116 results about "Motion perception" patented technology

The invention relates to a self-localization and motion perception method based on deep learning, and the method comprises the following steps: multi-modal perception: collecting RGB frames and event streams through employing a DAVIS346 event camera, obtaining texture and depth information through employing an RGB-D camera, and supplementing 3D structure data through a laser radar; hybrid optical flow driven motion perception: adopting a double-branch architecture of an improved RAFT network and an event optical flow private network; multi-modal 3D detection and trajectory management: fusing multi-modal data based on VoxelNeXt-Lite to generate a 3D detection frame, filtering false detection by combining point cloud density clustering and an event density threshold, then constructing a space-time diagram associated trajectory through a Spatio-Temporal Graph Transformer, and optimizing pedestrian trajectory prediction continuity by using an LSTM (Long Short Term Memory) model perceived by a gait cycle; multi-view semantic graph matching: constructing a 3D semantic voxel map containing vertical features, calculating scene similarity in combination with top view NetVLAD features and deformable graph matching, and dynamically adjusting semantic weight by using a situation encoder and knowledge graph reasoning; and carrying out multi-sensor fusion and robust positioning.

The invention is applicable to the technical field of medical image processing and mixed reality, and provides a laparoscopic surgery mixed reality navigation method based on deep learning and dynamic point tracking, which comprises the following steps: dynamically registering a three-dimensional model containing kidney, tumor and vessel with an initial frame of a laparoscope through a mixed reality alignment technology; the method comprises the following steps: constructing an operating forceps motion sensing model based on a time sequence deep neural network, realizing real-time control and parameter locking of a three-dimensional model pose, dynamically updating a two-dimensional feature point set by adopting a multi-feature-point combined tracker, constructing a candidate feature combination through a cross-quadrant sampling strategy, and constructing an operating forceps motion sensing model; a candidate feature combination is generated through a four-quadrant division and cross-regional sampling strategy, an optimal camera pose parameter is generated in combination with a re-projection error and pose continuity constraint, and an operation video is dynamically covered with a semitransparent three-dimensional model. The method can significantly enhance the spatial perception capability of the kidney anatomical structure, reduce the registration error of the kidney in the three-dimensional integrated kidney structure model and the laparoscope video, and improve the navigation precision.

The invention discloses a high-efficiency time sequence optical flow method, system and device for fusing event and image information and a medium. The method comprises the following steps: constructing a time sequence feature extraction module; constructing a double-branch collaborative iteration optical flow module of the event and the image; constructing a fusion modeling and iterative optimization optical flow module based on cross attention; training a dual-branch collaborative iteration optical flow module of the event and the image and a fusion modeling and iterative optimization optical flow module based on cross attention on a training set of an MVSEC public data set to obtain weights, and testing on a test set of the MVSEC public data set to obtain an optical flow calculation result; the system, the equipment and the medium are used for implementing the method. The method has the advantages of high time resolution, high dynamic modeling capability, fine multi-modal fusion, high estimation precision and the like, the motion sensing performance in a complex scene is remarkably improved, and powerful technical support is provided for an intelligent visual system.

The invention relates to a computer vision technology, in particular to an event camera target detection method based on Euclidean-hyperbolic motion perception, which comprises the following steps: down-sampling original event data, constructing an event graph by using the down-sampled data, and inputting the event graph into a double-space network for target detection; performing local perception feature extraction on the event graph by using a graph convolutional network based on a B-spline kernel function, and projecting the extracted features to a hyperbolic space; extracting global event dependency features from the local sensing features mapped to the hyperbolic space through a learnable curvature hyperbolic graph convolutional network; and decoding the global event dependency feature and inputting the decoded global event dependency feature into a sensing head, and outputting a target detection result by the sensing head. The method solves the problems of noise suppression, topological mapping and hierarchical expression in event data, has good application prospects and popularization value, and is particularly suitable for real-time target sensing tasks in complex dynamic environments such as automatic driving and robot navigation.

The invention relates to an action perception and reconstruction method and device, equipment and a storage medium. The method comprises the following steps: carrying out space-time synchronization and heterogeneous graph modeling on asynchronous data of an electromagnetic sensor and an inertial measurement unit (IMU) through multi-source sensor data preprocessing and fusion base construction; performing deep feature extraction and hierarchical fusion by using a gated attention time-space diagram neural network (ST-GNN); through a dynamic interference confrontation and feature purification mechanism, the robustness of the system in a complex environment is improved; generating a high-fidelity and physiologically reasonable action sequence by adopting a multi-scale space-time generation network (MSTGN) and combining biomechanical constraints; a physical information neural network (PINN) constraint is embedded, and high-precision six-degree-of-freedom (6DoF) pose estimation is realized through a decoupling collaborative framework. According to the method, the precision, robustness and real-time performance of motion sensing and reconstruction are effectively improved.

The invention discloses a teaching implementation method based on combination of motion perception and AI driving, and relates to the technical field of education and teaching. The implementation of the teaching method comprises the steps of multi-modal data expansion and acquisition, cross-dimensional dynamic evaluation and cognitive diagnosis, emotion adaptive feedback and intervention, cross-scene learning map construction and capability migration, and whole-process data closed loop and systematic evolution. And the multi-modal data expansion acquisition comprises the steps of dynamic data acquisition, physiological data acquisition, environmental behavior data acquisition and data preprocessing. The method has the advantages that a multi-modal data acquisition scheme of a common mobile phone camera and a built-in sensor is adopted; the system replaces the traditional dance teaching which depends on special three-dimensional modeling equipment and sensor gloves for boxing teaching, reduces the hardware cost, can realize full scene coverage of families, classrooms, outdoors and the like without professional site deployment, and solves the core problems of high hardware cost and limited scenes in the prior art.

The invention provides virtual reality-based dance teaching equipment and a teaching method thereof, and the equipment comprises a VR module which is used for providing multi-view immersive visual experience; the motion capture module is used for capturing the joint angle of the dancer in real time and the limb track feedback module and is used for simulating the ground counter-acting force in dancing through vibration or pressure, enhancing motion perception and providing a touch prompt for motion correction; the virtual scene module is used for creating a dance virtual scene; and the action analysis module analyzes the joint action deviation based on multiple view angles, generates an instant score according to the joint angle deviation and the rhythm synchronization rate, and records historical data to generate a progress curve. According to the method, the problems of resource shortage, feedback lag and other pain points in traditional dance teaching are solved, efficient, personalized and interesting dance learning experience is achieved through multi-modal interaction and a dynamic adjustment mechanism, and the application of the method is expected to promote dance education to develop in the intelligent and universal direction.

The invention relates to the technical field of computer vision and robot positioning, and discloses a dynamic 3DGS-SLAM method and system based on Gaussian pyramid and adaptive densification. The method comprises the following steps: generating a semantic mask through real-time dynamic target detection so as to distinguish a static background from a dynamic object; constructing a 3D Gaussian map model of the scene; rendering images in parallel on a full-resolution view and a down-sampling resolution view by adopting a motion perception Gaussian pyramid rendering method, and fusing rendering results by utilizing a dynamic mask so as to eliminate motion edge artifacts; a self-adaptive densification strategy based on a monotonic attenuation function is adopted, the initialization radius of the Gaussian ellipsoid is dynamically adjusted based on the training progress, and high-density initialization and progressive redundancy pruning are executed to optimize the geometric fidelity; and finally, performing coarse-to-fine camera pose optimization, and outputting a precise pose and a high-fidelity static map. According to the method, the problems of tracking drift and rendering artifacts under dynamic interference are effectively solved, and the positioning precision and the rendering quality are improved.

The invention discloses a satellite cloud picture sequence prediction method combining deformable convolution and pyramid pooling, and relates to the field of deep learning. The method comprises the following specific steps: (1) preprocessing satellite cloud picture sequence data and dividing into a training set and a test set; (2) building a satellite cloud picture sequence prediction model of a basic encoder-translator-decoder structure; (3) combining a motion perception loss function and an L2 loss function to supervise the model for training; and (4) predicting a satellite cloud picture sequence image. According to the method, a deformable volume operator based on dynamic sparseness is used in an encoder and a decoder to replace the traditional convolution operation, so that the network can adaptively select proper parameters according to the space structure of a cloud picture, softmax normalization in deformable convolution space aggregation is removed, and the robustness of the network is improved. The memory access is optimized to accelerate the running speed; a pyramid pooling structure is used in the translator, so that the calculation process can be simplified, and spatial-temporal characteristics of different scales in the cloud picture sequence can be captured; meanwhile, a motion perception loss function and a traditional L2 loss function are combined to supervise model training, and motion information between adjacent frames of a satellite cloud picture sequence can be further obtained. The method is not only suitable for all satellite cloud picture sequence images, but also can be applied to sequence image prediction in other complex scenes.

The invention relates to the technical field of intelligent vision and image fusion, and discloses an intelligent camera dual-light image fusion method and a dual-light integrated intelligent camera, and the method comprises the steps: collecting an original infrared image and an original visible light image; preprocessing the original infrared image and the original visible light image to obtain a double-path preprocessed image; adopting an improved optical flow algorithm to calculate an optical flow field for the two-way pre-processed image; generating an infrared band motion sensing feature set and a visible light band motion sensing feature set; generating a preliminary fusion image based on the infrared band motion sensing feature set and the visible light band motion sensing feature set; for the preliminary fusion image, obtaining an evaluation result including definition, motion consistency and detail richness evaluation indexes; dynamically adjusting a cross-modal attention weight and a gating value according to an evaluation result; and obtaining an optimized final fusion feature based on the dynamically adjusted cross-modal attention weight and the gating value. According to the invention, the fusion image has no dislocation when the target moves rapidly.

The invention discloses a dynamic scene reconstruction method, computer equipment and a program product. According to the method, picture information and camera poses of the same scene are obtained, a standard field composed of three-dimensional Gaussian primitives is constructed in a three-dimensional space, a deformation field is constructed, the standard field and the deformation field at different times are subjected to joint rendering in a three-dimensional Gaussian splashing mode, and various parameters are trained and optimized according to the difference between a synthetic image and a real image. Performing motion perception according to the motion condition of the three-dimensional Gaussian primitives, taking the primitives with smaller motion as static primitives, and performing partition modeling on the primitives with larger motion in time to obtain a layered dynamic scene model; and calling the model for rendering according to the given time and the camera pose in the reasoning stage, and outputting a target image. According to the method, the reconstruction precision of the high-dynamic target can be improved while the rendering efficiency is ensured, and the continuity and visual quality of the dynamic scene in the time dimension are improved by introducing the time consistency constraint at the boundary of the time subinterval.

The invention discloses an unmanned aerial vehicle target tracking method based on motion perception and feature enhancement, and relates to the technical field of unmanned aerial vehicle video processing. In order to enhance the response capability of a tracking network to a dynamic target and improve the robustness in a motion sudden change scene, a Transform architecture is adopted to extract global features, and a motion sensing module is introduced to perform modeling on target motion. A smooth deformation field is generated through a sparse offset prediction module to realize motion compensation, and the compensation precision is further improved by using a motion decoupling module. In addition, an adaptive motion attention enhancement module is designed, and the module adaptively adjusts the attention intensity according to the motion information. And finally, predicting a target bounding box by the classification regression network. According to the method, the feature expression is enhanced by effectively utilizing the motion information, and the tracking accuracy is remarkably improved.

The invention provides an occupied grid prediction method and device, electronic equipment and a readable storage medium. The method comprises the following steps: acquiring continuously acquired multiple frames of visual images and laser point cloud data; converting the visual image to a point cloud scene to obtain image point cloud data; fusing the image point cloud data and the laser point cloud data to obtain fused voxel features; performing motion perception prediction on the dynamic voxels in the fused voxel features through the occupancy flow to obtain motion perception features corresponding to the dynamic voxels; and performing state prediction based on the motion perception features to obtain an occupation prediction result. Visual images are converted into image point cloud data, and the image point cloud data and laser point cloud data are fused in a 3D space to obtain voxel features, so that information loss caused by multi-mode mutual projection can be avoided, and meanwhile, motion perception is directly performed on fused voxels through an occupation stream, so that complete space information can be effectively captured and utilized, and the accuracy of the motion perception is improved. Therefore, more accurate and comprehensive occupied grid prediction is realized.

The invention relates to a handle positioning method and device for sparse features, and the method comprises the steps: constructing a non-rigid multi-sensor joint state estimation model, and building a non-rigid connection through a camera and an IMU (Inertial Measurement Unit); performing joint optimization on the visual re-projection error and the IMU pre-integration error in the sliding window to obtain initial attitude estimation of the handle; based on the number of visible feature points, a geometric degradation state is analyzed through observability, a self-adaptive marginalization strategy of motion perception is established, and a sliding window is dynamically adjusted; based on a self-adaptive marginalization strategy, a prior quality monitoring and multi-hypothesis maintenance mechanism is implemented, initial attitude estimation is optimized, and an optimal handle attitude is output; equipment is realized based on the method. According to the method, high pose estimation precision can still be maintained in a complex interaction scene, long-term drift caused by inaccurate models is remarkably reduced, the survivability under the severe conditions of shielding, rapid rotation and the like is greatly improved, limited visual information in each frame is efficiently and accurately utilized, and the real-time operation capability of a system is ensured.

The invention relates to the technical field of medical information analysis, in particular to a thoracic surgery postoperative complication assessment method and system based on multi-dimensional analysis, and the method comprises the following steps: S1, obtaining a postoperative multi-dimensional monitoring data set; s2, performing standardized preprocessing on the multi-dimensional monitoring data set to generate standardized monitoring data; s3, classifying and integrating the standardized monitoring data into physiological function data, drainage feature data and motion perception data according to data types; s4, chest dynamic pressure change analysis is carried out, and pressure gradient change parameters are calculated; s5, obtaining an abnormal metabolite ratio index; and S6, generating a complication risk assessment report. According to the method, through standardized integration and weighted analysis of the multi-dimensional postoperative monitoring data, accurate quantitative evaluation and graded output of the thoracic surgery postoperative complication risk are realized, and the reliability of early warning and the timeliness of clinical intervention are improved.

The invention discloses a reference multi-target tracking method based on language decoupling and fine-grained multi-modal feature alignment, and relates to a computer vision technology. The method comprises the following steps: A, giving a training data set containing a video sequence and language description; b, inputting the video sequence in the step A into a backbone network to extract visual features, and inputting language description into a language model to extract text features; and C, performing multi-modal alignment and fusion through a cross attention mechanism according to the visual features and the language features extracted in the step B. And D, decoupling the language features extracted in the step B into local description and a motion state. And E, inputting the refined features extracted in the step C and the local description extracted in the step D into a static semantic enhancement module to extract target information. And F, associating the current frame target obtained in the step E with the existing trajectory by using a Hungary matching algorithm. G, inputting the matched target features in the step F and the motion state in the step D into a motion perception alignment module to enhance the target recognition capability; the tracking performance of the method is improved.

The present invention relates to the field of target recognition technology, and provides a moving target tracking method, device, electronic device, and storage medium. The method comprises: inputting a historical target bounding box sequence of a tracked target into a motion position representation module of a target tracking model to obtain a historical position embedding feature sequence of the tracked target; inputting a historical position embedding feature sequence of the tracked target into a target position prediction module of the target tracking model to obtain a current position embedding feature of the tracked target; inputting a regional image of a current search area into a visual recognition module of the target tracking model to obtain a current visual feature of the tracked target identified by the visual recognition module based on a target template; and inputting the current position embedding feature and current visual feature of the tracked target into a motion perception fusion module of the target tracking model to obtain a current target bounding box of the tracked target. The present invention improves the accuracy of tracking complex moving targets.

The invention relates to the technical field of computer vision, in particular to a motion perception rockfall intelligent monitoring method and device based on deep learning. The method comprises the following steps: receiving a monitoring video from a monitoring area; constructing a rockfall detection and tracking framework based on a YOLOv8 target detection network and a ByteTrack target tracking algorithm; preprocessing the monitoring image, inputting the monitoring image into a YOLOv8 detection network to carry out moving target feature extraction, and outputting a rockfall detection result; and track matching is carried out by using a ByteTrack algorithm in combination with Euclidean distance and motion direction constraint, and a rockfall motion track is generated. According to the method, a traditional image algorithm and a deep learning algorithm are combined, moving target detection and multi-target tracking are constructed, and rockfall in a complex rock background is accurately captured from multiple dimensions; according to the method, the space coordinate position of the rockfall target can be accurately obtained, a complete movement track can be generated for each rockfall target, statistical information of rockfall events is effectively obtained, and a reliable basis is provided for disaster grade evaluation and early warning.

The invention relates to the technical field of attitude evaluation, and discloses an attitude evaluation method and system based on motion perception. A posture evaluation method based on motion perception comprises the following steps: S1, collecting basic static parameters of a user's body, setting optical mark points on main joint points of the user's body, and arranging the user to execute a target motion in a predetermined dynamic capture space; s2, capturing a three-dimensional space coordinate sequence of the optical mark point in the process that the user executes the target action by using a plurality of optical cameras arranged in the dynamic capture space; and S3, constructing a personalized three-dimensional skeleton model according to the human body basic static parameters. According to the method, the deviation integral value of the actual motion track and the standard motion track in the whole motion period is calculated, so that the process continuous evaluation of the motion quality is realized.

The invention relates to the technical field of image segmentation, and provides a gastrointestinal tract endoscope image segmentation method, which comprises the following steps: carrying out time sequence preprocessing on an obtained intestinal tract video, eliminating fuzzy and shielded invalid frames, constructing a rhythm signal based on a multi-scale time sequence visual feature, and estimating a creeping phase; outputting a phase time sequence and a phase label of the video; selecting a representative phase key frame, performing pixel-level focus segmentation on the key frame, and calculating a pixel-level uncertainty score to obtain a key frame segmentation result with high confidence; carrying out time sequence consistency recovery on the segmentation mask of the key frame based on phase consistency mapping and phase alignment, and recovering an uninterpretable section caused by shielding with the assistance of adaptive segmentation correction of motion perception; and carrying out time sequence aggregation on a cross-frame segmentation result, extracting fragment-level statistical representation and confidence, forming a fragment-level suspicious area report, and outputting structured data for clinical auditing and index retrieval.

The invention discloses a vital sign monitoring method and system based on multi-modal perception and space-time restoration, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining an original video frame sequence of a target object, and constructing a multi-modal input tensor containing physical prior information based on the original video frame sequence; inputting the multi-modal input tensor into a feature extraction network, extracting a spatial-temporal feature and a stability feature through parallel processing, and fusing the spatial-temporal feature and the stability feature to obtain a basic feature; generating an attention weight based on motion information in the multi-modal input tensor, and performing dynamic weighting correction on the basic features by using the attention weight to obtain anti-interference features; and carrying out time sequence long-range dependence modeling on the anti-interference features, and carrying out logic repair and smoothing processing on the feature fragments which are subjected to motion interference currently by using historical feature fragments which are not interfered, so as to predict and obtain rPPG signals. According to the invention, industrial light environment adaptation and physical motion perception are realized, and meanwhile, long-range time sequence restoration can be realized when signals are lost.

The invention discloses a real-time pair practice training system and method based on a multi-modal large model, and relates to the technical field of large model pair practice. A multi-source sensor array comprising a visual capturing module, an auditory processing module and an action sensing module is deployed, and space-time synchronization of sensor data is realized by using a multi-modal alignment engine; generating a visual space-time vector reflecting a facial dynamic state and a limb time sequence rule; generating a fundamental frequency of the voice signal and an auditory feature vector of formant dynamic change; and generating a pressure sensitive feature vector fusing the hand motion trail and the contact force data. And carrying out weighted fusion on the feature vectors, calculating a comprehensive behavior evaluation index, comparing the comprehensive behavior evaluation index with a preset positive threshold value and a preset negative threshold value, and judging a user behavior state. And calculating a prediction error according to the actual state feedback of the user, and adjusting a positive threshold and a negative threshold in real time by using a dynamic threshold learning updating mechanism to realize adaptive optimization of an evaluation standard.

The invention discloses a long time sequence fusion target detection method for an automatic driving scene, and the method comprises the steps: carrying out the offset error prediction and alignment in an image space and a BEV feature grid, and achieving the high-precision time sequence modeling of a BEV space; the method specifically comprises the following steps: extracting multi-scale image features by adopting a backbone network; predicting an offset by using the historical image features and the current image features, and fusing the historical image features by using deformable convolution based on the predicted offset; generating BEV feature representation by using a view transformation module; designing a motion perception guide module, and resampling the historical BEV features; and fusing the historical BEV features and the current BEV features by using a gating circulation unit, and finally outputting BEV spatial expression with time sequence consistency. The method is suitable for a three-dimensional sensing system in an automatic driving scene, and the model performance is improved by using time sequence information.

The application provides a dynamic anti-aliasing rendering method and system based on edge detection and motion perception, and relates to the technical field of computer graphics rendering. The method comprises the following steps: performing 8-direction edge intensity detection on a pixel point in an original scene image by using a 3*3 sampling matrix and calculating an edge intensity value; calculating a motion factor and dynamically adjusting an edge detection threshold; if the edge intensity value of the pixel point is greater than the edge detection threshold, performing adaptive blur processing based on a 9-point weighted average algorithm to generate a blurred pixel value, calculating a mixing factor, mixing the original pixel value of the pixel point and the blurred pixel value to generate a final pixel value; if the edge intensity value of the pixel point is less than or equal to the edge detection threshold, keeping the original pixel value of the pixel point unchanged, and the final pixel value is the original pixel value, and finally outputting a final scene image by collecting the final pixel value, so that the anti-aliasing processing strategy can be adaptively adjusted according to the static and dynamic characteristics of the scene, the rendering performance is optimized while the image quality is ensured.

The application discloses an intelligent high-saturation snapshot method and system based on motion sensing and adaptive white balance delay, and belongs to the field of intelligent terminal image processing. The method is applied to an intelligent terminal comprising an inertial motion sensing module, an ambient light sensing module, a camera acquisition module and an end-side processing unit. Data is collected by the inertial motion sensing module, combined with dynamic threshold, trajectory continuity verification and false touch prevention condition identification effective swing. Based on ambient light data, a continuous mapping model is used to calculate the white balance delay time. Frame-level synchronous trigger shutter is adopted to collect images by using single camera continuous frames or double camera synchronous frames. After motion blur compensation and dynamic saturation enhancement, high-saturation clear images are output, and the output image metadata contains a special identification field. The application breaks through the industry technical prejudice, reduces the false trigger rate to below 0.5%, and improves the image forming rate to above 95%. Without adding new hardware, the application can be widely adapted to various intelligent terminals, solves the industry pain points of high false trigger, uncontrollable best imaging opportunity and swing blur of traditional snapshots, and reduces the difficulty of infringement and right protection evidence.

The invention discloses a lightweight inertial navigation positioning method and a lightweight inertial navigation positioning system based on a motion perception early-leaving mechanism, which are applied to mobile terminal positioning scenes such as pedestrian dead reckoning and the like. The method comprises the following steps: acquiring a three-axis acceleration and a three-axis angular velocity by using a built-in inertial measurement unit of a terminal, inputting preprocessed inertial data into an inertial navigation neural network, and sequentially outputting displacement estimation and covariance thereof at each layer by the network; and the early-backing judgment module takes the variance as a confidence index, compares the variance with a preset early-backing threshold value, terminates reasoning in advance at a current outlet and outputs a displacement result when the confidence meets the requirement, otherwise, continues to spread to a deeper layer, and realizes adaptive adjustment of the reasoning depth according to the motion complexity. According to the method, the reasoning efficiency is remarkably improved in low-dynamic scenes such as static scenes and slow scenes, the method has the advantages of being small in calculation overhead, low in energy consumption, suitable for real-time deployment on smart phones, wearable devices and augmented reality terminals and the like, and the endurance of a deep learning inertial navigation positioning system can be greatly improved.

This invention discloses a novel multi-sensory upper limb motion measurement system, comprising a soft felt garment (10), a motion measurement subsystem (12), and a sensory measurement subsystem (13) that respectively acquire multi-modal signals of motion and perception and transmit them to a multi-modal signal processing module (14) for amplification and filtering; a wireless signal concentrator (15) collects data from the multi-modal signal processing module (14); a computer client (16) integrates the multi-modal signals to construct an upper limb motion perception state information system, including an upper limb motion calculation model (11) and real-time multi-sensory information; the upper limb motion calculation model (11) calculates the data signals to form a motion model with real-time position coordinate data of each joint. This invention provides simple, accurate, and comprehensive real-time support for upper limb motion monitoring, which can effectively improve the accuracy and comprehensiveness of upper limb motion data acquisition and analysis in fields such as virtual reality interaction.

The application discloses a reference video object segmentation method based on semantic consistency and motion perception. The method comprises the following steps: 1. constructing semantic prompt information and video frame information of the reference video object segmentation dataset; 2. preprocessing the reference video object segmentation dataset; 3. establishing a reference video object segmentation model based on semantic consistency and motion perception: designing a double-branch decoupling strategy to decouple feature information at the semantic and visual levels, thereby extracting static and motion information of text description and visual features; designing a hierarchical motion perception module to capture and align motion information between different frames, analyze short-term and long-term motion information, thereby enabling the model to obtain the perception ability of long-term motion patterns; designing a semantic consistency module to align semantic description and video features, thereby improving the correctness of target selection and the integrity of the mask, and avoiding false detection of negative samples; designing a perception dynamic fusion mechanism to embed text information into the visual feature space, enabling the visual features to obtain text semantic information, thereby enhancing the cross-modal understanding ability of the model; 4. constructing a loss function, updating the model parameters, setting the training parameters, training, and obtaining the best weight; 5. detecting the test set image based on the best weight to obtain the final segmentation result. The application effectively decouples static and dynamic information, enhances the perception ability of object motion patterns, and improves the segmentation performance.

This invention discloses an adaptive temperature control method based on motion state perception, comprising the following steps: Step 1, the motion sensing module senses the user's motion intensity and transmits the motion intensity data to the main control module; Step 2, the main control module determines the motion intensity level based on the motion intensity data, matches a target heating level based on the motion intensity level, and sends a target heating level command to the heating module; Step 3, the heating module adjusts the temperature to the target temperature range according to the target heating level command. An adaptive temperature control system based on motion state perception is also disclosed, comprising a motion sensing module, a main control module, and a heating module; the temperature control system of this invention can automatically match the heating level temperature according to the motion intensity, so that the user does not need to manually adjust the temperature during exercise.

The invention discloses a gait recognition method based on local appearance and micro-motion perception, and belongs to the technical field of computer vision and biological feature recognition. The method comprises the following steps: firstly, preprocessing and aligning an input gait contour sequence; extracting spatial features by using a convolutional neural network, and performing non-uniform part division according to human body structure priori knowledge so as to focus on leg equal-height discriminant regions; further, an independent micro-motion sensing module is designed for each component, and unique short-distance space-time dynamic characteristics of each component are captured through a double-branch time sequence attention mechanism; and finally fusing the enhanced features of all the components to complete identity recognition. By emphasizing the local appearance difference and the micro-motion mode, the robustness and stability of the gait recognition model in a real complex scene are remarkably improved, and the method has important security monitoring and identity authentication application value.