comprises the following steps: S1, setting sea condition parameters of an operation sea area and geometric parameters and material parameters of a floating system; s2, initializing a hydrodynamic model and a multi-body model; s3, solving the barge motion equation under the current time step by using the hydrodynamic model to obtain barge motion data; s4, calculating an LMU contact force and a DSU contact force under the current time step according to the barge motion data in the step S3, feeding back the LMU contact force and the DSU contact force to the hydrodynamic model, and updating the external load of the barge; and S5, repeating the steps S3-S4 until simulation in the set time step is completed, and then outputting a simulation result. According to the method, precise calculation of dynamic movement of the floating body and precise simulation of complex contact-nonlinear restoring force of the LMU can be considered, and high-precision simulation analysis of the whole floating installation process of the offshore converter station is achieved.

The invention relates to the technical field of dynamic force measurement, and particularly provides an inertia force compensation method and system for a dynamic force measurement platform, and the method comprises the steps: arranging a plurality of acceleration sensors on the measurement platform, so as to collect the acceleration information of the platform in the dynamic motion in real time; by combining the mass of the dynamic force measurement platform and the inertia moment data of the dynamic force measurement platform, the error of the platform caused by inertia is calculated and compensated, so that the precision of dynamic force measurement is improved. The method provided by the invention is particularly suitable for small-amplitude disturbance force and micro-vibration measurement scenes, and the interference of the inertia force of the platform on the measurement result can be remarkably reduced.

The invention discloses a cooperative control method for controlling a suspension load by multiple unmanned aerial vehicles, and the method comprises the steps: firstly constructing a frame based on a trajectory, combining an online kinematics and dynamics motion planner with an airborne trajectory tracking controller, and generating a feasible trajectory considering dynamics coupling and constraint through a finite time optimal control problem (OCP); then, an extended Kalman filter (EKF) estimator is used for fusing data of multiple unmanned aerial vehicle inertial measurement units (IMU) and a system dynamics model, and the load attitude, torsion and the cable state are estimated in real time; and finally compensating the tension interference of the cable in real time through an INDI (Incremental Nonlinear Dynamic Inversion) controller. According to the invention, the agility and robustness of the suspension load system are cooperatively improved through online motion planning and the trajectory tracking controller, the low-speed limitation of traditional cascade control is broken through, an additional sensor does not need to be deployed at the load end, and the practicability and reliability of the system under the scenes of high-speed obstacle avoidance, complex trajectory tracking and the like are remarkably improved.

The invention belongs to the technical field of spacecraft ground simulation testing, and discloses a high-precision space GNC full-physical simulation test platform which is composed of a three-dimensional motion execution system, a distributed real-time control system, a capture interaction simulation system, a safety protection system and a three-dimensional motion simulation and path planning system. By integrating a three-dimensional track type large-range motion execution system and a double-grating ruler closed-loop detection structure, the simulation motion range is remarkably expanded, the technical bottleneck that an existing platform is limited by a single axis, a two-dimensional sliding rail and limited motion amplitude is broken through, high-precision cooperation of a motion control system and a track system is achieved through the platform, and high-precision cooperation of the motion control system and the track system is achieved. The space high-dynamic motion simulation of a large-size and high-inertia complex target can be met, autonomous tracking of a target flight path, flexible capture of a flying claw and small multi-degree-of-freedom control verification are supported through integrated design of a capture interactive simulation system, and complex tasks such as space rendezvous and docking, target capture and micro-force control can be effectively simulated.

The invention discloses a human-machine cooperation method based on human intention estimation and impedance adjustment and a related device. The human-machine cooperation method comprises the steps that multi-mode sensing signals of a robot tail end and human operation are acquired; inputting the multi-modal sensing signals of the robot tail end and the human operation into a human intention estimator to obtain a real-time human intention estimation result; inputting the real-time human intention estimation result into an adaptive force coupling enhanced dynamic motion primitive module to generate an expected motion track and a stiffness parameter; and the expected movement track and the rigidity parameter are input into an auxiliary impedance controller, the rigidity and the damping coefficient of the robot are adjusted through the auxiliary impedance controller, and the method and the related device can effectively improve the cooperation ability, the environment adaptability and the interaction safety of a robot system.

The invention discloses a mechanical arm teaching control method, device and equipment and a medium, and belongs to the technical field of mechanical control. The method comprises the steps that calibration operation is conducted on the basis of a teaching tool file, and a first conversion relation from a mechanical arm base coordinate system to an optical positioning coordinate system and a second conversion relation from a mechanical arm tail end coordinate system to a teaching tool coordinate system are obtained; based on the first conversion relation and the second conversion relation, obtaining a teaching track corresponding to the preset number of teaching operations; each teaching track is converted into a dynamic motion primitive model parameter, and a target teaching track is determined based on each dynamic motion primitive model parameter; and a target mechanical arm joint angle sequence is determined based on the target teaching track so as to control the mechanical arm. According to the technical scheme, complex parameter input or manual physical intervention is not needed in the teaching process, intuition, high precision and high efficiency are considered, and therefore precise track control and stable operation execution of the mechanical arm are achieved.

A highly dynamic motion simulator comprised of an occupant platform 22 which is supported and articulated between two vertically oriented multi-axis translation mechanisms 21 and 23. The five degrees of freedom embodiment produces motion with five degrees of freedom (FIG. 1). The six degrees of freedom embodiment adds a longitudinal translation assembly which produces motion with six degrees of freedom (FIG. 2). This design minimizes inertial mass providing high dynamic performance.

Described is a method for instilling the haptic dimension of texture to virtual and holographic objects using mid-air ultrasonic technology. A set of features is extracted from imported images using their associated displacement maps. Textural qualities such as the micro and macro roughness are then computed and fed to a haptic mapping function together with information about the dynamic motion of the user's hands during holographic touch. Mid-air haptic textures are then synthesized and projected onto the user's bare hands. Further, mid-air haptic technology enables tactile exploration of virtual objects in digital environments. When a user's prior and current expectations and rendered tactile texture differ, user immersion can break. A study aims at mitigating this by integrating user expectations into the rendering algorithm of mid-air haptic textures and establishes a relationship between visual and mid-air haptic roughness.

The invention relates to a human body multi-cycle gait track prediction method based on small sample space-time decoupling. The method comprises the following steps: firstly, obtaining a single-cycle gait sample through kinematics modeling and cycle segmentation, and extracting, expanding and normalizing parameters such as stride and stride frequency; secondly, establishing a gait parameter joint distribution model by using a Gaussian mixture model and Gaussian mixture regression, and predicting gait parameters according to the target speed; and then space and time decoupling reconstruction is carried out, a high-fidelity joint trajectory is generated by adopting an adaptive collaborative attention LSTM network in space, a zoom factor is calculated based on a predicted stride frequency and a target speed in time, a time sequence is adaptively adjusted through a dynamic motion primitive, and finally a multi-period and multi-speed high-fidelity gait trajectory is generated through fusion. The method has high precision, interpretability and strong generalization ability under the small sample condition, the generated gaits are natural and coordinated and accord with the biomechanical law, and the application cost of a humanoid movement device is remarkably reduced.

Provided are a biplanar image-based preoperative planning method and system for total knee arthroplasty. Based on a static biplanar lower-limb image of a knee joint of a target object in a standing position and a dynamic biplanar image of the knee joint during functional motion, as well as an imported three-dimensional model, functional parameters, morphological parameters, and a three-dimensional medial and lateral joint gap of the knee joint are calculated, and then iterative calculation is continuously carried out using a two-dimensional model and a three-dimensional dynamic model to obtain a dynamically optimized placement position, so that preoperative TKA prosthesis model selection for balanced joint space across full flexion angles can be achieved; meanwhile, TKA prosthesis position planning for correcting the mechanical axis of the knee joint under three-dimensional weight-bearing conditions is considered, thereby accurately restoring the real dynamic motion of a patient, and improving the prosthesis-patient matching degree.

The invention discloses a DSP-based robot dynamic environment moving object pose optimal estimation method and a related device, and the method comprises the steps: synchronously collecting real-time binocular video information of an environment according to a left camera and a right camera of a robot; acquiring feature points in the scene and SIFT descriptors thereof, and performing stereo matching on the SIFT features to obtain position information of the feature points in a three-dimensional space; dividing the scene into a static background and a dynamic moving object by adopting K-means clustering, and identifying a center position sequence of the moving object on the ground plane; predicting the state vector of the moving object through an extended Kalman filter to obtain a preliminary estimated value; and performing optimal estimation according to the initial estimation value to obtain optimal pose estimation of the moving object in the robot coordinate system. Optimal estimation is carried out through the visual sensor, the robot can more efficiently and accurately estimate the pose of the moving object in the working environment, meanwhile, the size of the system is reduced, and the portability of the system is improved.

The invention discloses a reduced-order modeling and integrated rapid forecasting method for dynamic motion and structural response of a floating fan. The method comprises the following steps: performing finite element modal analysis on a floating fan structure, extracting a rigid body modal and a main elastic modal, constructing a low-dimensional modal space, orthogonalizing mass and stiffness matrixes, and reducing the degree of freedom of a system; hydrostatic force recovery, mooring, gravity rigidity and external exciting force are calculated according to the initial position of the structure, a hydrostatic balance control equation is established through modal space transformation, and generalized displacement is solved through iteration; recalculating the water pressure at the Gaussian point of the structural grid based on the radiation diffraction theory, calculating the pneumatic load in combination with the blade element momentum theory, and obtaining a generalized external load through modal transformation after superposition; a motion equation is established in a modal space, generalized displacement is solved through numerical integration, a global displacement field is reconstructed, a stress time history is directly calculated, and integrated rapid forecasting is achieved. According to the method, the calculation efficiency is remarkably improved while the key dynamic characteristics and the response precision of the structure are kept, and reliable support is provided for near-real-time evaluation of operation and maintenance of the floating type draught fan.

The invention discloses a point cloud gait recognition method based on SE (3) motion modeling and a computer program product, and the method comprises the steps: obtaining the motion gait features of point cloud data according to the space-time neighborhood and the spatial neighborhood of the point cloud data, and the motion gait features are the SE (3) features of each point in the point cloud data; obtaining a preset interface set according to the point cloud data and the motion gait features, obtaining a plurality of optimal interfaces from the preset interface set, adjusting a space-time neighborhood and a spatial neighborhood according to the plurality of optimal interfaces, and obtaining adaptive hierarchical motion gait features; and projecting the adaptive hierarchical motion gait features to a camera plane to obtain preset feature representation, learning the preset feature representation by using a multi-scale feature fusion network, and outputting a gait recognition result. According to the method, the problems of shielding, cross-view and dynamic motion characterization in a complex scene can be solved, and the biomechanical consistency and scene adaptability of gait features can be remarkably improved.

This invention relates to the field of humanoid robot technology and discloses a humanoid robot torso waist rotation linkage adjustment device, including a torso lower mounting plate, a central rotation axis, an intermediate connecting bracket, a left hip connecting seat, and a right hip connecting seat. The torso lower mounting plate is horizontally disposed at the bottom of the robot torso. The central rotation axis is arranged vertically and its upper end is coaxially fixedly connected to the lower surface of the torso lower mounting plate. The intermediate connecting bracket is fitted into the middle of the central rotation axis and rotatably connected to the central rotation axis. The left hip connecting seat and the right hip connecting seat are symmetrically arranged on both sides of the lower end of the intermediate connecting bracket. This invention achieves synchronous adjustment of torso waist rotation and lower limb posture through a mechanical linkage structure, which can adjust the lower limb posture in time during waist rotation, effectively counteract the center of gravity shift caused by torso rotation, and improve the robot's balance stability in dynamic motion scenarios.

This invention discloses a personalized and adaptable spot positioning and filter integration design method. Addressing the technical problems of uneven film thickness and poor film formation consistency in existing large-area coating processes, it constructs a dynamic-static synergistic coating control system by combining the dynamic motion mode of the coating target with the static positioning optimization of the substrate and the precise control of static parameters of the coating environment. Simultaneously, it introduces zoned coating compensation, real-time film thickness monitoring feedback, and dynamic trajectory adaptation strategies to achieve uniform deposition of coating materials on large-area substrate surfaces, effectively improving the film thickness consistency, film formation uniformity, and film performance stability of large-area coatings. This invention is convenient to operate, provides precise control, and is compatible with various coating processes such as magnetron sputtering, vacuum evaporation, and ion plating. It can be widely applied to large-area coating production in fields such as flat panel displays, photovoltaic glass, semiconductor wafers, and large-area optical devices.

The application discloses a power maintenance method and device based on a hybrid dynamic motion primitive. The method comprises the following steps: obtaining a plurality of demonstration trajectories of an equipotential robot; modeling a dynamic motion primitive for each demonstration trajectory to obtain a corresponding initial hybrid dynamic primitive model; evaluating the fluency of the demonstration trajectory according to an evaluation index to determine the model weight of each initial hybrid dynamic primitive model; determining a hybrid dynamic primitive model according to the initial hybrid dynamic primitive model and the model weight; and writing the hybrid dynamic primitive model into the equipotential robot, so that the equipotential robot generates a maintenance trajectory when receiving a power maintenance task; and controlling the equipotential robot to perform a power maintenance operation based on the maintenance trajectory. The application solves the technical problem that the equipotential robot can only learn a single trajectory in the related art, and that the generated trajectory has too many inflection points in the process of power maintenance, which causes the swing of a boom truck, low maintenance efficiency and poor work quality.

The invention discloses a task-driven active and passive fusion mechanical arm motion evaluation method and system, and the method comprises the steps: guiding a mechanical arm task, and collecting a joint angle data sequence; performing time alignment and smoothing processing to obtain an ideal demonstration track; inputting into a dynamic motion primitive model for learning to obtain model parameters; generalizing a new joint angle sequence conforming to the trajectory shape; a real-time control instruction is generated by taking the real-time control instruction as a reference track and adopting a reinforcement learning controller, and the mechanical arm is driven to track the reference track; and a multi-dimensional evaluation index system is constructed, and quantitative evaluation is conducted on the motion performance, the energy consumption efficiency and the intelligent level of task execution of the mechanical arm. According to the method, the robustness and the self-adaptive capability of the mechanical arm in different task scenes are remarkably improved, so that the efficiency and the safety of automatic operation are improved while stable operation is ensured.

A tube-in-tube structure (100) for a linear motor machine is disclosed. The tube-in-tube structure includes (a) an outer dimensional frame (20) for responding to maintaining a payload (50) and allowing the payload (50) to move thereon; (b) an inner force frame (10) enclosed within a hollow cavity (22) of the outer dimensional frame (20) and responding to transmitting any reaction force applied by the payload (50) to a bottom of the inner force frame (10), wherein a bottom end of the inner force frame (10) is mounted to a bottom end of the outer dimensional frame (20), and such mounting is positioned to the ground. The inner force frame (10) will directly transmit any reaction force applied by dynamic motion generated by the linear motor machine to the bottom of the inner force frame (10), thereby minimizing or stabilizing vibration caused by linear motor motion.

The invention discloses a dynamic motion primitive coding method and system based on Lie group Lie algebra, and the method comprises the steps: uniformly representing the tail end pose and contact force spinor of a robot in a Lie group form, and constructing a unified mathematical framework; the method comprises the following steps: acquiring a pose and a force spinor sequence through a teaching phase, and converting the force spinor sequence into a virtual pose sequence through scaling mapping; establishing a dynamic motion primitive (DMPs) model based on the Lie group Lie algebra theory, and learning a model weight matrix and a primary function parameter by adopting local weighted regression; and according to the initial and target poses of the new task, generating a smooth trajectory through phase synchronization control and Euler integration, and outputting the smooth trajectory to a robot controller. According to the method, the problems of posture singularity and motion distortion caused by splitting of a posture channel in a traditional DMPs method are solved, the learning fidelity and reproduction stability of the six-degree-of-freedom posture track of the robot are improved, the smoothness and coordination of the whole motion process are guaranteed, and the debugging complexity of a force control system is reduced.

The invention provides a video generation method and apparatus, and an electronic device. The method comprises the steps of obtaining visual data including a dynamically moving target object; for each frame of image contained in the visual data, converting each pixel point in the image into a three-dimensional point to obtain a three-dimensional dot matrix; wherein the three-dimensional dot matrix comprises a plurality of three-dimensional points, and three-dimensional position and motion information corresponding to each three-dimensional point; determining control features related to video generation based on the three-dimensional dot matrix; wherein the control features comprise motion features and spatial features corresponding to the three-dimensional points; generating a target video containing the target object through the control feature guidance; wherein the motion of the target object in the target video accords with the motion feature indicated by the control feature. In the mode, the dynamic three-dimensional dot matrix is introduced as a control core of video generation, so that the three-dimensional perception capability and the control precision of a video generation technology are fundamentally improved, and the quality of video generation is improved.