1431results about How to "Improve training effect" patented technology

The invention provides a medical image segmentation method, which comprises the following steps: acquiring PET indirect development original image data, and preprocessing the image, including denoising, contrast enhancement and gray level normalization, to obtain a preprocessed PET image data set; designing segmentation middleware, integrating a plurality of segmentation algorithms including threshold segmentation, region growth and level set methods, and selecting proper segmentation algorithms for different types of lesion regions according to feature vector analysis results; pET image data with labels are generated through the simulation technology, a training data set is constructed in combination with real case data, the sample size is expanded through the data enhancement technology, and a large-scale data set used for algorithm training is generated; and the segmentation result of the U-Net is refined by using the segmentation middleware, the lesion area edge is focused on, and the lesion boundary is accurately positioned through an edge detection algorithm to obtain a fine segmentation result.

The invention relates to the technical field of power system operation training, and discloses a virtual reality-based power grid fault simulation training system and method, and the system comprises a virtual scene modeling module which is used for constructing a three-dimensional model of power grid equipment and an operation environment thereof, and providing a vivid training environment for subsequent modules; and the interaction and operation module is connected with the virtual scene modeling module and realizes interaction between the trainee and the virtual environment by using a VR head-mounted display, a handle and a motion capture device. By adopting a high-precision three-dimensional modeling technology, PBR material rendering and an LOD optimization strategy, high simulation of power grid equipment and an operation environment is realized, students can experience a vivid visual effect and immersion in a virtual reality scene, and meanwhile, by combining a gesture recognition technology realized by motion capture equipment and a convolutional neural network, the real-time simulation of the power grid equipment and the operation environment is realized. Operation actions of trainees can be accurately captured, a natural and visual interaction mode is provided, and the authenticity and operation experience of the training system are greatly improved.

The invention provides an HVAC system temperature control and energy consumption optimization method based on deep reinforcement learning. Through an entropy-based data screening method, a strategy evaluation module and a design reward function, HVAC system temperature control and energy consumption optimization are realized under external condition interference. According to the entropy-based data screening method, the value of data for agent training is calculated through an error formula, high-value data is repeatedly used for agent training, and the training effect is enhanced; the strategy evaluation module is used for further solving the problem of overestimation of a Q value on a double-Q structure, correcting the output of the Q value, reducing the influence of a random Q value, and adding regularization of a strategy and an actual action into a loss function updating process, so that the influence of random noise on the Q value is eliminated, and the accuracy of strategy output is ensured; a reward function for controlling the temperature and optimizing the energy consumption is designed, so that the intelligent agent can balance the contradiction among multiple tasks, and the optimal solution of multi-target task intersection is searched, so that the HVAC system can control the temperature to be always kept in the temperature range, and meanwhile, unnecessary energy consumption is reduced.

The invention relates to a method for evaluating the situation awareness ability of a pilot based on a Ti-SA model. The Ti-SA model comprises an eye movement trace multi-dimensional attention enhancement module, an AOI multi-scale comprehensive feature module and a fusion module. The evaluation method comprises the steps of collecting eye movement trace time sequence data and flight training videos of a pilot at continuous moments in real time, and obtaining corresponding AOI fixation time sequence data based on the eye movement trace time sequence data, the flight training videos and a preset AOI subdivision area; preprocessing the eye movement trace time sequence data and the AOI fixation time sequence data; inputting the preprocessed eye movement trace time sequence data into an eye movement trace multi-dimensional attention enhancement module to obtain a category probability value corresponding to the eye movement trace time sequence data; inputting the preprocessed AOI fixation time sequence data into an AOI multi-scale comprehensive feature module to obtain a category probability value corresponding to an AOI fusion feature; and inputting the two category probability values into a fusion module to obtain a predicted situation awareness level.

The invention discloses a driving skill training method constructed based on a diversified virtual examination room, and relates to the technical field of data processing, and the method comprises the steps: constructing a system based on the diversified virtual examination room, constructing a virtual examination room scene to carry out the driving training of a training user, outputting a driving behavior data set in real time, and carrying out the scoring, and analyzing the data according to the driving score, generating a personalized training path in combination with historical data when the driving score does not reach the standard, and training the user to carry out training step by step according to the training path until a preset skill scoring target is achieved. The technical problems that in the prior art, a traditional driving training method cannot dynamically adjust a training path to meet the actual skill requirements of trainees, consequently, training content lacks pertinence and is low in efficiency are solved, the individuation and pertinence of driving skill training are improved, it is ensured that the training content is highly matched with the actual skill requirements of the trainees, and the driving skill training efficiency is improved. And the training efficiency and effect are improved.

The invention relates to the technical field of medical simulation training, in particular to an anesthesia puncture virtual simulation operation training and analysis system which comprises a force feedback calibration unit, a force feedback adjustment unit and a force feedback output unit. A theoretical force feedback value is calculated based on a human tissue mechanical model library, tissue levels and characteristics are accurately considered, in a deviation correction link, a force perception deviation value is periodically recalculated, data are processed by applying a sliding window averaging method, overall deviation and local deviation are distinguished and correspondingly adjusted, and the overall deviation is uniformly corrected; functional fine adjustment is locally set according to puncture depth sections, time-dependent force feedback of tissue viscoelasticity is considered, components are calculated through convolution integral and synthesized into final values with elasticity and inertia components, force feedback accuracy is improved comprehensively, individual differences and operation details of trainees are fitted, and anesthesia puncture training quality and effect are improved efficiently.

The invention discloses a training and evaluation system and method based on the competency of a pilot, and relates to the technical field of educational training, and the system comprises a learning progress monitoring module which is used for collecting the learning data, including the learning time and the task completion condition, of the pilot in the training process in real time, and carrying out the deep analysis of the data, according to the method, the learning progress and the potential difficulty of the pilots are recognized, and a learning progress report is formed, the individual difference and the potential difficulty of each pilot are recognized by monitoring the learning progress and performance of the pilots in real time, and then the training content, the training method and the training progress are adjusted; the targeted training mode ensures that the training content is always matched with the actual requirements of the pilots, the potential of each pilot can be played to the maximum extent, the overall training effect is improved, in addition, real-time feedback and suggestions provided by the system also help the pilots to correct errors and optimize learning strategies in time, and the training effect is further improved.

The invention relates to the technical field of electrocardiogram monitoring, in particular to an electrocardiogram monitoring method and system based on heart rehabilitation data. The method comprises the following steps that electrocardiosignals in different motion states are collected through a multi-lead electrocardiograph and motion monitoring equipment, and fatigue electrocardio mapping is carried out to obtain fatigue signals. Then, heart rate changes of the fatigue electrocardiogram mapping signals are analyzed, the disordered scatter point trend of the heart rate is quantified, dynamic inflection point distribution is fitted, and trend dynamic data are obtained; multi-scale rule intersection analysis is carried out based on the data, a fatigue heart rate change rule is extracted, and rehabilitation deviation quantitative evaluation is further carried out. Finally, the rehabilitation deviation evaluation result is sent to the terminal for real-time electrocardiogram monitoring, and heart rehabilitation management is supported. According to the method, the electrocardiogram monitoring technology is optimized, so that the electrocardiogram monitoring technology is more perfect.

The invention discloses a polyp image segmentation method based on boundary clue depth fusion, and the method comprises the steps: inputting an intestinal polyp image into a backbone network through constructing a boundary clue depth fusion network, and extracting different levels of feature information, so as to guide the extraction of a polyp region; a boundary clue mining module is constructed to obtain boundary detail information by means of partial hierarchical features, and more accurate positioning of a polyp area is realized; constructing a foreground target enhancement module to enhance the learning of the network on the overall structure characteristics of the foreground target in combination with the reverse attention branch, so that the network focuses on the foreground target; and constructing a depth feature fusion module to integrate boundary information and foreground information, further improving the segmentation performance through cascade feature fusion and boundary information guidance, and finally obtaining a polyp segmentation result by using the trained model. According to the method, the limitation of the polyp in boundary uncertainty and morphological diversity is overcome, the weighted joint scale difference loss optimization model is adopted in training, and high-precision recognition and segmentation of the polyp region are realized.

The invention is applicable to the technical field of deep learning and data processing, and provides a time sequence prediction model training method, a time sequence prediction model comprises an input layer, a first branch network, a second branch network and an output layer, and the first branch network and the second branch network each comprise a MambaFFN module. The model training method comprises the steps of performing normalization processing on a training data set through an input layer, performing variable embedding processing on the normalized training data set through a first branch network to obtain a first representation, performing patch embedding processing on the normalized training data set through a second branch network to obtain a second representation, and performing patch embedding processing on the second representation to obtain a second representation. And performing fusion processing on the first representation and the second representation through an output layer to obtain a prediction result on each time step in the prediction time step, and performing iterative training on the time sequence prediction model by adopting a loss function based on the prediction results until the loss function converges to obtain the time sequence prediction model. Therefore, the training effect of the model and the precision and accuracy of the model on future prediction are improved.

The invention relates to a Transform and RF fused power load prediction method, which comprises the steps of obtaining short-term power load data to be predicted in a target area, inputting the short-term power load data into a trained random forest network RF after clustering and decomposition, and outputting a power load prediction result. The RF training process comprises the following steps: acquiring power load historical data and corresponding historical weather data of a specific area, and preprocessing to obtain a first training data set and a second training data set with labels respectively; clustering and decomposing the first training data set with the label and the second training data with the label based on the power load historical data and the time sequence characteristics of the corresponding historical weather data; respectively inputting the clustered and decomposed power load data into a multi-modal time sequence prediction converter to obtain a feature sequence; and the feature sequence is transmitted to RF for regression prediction, regression loss is calculated, and the random forest network is iteratively optimized through the regression loss. The accuracy of a power load prediction result can be obviously improved.

The invention relates to a cross-domain cold start article recommendation method based on triple contrast learning. The method comprises the following steps of: performing biased processing on article attribute information and knowledge graph information in a source domain and a target domain to obtain initial attribute representation and initial graph representation, and acquiring initial user representation from interaction information of a user and the article in the source domain by utilizing a multi-thermal coding method; obtaining the screened attribute representation by using a multi-layer perceptron, and obtaining the aggregated map representation by using a map attention network; on the basis of current attribute representation, graph representation and interaction information, a recommendation result is obtained through a trained cross-domain cold-start item recommendation model, and the model is constructed and optimized through a self-comparison learning loss function, a source domain comparison learning loss function, a cross-domain comparison learning loss function and a BPR loss function in a joint training mode. Compared with the prior art, the method has the advantages of fully utilizing hot article collaboration information, reducing field difference, improving article content characterization quality and the like.

The invention belongs to the technical field of virtual reality, and discloses a virtual rich real environment training system based on benign stress, comprising: a database module for storing cognitive tasks and virtual scenes; the acceptance testing module is used for testing a preset cognitive task of the user and generating a scene acceptance index; and the initial training task generation module is used for generating a training task comprising two groups of scene parameters and one cognitive task according to the index. And the training execution module is used for constructing a dynamically changing virtual training scene according to the two groups of parameters and executing a cognitive task in the virtual training scene. And the training task updating module is used for generating a new training task based on the execution result and the acceptability index. And the training planning module formulates an overall training plan and controls an execution process. Personalized training content is provided by evaluating individual differences of the users, it is ensured that the training difficulty is matched with the capabilities of the users, and the training effect is optimized while the user participation degree and the possibility of insisting in training for a long time are improved.

The invention relates to a visual attention mobility anti-attack method based on self-adaptive token fine tuning, which comprises the following steps of: constructing a feature important matrix by utilizing a proxy model middle layer, and generating a semantic-guided sparse mask; performing token-level point multiplication on the sparse mask and the initialized disturbance, and performing linear superposition on the sparse mask and the initialized disturbance and an original image to serve as initial input data; inputting the input data into the agent model for iterative training; in each round of iterative training, sparse masks are continuously updated, the number of discarded tokens is gradually increased, and gradient variance between layers in an agent module is reduced through gradient back propagation by using a gradient penalty factor and a balance factor. Meanwhile, self-adaptive truncation factor sets are designed for different modules of the agent model to weaken the attention mechanism of the model, and through T times of iterative training, a target loss function is maximized to obtain optimal mobility disturbance. Experiments prove that the method has better attack effect and mobility on the attack of the target model.

The invention discloses a flight space directional training auxiliary system and method, and the system comprises a flight task simulation module; the meteorological environment simulation module is used for simulating a meteorological scene corresponding to a flight task, and simulating the meteorological scene by establishing a multi-dimensional meteorological model and combining meteorological parameters; the meteorological environment parameter control module comprises a time step length control module which adopts a preset time step length to input meteorological parameter observation data collected in real time to the meteorological parameter setting module; the specific meteorological event control module is used for transmitting meteorological parameters of a specific meteorological event area to the meteorological parameter setting module when the airplane triggers a specific meteorological event; and the training scheme generation module is used for generating a proper training scheme according to the historical training information of the flight personnel. According to the system and the method, a real and personalized simulation training environment can be provided for the flight personnel, and the professional skills and the capability of coping with various flight conditions of the flight personnel are effectively improved.

The invention relates to a lower limb diagnosis training system, and belongs to the field of medical diagnosis and rehabilitation training. The system comprises a detection part used for collecting data information related to the lower limbs of a patient; and the control unit is used for analyzing and processing the data information so as to calculate and obtain a diagnosis result. The control unit can calculate a neural function index based on the electroneurographic signal parameters, calculate a joint activity index based on the joint activity, calculate a muscle strength index based on the muscle strength evaluation value, and further calculate an initial lower limb function comprehensive score; the initial lower limb function comprehensive score is used as an initial diagnosis result to output a control signal of a numerical range of a combination of traction forces in multiple directions, so that electromyographic signal parameters and blood flow parameters acquired by a detection part in real time in the training process of a patient can be used by a control unit for correcting the initial lower limb function comprehensive score; and obtaining a dynamic lower limb function comprehensive score for adjusting the specific numerical value of each traction force distribution.

The invention discloses a caesarean section training method and system based on mixed reality, and the method comprises the steps: obtaining the medical image data of the pelvis and abdomen of a pregnant woman, recognizing and marking the boundary and region of a key tissue structure, constructing a three-dimensional model of the key tissue structure, and generating key indexes of caesarean section training. Importing the constructed three-dimensional model into an MR system, collecting action data, generating a real-time feedback prompt, receiving a training ending signal, generating an evaluation result, recording operations which do not conform to an evaluation standard, and providing targeted improvement suggestions for trainees. According to the method, the quality and efficiency of operation training can be effectively improved, a safe, efficient and immersive training platform is provided for trainees through precise processing of medical images, efficient reconstruction of three-dimensional models, real-time feedback and personalized operation evaluation, personalized guidance can be carried out according to the performance of the trainees, and the training efficiency of the trainees is improved. And the training effect of surgical skills is further improved.

The invention discloses a Pick ball motion trail tracking and tactical decision-making auxiliary system which adopts a high-frame-rate camera, a micro sensor installed on a ball and an athlete motion sensor worn on an athlete to respectively collect video image data, ball motion data and athlete motion data in the Pick ball motion process. Precise positioning and trajectory tracking are carried out on a ball through a target detection and tracking algorithm based on deep learning, a tactical knowledge model is constructed by combining site information and athlete action data and applying a decision tree and a neural network algorithm, real-time tactical decision suggestions are provided for athletes and coaches, and the training efficiency is improved. And the competitive level and the training effect of the Pick ball sports can be improved.

The invention discloses a polyp segmentation method based on edge guidance and deep supervision PVT. The method aims at a colonoscope image with a polyp, and outputs the position and the shape of the polyp by using a constructed polyp segmentation network. The polyp segmentation network uses PVT-v2 as a backbone network for feature extraction, and further comprises a context fusion decoding branch, a boundary extraction branch, a boundary sensing module and a cascade feature fusion branch. And the context fusion decoding branch serves as an image decoder and receives a semantic feature map output by the PVT-v2 backbone network. And the boundary extraction branch enriches the content of the detail feature map by using a refining module to obtain an edge prediction result. And the boundary sensing module fuses the refined features and the output feature map of the context fusion decoding to obtain a boundary sensing prediction result. And the cascaded feature fusion branch fuses fusion features under different scales layer by layer according to a sequence from deep to shallow, and outputs a final polyp segmentation prediction result P5.

The invention discloses a transfer learning-based cross-data center storage and calculation cooperative training acceleration method. The method comprises the following steps of S1, deploying an initial version of a deep learning model in a plurality of heterogeneous data centers; s2, performing initial training on the deep learning model in the first data center to generate a first model parameter set; s3, generating a local deep learning model parameter set; s4, establishing a dynamic weight adjustment mechanism of the heterogeneous data center; s5, distributing different computing resources to each heterogeneous data center according to the participation weight; s6, scheduling storage resources and computing resources in real time according to resource use conditions and task requirements of the heterogeneous data centers; s7, performing multi-task parallel processing; s8, monitoring the resource utilization rate of each heterogeneous data center through a dynamic resource scheduling strategy; and S9, generating a final global optimization model parameter set, and applying the global optimization model parameter set to subsequent task deployment of each heterogeneous data center. According to the method, the collaborative training efficiency of the cross-heterogeneous data center is optimized.

The invention relates to the technical field of intelligent position information recommendation, in particular to an intelligent position recommendation method based on artificial intelligence, which comprises the following steps: collecting basic information and behavior data of a user, and constructing a user portrait; acquiring position information, extracting key information to establish a position information base, and introducing an expert system to supplement the position information; understanding implicit significance of position description through a natural language processing method and a semantic analysis method; performing feature extraction on the user portrait and the position information, and constructing a feature data set; training a recommendation model; user feedback is actively inquired through an active learning algorithm; and generating a position recommendation list for the user through the trained recommendation model. The method optimizes description of position information, is more convenient for users to understand, and solves the problem that accurate user portraits cannot be constructed due to lack of data for users and new positions in specific fields through an active learning algorithm and introduction of an expert system.

The invention discloses a VVC-SCC intra-frame coding method and device based on multi-stage irregular coding unit division, and relates to the field of video coding, and the method comprises the steps: designing and training a coding unit division prediction model based on a multi-stage convolutional neural network; secondly, performing coding unit division prediction model prediction on an input video frame through a trained coding unit division prediction model, and outputting prediction probabilities of mode types of all CUs of the current frame; and finally, setting different threshold values for different mode categories, and guiding a VVC-SCC encoder to perform encoding according to the prediction probability and the corresponding threshold values, thereby skipping unnecessary traversal of the mode categories. The problems that an existing VVC-SCC intra-frame coding method is long in coding time, and the calculation complexity of a coder is high are solved.

The invention discloses a model training method, a data processing method, electronic equipment, a computer readable storage medium and a computer program product, and relates to the technical field of large model technology and data processing. The model training method comprises the following steps: training an initial model by adopting first training data to obtain a first training result; evaluating the first training result from a plurality of evaluation dimensions to obtain feedback information; second training data is constructed based on the feedback information and multi-role information, and the multi-role information is used for setting a plurality of different roles in the training data construction process; and performing iterative adjustment on the initial model by adopting the second training data to obtain a target model. The technical problems that the model training efficiency is low and the effect is poor due to the fact that the degree of dependence of training data set construction on manpower is high are solved.

The invention discloses a 2D-CNN network wake flow electromagnetic field detection method based on wavelet transform time-frequency spectrum characteristics. The method comprises the steps of (1) generating an underwater vehicle wake flow electromagnetic field calculation data set through a multi-physics field coupling method of hydrodynamics and electromagnetism, (2) extracting time-frequency spectrum features of a wake flow electromagnetic field through wavelet transform, and (3) developing a wake flow electromagnetic field detection model based on a 2D-CNN deep neural network. And training the model by using the data after the wavelet transform for wake flow electromagnetic field detection. According to the 2D-CNN network wake flow electromagnetic field detection method based on the wavelet transform time-frequency spectrum characteristics, the detection capability of the wake flow electromagnetic field can be effectively improved, a new technical means is provided for detection of the wake flow electromagnetic field of an underwater vehicle, and the method has a good application prospect in the field of marine target electromagnetic field monitoring and early warning.

The invention relates to a training method and device of a quantum generative adversarial network, equipment and a storage medium. The method comprises the steps that the Rayleigh divergence relation between real quantum state data and forged quantum state data output by a generator is constructed, a first loss function is constructed based on the Rayleigh divergence relation, and the first loss function is used for enabling the difference between the forged quantum state data output by the generator and the real quantum state data to be minimum; constructing a second loss function of the discriminator, wherein the second loss function is used for maximizing the probability that the discriminator outputs a false first discrimination result and a true second discrimination result; and alternately executing the first training of the generator and the second training of the discriminator until a training cut-off condition is reached, and completing the training of the quantum generative adversarial network. By adopting the method, the quantum generative adversarial network can be prevented from falling into a barren plateau phenomenon, so that the training effect of the quantum generative adversarial network is improved.

The invention provides a force feedback rehabilitation finger separating plate which comprises a finger separating plate body, a measuring module, a signal processing module, a display module and a music feedback module, the finger separating plate body comprises a first flexible layer, a supporting plate and a second flexible layer which are stacked in sequence, and the supporting plate has certain hardness and can be bent; a plurality of sealing cavities are formed between the first flexible layer and the supporting plate and serve as sensing units, the sensing units are connected with a measuring module through pipelines to indirectly measure hand force data, and each sensing unit comprises a plurality of bags so that the sensing units can make contact with the hand to achieve deformation measurement; the signal processing module processes pressure signals obtained by the measuring module and then transmits the pressure signals to the display module and the music feedback module, training data are monitored in real time, rehabilitation training and music feedback are combined, the interestingness of the training process is improved, and the treatment enthusiasm of a patient is improved.

The invention belongs to the technical field of teaching video processing, and particularly relates to an interactive physical education system and method based on video image processing. According to the interactive physical education teaching method based on video image processing, training actions of a user can be captured in real time and are compared with standard actions for analysis, so that correct actions and deviation actions can be accurately recognized, continuous correct time periods and continuous deviation time periods can be determined through analysis of key nodes, and the accuracy of physical education teaching is improved. According to the method, the deviation action is subjected to orderliness analysis, independent deviation and progressive deviation are distinguished, targeted correction suggestions can be provided according to the type of the deviation action, the user is helped to improve the action and improve the training effect, in addition, the training action of the user can be fed back in real time according to the improvement amount, and the training effect is improved. The user can know the training effect in time, adjustment is carried out according to feedback, through continuous correction analysis and feedback, the user can gradually improve the motion quality, and the better training effect is achieved.

The invention discloses an intelligent simulation system and method based on multi-modal interaction and a dynamic teaching strategy, and the system comprises a multi-modal interaction emotion perception module, a dynamic teaching strategy generation module, a situational multi-post training module, an automatic guidance and evaluation module, and an integration and optimization module. Specifically, the multi-modal interaction emotion perception module comprises a visual information acquisition module and a voice information acquisition module; and the visual information acquisition module is used for capturing facial expression, eye movement and gesture information of the student through a camera. The invention belongs to the technical field of simulation training and intelligent teaching, and achieves the technical effects that the limitation of the prior art is broken through by organically combining the technologies of multi-modal emotion perception, dynamic teaching strategy generation, situational multi-post training, multi-dimensional intelligent evaluation and the like, and compared with the prior art, the system has the advantages that the system is simple in structure and convenient to use. The method has obvious technical advantages in the aspects of real-time emotion perception, self-adaptive teaching strategies, cross-role interaction, multi-dimensional evaluation and the like.

The invention discloses a spoken English man-machine conversation interaction training method and system, and relates to the technical field of spoken English training, spoken English audio signals of a user are monitored in real time through a microphone, and the frequency spectrum information of the audio signals is refined into each frame of spectrogram and cepstrum coefficient feature vectors by using Fourier transform and feature extraction technologies; therefore, the voice signal characteristics of the user can be completely captured; three core dimensions of semantic similarity, grammatical structure similarity and fluency are incorporated into an evaluation model, a comprehensive oral English training evaluation system is formed, a text sequence is generated by analyzing a time sequence dependency relationship through a convolutional neural network, and the oral English training evaluation model of a deep learning technology is combined, so that the oral English training evaluation accuracy is improved. According to the method, the expressions of the user in the aspects of semantics, grammar and fluency can be accurately measured, the oral training level of the user can be comprehensively judged by the system by comparing the comprehensive evaluation coefficient with the preset qualified threshold value, and a personalized learning guide is provided according to a result.

The invention provides a multi-task-based power transmission line stockbridge damper defect detection method, which comprises the following steps: acquiring an original power transmission line stockbridge damper image, and generating a data sample close to a real fault sample; a multi-task detection network model is established, a classification loss function and a segmentation loss function are constructed for the multi-task detection network model, and the established multi-task detection network model comprises a shared feature layer, classification branches and segmentation branches; based on a data sample, training the multi-task detection network model by adopting an alternate training mode to obtain a trained multi-task detection network model, and in the training process, adaptively adjusting weight parameters by adopting an adaptive dynamic weighting strategy; and inputting a to-be-detected image of the vibration damper of the power transmission line into the trained multi-task detection network model to obtain segmentation positioning and defect detection results of the vibration damper of the power transmission line. According to the invention, the problem of poor robustness and detection performance in the prior art can be solved.