1477 results about "Kaiman filter" patented technology

The invention discloses a building carbon emission monitoring system and method based on big data acquisition, particularly relates to the field of building carbon emission monitoring, and is used for solving the problems of energy use optimization and carbon emission dynamic regulation and control. Through combination of a recursive feature elimination algorithm and Kalman filter dynamic correction, efficient identification of an electric appliance operation state and accurate correction of load data are realized. Time-delay-free synchronization of distributed energy and load requirements is realized through a time sequence cross-correlation algorithm, and the efficiency and precision of energy allocation and carbon emission monitoring are greatly improved. The dynamic carbon load pressure index and the global energy consumption difference index cooperate to assess changes of load and carbon emission, rapid response to abnormal energy consumption and carbon emission is realized, generation and prediction optimization of a regulation and control strategy are carried out in combination with a fuzzy control algorithm and an autoregression model, the carbon emission of buildings is effectively reduced, and the construction efficiency is improved. And intelligent and dynamic load and carbon emission management is realized.

The invention provides a brushless motor attitude detection and balance control method and system based on a gyroscope, and relates to the technical field of control, and the method comprises the steps: collecting angular velocity and acceleration data through a six-axis gyroscope, carrying out the noise reduction through wavelet transform, and carrying out the attitude calculation through the combination of an extended Kalman filter and a quaternion algorithm. A rotor position signal is obtained through a magnetic encoder, nonlinear compensation is carried out, and rotating speed data are calculated. A motor state is modeled by adopting a long-short-term memory network, a double-layer adaptive fuzzy neural network controller is constructed, and attitude error compensation and rotation speed fluctuation suppression are realized. A controller model is optimized through particle swarm optimization and a genetic algorithm, a compensation current vector is corrected in real time, and the control precision and stability of the brushless motor are improved. According to the method, the operation efficiency and the dynamic response capability of the brushless motor are effectively improved.

The invention relates to the technical field of battery management systems, and particularly discloses a battery signal acquisition precision optimization method of a BMS (Battery Management System), which comprises the following steps of: dynamically adjusting a covariance matrix of process noise and measurement noise by monitoring the voltage, current and temperature of a battery in real time so as to cope with the change of the working state of the battery; according to the change trend of the charge and discharge rate and the temperature, noise parameters are predicted and corrected, and the adaptability of the system in an extreme environment is improved; in combination with data from a plurality of sensors, weighted average and a self-adaptive adjustment mechanism of a Kalman filter are adopted, so that evaluation of the system on signal quality of different sensors and effective suppression of noise are enhanced; aiming at the sudden change of the battery state, the system can be ensured to quickly respond by automatically adjusting the filtering algorithm, the accurate monitoring of the battery health state is ensured, the stability and the accuracy of the battery management system can be improved, and the problem of estimation error under the extreme working condition of a traditional method is solved.

The invention discloses a multi-unmanned aerial vehicle combined flight control system coping with wind power change, and relates to unmanned aerial vehicle flight control, and the system comprises a flight data collection module which is used for collecting unmanned aerial vehicle flight data in real time, and the unmanned aerial vehicle flight data comprises wind power data, attitude angle data and height data of each unmanned aerial vehicle; the data receiving and processing module is electrically connected with the flight data acquisition module, and the data receiving and processing module is used for receiving unmanned aerial vehicle flight data acquired in real time and performing noise removal on the acquired data. According to the multi-unmanned aerial vehicle combined flight control system provided by the invention, the wind field prediction model is constructed by combining the LSTM neural network with the time attention mechanism, wind power periodic change characteristics are accurately captured, real-time correction is performed in cooperation with the Kalman filter, the prediction precision and timeliness are both optimized, and the system is suitable for large-scale popularization and application. A multi-unmanned aerial vehicle cooperative kinetic model is constructed based on a prediction result, so that the formation trajectory tracking error is reduced.

The invention discloses an unmanned ship multi-agent collaborative obstacle avoidance method and system for complex sea conditions, and relates to the field of unmanned ship multi-agent collaborative obstacle avoidance, and the method comprises the steps: obtaining the real-time data of each unmanned ship and the surrounding environment; generating a candidate obstacle target point cloud cluster list based on a density clustering algorithm; on the basis of a Kalman filter, real-time absolute motion state estimation of the candidate obstacles is obtained, and an obstacle feature list is output; determining a safety radius compensation amount required by autonomous obstacle avoidance of each unmanned ship; obtaining the safe sailing space of each unmanned ship at the current moment; constructing a global synthetic potential field, and generating a group of optimal alternative paths of the unmanned ship from the current position to the target point; on the basis of adopting a consensus binding algorithm, an unmanned ship multi-agent collaborative obstacle avoidance path reaching a consensus is obtained. The method has the advantages that safe, efficient, cooperative and consistent intelligent obstacle avoidance of a multi-unmanned-ship cluster in a real complex marine environment is realized.

The invention provides a rotor temperature estimation method and device of a motor, a vehicle and a storage medium, and the method is applied to the field of motor temperature monitoring, and comprises the steps: obtaining the current environment temperature, the current shutdown time, the stator loss and the rotor loss of the motor, determining the initial rotor temperature of the motor according to the current environment temperature and the current shutdown duration, and obtaining the temperature rise of the motor by using a preset thermal network model and a preset double-Kalman filter based on the current environment temperature, the stator loss and the rotor loss; and obtaining the actual rotor temperature of the motor according to the initial rotor temperature and the temperature rise. According to the method, the high-precision estimation of the rotor temperature of the motor is realized by establishing the thermal network model and combining the double Kalman filters, the dependence on a hardware sensor is not needed, the cost is reduced, and the reliability and robustness of the system are improved.

The invention relates to the technical field of battery intelligent monitoring, and discloses an intelligent monitoring system and method for a new energy automobile battery. The method comprises the following steps: acquiring a target battery operation data set of a new energy automobile battery, and extracting time domain differential characteristics and frequency domain energy loss characteristics in a battery charging and discharging process to obtain a multi-dimensional state characteristic set; performing electrochemical characteristic analysis on the new energy automobile battery based on the multi-dimensional state feature set to obtain a micro degradation state judgment result; and dynamically adjusting the parameter configuration of a hybrid Kalman filter according to the micro degradation state judgment result, and generating a battery state-of-charge estimation value. According to the method, the problems of insufficient estimation precision and accumulative errors in a traditional method are solved, and the use safety and reliability of the battery are improved.

The invention relates to the technical field of underwater intelligent navigation and integrated navigation, in particular to a submersible integrated navigation multi-source data fusion method based on an elastic random model, which comprises the following steps: preprocessing multi-source navigation data output by an inertial navigation system (INS), a Doppler velocimeter, an ultra-short baseline / long baseline, an altimeter and a depthometer; constructing a state space model containing navigation parameter errors and sensor errors, and screening an optimal sensor combination mode; the method comprises the following steps: constructing an elastic correction parameter model and a dynamic adjustment state and observation noise covariance matrix by combining an elastic PNT framework, performing local filtering on multi-source data by adopting a plurality of parallel elastic self-adaptive robust volume Kalman filters (CKF), outputting a fusion resolving result, and realizing optimal estimation of a global navigation state through a dynamic information distribution strategy. The method has relatively high anti-interference performance, self-adaptive capability and combination switching robustness, and is suitable for navigation and positioning tasks of a high-precision submersible in a complex underwater environment.

The invention belongs to the technical field of unmanned aerial vehicle autonomous navigation, and relates to an intelligent unmanned aerial vehicle navigation method and system based on bimodal obstacle feature extraction. Aligning the information of the multi-modal sensor in time and space through a time-space alignment network; real-time pose information of the unmanned aerial vehicle is obtained through a Kalman filter fused with a positioning failure adaptive strategy; a double-flow obstacle feature extraction network is adopted to extract a composite environment vector containing dynamic obstacle features and static obstacle features from the multi-modal sensor data; training the SAC model based on the maximum entropy by adopting an improved reinforcement learning training method based on curriculum learning to obtain an SAC strategy network; and inputting the composite environment vector and the real-time pose information of the unmanned aerial vehicle into an SAC strategy network, and obtaining the optimal control of the unmanned aerial vehicle in an end-to-end manner, thereby having good path planning performance, and effectively improving the task execution capability and safety of the unmanned aerial vehicle in a complex environment.

The invention provides an electric power system natural disaster early warning and response method based on real-time monitoring, and relates to the technical field of electric power early warning response, and the method comprises the following steps: data analysis and processing: carrying out the analysis and processing of real-time monitoring data in an electric power system through combining a long short-term memory network with a Bayesian network; generating a scheduling scheme: updating the state of the power system in real time and generating an optimal scheduling scheme based on a multi-source data fusion and optimization method combining a Kalman filter and a genetic algorithm; generating an emergency response strategy: simulating the influence of natural disasters on the power system in real time in the virtual model through a digital twin technology and a self-adaptive dynamic planning method, and providing the emergency response strategy; and anomaly detection and self-healing: carrying out anomaly detection and self-healing on the operation state of the power system by combining big data analysis with edge calculation.

The invention discloses a lithium battery health state prediction method and system. The method comprises the following steps: acquiring lithium battery operation data; inputting the lithium battery operation data into a lithium battery health state prediction model for health state prediction to obtain a prediction result; wherein the lithium battery health state prediction model is obtained by collecting historical battery operation data as a sample set to train a basic model, a forward propagation algorithm, a back propagation algorithm and hyper-parameter tuning are adopted in the training process, and the basic model is a model formed by connecting an encoder of a Transform layer, a multi-layer KAN network as a decoder and an unscented Kalman filter; and outputting the prediction result. By implementing the method provided by the invention, the accuracy and stability of SoH prediction can be improved.

The invention relates to the technical field of battery state monitoring, and discloses a sightseeing vehicle battery state monitoring method, device and equipment and a storage medium, and the method comprises the steps: collecting the voltage, current and temperature data of a sightseeing vehicle battery, and carrying out the preprocessing, and obtaining a battery operation parameter data set; constructing a battery equivalent circuit model, and performing global and local two-stage optimization to obtain an optimal parameter set of the battery equivalent circuit model; constructing a state equation and an observation equation, and carrying out initialization setting on a covariance matrix of the triple forgetting factor adaptive extended Kalman filter to obtain a battery state observation model; priori state estimation is obtained through time updating operation, posterior state estimation is obtained through measurement updating operation, hierarchical estimation is carried out on the charge state and the polarization voltage of the battery, and the state characteristic quantity of the battery is obtained; according to the method, the reliability and the safety of the battery management system of the sightseeing vehicle are improved.

The invention discloses a power distribution network planning method and system considering distributed energy uncertainty, and relates to the technical field of power grid planning, and the method comprises the steps: collecting distributed energy node data, compensating space-time migration in combination with meteorological data, and carrying out the time sequence alignment through a dynamic time warping algorithm; constructing an improved Wasserstein generative adversarial network to generate a conventional scene, and injecting Gaussian noise through potential spatial disturbance to generate an extreme scene deviating from training distribution; inputting the mixed scene set into a mixed integer nonlinear programming model, and adopting a graph neural network to establish a topology-power flow agent model to accelerate solution; updating line impedance parameters through a Kalman filter, and collecting and checking actual output; and decomposing the corrected planning scheme into cloud global optimization and edge local control, and carrying out cloud-edge collaboration. According to the method, the adaptability of a power distribution network planning scheme in a complex and uncertain environment is improved by combining spatial-temporal feature alignment, adversarial network scene enhancement, a graph neural network and cloud edge collaborative optimization.

The invention discloses a lithium battery SOC (State of Charge) estimation method based on constrained optimization Kalman filtering. The method comprises the following steps: building a battery RC (Resistance-Capacitance) model; establishing a Kalman filtering model and initializing parameters; processing the collected battery data; iteratively updating Kalman filtering parameters continuously through constraint optimization to obtain more accurate values; mSE and RMSE are used for a Kalman filter to evaluate the performance of the filter. In order to overcome the limitation of the Kalman filter on parameter setting and the sensitivity of the Kalman filter on model change, the parameter adjustment process of the Kalman filter is optimized by introducing constraint conditions, and the parameters of the filter are dynamically adjusted in combination with an advanced optimization algorithm, so that the filter can keep the optimal estimation performance under different working conditions. Meanwhile, through limitation of constraint conditions, it is ensured that the estimation result conforms to the physical characteristics and actual working conditions of the battery, and the accuracy and reliability of SOC estimation are further improved.

The invention discloses a switch cabinet fault diagnosis method and system based on thermal imaging and voiceprint. The method comprises the following steps: acquiring historical operation data, voiceprint data and thermal imaging data of a switch cabinet and an operation site picture of the switch cabinet; carrying out noise reduction processing on the acquired voiceprint data and historical operation data by adopting an infinite Kalman filter; constructing a space-time attention mixed expert model to carry out training prediction on the voiceprint data set and the historical operation data set, and obtaining predicted voiceprint data and historical operation data at the next moment; constructing an improved MaxViT-DF model to classify the collected thermal imaging data and the operation site picture, and training to obtain a fault feature image and a fault-free image; and fusing fault feature data output by the space-time attention hybrid expert model and the MaxViT-DF model by adopting multi-modal orthogonal sequence fusion. According to the invention, the accuracy and efficiency of fault diagnosis of the switch cabinet are improved, operation and maintenance personnel can find and process potential problems, and the reliability and safety of a power system are improved.

The invention discloses an anesthesia early warning method based on multi-source signal fusion. The anesthesia early warning method comprises the following steps that physiological signals in the anesthesia process are collected in real time and preprocessed; a multi-modal Transform attention network is adopted to extract correlation features in the time sequence physiological signals, and multi-modal fusion features are obtained; predicting the fusion feature at the next moment by using a long-short-term memory neural network to obtain a prediction error; updating a noise and observation covariance matrix of the Kalman filter according to the prediction error; the dynamically adjusted Kalman filter calibrates the multi-modal fusion features in real time; according to the calibration characteristics, anesthesia depth and consciousness state prediction values are calculated in real time, and early warning is output. The real-time performance, the stability and the accuracy of anesthesia state monitoring are improved.

The invention relates to a multi-sensor fusion positioning method suitable for a multi-axle steering vehicle, and the method comprises the following steps: S1, installing a plurality of sensors on a multi-axle all-wheel steering vehicle platform, collecting IMU data, a laser radar real-time scanning point cloud, a GNSS signal, the reading of each wheel angle sensor, and the reading of each wheel speed meter, and carrying out the real-time scanning of the IMU data, the laser radar real-time scanning point cloud, the GNSS signal, the reading of each wheel angle sensor, and the reading of each wheel speed meter; and the timestamps of all the collected signals are aligned. And S2, establishing a kinematic model of a to-be-estimated state, and performing distortion removal on the laser radar point cloud data according to the IMU data to realize motion compensation of the point cloud. And S3, establishing an error state discrete model, a laser radar observation model, a multi-axis steering vehicle kinematics observation model and a GNSS observation model. And S4, considering the influence caused by multi-wheel non-Ackerman steering, and providing a method for adaptively adjusting the observation noise covariance of the multi-axle steering vehicle so as to ensure the correct weighting of chassis observation data. And S5, fusing IMU, GNSS, laser radar, and rotation angle and rotation speed observation information of each wheel of the multi-axis steering vehicle, constructing a sequential iterative error state Kalman filter, and obtaining an optimal positioning estimation result through iterative calculation.

A GNSS positioning slowly-varying fault detection method based on residual-SVR regression comprises the steps that an observation information sequence is acquired based on a Kalman filter, and a covariance matrix of the observation information sequence is calculated; accumulating multi-step information through a sliding window, and constructing chi-square statistics; based on the fault-free data, constructing a training set by taking an innovation sequence as input and chi-square statistics as output, and generating an innovation-statistics mapping function; and fitting a normal slope threshold value based on an SVR predicted value, carrying out least square fitting on an observation statistic curve by sliding a window in real time, and judging whether to start a slow change fault alarm or not. According to the method, the residual error sequence is directly used as model input, and the dynamic chi-square statistical magnitude is used for replacing a traditional dichotomy label, so that the detection delay is reduced; an SVR detection model based on grid search and cross validation collaborative optimization is utilized, and an optimal parameter combination of a minimum mean square error (MSE) is screened through logarithm uniform sampling, interval linear sampling and five-fold cross validation, so that the average absolute error of slowly varying fault detection is reduced.

The invention relates to the technical field of power transmission line anomaly recognition, and particularly discloses a power transmission line anomaly recognition model construction method combined with edge calculation, which comprises the following steps of: gradually increasing electromagnetic interference intensity test, acquiring current and voltage signals in real time, and monitoring electric field intensity and signal change by combining a high-precision sensor; dynamically calculating an influence value of electromagnetic interference; recursively estimating the dynamic influence of the electromagnetic interference through a Kalman filter, accurately modeling the influence of the electromagnetic interference by combining the system state and measurement data updating, and realizing the real-time evaluation of the influence of the interference on the system by calculating an electromagnetic interference coefficient and an electromagnetic shielding efficiency expected deviation; according to the method, a spectral analysis and high-frequency current and voltage fluctuation feature extraction method is adopted, and accidental interference and system faults are effectively distinguished by calculating a spectral ratio; through multi-level data analysis and optimization, real fault signals and electromagnetic interference signals can be accurately distinguished, and false alarms and missing alarms are reduced.

The invention discloses an intelligent sensing system for multi-modal data fusion, and relates to the technical field of intelligent traffic. The multi-modal data fusion intelligent sensing system comprises a data acquisition and preprocessing module, a multi-modal data fusion module, an intelligent sensing module, a target detection and behavior prediction module and a system verification and optimization module, and high-precision and high-stability target sensing and tracking are realized by integrating millimeter wave radar and camera data. The multi-modal data fusion module effectively combines physical features and visual features, and improves the accuracy and robustness of perception. The intelligent sensing module adopts a volume Kalman filter and a Hungary algorithm, state estimation and target association are optimized, and continuous and accurate tracking of the target is ensured. The target detection and behavior prediction module utilizes an advanced YOLOv5 model and a DeepSORT algorithm to realize multi-target accurate detection and trajectory prediction, and the intelligent level of the system is further improved.

The invention discloses a multi-camera cooperative visual tracking system and method for invading animals of a transformer substation, and the system achieves the precise monitoring and real-time tracking of the invading animals, such as birds, mice, cats and dogs, in a transformer substation region through the cooperation of a plurality of high-resolution cameras. The method comprises the following steps: firstly, acquiring a region video stream by a plurality of cameras, setting a main tracking camera (MTU), and extracting a dynamic target region through a frame difference method and a threshold segmentation technology; and then background noise is removed through area filtering, and a target contour is obtained. Target tracking adopts SIFT feature extraction and a brute force matching algorithm (BFMatcher) to realize cross-frame matching, and a Kalman filter is used to predict and update a target position. In order to ensure cooperative work of multiple cameras, a cross-view tracking technology is introduced into the system, and when a target crosses the view field of the cameras, adjacent cameras share target information, and a detection mode is activated in advance. The method overcomes the problem of a single visual angle blind area, effectively improves the monitoring efficiency, and is suitable for continuous monitoring and accurate tracking of animals around the transformer substation in a complex environment.

The invention discloses a thermal management strategy management method and system based on an energy storage battery system, and relates to the technical field of battery thermal management, and the method comprises the steps: collecting temperature data, carrying out the preprocessing, generating and recording a measurement result and a convective heat transfer coefficient, taking the measurement result and the convective heat transfer coefficient as the input of a Kalman filter, and outputting an optimal estimation result of a battery. The temperature data refer to the environment temperature and the core temperature of the single battery, based on the optimal estimation result of the battery, establishing a fuzzy rule base for iteration and updating, outputting a dynamic temperature difference threshold value, constructing a three-dimensional thermodynamic potential function, calculating the cosine similarity of a cooling driving force vector and a cooling mode, and correcting the cosine similarity to obtain a cooling driving force vector. The mode switching list is generated, the cooling equipment decision mode is generated, starting judgment and optimization of the cooling mode are carried out, and the intelligent level, safety and energy efficiency ratio of thermal management of the energy storage battery system are remarkably improved.

The invention relates to the technical field of vibration isolation control, discloses an air floating support quasi-zero stiffness active vibration reduction method based on model predictive control, and aims to solve the problems of high vibration isolation requirements of precision equipment in a low-frequency vibration environment, response lag of a traditional vibration isolation technology and insufficient control precision under uncertainty. The method comprises the following steps: step 1, carrying out system modeling on the vibration damping device to obtain a state space model of the vibration damping device, and introducing a Kalman filter into the state space model to predict state information in real time; step 2, constructing an optimization problem and a constraint condition of the MPC controller; step 3, carrying out robustness and stability analysis on the MPC controller; and step 4, fusing sensor data through the MPC controller after robustness analysis, solving an optimization problem through an optimization algorithm, calculating an optimal control input # imgabs0 # in real time, driving a voice coil motor, and suppressing low-frequency vibration.

The invention relates to the technical field of organic display data processing, in particular to a touch data processing method based on an organic display, which comprises the following steps of: monitoring a flexible deformation acceleration component and an environmental parameter change rate in real time through a dynamic trigger function, and triggering high-priority baseline calibration when a threshold value exceeds a limit; and generating a self-adaptive compensation amount matched with the current touch signal noise spectrum. And constructing a dielectric constant offset prediction model by using a long short-term memory neural network or a Kalman filter, deducing the dynamic change trend of the dielectric property of the material, and correcting a touch signal compensation coefficient in a grading manner according to the predicted offset. And a closed-loop error correction mechanism is formed based on the spatial distribution entropy reverse optimization model weight parameter and the trigger threshold boundary of the contact coordinate residual matrix. According to the method, the problem of time domain mismatch of fixed period calibration and organic material nonlinear offset is solved, the accumulation of touch coordinate analysis errors in a high-curvature deformation scene is effectively inhibited, and the touch positioning precision and the interaction response real-time performance are improved.

The invention discloses a chip test result real-time compensation method based on a PSO-KF algorithm, belongs to the technical field of integrated circuit testing, and solves the problem of how to improve the accuracy and stability of integrated circuit testing. Based on the real-time change of the measurement data and the environmental noise, a particle swarm optimization algorithm is adopted to optimize noise parameters in the Kalman filter; the particle swarm optimization algorithm is used for updating noise parameters after each measurement data is obtained, and the optimal parameters are applied to the Kalman filter again, so that real-time compensation of the measurement data is realized; and the yield is evaluated based on the compensated data, and the data is fed back to the particle swarm optimization algorithm for the next iterative optimization, so that the Kalman filter can be ensured to be adaptive to the noise change and keep the optimal compensation effect, and the technical scheme of the invention improves the precision, yield and product quality of the chip test.

The embodiment of the invention provides a construction hoist height measurement method and device based on double correction, and the method comprises the steps: building a pressure height mapping model for initial calibration through fusing the data of a high-precision air pressure sensor and a six-axis gyroscope; the weight of the sensor is dynamically adjusted by adopting a motion state evaluation matrix, and error correction is realized by combining null drift reference point records. And constructing a state transition equation and an observation equation by using an extended Kalman filter, and obtaining a high fusion result through a prediction updating step. Meanwhile, the system has the functions of abnormal state monitoring and sensor mode automatic switching, and double guarantees of measurement precision and reliability are achieved. The method effectively solves the problem that a traditional single sensor measurement scheme is unstable in precision in a complex environment.

The invention provides a BIM-based bridge construction progress monitoring method and system, and the method comprises the steps: obtaining an observation matrix based on BIM model data and sensor monitoring data of a target bridge; inputting the observation matrix into a preset Kalman filter to obtain a corresponding fusion matrix; carrying out Granger causal test by taking the overall progress deviation as a dependent variable and taking the stress ratio, the environmental index, the resource delay rate and the process progress deviation as independent variables to obtain a Granger causal test result; and constructing a corresponding DAG, and obtaining the reason causing the overall progress deviation based on the DAG. According to the scheme, BIM model data and sensor monitoring data are accurately fused, the DAG between the overall progress deviation and each piece of quantitative information is constructed, the quantitative information causing the overall progress deviation can be accurately obtained, and an effective basis is provided for subsequent decision making.

The invention relates to the technical field of book position retrieval, in particular to an intelligent library book retrieval method based on the Internet of Things, which comprises the following steps of: deploying an RFID (Radio Frequency Identification Device) dense read-write array on a bookshelf layer, arranging a UWB (Ultra Wideband) positioning base station in a library area channel, and embedding a composite label integrated with various sensors into a book cover; a dynamic position discrimination algorithm is introduced to identify the moving state of the book, an extended Kalman filter is utilized to fuse data to establish a motion trail prediction model, a hierarchical position display mechanism is adopted to provide detailed position information, and for temporarily placed books and overtime non-return books, the accuracy of positioning and tracking of the books is improved. And special processing and alarm prompting are respectively carried out, so that the problems of dynamic position updating lagging and inaccurate retrieval result in traditional library book retrieval are effectively solved, and the book retrieval efficiency of readers is remarkably improved.

The invention discloses a real-time multi-target detection system and method based on deep learning, and relates to the technical field of deep learning, and the method comprises the steps: collecting a video stream and an image sequence of a to-be-detected scene, and carrying out the preprocessing of the video stream and the image sequence; multi-scale feature extraction is carried out on the preprocessed input data by using a deep learning model, a multi-scale feature pyramid is constructed, and each layer of feature map represents target information of different scales; fusing the feature maps of different scales, introducing an attention mechanism, and performing weighting processing on the features of the target area; using an improved YOLOv5 target detection model to carry out target detection on the fused feature map, and outputting a bounding box and a category probability of a target; performing real-time tracking on the detected target by adopting a Kalman filter algorithm to generate a motion track of the target; carrying out post-processing on the detection result, wherein the post-processing comprises non-maximum suppression, target association and trajectory smoothing; and finally outputting a detection result and a tracking trajectory of the target.

The invention belongs to the field of aquaculture, and particularly relates to an underwater fish multi-target identification, tracking and motion analysis method, the method is based on a multi-target attitude tracking framework based on direction perception, and the framework comprises an improved YOLOv11 model (YOT model) and a directional bounding box attitude tracking module (OPBT module). The YOT model integrates directional bounding box detection and attitude estimation, adopts a task-aligned dynamic detection head structure, and combines a multi-task loss function to realize efficient direction perception detection and key point positioning. The OPBT module is based on an unscented Kalman filter, integrates rotation IoU, angle similarity and body length similarity, and improves identity continuity and occlusion recovery capability in multi-target tracking. The framework is suitable for complex scenes such as high-density fish schools, the real-time processing performance can be kept, meanwhile, the detection and tracking precision can be remarkably improved, and stable and reliable technical support is provided for individual behavior analysis.