62 results about "Joint coordinates" patented technology

The invention provides an AI-based nuclear power station high-risk operation compliance real-time monitoring method and system, and the method comprises the steps: processing human body joint point coordinate data of an operator through employing an AI dynamic contour modeling algorithm, and generating a limb movement track sequence; scanning the preset operation area to generate a relative motion vector between the special operation tool for the nuclear power station and the operator; detecting an action deviation mode of the operator based on the limb movement track sequence, and generating a detection result; fusing the relative motion vector and multi-channel video stream frame synchronization data obtained by applying an AI video analysis algorithm to generate a space intrusion relation between the nuclear power station special operation tool and the nuclear power station high-risk equipment; and when the action deviation mode exists in the detection result and the space intrusion relation exceeds a preset safety threshold, generating an operation compliance signal. According to the invention, real-time dynamic evaluation of human motion compliance and tool equipment safety distance in the high-risk operation process of the nuclear power station is realized, and the safety monitoring precision and real-time early warning capability of high-risk operation are improved.

The invention belongs to the technical field of computer vision and robot collaborative perception, and relates to a 3D human body posture estimation method and system based on space time sequence information fusion. The method comprises the steps that a multi-view image sequence is collected, and 2D joint coordinates of all view angles are extracted from the multi-view image sequence through a posture detector; position embedding, global embedding and edge embedding are carried out based on the 2D joint coordinates, and spatial features, including position information, global association information and skeleton edge information, of each joint point are obtained; fusing the features of each visual angle in the spatial features through a cross-visual-angle attention mechanism to obtain global feature representation; performing spatio-temporal feature enhancement and time sequence mixing on the global feature representation to obtain features integrating time sequence information, space information and channel information; and performing 3D posture regression on the features of the integrated time sequence information, the space information and the channel information to obtain a 3D human body posture. According to the invention, accurate and robust multi-view 3D human body posture estimation can be realized.

The invention discloses a three-dimensional attitude estimation method and device, a storage medium and electronic equipment, and relates to the technical field of three-dimensional attitude estimation, and the method comprises the steps: carrying out the joint estimation of a multi-view image of a target object, and obtaining a joint feature map and a joint heat map of each view; performing cross-view-angle alignment analysis on the multi-view-angle joint heat map to obtain an initial three-dimensional joint coordinate of each joint point; carrying out deformable cross-attention feature aggregation on the multi-view joint feature map to obtain multi-view fusion features of the joints; and performing three-dimensional attitude prediction based on the initial three-dimensional joint coordinates of the joints and the multi-view fusion features to obtain target three-dimensional joint coordinates of the joints. According to the invention, the estimation accuracy of the three-dimensional coordinates of the joint points during three-dimensional attitude estimation can be effectively improved.

The invention relates to the technical field of numerical control systems and five-axis calibration, in particular to a five-axis RTCP parameter calibration method based on an iterative algorithm, and the method comprises the steps: (1) calibration device deployment: fixing a standard ball on a workbench, installing a displacement sensor on a flange surface of a main shaft through a tool clamp, and correcting the deviation; (2) data acquisition: a mobile sensor is tangent to a standard ball, joint coordinates and displacement are recorded, and data cover a spherical surface; (3) mathematical model establishment: deriving a sensor tail end sphere center coordinate formula based on a five-axis kinematic model (double swing heads, double rotary tables and single rotary table single swing head); (4) constructing a target optimization function: according to the tangent geometrical relationship of the two balls, establishing an optimization function taking the RTCP parameter and the center coordinate of the standard ball as variables; and (5) iterative solution: carrying out iterative calculation by adopting a differential evolution algorithm, taking an optimization function value as fitness, and outputting an RTCP parameter after a precision requirement is met. The method is easy and convenient to operate, high in precision, high in universality and suitable for RTCP parameter calibration of the low-end five-axis machine tool.

The invention relates to the field of human body measurement automation, and discloses an automatic skeleton point recognition method and system for a digital human, and the method comprises the steps: firstly recognizing an RGB lens array corresponding to the digital human in a target scene, collecting an omnibearing image, and carrying out the three-dimensional reconstruction to generate image point cloud data; then noise points are detected, filtering parameters are analyzed, noise is reduced, and an optimized point cloud model is constructed; then multi-angle human body images are collected, and joint points are recognized to generate two-dimensional skeleton data; then determining a mapping relation of the two-dimensional skeleton in a three-dimensional space, converting the two-dimensional skeleton into a three-dimensional joint coordinate set, positioning an attitude registration point of the SMPL model and matching a skeleton topological structure; and finally, determining a size measurement value based on the topological structure, and generating an automatic identification result after calibration. According to the invention, the recognition accuracy, efficiency and automation degree of the digital human skeleton points can be improved.

The invention discloses an IMU (Inertial Measurement Unit) visual fusion motion analysis method which comprises the following steps: a multi-source data synchronous acquisition step: acquiring three-dimensional linear acceleration and angular velocity original data through an inertial measurement unit, and acquiring a visual image sequence at the same time; a dynamic zero offset compensation step: executing a preset action on the inertial original data to calculate a zero offset compensation value; a time-space synchronization triggering step: generating a synchronization signal when detecting that the calibration inertial data reaches a preset threshold value, and establishing a mapping relationship between the timestamp and the visual frame sequence; a motion attitude fusion step: taking an attitude derived from calibration inertial data as prediction input, taking a joint coordinate extracted from a visual image sequence as observation input, and updating a fusion state through Kalman filtering; and a quantitative index extraction step: calculating a joint motion range, a motion instantaneous rate and a coherence coefficient. The method solves the technical realization problem of multi-source data acquisition and fusion, improves the accuracy and real-time performance of motion analysis, and is suitable for the fields of motion teaching and rehabilitation training.

The invention discloses a multi-modal three-dimensional hand grid reconstruction method and system based on topology perception Transform, and belongs to the technical field of computer vision and intelligent interaction. According to the method, a topology perception hierarchical coding denoising network containing anatomy and topology constraints is introduced in the three-dimensional hand posture and grid reconstruction process; a real hand 3D joint coordinate is subjected to perspective projection, forward noise addition and feature fusion, and then is processed by a topology perception hierarchical coding denoising network composed of a hand part hierarchical coding module, a hierarchical feature aggregation mechanism and a topology perception Transform module to obtain a joint feature containing rich structure information; and finally, two-stage hand grid vertex construction from rough to fine is realized through a joint-guided grid reconstruction module. According to the method, the precision and robustness of three-dimensional hand posture and grid reconstruction are remarkably improved.

The present application belongs to the technical field of robot control, and particularly relates to a stability judgment method of a robot system variable impedance model using a variable augmented matrix, which enables the robot to verify its stability offline when the stiffness changes. Impedance control is a general force control scheme suitable for various manipulator tasks, one of which is human-robot collaboration: the manipulator performs force-position hybrid control on the object according to the impedance control law, while the operator applies control force to the object or the robot to realize human-robot collaboration and complete work such as assembly, carrying, etc. According to different manipulator tasks, variable stiffness parameters can be used to reflect flexibility and compliance. Not only can the dynamic relationship between external force and robot motion be controlled, but also these dynamic behaviors can be flexibly changed in a continuous manner during the task. Impedance control can be implemented in joint coordinates, Cartesian coordinates or other task-specific coordinates. Impedance targets can be achieved in different ways.

The present application relates to a kind of plane five-pole closed-chain rehabilitation robot elimination modeling and force mapping control method, belong to rehabilitation robot control technical field.The method is aimed at the coupling of active joint and passive joint in plane five-pole closed-chain mechanism, control implementation difficulty under active joint coordinates and the problem that driving torque is difficult to accurately map as end force, establishes closed-chain position constraint equation, obtains velocity constraint relationship by derivation to constraint equation, under the nonsingular condition of passive joint constraint jacobian matrix, passive joint variable is explicitly eliminated, constraint consistent jacobian matrix is constructed, reduced dynamics equation under active joint coordinates is established, end force estimate is solved, and motor driving torque instruction is generated according to the end force estimate.The method is suitable for end impedance control, trajectory tracking control and rehabilitation training boost control of plane five-pole closed-chain rehabilitation robot.

The invention discloses a three-dimensional human body posture comfort evaluation method and system based on an action sequence. The method comprises the following steps: acquiring video data; extracting any frame of image of the video; space coordinates of human joints, angle changes between adjacent joints and the overall gravity center position are recognized and calculated; performing RULA (fast upper limb assessment) analysis on the basis of the joint coordinates to generate an RULA score; calculating a joint angle score according to the joint angle; calculating a gravitational potential energy score through the center-of-gravity shift amplitude; and the overall comfort level of the action is judged by integrating the three scores. The system comprises a data interface and extraction module, a kinematics calculation module, a joint displacement risk calculation unit, a joint angle risk calculation unit, a gravitational potential energy change risk calculation unit and a comfort level evaluation calculation module. By using the method and the device, continuous comfort scores which change along with time can be generated. The method can be widely applied to the field of video analysis.

The invention relates to the technical field of non-standard automation equipment control, in particular to an industrial robot machining program migration method based on a configuration description file, and solves the problem that a non-standard industrial robot cannot migrate a machining program. The method comprises the steps that S1, a configuration description file of a prototype industrial robot is constructed, and based on object-oriented abstract modeling, kinematics, peripheral and communication modeling and a machining program are included; s2, the target robot is assembled according to the prototype physical structure, and it is ensured that the same configuration, component assembly, the axis movement direction and the origin coordinate are consistent; s3, loading the configuration file to the target robot; s4, correcting assembly errors through visual sensing, obtaining accurate coordinates among the parts, and then generating joint coordinates through inverse solution; and S5, the target robot executes the adaptive machining program, and program migration is achieved. According to the method, prototype machine information is packaged through the configuration description file, after the target machine is loaded, errors are corrected only through visual calibration, rapid adaptive machining can be achieved through kinematic chain inverse solution, and rapid copying of robots of the same configuration is achieved.

The invention provides a control method for a fitting gap of an exterior curved surface of a repair welding position of a storage tank, which comprises the following steps: S1, extracting spatial three-dimensional coordinates of n points uniformly distributed around a defect area from three-dimensional scanning point cloud data of the exterior curved surface of the storage tank, and defining the geometric center of the defect area as a defect center; s2, based on the defect center, mapping the space three-dimensional coordinates of the n points extracted in the step S1 to a robot tool coordinate system through a coordinate conversion relation to obtain converted coordinates of n target points; s3, calculating a three-dimensional space error value between the converted coordinates of the target point and the coordinates of the corresponding point on the supporting tool; s4, iteratively optimizing the error value in a robot joint kinematics constraint range, solving an optimal support tool space coordinate, and taking the coordinate as a robot motion target point; and S5, joint coordinates corresponding to the moving target point of the robot are transmitted to a robot controller through the real-time industrial bus communication technology, and the mechanical arm is driven to drive the supporting tool to move to the target point.

This invention discloses a method and system for martial arts movement instruction and digital dissemination based on digital twins. It acquires three-dimensional joint coordinates and motion parameter data from the practitioner's limbs using motion sensors. A preset coordinate transformation algorithm processes the data to form a standardized input sequence, resulting in standardized limb joint data. This constructs a digital twin model of the practitioner. Simultaneously, a standard template model of the corresponding martial arts movement is retrieved from a preset database. A three-dimensional mapping algorithm is used to overlay and compare the two models to determine the initial joint position matching degree. If the matching degree is lower than a preset threshold, a deviation detection mechanism is triggered. Joint angle deviation and movement rhythm error indicators are extracted from the comparison results, and quantified values ​​are calculated using machine learning algorithms to obtain the deviation analysis results. This invention improves the accuracy and safety of training, significantly enhances the overall instruction effect, and provides an innovative technical path for the digital presentation, immersive dissemination, and cultural inheritance of traditional martial arts.

The application discloses a three-dimensional human body posture estimation method based on dynamic modeling and bone topological consistency, multi-modal input construction and feature extraction, feature extraction is respectively performed on input RGB images, two-dimensional joint coordinates and skeleton structure diagrams, and each mode feature after extraction is uniformly mapped to the same dimension, thereby providing a basis for subsequent multi-modal fusion; by sequentially executing five core processes of multi-modal input construction and feature extraction, dynamic credibility modeling and adaptive fusion, bone topological guided repair, multi-scale detail enhancement and three-dimensional regression, and loss function joint optimization, the problem that a static strategy is adopted for multi-modal fusion of an existing three-dimensional human body posture estimation method and the mode contribution proportion cannot be dynamically adjusted according to a scene is effectively solved.

The application discloses a kind of based on the joint coordinate measuring machine calibration method of virtual calibration member construction.It is as follows: one, the coordinates of 12 sampling points are determined;Two, 12 sampling points are used to construct selected type virtual calibration member.Three, the spindle of precision machine tool drives calibration tool to move to the sampling point position on all virtual calibration members in turn.Each time calibration tool reaches a sampling point, the coordinates of calibration tool are measured using the spherical probe of joint coordinate measuring machine.Four, according to the measurement results, joint coordinate measuring machine is calibrated.The application constructs a large number of different types of virtual calibration members by setting 12 sampling points in precision machine tool, and the size and spatial pose of calibration member can change according to the change of sampling point;Compared with the existing calibration method using entity calibration member that can only change spatial pose, the calibration effect and calibration efficiency of the application are significantly enhanced.

The application discloses a monocular three-dimensional human pose estimation method fusing part perception and bone topology prior, and belongs to the field of computer vision. The application aims to solve the problem that existing methods ignore human motion heterogeneity and lack explicit anatomical constraints. The method comprises the following steps: obtaining 2D joint coordinates of a monocular video sequence and generating initial features; extracting features by using a parallel double-path time attention mechanism, wherein a global path captures overall motion, and a local path independently models features divided into five semantic component groups, and a cross-part graph convolution network is used to capture limb coordination, and two features are fused through an adaptive gating unit; and finally, features are transformed to a bone domain through a bone structure refining module, a graph convolution network is used to explicitly correct bone length and connection relationship based on a topology structure, and a refined three-dimensional pose is generated. The application enhances the physical rationality and accuracy of the predicted pose.

The invention relates to a human body foot point cloud completion method based on joint guidance, and belongs to the technical field of three-dimensional human body data processing. The problem that a reasonable anatomical structure conforming to biomechanical constraints is difficult to recover when foot data is missing in an existing point cloud completion method is solved. According to the technical scheme, foot key joint coordinates are predicted and features are extracted through a joint enhancement encoder, a sparse seed point cloud is generated by a generator, and three-level progressive refining reconstruction is carried out by a decoder. The method has the technical effects that the anatomical accuracy of a completion result and the robustness to noise are remarkably improved, detail recovery from coarse to fine is realized, and a high-quality data basis is provided for medical application.

The invention provides a human motion sequence prediction method and device, equipment, a medium and a program product. The method comprises the following steps: acquiring historical motion data and motion types of a target user in a first time period; preprocessing the historical motion data to obtain a joint coordinate sequence of the target user; extracting spatiotemporal features of the joint coordinate sequence, and performing coupling processing on the spatiotemporal features to obtain first motion features; according to a first constraint parameter, physiologically evaluating and correcting the first motion feature to obtain a second motion feature, the first constraint parameter being used for constraining a physiological activity interval of the joint of the target user; and according to the second motion feature and the motion type, generating a joint coordinate prediction sequence of the target user in a second time period in the future. The joint coordinate prediction sequence predicted by the embodiment of the invention is high in accuracy, conforms to the physical law of human motion and is high in adaptability.

The invention discloses a joint type coordinate measuring machine calibration method and system based on collaborative kinematics and measuring head deviation parameters. The joint type coordinate measuring machine calibration method comprises the following steps: acquiring space coordinates of an actual measuring head of a joint type coordinate measuring machine based on a mathematical model of the joint type coordinate measuring machine; synchronously solving kinematics-probe parameters through a unified optimization framework, and constructing an experience playback buffer area by using probe real-time deviation data; the experience playback buffer area comprises the state, the action and the reward of the current moment and the state of the next moment. In combination with the dynamic decision-making capability of deep reinforcement learning, training the constructed deep neural network through an experience playback buffer area; and the trained deep neural network is used to obtain the optimized joint structure parameters and offset based on the state of the current moment, and the parameters and offset are input into the joint type coordinate measuring machine to complete calibration, so that the calibration precision and efficiency are improved.

The invention provides a three-dimensional human body posture estimation method based on multi-scale geometric constraint of graph convolution hybrid MLP, and relates to the technical field of posture estimation. The method comprises the following steps: acquiring a two-dimensional human joint coordinate sequence; mapping the two-dimensional human joint coordinates to a high-dimensional feature space through an embedded layer to obtain initial high-dimensional features; inputting the initial high-dimensional features into a lightweight spatial attention module, and performing adaptive weighting on the joint features to obtain attention enhancement features; the attention enhancement features are input into a multi-layer stacked MLP hybrid module, spatial topological relation modeling and channel feature dependence modeling are achieved through alternate processing of spatial paths and channel paths, and fusion features are output; performing training optimization on the model based on a multi-scale geometric constraint loss function; and decoding the fused features into three-dimensional human body joint coordinates through an output layer. According to the method, the problems that space structure modeling is insufficient and skeleton constraint is insufficient when an existing method is used for processing complex postures are solved.

The invention relates to the technical field of visual analysis, in particular to a physical training posture real-time correction and feedback system based on machine vision, which comprises a data acquisition module, a posture deviation calculation module, an evaluation decision module, a multi-modal feedback generation module and a real-time correction guide module. Human body posture estimation is carried out to obtain joint coordinates of the athlete; mapping the joint coordinates to a historical personalized posture reference library of the athlete to obtain a dimension posture deviation quantification result; inputting the dimension attitude deviation quantification result into a preset dynamic evaluation rule engine to obtain a structured evaluation result; performing strategic analysis on the structured evaluation result to obtain a composite feedback instruction; based on the composite feedback instruction, conducting guiding regulation and control on the real-time posture so as to correct the physical training posture; according to the method, the personalized deviation judgment and real-time correction feedback efficiency of the physical training postures can be improved.

This invention relates to the field of intelligent service robot perception technology, and discloses an automatic generation system for tourist interpersonal interaction body language based on deep learning. The system acquires scene images through an image acquisition and skeleton extraction module and extracts the coordinate sequences of all two-dimensional skeletons of people, storing them in a database. A joint motion feature extraction module calculates the changes in joint coordinates frame by frame and inputs them into a trained deep learning model. An action category differentiation module classifies continuous actions into discrete actions such as waving, approaching, standing, leaving, and passing by. A depth information acquisition module simultaneously acquires distance information of people. A target service priority calculation module integrates action category, depth information, and action sequence to calculate and sort the service priority score for each person, generating a target service priority queue. This invention can automatically parse tourist interaction intentions and intelligently prioritize services.

The present invention relates to a system and a method for evaluating motion imitation. According to an embodiment of the present invention, a user's motion performance accuracy is evaluated in real time and the result is converted into a score and provided to the user, thereby promoting motivation for motion performance and improving physical activity. In addition, according to the present invention, a user's motion performance state can be more accurately and quickly analyzed on the basis of image analysis and joint coordinate extraction technology, thereby stably evaluating a motion regardless of physical conditions and motion environments of various users such as seniors and infants.

This invention provides a multi-view 3D pose reconstruction method, comprising: constructing a virtual simulation environment; setting the parameters of the virtual simulation environment using a domain randomization strategy; acquiring multi-view images and extracting 2D joint coordinates and confidence scores; normalizing the 2D joint coordinates based on the parameters of each virtual camera to obtain normalized coordinates; concatenating the normalized coordinates with the confidence scores and inputting them into a shared-weight attention encoder; learning the joint topology within a single view through self-attention and outputting view features; concatenating the rotation matrix and translation vector of each view and mapping them to an explicit extrinsic semantic vector via a multilayer perceptron; inputting the view features and extrinsic semantic vector together into a global attention module; performing geometric consistency evaluation through an attention mechanism; and outputting 3D joint coordinates. This method solves the problems of weak cross-view generalization ability, large interference from differences in camera intrinsic parameters, and insufficient robustness against occlusion in existing technologies for multi-view 3D pose estimation.

This invention relates to a method for elimination modeling and force mapping control of a planar five-bar closed-chain rehabilitation robot, belonging to the field of rehabilitation robot control technology. This method addresses the problems of coupling between active and passive joints in planar five-bar closed-chain mechanisms, difficulties in achieving control in active joint coordinates, and the difficulty in accurately mapping driving torque to end-effector force. It establishes closed-chain position constraint equations, obtains velocity constraint relationships by differentiating the constraint equations, explicitly eliminates passive joint variables under the condition that the passive joint constraint Jacobian matrix is ​​non-singular, constructs a constraint-consistent Jacobian matrix, establishes reduced dynamic equations in active joint coordinates, obtains the estimated end-effector force, and generates motor drive torque commands based on the estimated end-effector force. This method is applicable to end-effector impedance control, trajectory tracking control, and rehabilitation training assistance control of planar five-bar closed-chain rehabilitation robots.

The application provides a three-dimensional winding imaging denoising method and system based on a physical constraint optimization algorithm, and belongs to the field of ultrasonic imaging denoising, and comprises the following steps: collecting ultrasonic echo signals of a transformer winding and performing pretreatment to generate a joint coordinate matrix; reorganizing the joint coordinate matrix into a three-dimensional winding data parameter matrix, constructing a physical constraint term based on the amplitude-energy coupling relationship of ultrasonic waves in the winding medium, and constructing a fitness function for evaluating the quality of coordinate data in combination with the physical constraint term; initializing iteration parameters and solution space boundary parameters of the physical constraint optimization algorithm, taking the three-dimensional winding data parameter matrix as initial data, performing multiple rounds of iteration optimization until the convergence condition is met; and generating a denoised three-dimensional winding image by using the optimized coordinate parameters, through mutual checking of coordinate systems and three-dimensional point cloud reconstruction. The application effectively suppresses the interference of specific frequency band electromagnetic pulse noise, and improves the precision and clarity of three-dimensional winding imaging.

The invention provides a 3D human body posture recognition method and device based on a monocular camera, and the method comprises the steps: carrying out the preprocessing of an RGB image collected by the monocular camera, and carrying out the detection, and obtaining a two-dimensional key joint point coordinate of a human body; predicting a depth map corresponding to the RGB image through a monocular depth estimation model; performing coordinate mapping on coordinates of the two-dimensional key joint points and the depth map, and calculating three-dimensional space coordinates of each joint point by combining internal parameters of the camera; and performing time sequence smoothing and posture optimization processing on the three-dimensional space coordinates, and outputting a three-dimensional human body posture. According to the technical scheme provided by the invention, high-precision three-dimensional attitude reconstruction is realized on the basis of low-cost monocular hardware, the problems of plane ambiguity, scale ambiguity and shielding misjudgment are effectively solved, and the method is suitable for various civil scenes such as intelligent monitoring, somatosensory interaction and medical rehabilitation.

The invention discloses a shadow puppet digital human interaction method, system and device and a medium. The method comprises the steps of firstly obtaining instruction information input by a user; a corresponding shadow puppet action sequence is inquired from an action mapping table according to the instruction information, the action sequence comprises an initial action section and a subsequent action section, and each action section comprises a plurality of continuous action frames; a target reference joint of a first frame in an action sequence is used as a baseline, after the coordinate difference of the target reference joints of starting and ending frames of action segments is determined, translation alignment is performed on joint coordinates of the action segments according to the baseline and the coordinate difference, then a corrected action sequence is obtained according to the joint coordinates, the action sequence is rendered, and the target reference joint of the first frame in the action sequence is rendered. Therefore, a dynamic deduced picture is obtained. According to the embodiment of the invention, through the instruction driving action matching and the datum joint-based action section translation alignment design, the interaction between the shadow puppet digital human and the user and the coherent and stable presentation of action cross-section switching are realized, and the viewing experience is effectively improved.

The invention discloses a cerebral apoplexy lower limb intelligent rehabilitation training system, which relates to the technical field of rehabilitation instruments and comprises a walking module, a body fixing bracket, a foot fixer, a leg adjusting module, a waist and shoulder auxiliary fixing module and a training posture recognition and correction module. According to the training posture recognition and correction module, a binocular 3D camera is innovatively adopted to collect data, a Voxel CNN network and multi-constraint energy function optimization are combined, high-precision joint coordinate recognition, joint angle calculation and motion trajectory similarity evaluation are achieved, and a self-adaptive admittance control algorithm based on performance is matched, so that the precision of the robot is improved. The training load and the auxiliary force can be dynamically adjusted in real time according to the rehabilitation ability of a patient, customized rehabilitation training of one person and one strategy is achieved, the one-step training mode of a traditional instrument is thoroughly changed, rehabilitation training better meets the individual requirements of the patient, and the rehabilitation effect and the training efficiency are remarkably improved.