51 results about "Visual occlusion" patented technology

The invention relates to the technical field of multi-modal posture recognition, and relates to a multi-modal posture recognition system based on Transform fusion, which fully excavates the complementary advantages of visual information in spatial detail representation and inertial information in time sequence dynamic capture through space-time alignment of multi-modal data and deep fusion based on an attention mechanism. The accuracy and the stability of attitude estimation under the conditions of visual occlusion, rapid movement and complex illumination are obviously improved; posture optimization is carried out by introducing physical constraints such as bone length constancy and joint movement range limitation, and time domain smoothing and contact state correction are applied, so that the precision of the generated three-dimensional posture sequence is ensured, a solid technical foundation is laid for improving the naturalness, safety and intelligent level of man-machine interaction, and the method is suitable for popularization and application. And meanwhile, reliable application and deep development of the intelligent robot in key fields of service robots, virtual reality, remote cooperation and the like are powerfully promoted.

The invention discloses an intelligent security and protection monitoring method, system and device and a storage medium, and the method comprises the steps: carrying out the cross-modal fusion analysis of a roughly extracted first visual feature and first radar feature of a target, and generating a dynamic fuzzy intensity coefficient based on an abnormal behavior result after the fusion analysis, so as to dynamically adjust the fuzzy intensity of a target face region; according to the change of the target dynamic fuzzy intensity coefficient, judging whether a high-risk early warning signal is output or not; if the high-risk early warning signal is output, cross-modal deep fusion analysis is carried out on the extracted fine second visual feature and second radar feature of the target, and a target abnormal behavior judgment result is output based on the features after deep fusion; and tracking the target based on the judgment result, associating a visual detection target with a radar tracking target when the target is visually shielded, predicting the future motion trajectory of the shielded target, and recovering visual tracking when the shielding is removed. Accurate detection of abnormal behaviors is realized.

The invention discloses a visual sparse observation and physical model fused flexible body deformation tracking method, which comprises the following steps of: firstly, constructing a flexible body co-rotation finite element model considering geometric nonlinearity, and establishing a low-dimensional flexibility coupling model for describing a mechanical transfer relationship between a virtual actuating source and a characteristic observation point by utilizing a static condensation technology; in the real-time tracking process, a binocular vision system is used for collecting images and resolving three-dimensional coordinates of sparse feature points in a real physical world; constructing and solving a reverse optimization problem based on the low-dimensional flexibility coupling model, and calculating a virtual correction force vector capable of minimizing an observation error; and applying the virtual correction force vector to the finite element model to drive the finite element model to deform, and updating a full-field geometric node position and a tangent stiffness matrix in real time. Through deep fusion of a physical model and sparse visual observation, the problems of visual occlusion and virtual-real model deviation in flexible body assembly are effectively solved, and online real-time full-field high-precision deformation state estimation is realized.

The invention discloses a machine tool accessory automatic adjusting control system based on visual recognition, and relates to the technical field of automatic adjusting systems.The machine tool accessory automatic adjusting control system comprises a machine tool body, a multispectral visual module and three-axis driving mechanisms, and the three-axis driving mechanisms are fixedly installed on the periphery of the outer wall of the machine tool body; and a multispectral vision module is fixedly mounted on the outer wall of the three-axis driving mechanism. According to the invention, through the multispectral vision module, fitting multidimensional information fusion is realized, the problem of incomplete information acquisition caused by vision shielding is solved, the accuracy and reliability of machine tool fitting recognition are improved, and richer and more accurate data support is provided for subsequent adjustment and control.

The invention relates to the technical field of image processing, in particular to a numerical control milling machine workpiece three-dimensional modeling method and system. The method comprises the following steps: clamping a workpiece on a numerical control milling machine, controlling the workpiece to rotate at a constant speed, synchronously performing axial feeding, and collecting a correction spiral image sequence and a groove bottom image sequence; performing space-time registration on the corrected spiral image sequence and the groove bottom image sequence to extract workpiece modeling feature data; performing conversion from polar coordinates to rectangular coordinates based on the workpiece modeling feature data, resolving three-dimensional space coordinates and corresponding surface attribute values, and generating workpiece three-dimensional point cloud data; and inputting the three-dimensional point cloud data of the workpiece into three-dimensional modeling software, and carrying out gridding processing and texture rendering to obtain a three-dimensional digital model of the workpiece. According to the method, the workpiece is driven to execute self-adaptive tracking spiral motion, so that visual shielding caused by dense threads is avoided, and high-precision three-dimensional modeling of the small-helix-angle deep-groove workpiece is achieved.

A dynamic projection visual layout method based on nonlinear programming and multi-objective optimization optimizes a visual presentation effect of a projection result from two aspects of reducing visual occlusion of the projection result and enhancing sample stability in a dynamic iteration process. In this way, a user can be effectively helped to track a dynamically changing data mode, and continuous data exploration and analysis decision are supported. Besides, the method also adopts a mathematical optimization method of nonlinear programming and multi-objective optimization, aims of reducing visual overlapping or enhancing a data visual mode and the like are fulfilled, and the visual definition and the user interaction experience are improved.

The invention discloses a self-adaptive federated Kalman filtering unmanned aerial vehicle cluster laser communication method based on exponential sliding maximum likelihood estimation. The method specifically comprises the steps that firstly, a visual sub-filter and a laser sub-filter are constructed through an ESMLE algorithm, then information of the two sub-filters is fused on the basis of a federated Kalman filter frame, optimal relative position estimation of a friend unmanned aerial vehicle is output, and finally the expected angle of the photoelectric pod is calculated according to the estimation. And a servo mechanism is driven to realize accurate tracking of a target and continuous maintenance of a laser communication link. The method can effectively deal with the problem of visual occlusion, overcomes the defects that a single visual sensor in an existing unmanned aerial vehicle photoelectric tracking system is easily interfered by the environment and lacks a fault isolation mechanism, and improves the stability and robustness of an unmanned aerial vehicle laser communication link under the condition that the sensor fails or the environmental noise is suddenly changed.

The present application relates to the technical field of personnel positioning tracking, and provides a personnel trajectory positioning tracking method and system based on RFID and AI video recognition, which comprises the following steps: acquiring video data in an industrial environment; judging whether an event of a personnel target being blocked by an automated device occurs based on the video data; if the event of the personnel target being blocked by the automated device occurs, acquiring predetermined motion information of the automated device; calculating a predicted reappearance space-time window of the personnel target based on the predetermined motion information and a state of the personnel target before being blocked; processing the video data within the predicted reappearance space-time window, searching for and re-recognizing the personnel target, so as to realize personnel trajectory positioning tracking. The present application has the advantages that it can effectively deal with visual blocking of the personnel target by the automated device, and realize rapid and accurate re-recognition and trajectory tracking after the blocking, thereby improving the continuity and accuracy of personnel positioning tracking.

The invention discloses an industrial robot and environment contact event identification method based on cross-modal learning, and the method comprises the steps: collecting a visual image and a force sense time sequence of the tail end of a robot, and extracting visual features and force sense features through convolution operation and continuous wavelet transformation; a force sense feature driving offset generation network is utilized to calculate a time offset scalar, time dimension deformable convolution is executed on visual features, and time-space accurate alignment of visual and tactile features is achieved; according to kinematics parameters of the robot, kinetic theoretical stress is calculated, a physical residual vector is obtained through difference between the kinetic theoretical stress and actually-measured force, and inertia force interference generated by motion of the robot is eliminated; mapping the physical residual error, fusing the mapped physical residual error with the aligned visual tactile features, and inputting the fused physical residual error into a neural network to identify a contact type; according to the method, a physical consistency constraint and anti-occlusion reconstruction mechanism is introduced in a model training stage, so that the problems that multi-modal data are asynchronous, and the model lacks physical interpretability and visual occlusion are effectively solved, and the accuracy and robustness of event recognition are improved.

The application discloses a kind of dynamic detection methods of vehicle, the present application relates to the technical field of dynamic detection, based on the compensation detection of this visual occlusion area to determine occlusion dynamic target;Based on the occlusion dynamic target and dynamic driving stereogram determines enhanced panoramic view;Mark the multiple key geometric features of unmanned vehicle, according to multiple key geometric features, the current driving data of unmanned vehicle and enhanced panoramic view constructs dynamic detection system, guarantees the dynamic detection of unmanned vehicle in driving process.According to the area content of each dynamic detection area, corresponding dynamic early warning level and the driving state of unmanned vehicle determine the early warning control logic of unmanned vehicle;Based on the relative swing angle and the current driving data of unmanned vehicle determine corresponding variable characteristics, according to the variable characteristics and the multistage reasoning of early warning control logic of unmanned vehicle determine corresponding multi-modal emergency measures, improve the accuracy of multi-modal emergency measures.

The embodiment of the invention provides a sound effect playing method and device, a voiceprint identifier display method and device and terminal equipment, and relates to the technical field of application program development. The sound effect playing method comprises the following steps: displaying a virtual environment containing a first virtual object; under the condition that the first virtual object is located in a visual occlusion area in the virtual environment and triggers at least one sounding event in the visual occlusion area, the sound effect obtained after weakening processing of the at least one sounding event is played, the visual occlusion area is used for conducting visual occlusion on virtual elements in the area, and the first virtual object triggers the at least one sounding event in the visual occlusion area. The weakening processing is used for weakening the effect of the sound effect on distinguishing the position of the first virtual object. According to the technical scheme provided by the embodiment of the invention, the richness of sound effect playing in the game application program can be improved.

The application belongs to the technical field of passenger flow monitoring and management, and particularly relates to a subway station area passenger flow detection analysis method based on multi-modal sensing capability, which comprises the following steps: calculating a visual effectiveness coefficient according to the effective vertical distance of a pixel point in a video frame to a near-end reference line, and then dividing the video frame into a near-field anchoring area and a far-field area to be reconstructed; calculating a dynamic radio frequency penetration factor according to the number of passenger targets in the near-field anchoring area and the number of near-field radio frequency terminals; and obtaining a reconstructed passenger flow total amount by using the visual effectiveness coefficient of the far-field area to be reconstructed to perform weighted fusion according to the number of passenger targets in the near-field anchoring area, the dynamic radio frequency penetration factor, the number of far-field radio frequency terminals and the number of passenger targets in the far-field area to be reconstructed. The application can adaptively change the crowd structure, effectively solve the problem of missed detection caused by far-end visual occlusion and improve the passenger flow detection precision.

The invention relates to an interactive object recognition and posture estimation method and device based on humanoid body perception and a medium. The method comprises the steps that S1, body perception data of an actuator installed at the tail end of a robot are read in real time; s2, preprocessing the collected body sensing data based on a physical model to recover spatial features of the body sensing data, and encoding a preprocessing result by using a series-parallel connection body sensing data encoding module; s3, analyzing the body sensing data coded by the series-parallel connection body sensing data coding module by using an object recognition and attitude estimation model constructed based on a diffusion model to obtain semi-structured hidden space representation representing an interactive object state; and S4, identifying the interactive object from the semi-structured hidden space representation of the interactive object state and decoding attitude information. Compared with the prior art, the method provided by the invention improves the accurate recognition and attitude estimation of interactive objects in visual occlusion and narrow space.

The application provides a human posture estimation device based on global pressure and a method thereof. The device comprises: a spatial feature encoder configured to extract human global pressure spatial features from a global pressure frame sequence of human motion; a long short-term time sequence attention module configured to extract human global pressure time sequence features from the human global pressure spatial features and fuse the human global pressure time sequence features with the human global pressure spatial features to obtain pressure space-time features; and a motion regressor configured to perform nonlinear regression calculation on the pressure space-time features to obtain posture and displacement parameters of a human parameterized model. The application replaces visual information with pressure information which is independent of lighting conditions, does not infringe privacy and is not affected by visual occlusion, thereby achieving non-invasive expansion of human posture estimation in special scenarios.

The present invention provides an automated injection mechanism integrating tactile perception, comprising a tactile sensor and an automated injection mechanism. The tactile sensor is mounted on the front end of the mechanism and can read information on the surface of human skin. The main body of the injection mechanism controls the linear motion of the needle through a screw motor and a slider, controls the rotational motion of the needle through a reduction DC motor, and detects the mechanical information applied to the needle through a torque sensor. The needle passes through the frame and gel portion of the tactile sensor and pierces out at the center. When it approaches the skin surface, it is driven by an air cylinder to generate acceleration, quickly penetrates the skin, and then the electric push rod is activated to push the syringe to complete the injection operation of the drug. The automated injection mechanism integrating tactile perception described in the present invention can be combined with robotic arm technology to perceive the deformation and movement of internal blood vessels during the injection process in real time through external deformation, thereby optimizing the injection strategy in real time, avoiding visual occlusion, and improving the success rate of injection. The structure is simple, the cost is low, and it is easy to manufacture and promote.

The invention discloses a visual marking method and equipment based on fusion positioning, and a medium, belongs to the technical field of airport scene monitoring and management, and aims to solve the problems that the recognition accuracy is reduced, the high computing power requirement is needed, the real-time monitoring adaptability is poor, and the recognition efficiency is high due to the fact that existing airport scene monitoring is easily shielded by vision. And accurate geographical location information is difficult to obtain. The method comprises the following steps: performing coverage calculation of a maximum monitoring area on a visible area on an airport map to obtain a polygonal visible area; acquiring positioning terminal signals of key equipment and workers in an airport range; carrying out conversion processing of related camera point locations on a real-time geographic coordinate system in the real-time position information to obtain a camera coordinate system; performing conversion processing on the camera coordinate system to obtain a two-dimensional image coordinate system; mapping the two-dimensional image coordinate system into a polygonal visible area; and performing shielding labeling display on part of shielding targets in the initial labeling display information to obtain fused labeling display information.

A system and method for synchronizing and spatializing media between client device and server device. Client-side artificial intelligence application determines the position of server device by fusing data from at least two sources: visual data obtained from a camera and acoustic data obtained from a predetermined reference signal emitted by server device. This data fusion enables continuous spatial tracking, including during periods of visual occlusion. A timestamp-based protocol is used to synchronize media playback. Based on the tracked position, client device applies directional audio filters, such as Head-Related Impulse Response (HRIR), to render audio that appears to originate from server device's physical location. The method further compensates for acoustic propagation and signal processing delays to ensure accurate spatiotemporal alignment.

The invention provides a global pressure-based human body posture estimation device and a global pressure-based human body posture estimation method. The device comprises a spatial feature encoder used for extracting human body global pressure spatial features from a global pressure frame sequence of human body actions; the long and short term time sequence attention module is used for extracting human body global pressure time sequence features from the human body global pressure space features and fusing the human body global pressure time sequence features with the human body global pressure space features to obtain pressure space-time features; and the action regression device is used for performing nonlinear regression calculation on the pressure spatial-temporal characteristics to obtain posture and displacement parameters of the human body parameterized model. Visual information is replaced by pressure information which does not depend on illumination conditions, has no privacy invasion and is not affected by visual occlusion, and non-invasion expansion of human body posture estimation in special scenes is achieved.

The invention discloses an unmanned aerial vehicle attitude control method based on multi-scene cruise, and particularly relates to the field of unmanned aerial vehicle flight control. Collecting a wind field change deviation rate, a magnetic interference intensity index and a visual occlusion density abnormal value to form a disturbance characteristic parameter set; constructing a disturbance influence index matrix in each preset cruise scene; acquiring a scene sequence in an actual cruise path, calculating a scene switching frequency and constructing a fusion disturbance vector; inputting the fusion disturbance vector into an attitude control model constructed based on a long short-term memory network, and outputting an attitude control parameter group; dynamically adjusting an unmanned aerial vehicle attitude instruction according to the control parameter group; the control strategy can be dynamically adjusted according to environment disturbance change and scene switching characteristics, attitude stability control of the unmanned aerial vehicle in a complex and changeable environment is achieved, and flight safety and task reliability are improved.

A sound effect playback method and apparatus, a voiceprint identifier display method and apparatus, and a terminal device, relating to the technical field of computers. The sound effect playback method comprises: displaying a virtual environment containing a first virtual object; when the first virtual object is located in a visual occlusion area of the virtual environment and triggers a sound production event in the visual occlusion area, playing back a sound effect of the sound production event obtained after the sound production event undergoes attenuation processing, wherein the visual occlusion area is used for generating an occlusion effect on a virtual element in the area, and the attenuation processing is used for attenuating the effect of the sound effect on identification of the location of the first virtual object. The technical solution provided in embodiments of the present application can enhance the richness of sound effect playback in a game application program.

The invention discloses a small window interaction method and device of vehicle-mounted equipment, electronic equipment and a storage medium, and relates to the technical field of automobile information entertainment systems.The method comprises the steps that when non-task small window operation is triggered, a task small window is switched to be in a preset transparency suspension state or an edge parking state, so that the task small window continues to run in the background, task interruption is avoided, and user experience is improved. Multi-task parallelism is realized; furthermore, the preset transparency suspension form reduces visual shielding, and the edge stopping form fixes the position of the small window, so that mistaken touch can be reduced; the small windows switched into the preset form are centrally stored in the preset area and arranged in sequence, a user can trigger the target small window to quickly switch, and the small windows in the preset form are centrally stored in the preset area, so that the operation efficiency is improved, the switching, storage and trigger recovery logic of the task small windows are unified, and the development and maintenance cost is also reduced.

The invention discloses a motor load attitude observation method based on multi-source heterogeneous signal fusion. The method comprises the following steps: extracting an absolute position predicted value and observation uncertainty of a load end through a visual encoder; synchronously utilizing a signal encoder to extract dynamic characteristics from the electric signals of the driving motor, and executing integral deduction based on physical constraints to obtain a prior track of a load end; kalman gain is dynamically generated in combination with time sequence characteristics, observation uncertainty and innovation, and nonlinear correction of a prior track of a load end is achieved. In addition, a fault detection mechanism based on normalized innovation square is introduced, the gain is automatically adjusted under the working condition of visual occlusion or interference, and the observation robustness is ensured. According to the method, the problems of visual processing delay, load end transmission nonlinear errors and large accumulative errors of a traditional observer are effectively solved, and the precision, the real-time performance and the anti-interference capability of load end position observation are remarkably improved.

The invention relates to the technical field of dish-washing machines, in particular to an intelligent dish-washing machine based on time sequence visual perception, which comprises an image acquisition module configured to be activated in response to an extraction action of a target dish basket, continuously capturing a time sequence image sequence of the target space at a preset sampling frequency; wherein the target space is defined between the inner surface of the opened door plate and the bottom surface of the bowl basket on the last but one layer; the bowl basket below the last but one layer of bowl basket is a target bowl basket; the image acquisition module is further configured to stop capturing in response to the pushing action of the target bowl basket; the image processing module is connected with the image acquisition module and is used for receiving the time sequence image sequence, processing the time sequence image sequence by applying a preset image processing algorithm and generating a load signal representing a target space state; a control system; according to the invention, the visual shielding probability caused by tableware stacking can be obviously reduced, and accurate identification of tableware loads is realized.

The invention relates to an interactive object recognition method, system and device and a medium. The method comprises the steps that S1, body sensing data of an actuator installed at the tail end of a robot is read in real time; s2, preprocessing the collected body sensing data based on a physical model to recover spatial features of the body sensing data, and encoding a preprocessing result; s3, analyzing the encoded body sensing data by using an object recognition model constructed based on a transformer model to obtain a hidden space representation representing an interactive object category; in the training process of an object recognition model constructed based on a transformer model, body sensing data is enhanced through a data enhancement module based on knowledge and physical joint constraint, so that embedding of a basic physical rule is realized; and S4, decoding the hidden space representation representing the category of the interactive object, and identifying the interactive object. Compared with the prior art, the method improves the accurate recognition of interactive objects in visual occlusion and narrow space.

Disclosed in the present invention is a multi-view fusion human motion estimation method based on distributed progressive Gaussian filtering. The method comprises: firstly, performing data collection by means of two Azure Kinect DK cameras, and determining initial data; then, performing data filtering fusion processing; performing classification processing on the initial data of the two Azure Kinect DK cameras by using a Mahalanobis distance, and screening and discarding measurement information of the Azure Kinect DK camera highly susceptible to visual occlusion; for measurement information of the Azure Kinect DK camera which is determined to be less affected by visual occlusion, guiding, by means of measurement information of another Azure Kinect DK camera, to perform progressive filtering fusion on the measurement information of the Azure Kinect DK camera, so as to achieve the effect of implicit compensation; and finally, performing global fusion, so as to further improve the precision of human body pose estimation. The method has the advantages of low cost and high precision, and it is not necessary for a tester to wear a joint point marker, and thus, no inconvenience is caused to the tester.

The invention belongs to the technical field of constructional engineering, and particularly relates to a building bearing column sight shielding perspective method, which comprises the steps of S1, installing a camera, a transparent display screen and a computer system on a bearing column, S2, processing key points of a shielded picture image, S3, carrying out perspective processing on the key points of the shielded picture image, and S4, carrying out perspective processing on the key points of the shielded picture image. S3, performing feature extraction on the image; and S4, synchronously displaying the image. According to the visual line shielding perspective method for the building bearing column, the shielded picture can have the effects of being real-time, free of time delay, high in picture fitting degree and high in brightness and actual brightness fusion in a dead-angle-free mode.

The invention discloses a robot navigation system for a agaricus bisporus planting shelf, the agaricus bisporus planting shelf comprises a support column and a planting groove, and the robot navigation system comprises a robot body, a navigation module and a composite shelf terrain; the robot body comprises a robot body, a four-wheel driving module and a controller; the navigation module comprises a laser range finder array and an acceleration sensor; the laser range finder array comprises a plurality of laser range finder sensor modules installed on the robot body. The method solves the problems of GPS failure, visual occlusion and high deployment cost in a shelf planting scene, has the characteristics of high environmental adaptability, high robustness and accurate operation, and is suitable for automatic management of multilayer planting of the agaricus bisporus.

The invention discloses a dinner plate dish automatic identification and volume estimation system based on RGB-D vision, and belongs to the technical field of RGB-D vision. Comprising a multi-modal data acquisition and preprocessing module, a dinner plate sensing stability judgment module, a dish instance segmentation module, a space coordinate inversion point cloud generation module, a dish geometric model reconstruction module, a multi-strategy adaptive volume calculation module and a data output module. Volume estimation is converted from plane approximation to direct calculation based on real space geometric morphology; through construction and processing of the three-dimensional point cloud of the dishes, the system can better deal with complex conditions of dish stacking, uneven thickness, placement posture change and the like, errors caused by a shooting angle, illumination change and visual occlusion are remarkably reduced, and a volume estimation result is more stable and reliable.

The invention discloses a three-dimensional flow field dynamic adaptive visual sampling method based on deep reinforcement learning, and belongs to the technical field of scientific visualization and computer graphics. According to the method, an interaction environment fusing multi-dimensional physical characteristics and real-time particle density is constructed; the method comprises the following steps: designing an intelligent agent based on an Actor-Critic architecture, and capturing space-time correlation characteristics of a flow field structure and particle distribution by adopting a three-dimensional U-Net combined with a ConvGRU as a strategy network; dynamically outputting newly added particle quantity budget and spatial probability density distribution through a double-head action decision mechanism; and finally, in a GPU-based parallel simulation environment, a particle trajectory is updated by using high-order numerical integration and adaptive life cycle management, and animation rendering is carried out. According to the method, a threshold value does not need to be set manually, the sampling strategy can be automatically adjusted according to the flow field features and the current visual density, the contradiction between visual occlusion and feature loss is effectively solved, and clear and dynamic visualization of a large-scale complex flow field is achieved.

The invention relates to an escalator passenger tumble monitoring method and device, escalator monitoring equipment and a medium, and the method comprises the steps: responding to a first tumble event triggered based on a first video stream collected by first image collection equipment, and obtaining a first timestamp and a first tumble confidence coefficient corresponding to the first tumble event; according to the first timestamp, a second video stream of the corresponding time period is obtained from at least one second image collection device deployed on the target escalator, the second falling confidence coefficient of the second video stream is determined, and the shooting view angle of the second image collection device is different from that of the first image collection device; calculating a comprehensive confidence coefficient according to the first fall confidence coefficient and the second fall confidence coefficient; and when the comprehensive confidence coefficient is larger than a preset confidence coefficient threshold value, an escalator passenger falling event is triggered. According to the method, the target escalator is monitored in an all-around mode through the image collecting devices with different shooting visual angles, and the problem that the passenger falling event is not detected in time or missed due to visual shielding is solved.