671 results about "Motion detection" patented technology

A multimodal intelligent agent system for dynamic environmental monitoring and user-centered support, consisting of: a multimodal sensor module configured to continuously acquire environmental and behavioral data from multiple input modalities, including at least one visual sensor, at least one acoustic sensor, at least one environmental conditions sensor, and at least one proximity or motion detection sensor, each generating modality-specific data streams representing visual images, audio waveforms, physical environmental parameters, and motion signatures within a monitored environment; a data preprocessing and fusion subsystem that is operationally coupled with the multimodal sensor module and configured to normalize, temporally align, and transform the modality-specific data streams into high-dimensional feature embeddings using a variety of encoders, wherein the visual encoder uses convolutional or vision transformer architectures, the audio encoder uses a spectral-temporal feature extractor, and the sensor encoder transforms raw analog data into context vectors suitable for multimodal alignment; a multimodal processing unit consisting of a transformer-based large language model (LLM) trained on paired multimodal datasets and configured to perform semantic fusion, context abstraction, and inference across the aforementioned aligned multimodal feature embeddings to generate a contextual understanding of environmental and behavioral states; an adaptive agent controller coupled to the multimodal inference processing unit and configured to instantiate, manage, and terminate a variety of task-specific intelligent agents, each agent being a software unit configured to perform a specialized function selected from meeting summarization, behavioral analysis, misplaced object detection, or environmental anomaly identification, with the agents dynamically interacting with the inference engine to retrieve contextually relevant multimodal embeddings for task execution; a personalization and adaptive learning subsystem consisting of a user preference database and a neural memory structure configured to update and refine model parameters based on user-specific interaction history, thereby enabling personalized output generation, prioritization of recommendations, and long-term behavioral adaptation; and An output generation interface is operationally connected to the adaptive agent controller and configured to produce multimodal output in textual, visual, and auditory form. The interface is capable of displaying human-readable summaries, notifications, and visual reconstructions of identified entities or environmental states.

The invention relates to the technical field of motion detection, in particular to an old people falling early warning system based on motion trail analysis, which comprises a gravity center identification module, a trend identification module, an inertia pushing module, a phase judgment module and a grade response module. According to the method, the vertical coordinates of the key points of the waist and the two feet in the image frame are acquired, and the height difference change is continuously tracked, so that the potential falling initial state can be identified from the static posture, the vector included angles among the three groups of bone points of the hip knee, the knee ankle and the shoulder pelvis are subjected to sequence analysis, and whether synchronous disintegration occurs in the coordination action is identified; the identified trend sequence is mapped according to the risk level, a differential response mechanism is applied, a high-density corresponding relation is constructed between the behavior trend and the alarm level, graded response and time sequence triggering of the falling risk are achieved, the accurate identification capability of the falling precursor action is enhanced, the sensitivity and discrimination of the alarm response are also improved, and the safety of the falling risk is improved. And misjudgment and missed judgment are effectively avoided.

The invention discloses a PDMS / MWCNTs flexible resistive pressure sensor with a double-layer corrugated structure and a preparation method of the PDMS / MWCNTs flexible resistive pressure sensor, and belongs to the technical field of flexible piezoresistive sensors. The PDMS / MWCNTs flexible resistive pressure sensor with the double-layer corrugated structure sequentially comprises a silica gel packaging layer, a copper foil, a PDMS / MWCNTs conductive thin film with a high-amplitude corrugated structure, a PDMS / MWCNTs conductive thin film with a low-amplitude corrugated structure, a copper foil and a silica gel packaging layer from top to bottom, the cross section of the PDMS / MWCNTs conductive thin film with the high-amplitude corrugated structure and the cross section of the PDMS / MWCNTs conductive thin film with the low-amplitude corrugated structure are both in a wave shape, and the wave length of the wave shape of the PDMS / MWCNTs conductive thin film with the high-amplitude corrugated structure is the same as that of the wave shape of the PDMS / MWCNTs conductive thin film with the low-amplitude corrugated structure. And the wave height of the wave shape of the PDMS / MWCNTs conductive thin film with the high-amplitude corrugated structure is higher than that of the wave shape of the PDMS / MWCNTs conductive thin film with the low-amplitude corrugated structure. The flexible resistance type pressure sensor realizes high linearity and good sensitivity in a wide pressure range, has excellent mechanical stability, rapid responsiveness and good durability, and has great potential in the fields of human wearable health monitoring, motion detection and the like.

The invention provides a mechanical arm motion precision detection system and method based on moire fringes, and relates to the field of image data processing, and the system comprises a grating projection module which is used for determining grating parameters based on motion information of a mechanical arm and projecting a grating pattern to a detection surface of the mechanical arm based on the grating parameters, the grating parameters at least comprise grating density, grating angle and grating type; the image acquisition module is used for acquiring continuous frames of moire fringe images in the movement process of the mechanical arm; and the data processing module is used for carrying out phase calculation on the continuous frames of moire fringe images and resolving the displacement error, the rotation angle error and the vibration error of the mechanical arm, and the motion detection precision of the mechanical arm is improved.

The invention relates to the technical field of reflected waves, in particular to a coal mine disaster comprehensive evaluation system based on three-component micro-motion detection, which comprises a frequency direction extraction module, a propagation included angle calculation module, a sensitive path identification module, a critical turning identification module and a structural disturbance mapping module. According to the method, the continuous sequence is formed in the space through the direction vector constructed through the main frequency value of the measuring point, an included angle screening strategy is used for eliminating incoherent direction sections, it is ensured that the path has trend consistency, the frequency amplitude difference value and symbol inversion are used for detecting a mutation point, and the response recognition capacity for directional changes is enhanced; the normalized frequency ratio is combined with a standard deviation and a difference value to carry out joint calculation so as to quantify a structure jump degree and improve the accuracy of reflex position expression, disaster boundary points form an image set, and a structure identification channel is established by taking a frequency direction as a principal line; and the spatial expression definition and precision are improved in the aspects of path aggregation, boundary precision, disaster recognition and partition and continuous structure drawing.

The invention relates to an agricultural robot, and discloses an animal husbandry pushing robot autonomous navigation method based on a laser radar, which comprises a microsecond-level multi-sensor time synchronization architecture based on an FPGA (Field Programmable Gate Array); in a navigation layer, seamless switching between the GNSS and the laser SLAM is realized by constructing a virtual fusion observation value by utilizing extended Kalman filtering and a self-adaptive dynamic weight algorithm; in the sensing layer, an RGB-I multi-mode model fused with laser reflectivity and a biological micro-motion detection technology are adopted, and feed and biological obstacles are accurately distinguished; and in the execution layer, flexible self-adaptive material pushing is achieved in combination with material groove edge fitting and force-position hybrid control, and the navigation precision, the operation quality and the safety of the material pushing robot in the all-weather and unstructured pasture environment are effectively improved.

The invention discloses a screen structure, an intelligent ground screen, a monitoring method, a medium and a computer product. The stress layer is suitable for being in contact with a target object, the circuit board layer is arranged below the stress layer, the force conducting layer is arranged below the circuit board layer, the force sensing layer is arranged below the force conducting layer, the force sensing layer comprises a pressure sensor, the force sensing layer abuts against the force conducting layer so as to be suitable for monitoring pressure from the force conducting layer, and the force sensing layer is connected with the base. Wherein the stress layer comprises a mask structure and a force conduction structure, the force conduction structure is provided with two opposite ends in the direction from the base to the stress layer, the first end is connected with the mask structure, the second end is connected with the force conduction layer so that the pressure of the mask structure can be conducted to the force conduction layer, the circuit board layer is provided with a through hole, and the force conduction structure penetrates through the through hole. According to the invention, the motion state of the target object can be accurately monitored in real time, and a high-sensitivity and high-precision circuit system can be protected from being damaged in frequent high-load motion detection.

The invention discloses a video restoration data set construction and restoration method combined with generation of a large model, and belongs to the technical field of video restoration, and the method comprises the following steps: S1, slicing; s2, screening: performing aesthetic scoring and motion detection on the slices, and screening and retaining high-dynamic and high-image-quality fragments; s3, segmenting a core object; s4, extracting text features; s5, model input; s6, performing feature fusion; s7, model training: only training parameters of the cross attention layer, and locking gradients of other layers to reduce the calculation cost; s8, calculating a loss function; s9, video input: inputting a to-be-restored video, and generating an object Mask and text description of the to-be-restored video; and S10, outputting the restored video: inputting the Mask video, the object Mask and the text description into the improved generated large model, and outputting the restored video. According to the staged feature injection strategy, the global naturalness and the local reality sense are considered, and the visual consistency of the repaired content and the original video is remarkably improved.

The invention discloses a port container cargo loading and unloading state identification method based on computer vision, and the method comprises the steps: collecting visible light and infrared video images in a port, and carrying out the preprocessing of the images through combining with a defogging algorithm; carrying out the recognition and classification of port objects in the video based on an improved OfficientDet target detection model, and dynamically adjusting the feature fusion weight; in combination with a motion detection technology, if a target is detected to move, the target is divided into a truck type; otherwise, classifying into a'container 'class; further optimizing and classifying the detection target based on an ST-MRF image segmentation technology, and extracting a container class; using a DINO-X target detection model to identify whether a container cover exists or not, and marking a container without a cover as'no goods, to be loaded and unloaded '; and through a multi-feature fusion loading and unloading state probability model MFSM, the container cover coverage rate, the cargo height and the inclination angle are integrated to judge the state of the container. The accuracy and efficiency of container loading and unloading state recognition are improved, and the intelligent level of port container loading and unloading management is improved.

The invention discloses an epilepsy signal identification method and system based on multi-modal information of wearable equipment. The method comprises the following steps: firstly, preprocessing acquired multi-modal physiological signal data; calculating a time-frequency feature through a sub-band analysis coefficient, extracting a time domain feature, a frequency domain feature and a nonlinear feature after reconstruction, and constructing a multi-modal feature matrix; obtaining an action feature matrix through a time-frequency feature threshold classifier; performing T test, correlation analysis and recursive feature elimination, and screening to obtain a final training set and an optimal feature list; respectively training an LSTM attack detection model and a random forest motion detection model, and constructing a serial ensemble learning framework; and finally, outputting a risk value based on a sliding window, triggering an alarm when the risk value continuously exceeds a threshold value, and starting a misinformation suppression mechanism in a non-response period. According to the method, non-attack motion interference is filtered through a designed serial integrated learning framework, the false alarm rate of the system is remarkably reduced, meanwhile, high detection sensitivity and accuracy are kept, and the real-time processing efficiency and practicability of the system are improved.

The invention relates to a pressing plate operation behavior analysis and error prevention method and system based on image recognition and machine learning. The method comprises the following steps: acquiring a pressing plate operation area video stream through image acquisition equipment, and segmenting a video into independent operation event segments based on motion detection and trajectory analysis; extracting key point time sequence data of hands of an operator by using a human body posture estimation model, and constructing a dynamic feature sequence fusing a spatial relationship, kinematics and posture semantics; performing multi-level similarity comparison on the dynamic feature sequence and a standard operation template, and judging operation compliance by a machine learning model in combination with a dynamic threshold value; triggering graded early warning and intervention according to a judgment result; and an incremental learning mechanism is adopted, and the template and the threshold value are adaptively optimized based on historical data. According to the invention, accurate and intelligent analysis and active error prevention of the whole operation process of the pressing plate are realized, and the safety level of electric power operation is effectively improved.

The invention relates to the technical field of motion detection alarm, in particular to a multi-agent fusion perception area intrusion abnormity alarm method and system, and the method comprises the following steps: collecting a target coordinate and a timestamp, integrating and sorting tracks, analyzing the direction and time change, judging the stability, recognizing an induction behavior, and screening and verifying the abnormity. According to the method, through integration of the spatial position and time information of the multi-source sensing intelligent agent, the continuous track of the target is reconstructed, the integrity identification of the motion behavior is enhanced, and the track direction change and time interval linkage analysis is carried out, so that the sensing depth of the behavior trend is improved; multi-dimensional recognition of abnormal behaviors is achieved through track stability and regional stay state linkage judgment, continuous conditions are introduced into abnormal verification to filter accidental interference, alarm output is matched with a regional strategy based on a verification result, the accuracy and robustness of intrusion detection are effectively improved, and the practical requirement in a high-dynamic scene is met.

Systems and methods for enrolling and authenticating a user in an authentication system via a user's camera of camera equipped mobile device include capturing and storing enrollment biometric information from at least one first image of the user taken via the camera of the mobile device, capturing authentication biometric information from at least one second image of the user, capturing, during imaging of the at least one second image, path parameters via at least one movement detecting sensor indicating an authentication movement of the mobile device, comparing the authentication biometric information to the stored enrollment biometric information, and comparing the authentication movement of the mobile device to an expected movement of the mobile device to determine whether the authentication movement sufficiently corresponds to the expected movement.

Data obtained from synthetic aperture radar "SAR" image acquisition apparatus on a satellite, operating in orbit around Earth, is processed to form looks at a target area on Earth at different points in time. Dense optical flow is computed to identify apparent motion between successive looks and to determine apparent velocity for at least some pixels between successive looks. The apparent velocity is converted into actual velocity by correction for the relative motion between the platform and the target. A 2D velocity map of motion in the target area may be generated.

A solid-state lighting fixture assembly having a lighting fixture with a socket configured to receive a plug associated with one or more accessories to allow for easy in-field mounting of accessories, e.g., controls, onto installed lighting fixtures. The socket may be internally electrically connected to an auxiliary power output of a driver and / or to a battery power pack within the lighting fixture assembly, thereby providing direct-current voltage power for the accessory and, also, allowing for signal transmission to and from the accessory. Each accessory includes one or more sensors and communication components to provide the connected lighting fixture assembly with specific capabilities including, but not limited to, motion detection, ambient light level detection, ambient temperature measurement and wireless communications. The wireless communication can also be used to control one or a group of lighting fixtures and transmit sensor data associated with, for example, monitoring space utilization and asset tracking.

Systems and techniques are described for image signal processor motion detection. In various examples, first image data including pixel values may be generated, where each pixel value is associated with a respective color tag of a color filter array. A first set of the pixel values corresponding to a first block may be determined. A second set of pixel values may be determined based on a respective pixel value of the first set and the respective color tag associated with that respective pixel value. A first intensity value associated with the first block may be determined based on the second set of pixel values. A second intensity value associated with a reference block may be determined. A comparison value representing a difference between the first intensity value and the second intensity value may be determined. A signal indicating detected motion may be sent based on the comparison value.

Provided are an animal monitoring system, an animal monitoring server, an animal monitoring method, and an animal monitoring program capable of watching a behavior of a monitored animal with high accuracy while reducing power consumption. An animal monitoring system 1 that monitors a monitored animal 6 includes: a motion detection unit 502 attached to the monitored animal 6 and configured to detect motion information of the monitored animal 6; a behavior estimation unit 212 configured to estimate a behavior of the monitored animal 6 on the basis of the motion information detected by the motion detection unit 502; and a power saving control unit 508 configured to start control of reducing power consumption in the motion detection unit 502 when the behavior estimation unit 212 estimates that the behavior of the monitored animal 6 is in an inactive state.

The invention discloses an intelligent house home safety protection system, and relates to the technical field of home safety protection. Comprising an environment data acquisition module, a data preprocessing module, an abnormal behavior detection module, a multi-sensor data fusion module, an abnormal event verification module and an alarm feedback and user interaction module, the environmental parameters include but are not limited to temperature, humidity, illumination, motion detection data and sound data. According to the invention, the accuracy of smart home safety protection is improved through multi-sensor data fusion and a deep learning technology, and false alarms are reduced. The system analyzes sensor data in real time and accurately identifies security threats and conventional activities. The self-adaptive verification mechanism adjusts a strategy according to user behaviors and combines a multi-level alarm mechanism to ensure that the user responds in time in an emergency, so that the safety protection effect is improved.

The present disclosure provides methods, systems, and mediums for identifying an active speaker within an online conferencing session. The method comprises the steps of receiving an audio / video stream from a client device during an online conferencing session. Upon a participant speaking: identifying, a voiceprint representing the participant, wherein the voiceprint represents one or more unique vocal characteristics of the participant. Detecting a spatial position of the participant based upon movement of one or more markers of interest on the first speaker. Generating a mapping between the voiceprint and the spatial position of the participant. Using the voiceprint and the spatial position of the participant to identify the participant as a first active speaker. The method further comprises generating instructions to adjust positioning of a camera so that the first active speaker is centered within a video stream.

The invention relates to the field of target micro-motion detection, and provides a Rydberg atom-based target micro-motion detection system and method, and the system comprises a microwave generation device, a Rydberg atom sensor, a photoelectric detector, and a signal collection and processing module. The microwave generating device comprises a local oscillation microwave source for generating a local oscillation microwave field and a signal microwave source for generating a signal microwave field; the signal microwave field penetrates through the Rydberg atom sensor and is emitted to a to-be-detected target, an echo signal field of the to-be-detected target returns to the Rydberg atom sensor and is subjected to three-field frequency mixing with the local oscillation microwave field and the signal microwave field, and a target micro-motion signal carrying frequency mixing information is obtained; the photoelectric detector is used for detecting the target micro-motion signal carrying the frequency mixing information and processing the target micro-motion signal to obtain an electric signal, and the electric signal is input to the signal acquisition and processing module to complete target micro-motion detection. Target micro-motion detection is carried out based on the Rydberg atom sensor, and the detection capability of an existing radar system on weak and small targets can be improved.

To provide aerosol-generating devices configured to generate an aerosol during a usage session, and to provide methods of using such devices.SOLUTION: An aerosol-generating device for generating an aerosol from an aerosol-forming substrate is configured to generate the aerosol during a usage session. The aerosol-generating device comprises an indicator system (60), a motion detection unit (440) configured to detect movement of the device and identify a gesture associated with the detected movement, and a controller (420). The motion detection unit is configured to detect movement of the device and identify a first gesture and a second gesture that is different from the first gesture.SELECTED DRAWING: Figure 4

The invention discloses a massage quantization glove based on multi-signal fusion and a real-time feedback control system thereof. The massage quantization glove comprises a glove body, a double-layer differential sensing unit, a perturbation excitation unit, a motion detection unit, a signal processing and fusion unit and a feedback interaction unit. The double-layer differential sensing unit is composed of an outer-layer pressure sensing film, an isolation elastic layer and an inner-layer pressure sensing film and is used for collecting overall contact force and differential response signals. The perturbation excitation unit applies low-amplitude disturbance in the massage process, so that soft tissue viscoelasticity characteristics are extracted; the motion detection unit collects hand motion information and performs inertia compensation; the signal processing and fusion unit is used for denoising and analyzing the multi-source signal to generate a dose index and a hardness grade; and the feedback interaction unit prompts an operator through visual sense, auditory sense or tactile sense. Real-time quantification and feedback of massage dosage are realized, tissue hardness change can be reflected, excessive massage is avoided, curative effect consistency is improved, and massage operation standardization is promoted.

The invention discloses an electric power overhaul video action detection method based on a multi-scale state space, and the method comprises the steps: firstly collecting long video data in an electric power overhaul process, segmenting the long video data into a plurality of time sequence segments, and carrying out the spatial feature coding and high-dimensional mapping of the plurality of time sequence segments, and obtaining a time sequence token sequence; then, an electric power overhaul video action detection network is constructed and trained, and the electric power overhaul video action detection network comprises a time sequence multi-scale coding module, a scale perception state fusion device and a multi-label prediction layer; and finally, performing action detection on a time sequence token sequence corresponding to the electric power overhaul long video data to be detected by adopting the trained electric power overhaul video action detection network to obtain an action category detection result. According to the method, timing sequence multi-scale coding, state space modeling and a scale perception feature fusion mechanism are fused, the short-time sudden action and the long-range dependency relationship can be captured at the same time, and the recognition capability and the positioning precision of the complex concurrent action in the electric power overhaul video are remarkably improved.

The invention relates to the field of image data processing, and provides a steam boiler gas leakage detection method based on image processing, which comprises the following steps: setting an image acquisition unit, covering a predetermined leakage risk area together, acquiring corresponding reference images under different production working conditions, and establishing an image reference library covering a target area; the method comprises the following steps: dividing a boiler operation time axis into a plurality of production states, calling corresponding reference images according to different states, and carrying out difference, partition comparison and motion detection on the real-time acquired images to obtain suspected leakage cloud cluster areas in the corresponding states; arranging a condensate water indication patch, collecting a patch image, extracting through color, brightness and texture features to form a patch humidity change curve, carrying out multi-source information fusion on the patch humidity change curve, time sequence features and production time states, and outputting a corresponding steam leakage risk level; and generating alarm information according to the steam leakage risk level, and recording and archiving the alarm information in a monitoring and management system for energy consumption analysis and risk tracing.

The invention belongs to the field of dangerous rock monitoring, and provides a dangerous rock monitoring method, device, equipment and medium based on wide-area electric shock amplitude frequency, and the method comprises the following steps: carrying out high-frequency wide-area electromagnetic method detection, electromagnetic radiation monitoring, micro-motion detection and vibration signal monitoring on a dangerous rock monitoring area; based on the high-frequency wide-area electromagnetic data at different time points, whether resistivity difference abnormity exists or not and whether an electromagnetic radiation abnormal event occurs or not are determined; based on the micro-motion data at different time points, whether speed structure difference abnormity exists or not and whether a vibration abnormal event occurs or not are determined; and when the electromagnetic radiation abnormal event and the vibration abnormal event are synchronous in occurrence time and consistent in spatial positioning, determining that the event is a rock crushing event. Through multiple detection means, broadband range real-time monitoring of electromagnetic signals and vibration signals is realized, abnormal signals generated by internal changes of rocks can be detected and identified to the maximum extent, the dangerous rock disaster-pregnancy process is effectively monitored, ground disaster missing report is reduced, and the early warning success rate is improved.

The invention relates to the technical field of image recognition, in particular to an intelligent recognition and detection method for preceding vehicle starting. The method comprises the following steps: acquiring an image of an automobile data recorder, and judging whether an automobile is in a static or moving state; extracting a plurality of view angle regions and carrying out region separation according to texture density; detecting the area smoothness to determine a motion detection sensitivity; and according to the state and sensitivity of the vehicle, front vehicle starting identification is carried out on each area, starting early warning response is sent out based on an identification result, and the detection accuracy and real-time performance are improved. According to the method, the driving video and the motion sensing data are fused, and multi-view region segmentation, adaptive sensitivity adjustment and a dynamic early warning mechanism are combined, so that high-robustness and high-precision recognition of the starting behavior of the front vehicle under various illumination and working conditions is realized, and the method has good universality and compatibility.

The invention discloses a thermal infrared imager intelligent temperature measurement platform management method and system. The method comprises the following steps: S1, information acquisition and multi-modal data binding; s2, encoding and encrypting data; s3, wireless transmission and decryption are carried out; s4, the server side performs intelligent processing; s5, result distribution and storage; according to the invention, through deep fusion of multi-modal data binding and environment interference parameters, the motion detection and temperature pre-judgment process can dynamically adapt to interference changes in a complex scene; a multi-modal fusion dynamic threshold algorithm, a space-time attention enhancement model and a layered time sequence temperature prediction model are integrated, the full-dimension requirement from real-time monitoring to long-term trend analysis is covered, and the intelligent level and scene adaptability of the system are remarkably improved; and a whole-process system architecture of sensing, transmission, processing, storage and display is constructed, so that the platform has high efficiency of data processing and convenience in use.

Implementations described herein relate to a method, system, and computer-readable medium for generating animation for a 3D avatar from input video and audio captured on a client device. A camera may capture facial video, and a trained face detection model and a trained regression model output a set of video FACS weights, head pose, and facial landmarks that will be converted into animation for the 3D avatar. In addition, a microphone may capture audio emitted by a user, and a trained face motion detection model and a trained regression model output a set of audio FACS weights. In addition, a blending term is provided for identifying time progression in the audio. A modular blending component blends the video FACS weights and audio FACS weights based on the blending term to create final FACS weights for animating the user's avatar, a character's clothing, or another animation-enabled construct.

The invention provides a water conservancy twinborn model rendering optimization method for dynamically generating gray level image frames based on a view cone. The method comprises the following steps: obtaining visible area model data after the current view cone is dynamically cut; outputting the visible area model data to a multi-layer rendering target comprising a color buffer area, a vertex buffer area and a depth buffer area through off-screen rendering so as to generate intermediate frame data isolated from screen display; performing gridding processing on the intermediate frame data, and displaying the equipment resolution according to the physical range size of the water conservancy project or the drainage basin scene; the average color value of each grid unit is calculated, desaturation processing is carried out to generate a basic grey-scale map, and adaptive feature enhancement processing is carried out to generate an optimized grey-scale map or a gradient map; and a GPU rendering pipeline is input and acts on hydrological feature analysis of the post-processing stage of the current frame and a preloading data source rendered by a twinborn model of the next frame at the same time, historical frame data are reused in a static region through motion detection, and real-time rendering is triggered in a dynamic region.

The invention provides a personnel action visual detection method based on end-side cooperation, and relates to the technical field of visual detection. The personnel action visual detection method based on end-side cooperation specifically comprises the following steps: S1, system deployment; s2, end-side data acquisition and processing; s3, dynamic data transmission; s4, edge side feature fusion and fine detection; and S5, result feedback and model updating. According to the invention, through cooperation of the end-side lightweight model and the edge-side high-precision model, the detection delay is reduced; the computing power is optimized through self-adaptive weight distribution, and stable data transmission is ensured through a dynamic transmission strategy; multi-scale feature fusion improves the detection precision, and a dynamic pruning mechanism gives consideration to light weight; the cloud end regularly updates and adapts to scene changes through knowledge distillation and federal learning optimization models; the system supports various action detection tasks, is suitable for the fields of intelligent security, medical monitoring, industrial monitoring and the like, is superior to the prior art in the aspects of precision, real-time performance, resource efficiency and applicability, and has wide application prospects.