19results about How to "Accurate understanding" patented technology

This invention relates to the field of named entity recognition technology, and discloses a multimodal named entity recognition method and apparatus based on deep interaction of text and image features. The method includes: acquiring original text and an original image; extracting multimodal features from the original text and the original image respectively to obtain a text feature sequence and an image feature sequence fused with multi-scale features; performing cross-modal deep interaction fusion on the text feature sequence and the image feature sequence to obtain an interacted multimodal feature sequence; performing context-aware adaptive gating fusion on the interacted multimodal feature sequence to obtain a fused feature sequence; and performing conditional random field sequence decoding on the fused feature sequence to obtain an entity label sequence. This invention achieves deep bidirectional interaction and dynamic adaptive fusion of multimodal information, effectively solving the text ambiguity problem and improving the accuracy of named entity recognition.

The utility model belongs to door lock technical field, especially relate to a door lock, including inner mounting seat and outer mounting seat, and the inner mounting seat is rotatably installed with inner door handle and outer door handle on outer mounting seat respectively, and the inner door handle and outer door handle are linked through square shaft, still include: observation window, indicating piece, slide and install in the inner side of inner mounting seat, with observation window adjacent, can move back and forth between the anti -lock position and the unlocking position, have the first position of unlocking indication when unlocking, have the second position of anti -lock indication when anti -lock, wherein, the inner door handle head one end of inner door handle passes through inner mounting seat and is linked with indicating piece through transmission structure to realize indicating piece back and forth movement. The utility model has the advantages that: the setting of indicating piece cooperation transmission structure can realize indicating piece and door handle open anti -lock linkage, and the setting of observation window can understand door lock state quickly, accurately, unambiguously, effectively prevent user from being mistaken and think that the door has been locked.

This invention discloses a robot autonomous configuration service system built around a large-model agent framework, including a robot autonomous task system, a large-model agent framework, an application system, and an evaluation system. The large-model agent framework receives user instructions, parses task requirements, and generates task execution strategies. The robot autonomous task system drives the robot to complete physical actions based on the task execution strategies and introduces an agent extension mechanism to adapt the large-model agent framework to different types of robots and task environments. The application system is used for data transmission, storage, and visual interactive operation. The evaluation system collects task execution data in real time and outputs quantitative evaluation results to optimize the execution strategy. This system effectively reduces the threshold of human-machine interaction and the difficulty of understanding two-way intentions, significantly improving the robot's task adaptation flexibility and execution reliability, and is suitable for various robot autonomous operation needs in multiple scenarios such as inspection, delivery, and search.

The application discloses a kind of intelligent physiotherapy control systems of physiotherapy equipment, belong to the technical field of physiotherapy instrument control system, the system includes integrated in physiotherapy comb body physiotherapy execution module, for real-time acquisition user physiological parameter physiological parameter sensing module, and with the communication connection of both intelligent control unit.State analysis module is included in intelligent control unit, for the current physiological state of user is analyzed according to the physiological parameter of acquisition;Strategy generation module is used to match or generate target physiotherapy strategy from preset physiotherapy mode library based on current physiological state;And drive control module is used to drive physiotherapy execution module to work according to target physiotherapy strategy.The application is fused by multi-source physiological signal, and the state of user is analyzed, and using the intelligent decision mechanism including reinforcement learning, generative algorithm, dynamically generates or optimizes physiotherapy strategy, realizes the real-time self-adaptation and high personalization of physiotherapy process, improves the safety and effectiveness of physiotherapy.

The application relates to the technical field of artificial intelligence and industrial simulation, in particular to an evolvable industrial simulation question and answer intelligent agent system and a simulation question and answer method thereof. The evolvable industrial simulation question and answer intelligent agent system comprises a knowledge base module, a tool calling module, an intelligent agent core module, a user interaction module and a log and audit module; the knowledge base module comprises a multi-element knowledge collection unit, a document loading and cutting unit, a vectorization embedding and storage unit, a semantic retrieval unit, a knowledge evolution unit, a rule base and a case base; the application has the beneficial effects that (1) knowledge is dynamically evolved and updated autonomously and timely; (2) the system has the capability of autonomously refining knowledge from simulation results; (3) tool calling is intelligentized, and the system realizes "question and answer as calculation"; (4) the system has strong capabilities of complex problem disassembly and iterative solution; and (5) the system has excellent adaptability and expandability.

This application discloses a vehicle driving path planning method and electronic device, relating to the field of vehicle control technology. The method includes: acquiring image data of the vehicle driving environment from at least two angles; processing the image data using a preset image processing strategy to obtain the location information of pothole regions. The preset image processing strategy processes the image data in the following order: image closing operation, image opening operation, edge detection, color segmentation, and contour detection; calculating the depth information of the pothole regions based on the image data and the location information of the pothole regions; and determining adjustment parameters for the vehicle driving path based on the location information of the pothole regions, the current vehicle speed, and the depth information of the pothole regions, thereby planning the vehicle driving path based on the adjustment parameters. This application's method achieves low-cost, high-real-time, and high-precision intelligent driving path planning in complex road environments, significantly improving vehicle driving safety and passenger comfort.

The utility model provides a rotatable spreadlight lens street lamp module, which comprises a light source assembly, a plurality of light-emitting units and a plurality of light-emitting units, the two opposite sides of the light source assembly are provided with a movable heat dissipation assembly and a movable polarization assembly respectively. The polarization assembly is provided with an indication assembly; the polarization assembly is located on a light emitting path of the one or more light emitting units; the polarization assembly is rotationally connected with the heat dissipation assembly; the polarized light assembly moves relative to the light source assembly after being subjected to external force, the polarized light assembly drives the indicating assembly located on the polarized light assembly to correspondingly move so as to indicate the change of the polarized light direction, the irradiation range and angle of light rays are accurately adjusted according to the actual illumination requirement, light ray waste is avoided, the light rays can more evenly cover the illumination area, and the illumination effect is improved. The problems that part of the area is too bright and part of the area is too dark due to the fixed illumination direction of a traditional street lamp are solved, so that a more comfortable illumination environment is provided, and the illumination efficiency is improved.

This invention discloses a method for fitting a 3D human body model based on 2D contour interaction and semantic segmentation. The method first parses a user-drawn 2D stick figure skeleton using a JPN module and drives an SMSLX model to generate an initial 3D human body. Then, the initial model contour is rendered from multiple fixed perspectives, generating corresponding bounding voxel spaces and point cloud projections. A prior network combining UDF and positional information is introduced to guide the training of the multi-view semantic segmentation network module UPS, achieving accurate segmentation of each part of the contour image. Based on the segmentation results and voxel point cloud projections, surface constraint point clouds for each body part are calculated. Finally, using the constraint point clouds as targets, the SMSLX model parameters are optimized for each part using a CFM module, ultimately generating a 3D human body shape that matches the user's editing intent. This invention allows users to indirectly adjust complex 3D models through intuitive 2D contour editing, significantly improving interaction efficiency and avoiding direct manipulation of complex 3D data.

Provided in one scenario are an information processing method, device, equipment, and storage medium. The method comprises: in response to a preset event trigger, obtaining interface information of a graphical interaction interface, wherein the interface information comprises a first interaction object and an interaction focus, and the interaction focus represents an interaction position in the graphical interaction interface; based on spatial distribution information of the first interaction object relative to the interaction focus, obtaining context information of request information; and based on the context information and the request information, generating response information corresponding to the request information. The technical solution in one scenario accurately understands the request intention corresponding to a fuzzy instruction, solves the problem in the related art that the response information is inaccurate due to an inability to accurately understand the request intention, improves the accuracy of the response information, and improves the user experience.

The application relates to the field of water quality monitoring and discloses an intelligent water quality monitoring method based on multi-dimensional spectral technology, which comprises the following steps: synchronously collecting the absorption spectrum and the three-dimensional fluorescence spectrum of a water sample to be measured; correcting the three-dimensional fluorescence spectrum by using the absorption spectrum to obtain the corrected three-dimensional fluorescence spectrum; constructing a double-path deep neural network to extract absorption spectrum features and three-dimensional fluorescence spectrum features; inputting the two kinds of spectrum features into a self-attention module for internal correlation weighting; generating a fusion coefficient through a gate fusion unit, and dynamically fusing the two kinds of weighted spectrum features according to the fusion coefficient to obtain a fusion spectrum feature vector; and performing water quality parameter analysis based on the fusion spectrum feature vector. The application eliminates the fluorescence distortion of high-concentration samples through internal filter effect correction, deeply extracts heterogeneous spectrum features through a double-path convolutional neural network, and realizes adaptive weighting of the two kinds of spectrum information through dynamic gate fusion.

The application discloses a question and answer method based on video understanding, related equipment and a computer program product, and relates to the technical field of question and answer. The application extracts global scene features and role emotion features of each video frame in a target video to obtain global scene feature extraction results and role emotion feature extraction results of each video frame; fuses the global scene feature extraction results and the role emotion feature extraction results belonging to the same video frame to obtain single-frame fusion features of each video frame; fuses the single-frame fusion features of all video frames according to a dependency relationship to obtain video content features of the target video; and calls a configured large language model to generate reply content for a target question according to semantic information of the video content features of the target video. The application helps the large language model to fully understand the video content, realizes comprehensive and accurate analysis of role emotions, character relationships and other content in the video, and further improves the accuracy of replying to questions.

The application provides an intelligent video analysis method based on large model scheduling, which comprises the following steps: video data acquisition and preprocessing, video content feature extraction and analysis, large model dynamic scheduling and task allocation, video analysis based on the large model and result generation, result fusion and post-processing, online incremental learning and model updating, and the like. The method can efficiently and automatically analyze, has the advantages of quantifying behavior reliability, flexible adaptation to scenes, and support for complex decisions. Meanwhile, multi-dimensional information fusion improves the accuracy of behavior description and the practical application value of the analysis method.

This invention proposes a cross-modal semantic perception-based image and text retrieval method and system, relating to the field of image and text retrieval technology. The specific scheme includes: acquiring images and text to be matched; inputting the images and text into a trained cross-modal semantic perception network to obtain the similarity between the images and text; matching the images and text based on the similarity to obtain the image and text retrieval results; wherein the cross-modal semantic perception network extracts text features and image features from the text and images respectively, and performs binary relation reasoning and multi-variable relation reasoning on the fused text features and image features respectively to obtain the final image features, and calculates the similarity between the final image features and text features; this invention, based on a multi-level information dynamic fusion module and a relation perception module, deeply explores the semantic association between vision and language, and provides a more accurate estimation of the similarity between images and text.

This invention discloses a system and method for predicting and evaluating the dynamic performance of fabrics and garments based on multi-dimensional sensing. The system includes: a control module for generating composite motion commands to drive an actuator to apply a composite dynamic load, including axial tension, radial torsion, and normal impact, to a fabric sample; a multi-modal sensing module for simultaneously acquiring image signals, acoustic signals, and mechanical signals from the fabric sample; a feature extraction module for preprocessing the multi-modal signals and extracting quantized feature vectors; an AI prediction module with a built-in deep learning model for inputting the quantized feature vectors and outputting the predicted dynamic performance of the garment; and an output module for visualizing the prediction results. This invention can simulate the complex mechanical states exerted on fabrics by real human activity, achieve objective quantification of subjective indicators such as feel and style, and accurately predict garment performance through a deep learning model. It can be widely applied in fabric research and development, garment design, and quality control.

This invention relates to the field of equipment monitoring and fault diagnosis technology, and discloses a method and system for equipment fault diagnosis based on knowledge graphs and large language models. The method includes: acquiring multiple types of data sources for the target equipment, preprocessing and extracting features to obtain a structured dataset; extracting knowledge triples from the dataset based on a natural language processing model and constructing an equipment health status ontology; constructing evaluation functions for each fault phenomenon to quantify the adaptability of monitoring methods, forming fault phenomenon-monitoring method mapping triples; merging the knowledge and mapping triples and storing them in a graph database to form a knowledge graph and optimizing it; using a pre-trained semantic embedding model to encode triples and construct a retrieval index library, receiving user query codes and retrieving and constructing a semantic subgraph; combining the semantic subgraph and the query into prompt words and inputting them into a large language model to generate recommendation results, updating the knowledge graph or index library based on user feedback. This invention improves the accuracy and adaptability of fault diagnosis and reduces operation and maintenance costs.

This invention discloses a traffic flow prediction method based on graph structure feedback and spatial prior. The method includes: acquiring target area data; embedding features into historical traffic flow time series data to generate an initial spatiotemporal feature matrix; inputting the initial spatiotemporal feature matrix into a temporal self-attention module to calculate the global dependencies between different time steps in the time dimension and outputting time-encoded features; constructing a bidirectional adjacency matrix through graph structure spatial modeling to output preliminary spatial features; performing spatial linear feedback processing on the preliminary spatial features and time-encoded features to obtain spatial linear feedback features; combining the spatial linear feedback features with spatial prior constraints and aggregating local neighborhood information through graph convolution operations with shared parameters to obtain spatial prior constraint features; and outputting traffic flow prediction values ​​for future preset time steps through a linear mapping layer based on the spatial prior constraint features. This invention achieves high-precision traffic flow prediction.

The invention relates to the technical field of artificial intelligence and game development, in particular to a game engine-oriented AI Agent tool calling method and virtual equipment, and the method comprises the following steps: receiving a natural language tool calling request sent by an AI Agent through a standard protocol; and tool matching and parameter analysis are carried out based on a pre-established tool definition system adopting natural language semantic description and multi-dimensional tags. The general request parameters are automatically converted into strong data types specific to the game engine, and all converted instructions are ensured to be safely executed in a main thread of the game engine through a task queue scheduling mechanism of the main thread. In addition, the method also supports the quick retrieval of the tool through the reverse index and the hot update of the tool based on file monitoring. According to the method, the problem that a large language model is difficult to directly understand and call a complex game engine API is effectively solved, and the technical threshold and the operation risk of participation of AI in game development are remarkably reduced.

The present disclosure provides a model training method, a rendering image generation method and a device thereof. The method comprises: obtaining a target training sample; the target training sample comprises: an initial reference image, initial noise corresponding to each view angle in N view angles, and M initial material channel images determined based on an initial material of an initial three-dimensional model in each view angle in N view angles; performing fusion processing on the M initial material channel images in each view angle and the initial noise corresponding to each view angle to obtain an initial noisy image of each view angle; inputting the initial reference image and the initial noisy image of each view angle into a to-be-trained generative model to obtain M estimated material channel images in each view angle and estimated noise corresponding to each view angle; obtaining a target loss value based on the estimated material channel image and / or the estimated noise; and performing model training on the to-be-trained generative model using the target loss value to obtain a target generative model.

The application provides a juicer human-computer interaction control method, system, device and medium. The hand region point cloud of a target user is separated from the hand image data of the target user when the target user controls the juicer by gestures. The gesture features of the target user when interacting with the juicer are determined based on the hand region point cloud. The context features of the historical interaction tasks of the target user and the juicer are determined. Then, adaptive gesture recognition is performed based on the gesture features and the context features to obtain the interaction intention vector of the target user and the juicer. The compound operation instruction of the target user to the juicer is determined according to the interaction intention vector and the current working state of the juicer. The working mode of the juicer is dynamically adjusted by converting the compound operation instruction into the machine control instruction of the juicer. The scheme of the application can realize dynamic context perception compound operation in the juicer gesture interaction control in a complex kitchen environment.