978 results about "Perception model" patented technology

The invention provides a vehicle collision risk early warning method, system and device and a medium, and belongs to the technical field of vehicle safety control. The method comprises the steps that surrounding environment information of a vehicle is collected in real time through a vehicle built-in sensor and a GPS module; a multi-sensor fusion device technology is used to carry out comprehensive processing on surrounding environment information of a vehicle, and an environment perception model is generated after data integration; based on the environment perception model, target identification is carried out by using a computer vision technology based on a convolutional neural network so as to identify travelers, vehicles, traffic signs and obstacles; performing behavior prediction based on a deep learning model to predict a behavior track of a front vehicle or pedestrian; evaluating a potential collision risk in the current scene in combination with the target recognition result and the behavior prediction result; when a potential collision risk is detected, an early warning signal is sent to a driver, and when the driver does not take corresponding measures, active safety measures are automatically triggered to avoid collision or reduce accident damage.

The invention relates to the field of data backup, in particular to an intelligent data backup method and system based on an AI large model. The method comprises the following steps: acquiring an enterprise global data list, performing intelligent data structure deconstruction and dynamic attribute mapping modeling, and constructing a holographic data semantic perception model; performing real-time transient risk mutation detection on the holographic data semantic perception model, and constructing an intelligent backup triggering mechanism; carrying out storage resource demand prediction based on an intelligent backup trigger mechanism, carrying out multi-storage cloud environment resource dynamic scheduling, and constructing an elastic backup storage resource pool; carrying out incremental backup analysis and self-adaptive compression coding to obtain an incremental backup coding packet; and performing dynamic backup sequence adjustment and intelligent incremental backup decision on the incremental backup coding packet based on the elastic backup storage resource pool, and constructing an intelligent incremental backup execution engine. According to the method, the reliability, the accuracy and the traceability of a backup result are improved through self-adaptive intelligent incremental backup.

The invention provides a flood peak evolution path and dam break risk early warning method and system. According to the method, a time-space coupling water conservancy data set is constructed by fusing multi-dimensional water conservancy monitoring data and regional rainfall prediction information, and a basin topology perception model is established. And further performing mode matching on the parameters and a historical dam break event characteristic spectrum, and generating a dynamic response strategy set including a flood storage and detention area capacity allocation scheme by correcting a space weight of a matching result. And finally, based on a watershed topology perception model, matching the strategy set with historical dam break features, realizing accurate identification of a flood peak evolution path and graded early warning of dam break risks, outputting a water conservancy analysis report containing risk grades, and realizing closed-loop management from data fusion and dynamic prediction to risk decision. According to the technical scheme provided by the invention, the efficiency and accuracy of intelligent analysis of the water conservancy data can be improved.

The invention relates to the field of crop monitoring and analysis, in particular to a crop growth monitoring method and system based on multi-sensor fusion. The method comprises the following steps: collecting multi-dimensional crop growth environment monitoring parameters, carrying out environment multi-dimensional perception fitting, and constructing a growth environment multi-dimensional perception model; acquiring a crop full-cycle growth monitoring image and a crop physiological state original data set; performing growth state deep evolution according to the crop physiological state original data set, and constructing a multi-mode growth state evolution graph; and carrying out adaptive region filtering and denoising on the crop full-cycle growth monitoring image flow, and carrying out crop three-dimensional morphological evolution analysis to obtain a crop three-dimensional morphological evolution rule. According to the invention, the virtual simulation model is perfected by collecting real-time parameters, intelligent analysis and real-time decision making are carried out, and the stability and accuracy of crop monitoring analysis are improved.

The invention discloses a multi-perception interaction method and system based on virtual reality. The method comprises the following steps: creating a virtual scene and an intelligent agent based on an interaction demand, and constructing a perception model of the virtual scene and the intelligent agent; the intelligent agent perceives visual, auditory and tactile environments, extracts perceived feature time sequence data, divides instructions according to the perceived feature time sequence data, and calculates instruction priorities. And generating a decision network according to a preset knowledge graph, and forming an instruction chain in combination with the instruction priority, so that the intelligent agent obtains a feedback strategy. A dynamic weight value is distributed to a feedback strategy, a virtual scene is updated through fusion of an attention mechanism and a perception model, an interaction mode is obtained, meanwhile, physical and psychological parameters of a user are recorded in real time, abnormal fluctuation segments are recognized, and emotional tendency of the user is learned from corresponding interaction, so that the interaction mode is continuously optimized. According to the method, the interaction experience can be improved, the intelligent agent can better meet the user requirement, the interaction effectiveness and adaptability are enhanced, and meanwhile good interpretability is achieved.

The invention relates to the technical field of unmanned aerial vehicle cluster control, and particularly discloses an intelligent cooperative flight path planning method and system for an unmanned aerial vehicle cluster system, and the method comprises the steps: firstly constructing a dynamic environment perception model, collecting data through all unmanned aerial vehicle sensors, and carrying out the preprocessing, attention feature extraction and federal learning fusion, and generating a global environment situation map; planning and screening a candidate track set by adopting a particle swarm-genetic hybrid optimization algorithm for adaptive weight adjustment on the basis of the image; a global optimal track consensus is achieved through an improved consensus algorithm and conflict resolution through a distributed collaborative negotiation mechanism and no human-computer interaction evaluation indexes; and finally, monitoring the environment in real time during execution, triggering dynamic re-planning when the environment is abnormal, and ensuring track adaptation through multi-level threshold and incremental planning. According to the method, the unmanned aerial vehicle cluster can quickly respond to the environment change and adjust the flight path, so that the task execution efficiency and success rate of the unmanned aerial vehicle cluster in the complex dynamic environment are improved.

The invention discloses a heterogeneous computing multi-target adaptive task scheduling method based on deep reinforcement learning, and the method comprises the following steps: S1, constructing a multi-dimensional dynamic perception model of a heterogeneous computing environment, and collecting and computing node performance indexes, task feature parameters and network states in real time; s2, defining a reward function as a multi-target weighted combination, fusing task completion time, energy consumption, resource utilization rate and cost, and dynamically adjusting the weight by a fuzzy comprehensive evaluation algorithm; s3, establishing a dual-channel deep reinforcement learning network architecture based on an attention mechanism; s4, establishing an adaptive exploration mechanism, combining an epsilon-greedy strategy and entropy regularization, and balancing exploration and utilization; the method has the beneficial effects that dynamic balance of multiple indexes such as task completion time, energy consumption and resource utilization rate is realized through combination of deep reinforcement learning and multi-objective optimization, a dual-channel network and a cross attention mechanism are adopted, and a task time sequence characteristic and a topological dependency relationship are modeled at the same time, so that a scheduling strategy is more accurate.

The invention discloses an AI-based conference record data searching method and system, and the method comprises the steps: carrying out the feature extraction and alignment through a multi-mode fusion neural network according to the voice, text, image and video data collected in a conference process, and obtaining a semantic representation vector; according to the semantic representation vector, combining context information of the conference scene, and utilizing a pre-trained context perception model to perform semantic enhancement processing to generate an enhanced semantic vector with context association; according to the enhanced semantic vector, combining with an external knowledge base, and utilizing a dynamic knowledge graph construction algorithm to generate a knowledge graph related to the conference theme in real time; and according to the knowledge graph, intelligent retrieval and recommendation of conference record data are carried out by using a graph neural network. By utilizing the embodiment of the invention, the intelligent retrieval efficiency and accuracy of the conference record can be improved.

The invention discloses a high-performance low-cost multi-modal data fusion model and equipment for complete driving environment state perception, and the model comprises a feature extraction module which is used for extracting initial potential features of a camera image and a laser radar point cloud; a probability cross attention mechanism constructs a feature fusion model of uncertainty perception; a query initialization model capable of reducing uncertainty is adopted, and the initialization process of query is enhanced by utilizing two-dimensional target detection priori; and based on a robustness optimization training strategy of query denoising, stable optimization is realized under uncertainty. A nuScenes data set is selected for training, verifying and testing the comprehensive performance of the model, and the TensorRT realizes CUDA acceleration and is quantitatively deployed on a vehicle-mounted calculation experiment platform. According to the multi-modal data fusion system, a perception state uncertainty explicit measurement calculation theory based on data and a model is innovated, a perception model for coping with environment uncertainty is established, the completeness, reliability and reconfigurability requirements of a driving environment are met, and driving safety and decision accuracy are ensured.

The invention discloses a building robot multi-machine collaborative operation optimizing and monitoring method and system based on digital twinning, and belongs to the technical field of building construction intellectualization. The method comprises the following steps: constructing a digital twinborn body comprising an environment perception model and a kinematics / dynamics model based on a building information model and robot physical parameters; performing multi-machine task allocation and path planning containing static / dynamic obstacle avoidance in the virtual environment; synchronizing physical environment data in real time through a multi-mode sensor; dynamically optimizing operation parameters by adopting a genetic algorithm or a particle swarm algorithm and realizing closed-loop control; and generating a safety early warning and emergency scheme based on machine learning. According to the method, Markov decision path planning of reinforcement learning and multi-agent game task allocation are creatively fused, laser radar-vision-inertial navigation multi-source data fusion is adopted, the technical problems that in a traditional method, digital twinning precision is insufficient, and dynamic cooperation efficiency is low are solved, and the method is suitable for large-scale popularization and application. And the construction efficiency, the safety and the man-machine interaction experience are remarkably improved.

The invention discloses a non-key unit fire-fighting early-warning management method based on dynamic risk grading, and the method comprises the steps: collecting multi-dimensional parameters, such as operation of fire-fighting facilities and activity of personnel, in real time through a multi-dimensional fire-fighting abnormal parameter perception model, carrying out the structural storage, dividing time windows through a dynamic risk time sequence management model, and extracting features; and fire risk values are calculated and graded by combining the two models. And determining an early warning trigger condition according to a risk grading result and fire-fighting early warning parameters, continuously monitoring and updating risk values and grades after early warning is started, and dynamically adjusting early warning information. And establishing a feedback mechanism, and feeding back a response measure effect to the model to optimize parameters and calculation rules. According to the invention, dynamic and precise early warning management of fire risks of non-key units is realized.

The invention provides an efficient remodeling production scheduling method, medium and system based on AI multi-agent dynamic negotiation, and belongs to the technical field of agents. A residual connection mechanism is used for processing time sequence correlation characteristics under a nonlinear working condition to establish a remodeling time prediction basis, a dynamic layered negotiation architecture is established to realize information interaction and decision transmission, and an equipment agent predicts remodeling time based on a multi-scale time sequence perception model and generates a bidding scheme. A coordination agent processes a bidding scheme by adopting a Pareto leading edge multi-objective optimization algorithm to balance multiple objectives, dynamically adjusts model parameters through a hybrid similarity evaluation mechanism to guarantee prediction stability, and automatically identifies an affected order subset to trigger an incremental re-negotiation process when production disturbance occurs. The technical problem that the production scheduling plan is frequently adjusted due to inaccurate remodeling time prediction is solved.

The invention discloses a bearing fault diagnosis method based on multi-scale frequency sensing dynamic enhancement, and the method comprises the steps: collecting a vibration signal in a bearing operation state, carrying out the preprocessing of the vibration signal, obtaining a time-frequency matrix, and dividing the time-frequency matrix into a training set and a test set; building a multi-scale frequency network sensing model, inputting a time-frequency matrix in a training set into the model to realize extraction of multi-scale features, then performing pooling, time sequence compression, flattening and dimension reduction on the extracted multi-scale features, and then outputting fault category probability distribution through a classifier; and finally, testing the trained model by using a test set, and calculating evaluation indexes such as accuracy, a confusion matrix, an ROC curve and the like. The method has high accuracy while keeping light weight, breaks through double limitations of fixed frequency band and sensitive rotating speed of a traditional method, and can provide a high-precision and low-cost light-weight solution for engineering application of variable-rotating-speed mechanical fault diagnosis.

The invention relates to the technical field of micro-nano motion control, and discloses a self-adaptive control method of a micro-nano high-precision motion platform. The method comprises the following steps: acquiring real-time pose feedback data and target trajectory data of a motion platform, and extracting dynamic response features; calling the trained motion feature analysis network to perform multi-modal feature separation, and generating a platform pose feature set; based on the set, performing space-time coupling analysis on the working environment parameters through an environment disturbance perception model to obtain a fusion result containing mechanical deformation characteristics and environment disturbance characteristics; inputting a fusion result into a dynamic compensation model to calculate a track correction amount, and outputting a driving compensation instruction; and calibrating the target trajectory data in real time according to the compensation instruction, and generating an actual control signal. According to the method, through multi-modal feature analysis, space-time coupling perception and dynamic compensation, the control precision and stability of the motion platform in a complex environment are improved, and the method is suitable for a micro-nano high-precision motion control scene.

The invention discloses an eVTOL multi-camera cooperative video compression coding method based on multi-source perception and intelligent partitioning. The method comprises the following steps: constructing a scene three-dimensional perception model by fusing multi-source data of visible light, infrared and depth sensors; the method comprises the following steps: realizing dynamic video partitioning based on motion vectors and semantic analysis, and dividing a picture into a core region, a secondary region and a background region; establishing a parallax compensation motion prediction model by adopting a cross-camera reference frame sharing mechanism; and high-fidelity compression of the key area is realized through layered entropy coding and a dynamic quantization parameter distribution strategy. And the decoding end reversely executes multi-source data fusion and partition reconstruction according to the coding metadata. The method is suitable for eVTOL multi-camera video real-time transmission scenes such as polling, surveying and mapping.

The invention discloses an immersive robot teleoperation method and system, and belongs to the technical field of robot teleoperation. An existing teleoperation scheme lacks an effective and timely feedback mechanism, is poor in operation intuition and low in control precision, and is difficult to compete with complex and fine operation tasks. According to the immersive robot teleoperation method, by constructing the first person mapping model, the interaction data acquisition model, the teleoperation processing model, the execution perception model and the immersive visual angle simulation model, real-time mapping can be performed on posture actions of an operator, and a real-time response three-dimensional image capable of being displayed in VR equipment is generated; an effective and timely immersive feedback mechanism from a first person perspective is formed, an operator can execute a task as like being personally on the scene, and the immersion, intuition and naturalness of interaction are improved, so that the control capability and the task execution effect can be remarkably improved, the operation efficiency and precision are improved, and then the operator can competent for complex and fine operation tasks.

The invention provides a depth estimation enhanced semantic segmentation method and device, equipment and a medium. The method comprises the following steps: extracting a corresponding depth image from an acquired RGB image by using a depth estimation large model; constructing a scene perception model, wherein the scene perception model comprises a coding layer and a decoding layer; the coding layer comprises an RGB image coding branch and a depth image coding branch, and is used for correcting and fusing the characteristics of the RGB mode and the depth mode output by the RGB feature extraction layer and the depth feature extraction layer based on data fusion modules arranged between the coding layers respectively to obtain the first fusion characteristics of each layer; a first fusion feature obtained after fusion of the data fusion modules except the last layer of data fusion module is input to a multi-scale fusion decoding module, step-by-step recovery of a feature map is achieved through a multi-scale feature fusion module and a frequency perception feature fusion device of the multi-scale fusion decoding module, and finally a semantic segmentation result is obtained. According to the method, scene perception is accurate, and meanwhile, only the minimum cost is needed.

The invention discloses a deep sea deduction scene course path intelligent planning system based on an optimal control theory, and relates to the technical field of course path planning. The data management platform is in communication connection with an environment perception modeling module, a constraint condition definition module, a global path planning module, a simulation verification module, a control execution module and a feedback adjustment module. By receiving and fusing deep sea environment data and equipment state data in real time, position deviation, speed deviation and course deviation in the sailing process can be accurately detected and analyzed, and based on the optimal control theory, the system can quickly formulate an adjustment strategy to ensure that equipment is always kept in a safe sailing range, so that the sailing safety is improved. The real-time feedback and adjustment mechanism greatly improves the safety and reliability of navigation, effectively avoids the risk of colliding with obstacles or entering a dangerous area, and provides a solid guarantee for deep sea operation.

The invention relates to a transfer learning-based power distribution network measurement data super-resolution sensing method. The method comprises the steps that a high-frequency time sequence is acquired and preprocessed, and training sample data is obtained; a cascade model is constructed based on a graph attention network and a transformer network, the training sample data and an incidence matrix are input into the cascade model, the cascade model is pre-trained, and the incidence matrix represents the topological connection relation of the power distribution network; the cascade model after pre-training is divided into a learning device part and a decision-making device part, transfer learning training is carried out, model parameters reaching a training target are reserved, a final super-resolution perception model is obtained, the learning device part comprises all layers except a transformer linear output layer, and the decision-making device part comprises a transformer linear output layer; and inputting real-time measurement data into the super-resolution sensing model to obtain a super-resolution measurement generation result.

The invention discloses a wind power generation abnormal data analysis method and system, and relates to the technical field of data analysis, and the method comprises the following steps: constructing a wind direction change rate enhanced perception model, analyzing the potential omen of wind direction abrupt change based on an ultra-short time scale wind direction change trend curve and a wind speed fluctuation coupling index collected at multiple measurement points, and determining the wind direction abrupt change. Generating a risk early warning label of wind wheel pointing deviation; and based on the risk early warning label, executing a high-frequency yaw disturbance prediction mechanism, and predicting an inflow angle continuous offset window caused by yaw response lag by using a nonlinear time sequence evolution trend and a short-term wind direction reversal probability curve. The wind direction sudden change early warning is realized through multi-measuring-point wind direction enhanced perception and wind speed coupling analysis, the response precision is improved and the energy consumption is reduced in combination with predictive yaw compensation and torque balance, the yaw parameters are dynamically optimized by using adaptive closed loop and reinforcement learning, the inflow angle is kept stable for a long time, and the power generation efficiency and the structural safety are improved.

The invention discloses a sleep thermal comfort perception and indoor sleep environment regulation and control method and device, and belongs to the technical field of computer vision and heating ventilation air conditioning control, and the method comprises the steps: obtaining video data, physiological data and environment data, which are measured in advance, of a subject during sleep; performing data annotation, feature extraction and data fusion on the video data, the physiological data and the environmental data to make a data set; inputting the manufactured data set into a pre-constructed sleep thermal comfort perception model to obtain a sleep thermal comfort perception prediction result; regulating and controlling the indoor sleep environment according to the obtained sleep heat comfort perception prediction result; according to the method, by predicting the sleep thermal comfort of the person, the operation parameters of the heating ventilation air conditioning system can be correspondingly adjusted, the influence of contact type equipment on a sleeper is avoided, the thermal comfort of the sleeper can be ensured, the energy use efficiency can be remarkably improved, and therefore the total energy consumption of a building is reduced, and energy conservation and emission reduction are achieved.

The invention relates to the technical field of track foreign matter real-time detection, and particularly discloses a track foreign matter real-time detection method based on YOLOv5 improvement. Comprising the steps of video vibration synchronous acquisition, space offset compensation parameter generation, dynamic feature enhancement processing, feature weight thermodynamic diagram generation, hierarchical perception feature aggregation, cross-scale target verification, foreign matter positioning instruction packaging and multistage early warning execution control. Image distortion caused by mechanical vibration and illumination fluctuation is effectively overcome by synchronously acquiring track video and vibration time sequence signals, constructing a spatial offset compensation parameter set and combining with a dynamic feature enhancement template to correct video frames, a feature weight thermodynamic diagram is generated through visual saliency detection, features are aggregated through a layered perception model, and the visual saliency detection accuracy is improved. Early warning levels are dynamically matched according to the sizes and the positions of the foreign matters, and a sound-light alarm and braking system is linked; according to the invention, the reliability of data transmission is ensured, the operation and maintenance cost is greatly reduced, and the safety requirement of millisecond response of modern rail transit is met.

The invention discloses a recalculation perception model splitting method and system oriented to assembly line parallel training, and belongs to the field of artificial intelligence of computer science. The method specifically comprises the steps that a dijkstra algorithm and two-layer binary search are used, and the minimum value of the sum of forward propagation time and reverse propagation time of assembly line microbatch under re-calculation perception model splitting is calculated; and an intermediate result obtained by a dijkstra algorithm in a search process is used to calculate and output an approximately optimal model splitting strategy for re-calculation perception. According to the method, the model splitting method and system are designed by utilizing the model characteristics of a Tasformer large model and the characteristics of an assembly line parallel mode, the model is non-uniformly split to each stage of the assembly line, so that the calculation load of each stage of the assembly line is balanced when re-calculation is started, and the calculation efficiency is improved. Therefore, the overall performance of large model pipeline parallel training and the utilization efficiency of cluster equipment are improved.

The invention provides a digital mechanical equipment data management system based on cloud computing. The digital mechanical equipment data management system comprises a data acquisition module for acquiring equipment state and environment information; the feature coding module is composed of a gating circulation unit and a multi-layer perceptron and is used for extracting high-dimensional features; the feature fusion module fuses the high-dimensional features and outputs sparse joint codes; the cloud computing model construction module constructs a model and introduces a trainable enhancement layer; the state interaction module enables the code to obtain an advanced interaction state based on a model; the state synchronization module synchronizes a device state and environment information through MechLayer; the implicit state generation module obtains an implicit state by combining normalization and enhanced connection; operating an environment perception modeling module to improve perception estimation modeling; the perception learning vector generation module obtains a learning vector; the state prediction module predicts an operation state; and the track decoding module processes the input to obtain an equipment running track. The equipment management efficiency and the operation reliability can be improved.

The invention relates to the technical field of electric vehicle control, in particular to an AI-based electric vehicle intelligent starting and acceleration control method and system. The method comprises the following steps: acquiring a driving log of a vehicle owner, performing intelligent vehicle optimization after starting, and constructing an intelligent starting mode selection strategy; physiological state data of a driver are obtained, state change evolution analysis is carried out, and a personalized driving portrait is constructed; holographic state sensing and vehicle virtual simulation are carried out, and a vehicle synchronous digital simulation model is constructed; radar feedback sensing information and a real-time monitoring image in front of a vehicle are collected, traffic situation prediction and vehicle scene prediction and perception are carried out, and a vehicle scene prediction and perception model is constructed. According to the method, the driving starting mode of the vehicle is intelligently selected, vehicle owner requirements and environment available power constraints are matched, and vehicle acceleration efficiency maximization and safety optimization are achieved.

The invention provides a road environment perception method and a perception model training method, and relates to the technical field of image processing, in particular to the fields of automatic driving, computer vision, deep learning and neural network models. The specific implementation scheme of the road environment sensing method comprises the following steps of: obtaining image features from a look-around image of a vehicle by utilizing a feature extraction network of a first sensing model; wherein the look-around image is obtained according to a plurality of image acquisition devices at different angles; determining a bird's-eye view feature map according to the image features and the bird's-eye view query vector by using a coding network of a first perception model; and respectively determining output results of the plurality of perception tasks according to the bird's-eye view feature map by using the plurality of head networks of the first perception model. According to the technical scheme provided by the invention, computing resources consumed by simultaneous execution of multiple visual tasks can be effectively reduced, and the real-time performance and task effects of the multiple visual tasks are ensured.

The invention discloses a data quality treatment method and system based on an AI Agent, and belongs to the technical field of artificial intelligence and data treatment. An initial data semantic distribution map is constructed, cross-dimension correlation feature factors are extracted, a multi-scale quality anomaly sensitive factor matrix is constructed, time sequence evolution weights are embedded, and a dynamic feature evolution trajectory is formed; performing perception modeling on the evolution trajectory by using an AI Agent, generating a multi-level quality risk thermodynamic diagram, extracting a deviation dense region and constructing an anomaly propagation path set; in combination with upstream and downstream data links and task flow information, calculating a potential impact factor weight, constructing a causal traceability map, and injecting a correction strategy label for a key field node; the AI Agent autonomously selects an adaptive strategy combination according to the target data segment, and performs online intervention on the target data segment; according to the method, closed-loop treatment of the data quality problem from perception and judgment to intervention and feedback is realized, and the method has the advantages of self-adaption, high interpretability and the like.

The invention relates to a self-adaptive task allocation and cooperative control method for a distributed unmanned ship cluster, and the method comprises the following steps: enabling an unmanned ship to collect the state information and environment information of the unmanned ship, constructing a state perception model with integrated cognition, and evaluating the navigation safety level and cruising ability of the unmanned ship; the unmanned surface vehicle determines an evaluation index according to a task type according to a state sensing model of the unmanned surface vehicle and information interaction between the unmanned surface vehicle and an adjacent unmanned surface vehicle, calculates fitness to different tasks, and dynamically negotiates and distributes the tasks based on a game theory and a negotiation mechanism; a reinforcement learning algorithm is introduced, the unmanned surface vehicles interact with the environment to adjust an action strategy according to reward feedback, meanwhile, learning experience is shared among the unmanned surface vehicles, and a cooperative control strategy is optimized. Compared with the prior art, the method fully considers the individual difference of the unmanned surface vehicles and the dynamic change of the marine environment, realizes the adaptive distribution and efficient cooperative control of tasks, and improves the operation efficiency, reliability and safety of the unmanned surface vehicle cluster in the complex marine environment.

The invention discloses a multifunctional intelligent electronic ear tag device. The ear tag adopts a head hoop and an arc-shaped ear hook structure, a connecting buckle is of a magnetic attraction type, punching is not needed, and the head hoop is made of medical silica gel, so that it is ensured that the ear tag is comfortable to wear by animals and convenient to use repeatedly. The power management module guarantees power supply, and the data monitoring module can collect multi-dimensional information such as animal body surface temperature, environment and behavior data. During working, breeding areas are divided to construct a wireless network, data are acquired and then preprocessed, and the animal growth state sensing model is utilized to analyze and upload a result. The power consumption adaptive algorithm dynamically adjusts the working mode of the ear tag according to the equipment state and the like, such as adjusting the sampling frequency and the local model complexity, so that low-power-consumption edge calculation and accurate reasoning are realized. The user terminal supports data display, interaction control and remote alarm, and helps a manager adjust a breeding strategy based on an analysis result. The system has the characteristics of high comfort, low power consumption and high precision, and provides technical support for digital transformation of animal husbandry.

Embodiments of this disclosure can provide a system and method for training a perception model to perform an autonomous driving task. During operation, the system can obtain labeled training data comprising images captured by multiple cameras mounted at different locations on a vehicle, and the perception model can generate, in parallel, a prediction output associated with the task and a confidence score based on the labeled training data. The confidence score can indicate a level of uncertainty associated with the prediction output. The system can generate an uncertainty-weighted prediction based on ground truth indicated by the labeled training data, the prediction output, and the confidence score; compute a loss function based on the uncertainty-weighted prediction; and update the perception model based on the loss function.