54 results about "Edge model" patented technology

The invention discloses an image processing method and device, a medium and a computer program product, and relates to the technical field of semiconductor manufacturing. The image processing method comprises the following steps: acquiring target image samples of a plurality of wafer samples through a scanning electron microscope, and obtaining a first contour image label of each target image sample; based on the target image sample, a first contour image result is generated through a generator of a preset image edge model, iteration training is carried out on the preset image edge model through pixel difference loss, adversarial loss and edge supervision loss until a convergence condition is met, and a target image edge model is obtained; the generator of the target image edge model can output an edge contour image based on the target image of the target wafer. Through multi-dimensional constraints of pixel difference loss, edge supervision loss and adversarial loss, the preset image edge model is trained in the direction of generating an actual edge contour, and the accuracy of the edge contour image is improved.

The invention discloses a low-altitude aircraft intelligent identification system based on an edge network, and belongs to the technical field of intelligent identification. Comprising a data acquisition module used for acquiring time sequence sensor data; the knowledge base construction module is used for constructing a physical rule knowledge base; the penalty value generation module is used for generating a physical violation penalty value; the model generation module is used for obtaining a physical rule embedded model; and the result output module is used for outputting an aircraft intelligent identification result conforming to the physical motion law. According to the method, the physical rule knowledge base is introduced into the edge network, and the lightweight model is trained by using the dual-target collaborative loss function, so that the effect of still outputting a physically consistent aircraft recognition result under the condition of limited computing power is achieved; the problem that in the prior art, a marginal small model generates jumping, over-limit or abnormal tracks due to lack of motion law constraints is solved.

An edge device provides an agentic framework to manage apps and artificial intelligence (AI) models used by the apps. An app and app metadata are downloaded from a cloud of servers to the device. The app metadata describes requirements of the app for AI models to be used by the app. The agentic framework performs a search in an on-device database that stores the app metadata and model metadata of edge models installed on the device. The search is performed to determine whether one of the edge models satisfies the requirements of the app. Following the search, the agentic framework sets a given edge model already installed on the device as a target model of the app, where the target model satisfies the requirements of the app. The agentic framework then directs the app to use the target model in response to a request for service.

The invention discloses a low-altitude unmanned aerial vehicle autonomous cruise method and system based on cloud edge basic model collaboration, and the method comprises the following steps: S1, collecting an environment RGB image through an airborne monocular camera of an unmanned aerial vehicle, and carrying out the preprocessing of the image, and obtaining a preprocessed gray image; s2, a neural scheduler based on deep reinforcement learning generates a scheduling instruction according to the environment data and the network state reasoned by the navigation model at the previous moment; s3, generating a preliminary flight instruction; s4, generating an optimized flight instruction; s5, generating a structured flight instruction; and S6, the unmanned aerial vehicle executes the preliminary flight instruction or the optimized flight instruction or the structured flight instruction. According to the invention, through dynamic on-demand cooperation and intelligent scheduling of the end-edge-cloud three-level model, resource consumption and delay are substantially reduced, and high-robustness and high-safety autonomous cruise of the unmanned aerial vehicle in a complex open environment is realized.

An agentic framework is provided on a device that interacts with a cloud and runs apps and artificial intelligence (AI) models. An app and app metadata are downloaded from the cloud to the device. The app metadata describes requirements of the app for AI models to be used by the app. A search is performed in a database on the device that stores the app metadata and model metadata of edge models that are installed on the device to identify a target model among the edge models that satisfies the requirements. The target model is downloaded from a collection of downloadable models in the cloud in response to a determination that the target model is not already installed on the device.

This invention discloses a large-scale visual-language model collaborative inference method for cloud-edge-device systems. Its features include: deploying a high-performance cloud-based LVLM model and a cloud-based archive in the cloud; deploying an edge-based LVLM model and an edge-based archive at the edge, wherein the edge LVLM model incorporates a retrieval-enhanced generation algorithm, and the edge archive and the cloud archive can dynamically interact in terms of knowledge; a task scheduler evaluates the complexity of received query tasks based on a task allocation algorithm, and allocates the query tasks to the edge for local processing or to the cloud for processing based on the complexity evaluation results. This invention can improve the inference efficiency within cloud-edge-device systems while maintaining high accuracy, effectively solving the technical problem of low inference efficiency caused by limited computing resources, intermittent communication windows, and bandwidth limitations of edge devices.

The invention discloses an overhead line system dropper defect identification method based on lightweight YOLOv12s. The method comprises the following steps: acquiring an overhead line system dropper image to be identified; and inputting a to-be-identified catenary dropper image into the pre-trained YOLOv12s-Lite, and outputting a catenary dropper defect detection result. The YOLOv12s-Lite is a lightweight network obtained by carrying out lightweight improvement on the original YOLOv12s; the lightweight improvement comprises the following steps of: replacing a backbone network in the original YOLOv12s with the OfficientNetV2s; an efficient multi-scale attention module EMAttention is introduced into a neck network in the original YOLOv12s. According to the method, the model complexity is remarkably reduced while the model defect identification precision based on the YOLOv12s algorithm is maintained, lightweight detection of catenary dropper defect identification is realized, and technical reference is provided for edge end model deployment.

When retraining an AI model used for object detection processing on edge devices to add identifiable classes, the aim is to improve the class identification accuracy of the AI ​​model while reducing the workload on the user. [Solution] The information processing device comprises an annotation processing unit that performs annotation processing on an input image using a large-scale AI model that performs object detection processing and is configured to output region information indicating an object detection region even for objects that cannot be classified as a class; a reception processing unit that accepts corrections from the user regarding the annotation results from the annotation processing unit; a large-scale model retraining processing unit that performs retraining processing on the large-scale AI model using the annotation information corrected by the user; and an edge model retraining processing unit that performs retraining processing on an edge model, which is an AI model used in an edge device that performs object detection processing on captured images, by knowledge distillation using the retrained large-scale AI model as the training model.

The invention provides precast pile upper surface modeling equipment. The precast pile upper surface modeling equipment comprises a rack, one or more trowelling mechanisms and one or more blunt edge modeling mechanisms. The trowelling mechanism is mounted on the rack; the blunt edge modeling mechanism is mounted on the rack, and in the working advancing direction, the blunt edge modeling mechanism is located behind the trowelling mechanism; the blunt edge modeling mechanism comprises at least one pair of blunt edge modeling assemblies with adjustable combination width in the transverse direction, and the two blunt edge modeling parts of the same pair of blunt edge modeling assemblies are oppositely arranged. A longitudinal walking mechanism is arranged at the bottom of the machine frame to drive the machine frame to move longitudinally, so that a trowelling mechanism and a blunt edge modeling mechanism on the machine frame are driven to move longitudinally, and in the moving process, the trowelling mechanism and the blunt edge modeling mechanism conduct trowelling operation and blunt edge modeling operation on the upper surface of the precast pile. Due to the fact that the combination width of the blunt-edge modeling assemblies in the same pair is adjustable in the transverse direction, the blunt-edge modeling mechanisms in the same pair can adapt to the width changes of the thick cavity section and the thin cavity section of the long mold cavity of the modeling mold in the longitudinal direction.

The invention discloses a traffic video edge analysis system based on Yolov5-Edge and Prewitt algorithms, and the system achieves the real-time collection, intelligent recognition and encrypted transmission of a vehicle-mounted video through the construction of a layered architecture of an object-side equipment layer, an edge calculation layer and a cloud service layer. According to the system, a lightweight Yolov5-Edge model and a Prewitt edge detection algorithm are fused in edge equipment, multi-channel video parallel processing is realized under finite computing power, and the video inspection efficiency is effectively improved; the frame rate and the number of processing paths are dynamically adjusted through the resource monitoring and load control module, and the real-time performance and stability of AI reasoning are guaranteed; a data management and encryption mechanism is adopted, only abnormal data are stored at an edge end and clouded through RSA and AES encryption modes, and security and compliance of sensitive behavior information in the transmission and storage process are ensured. The system can identify and classify abnormal behaviors of drivers and passengers without depending on a high-performance server.

The invention discloses a coal pile weight calculation and dynamic weighing method based on an intelligent sensor. The method comprises the following steps: sparsely deploying intelligent sensing nodes at the bottom of a coal pile and acquiring initial data; constructing a double-loop model of an edge online sequence extreme learning machine and a cloud physical information neural network; the edge model processes the real-time data stream and outputs a fast weight estimation value and a virtual observation residual error; the cloud model reconstructs full-field stress distribution by using the data and calculates a virtual sensing value; adjusting an edge model learning target through virtual value feedback between double rings; carrying out integration on the reconstructed stress field to obtain accurate total mass and local mass, and carrying out visualization; and monitoring residual features to trigger early warning and adaptively adjust physical parameters of the model. According to the method, high-precision, continuous, dynamic and self-adaptive direct weighing of the mass of the coal pile is realized under the condition of low-cost sparse sensing deployment, and the contradiction among dynamics, precision and engineering feasibility of a traditional method is effectively solved.

The application discloses a data blood relationship analysis method and device and a computer readable storage medium. The method comprises the following steps: obtaining job blood relationship data based on a directed graph corresponding to a scheduling task in a scheduling system, and obtaining a first application program identifier based on an execution log corresponding to the scheduling task; when a structured query language (SQL) statement is detected, obtaining field blood relationship data corresponding to the SQL statement and a second application program identifier corresponding to the field blood relationship data; determining a job blood relationship subgraph based on the job blood relationship data and a Neo4j point-edge model, and determining a table field blood relationship subgraph based on the Neo4j point-edge model and the field blood relationship data; and connecting the job blood relationship subgraph and the table field blood relationship subgraph based on the first application program identifier and the second application program identifier. The application can obtain blood relationship data of different types of data systems, adapt the data blood relationship analysis method to different types of data systems, and improve the expansibility of the data blood relationship analysis method.

According to the self-adaptive control method for the maintenance robot based on curvature edge feature reconstruction and cascade active-disturbance-rejection control, multi-source sensing is constructed, and image information is collected through sensors such as a high-resolution RGB camera and a depth camera. A real-time adaptive Canny edge detection method is adopted to process an RGB image so as to extract an edge, point cloud data of a depth camera is converted to a robot-based coordinate system through point cloud coordinate conversion, registration is completed, and meanwhile, the principal curvature and the normal vector of each point are estimated. And realizing automatic generation and dynamic adjustment of a spraying path based on the edge model, including spray gun attitude calculation and distance speed flow control. The system adopts a three-stage cascade controller architecture, an outer ring generates an expected position instruction, a middle ring performs position compensation to output a speed instruction, and an inner ring realizes speed and current regulation and integrates a hardware protection mechanism. The invention aims to improve the spraying precision, efficiency and coating quality of the robot and realize high-quality spraying operation on the surface of a complex workpiece.

The application discloses an emulsified asphalt demulsification speed determination method, relates to the online detection and intelligent monitoring technical field of road engineering materials, and remarkably improves the practicability and reliability of emulsified asphalt demulsification speed determination through integration of a lightweight time sequence prediction model and a self-adaptive processing mechanism; in terms of portability, an edge model based on a gated recurrent unit structure is adopted, calculation resource dependence is reduced, the method can be run on a handheld device, the problem of clumsiness of traditional laboratory equipment is avoided, and construction personnel can be conveniently and rapidly deployed on site; real-time performance is enhanced, a self-adaptive sliding window dynamically adjusts window length and step length according to data stability, the inference frequency is ensured to match the demulsification process, an online learning mechanism is combined, the model can instantaneously respond to environmental changes such as temperature fluctuations, and a demulsification speed estimation value can be outputted without obvious delay, so that paving and compaction time is guided, and construction interruption is reduced.

The invention discloses a micro-morphology measuring and modeling method for a cutting edge of a light shielding ring of a light shielding cover, and belongs to the technical field of precision measurement and three-dimensional modeling. The method comprises the following steps: firstly, acquiring three-dimensional point cloud data of a cutting edge of a light blocking ring through a laser scanning confocal microscope; carrying out denoising processing on the data by adopting wavelet threshold denoising and five-point secondary smoothing filtering; the data precision is improved by combining clamp installation optimization and inclination correction technologies; data pruning is achieved through curvature analysis, and a cutting edge feature area is extracted; missing data are repaired by using cubic B-spline interpolation; and finally, a high-precision cutting edge model is established through trapezoid fitting, arc fitting and NURBS curved surface reconstruction methods. According to the method, the problem that the micro-morphology representation accuracy of the cutting edge of the light shielding ring of the light shielding cover is insufficient in the prior art is solved, the accuracy and efficiency of the morphology representation of the cutting edge are remarkably improved, and technical support is provided for the surface morphology design of a high-precision light shielding cover.

The invention discloses an application construction and management system based on session type interaction and an implementation method thereof, belongs to the technical field of application construction management, and aims to unify multiple expressions of the same intention through small-sample causal association generation of a large language model and digest ambiguity of the same sentence through scene causal dependence intensity calculation. Semantic enhancement is realized by fusing semantic vectors of text and causal atlas features, richer intention and scene associated information can be contained, and accurate semantic input is provided for subsequent steps; diversified samples are generated through digital twinning, so that the recognition accuracy of an edge end model on unseen overexpression can be effectively improved; by deploying a lightweight large language model, sub-second response is realized, and the problem of cloud transmission delay is solved; by implementing a federal incremental learning and differential privacy synchronization mechanism, the user privacy is protected, and the real-time performance of a model semantic rule is ensured.

This invention relates to the field of edge-cloud collaborative robot technology, and discloses a method, system, and storage medium for constructing edge-cloud collaborative robot memory based on semantic consistency verification. Addressing the problem of difficulty in evaluating the semantic fidelity of edge models under limited bandwidth, which easily leads to long-term memory fragmentation, this invention generates a question-and-answer database from the cloud and distributes it. The edge device answers based solely on its own generated initial subtitles under image masking conditions, and the semantic consistency score is calculated by comparing the answers. Simultaneously, a dynamic trigger threshold is calculated by combining a bandwidth debt queue and real-time channel gain. When the score falls below this threshold, the neighboring sequence of abnormal frames is extracted and sent back to the cloud for batch reconstruction, overwriting the original memory of the edge device. This invention achieves online unsupervised quantization of edge device cognition under constrained networks, adaptively eliminates memory gaps, and achieves a globally optimal balance between memory quality and communication overhead, enabling the construction of high-quality contextual memories for robots in complex task scenarios.

The application discloses an edge inference driven intelligent auxiliary decision system for medical image recognition, relates to the technical field of medical image recognition auxiliary decision, and comprises a multi-modal adaptation module, an edge inference module, a cloud collaborative module, a knowledge distillation module, a resource scheduling module and a decision output module; the multi-modal adaptation module is used for extracting modal features of CT, MRI and X-ray images through a differentiable neural architecture search method to generate a lightweight edge model; the edge inference module is used for performing real-time inference on the medical images by using the lightweight edge model to output preliminary diagnosis results and confidence scores; the cloud collaborative module is used for setting a threshold value; when the confidence score is lower than the threshold value or the lesion area is smaller than a preset value, the medical images are transmitted to the cloud for deep analysis to output accurate diagnosis results and a lesion segmentation mask; and the knowledge distillation module is used for taking the lesion segmentation mask as a spatial constraint.

The invention relates to an edge-end model acquisition method and device based on multi-protocol data and computer equipment. The method comprises the following steps: loading a protocol feature library corresponding to a production system, determining a target transmission protocol from the protocol feature library based on a to-be-collected data type and network state data, collecting multi-protocol data of the production system in parallel based on the target transmission protocol and a pre-configured data access interface, carrying out data preprocessing on the multi-protocol data, and sending the pre-processed multi-protocol data to the production system; the method comprises the steps of obtaining standardized data, performing classification labeling on the standardized data to obtain a labeled sample set, dividing the labeled sample set according to a business scene to obtain a training sample set, and training a candidate business model based on a business task and the training sample set to obtain a trained target business model; and converting the target service model into a compatible format of the edge end device to obtain an edge end model, and deploying the edge end model to the edge end device. By adopting the method, the stability of the side end AI equipment can be improved.

The invention discloses a method for realizing an agent framework on a device and an agent framework providing device, and relates to an edge device for providing an agent framework to manage an application program and an artificial intelligence (AI) model used by the application program. Application and application metadata are downloaded from the server cloud to the device. The application metadata describes requirements of the application on the AI model for use by the application. A proxy framework performs a search in an in-device database storing application metadata and model metadata of an edge model installed on the device. The search is to determine whether one of the edge models meets the requirements of the application. After the search, the proxy framework sets a given edge model installed on the device as a target model of the application, where the target model meets the requirements of the application. The proxy framework then instructs the application to use the target model upon receipt of the service request.

The application provides a lithium ion battery pulse charging optimization method and system based on LSTM, and belongs to the technical field of battery management systems. The method comprises the following steps: obtaining time sequence slice caching of a battery multi-dimensional physical state and a global health feature; dynamically generating a pulse instruction set to be evaluated, using a broadcast mechanism to perform feature dimension alignment and splicing, inputting a pre-trained model to perform parallel prediction of an end voltage; combining a physical extreme value and a model prediction error to calculate a dynamic confidence penalty term, performing adaptive funnel screening to generate a safe candidate set; based on the set, issuing an optimal instruction, or triggering a multi-level abnormal exit state machine containing a hot start detection when the set is empty; obtaining actual execution feedback, filtering through a sample cleaning gateway, and fine-tuning a full connection layer under feature layer freezing. The application takes into account the extremely fast charging efficiency and absolute safety of lithium precipitation prevention, effectively solves the high-frequency time sequence fault and long-term aging misalignment problem of edge models under resource constraints, and significantly improves the control robustness.

This application provides a prediction method and related equipment for a lightweight edge model of the Industrial Internet, relating to the field of artificial intelligence. The method includes: acquiring state data of industrial equipment; tensorizing the state data to obtain a state tensor; generating an encoded result tensor of the state tensor through an encoding model, the encoding model including multiple encoders, each encoder including a self-attention layer and a position feedforward layer. In the self-attention layer, the input tensor is mapped to a query tensor, a key tensor, and a value tensor through query mapping tensors, key mapping tensors, and value mapping tensors. The output tensor of the self-attention layer is generated based on the product of the nonlinear mapping tensors of the query tensor and the key tensor, and the value tensor; and a second multilayer perceptron is used to predict the state analysis results of the industrial equipment based on the encoded result tensor. This application reduces computational complexity and improves state analysis efficiency by replacing similarity with the product of the nonlinear mapping tensors of the query tensor and the key tensor.

The invention relates to a system (10) for detecting at least one mobile head-mounted display (HMD) (11), comprising at least one vehicle (12), wherein a control unit (16) of the head-mounted display (11) is designed to perform an inside-out tracking based on a six degrees of freedom (6DOF) algorithm by means of at least one camera image recorded by at least one camera (15) and to determine a translation of the at least one mobile head-mounted display (11) based on the six degrees of freedom algorithm, wherein an application (17) is additionally designed to create a model of the edges of the interior of the vehicle (12) and to provide an automated computer vision-based six degrees of freedom tracking of the at least one mobile head-mounted display (11). Furthermore, the invention relates to a method for determining a 6DOF tracking of a mobile head-mounted display (11) with the system (10) described above.

This invention discloses an automated SD-WAN zero-contact deployment system based on an edge AI model, belonging to the field of information networks. To address the problems in existing technologies, such as single-mode system authentication, susceptibility to forgery and replay attacks, insufficient AI decision reliability, lack of a safety net, coarse configuration verification, coarse rollback granularity, and susceptibility to model training contamination, lack of gradient anomaly detection, this invention effectively resists device forgery and replay attacks by using multimodal fusion verification of hardware fingerprints, visual recognition, and digital certificates, combined with timestamps, random number anti-replay, and certificate chain verification. Furthermore, it introduces a comprehensive confidence assessment; when the confidence level falls below a threshold, it automatically switches to a safety net rule generation mode, ensuring that devices can still obtain the minimum operational configuration in unknown scenarios and preventing network paralysis due to AI misjudgment.

The invention provides an industrial internet lightweight edge model prediction method and related equipment, and relates to the field of artificial intelligence. The method comprises the following steps: acquiring state data of industrial equipment; performing tensorization on the state data to obtain a state tensor; generating an encoding result tensor of the state tensor by an encoding model, the encoding model comprising a plurality of encoders, the encoders comprising a self-attention layer in which the input tensor is mapped into a query tensor, a key tensor, and a value tensor by a query mapping tensor, a key mapping tensor, and a value mapping tensor, and a position feed-forward layer in which the input tensor is mapped into the query tensor, the key tensor, and the value tensor; generating an output tensor of the self-attention layer based on the product of the query tensor and the nonlinear mapping tensor of the key tensor and the value tensor; and predicting a state analysis result of the industrial equipment based on the coding result tensor through the second multi-layer perceptron. According to the method, the product of the nonlinear mapping tensor of the query tensor and the nonlinear mapping tensor of the key tensor is used for replacing the similarity, so that the calculation complexity is reduced, and the state analysis efficiency is improved.

The application discloses an edge model switching method and device, equipment, a storage medium and a program product, relates to the technical field of edge computing, and the method is applied to an edge device and includes the following steps: in a reasoning period in which a first model currently running continuously performs an inference task, incrementally loading a model file of a second model and asynchronously initializing the second model; in the case where the initialization is completed, entering a transition window period, performing parallel inference on input data by using the first model and the second model, and performing weighted fusion on inference results of the double models by using a smoothing weight function to obtain a final inference result; a first derivative of the smoothing weight function at a starting moment and an ending moment of the transition window period is zero; in the case where the transition window period ends, releasing resources of the first model, and switching the first model to the second model. The online switching of the edge device model is completed without causing business service interruption or output jump and other business disturbances.

The embodiment of the invention provides a distributed new energy equipment anomaly detection method and system based on edge calculation, and the method comprises the steps: precisely recognizing suspected drift high-value samples outside a model cognitive boundary through a dual evaluation mechanism of an abnormal score and an uncertainty score, and only uploading the samples based on an active learning strategy, thereby achieving the detection of the anomaly of the distributed new energy equipment. Therefore, a low-efficiency mode of uploading total data is abandoned, and the bandwidth overhead of cloud side communication is greatly reduced. At the cloud, after the master model completes incremental training by using the high-value samples, a model incremental updating technology is adopted, and only parameter difference values between the new and old models instead of a complete model file are calculated and issued. According to the updating mode, the limitation of the edge device on bandwidth, storage and computing resources is effectively overcome, low-cost, automatic and accurate self-adaption of the edge model is realized, and the long-term effectiveness and reliability of the detection system in a dynamic environment are ensured.

The invention discloses an ECA lightweight facial expression recognition method based on edge cloud collaboration, and relates to the technical field of computer vision and artificial intelligence, and the method comprises the steps: S1, training a universal model at a cloud end; s2, edge end model training: S21, dividing an expression data set into a training set and a test set, and preprocessing the data set; and S22, the edge end loads and freezes the first four convolutional layers pre-trained by the cloud end, and only the subsequent network layers are finely adjusted in the training process. And S23, introducing an ECA attention mechanism on the basis of transfer learning, highlighting a key expression area, and inhibiting background interference. And S24, adopting a depth separable convolution and H-swsh activation function in a subsequent layer to reduce the parameter quantity and the calculation quantity. And S25, using a loss function FocalLoss to solve the problem of number imbalance among expression categories in the data set. Finally, expression feature extraction and classification are independently completed at the edge end, and light-weight and efficient expression recognition is achieved.

This invention proposes a collaborative continuous test-time adaptation method for edge environments, belonging to the fields of edge intelligence and model continuous test-time adaptation technology. Through cloud-edge collaboration and two-stage learning, it achieves continuous and stable adaptation of edge device models to dynamic data distributions under resource-constrained conditions. Image inversion generation technology is used in the cloud to synthesize images and preserve historical knowledge; a pre-trained model including a frozen backbone and lightweight branch networks is deployed on the edge device. A two-stage replay mechanism is introduced, decoupling active forgetting and knowledge integration into two independent stages. This allows the edge model to first focus on learning new task information and then collaboratively integrate it with historical knowledge, thus balancing high plasticity and stability during continuous adaptation. This invention achieves continuous test-time adaptation in a storage-efficient manner while maintaining high plasticity and stability during continuous adaptation, supporting the long-term stable performance of edge models in dynamic environments.