9results about How to "Reduce data dependence" patented technology

The invention discloses an improved YOLOv8 fish image recognition method based on transfer learning, and the method comprises the following steps: collecting a fish image data set, and carrying out the preprocessing of the data set; correspondingly marking category names on the fish image data set according to fish categories, and dividing the fish image data set into a training set, a verification set and a test set in proportion; a standard YOLOv8 model is improved; all fish images in the ImageNet data set are adopted to pre-train the improved YOLOv8 model, and obtained weight parameters are migrated to the model; performing optimization parameter training on the improved YOLOv8 model after transfer learning by using the training set and the verification set; and performing target detection on the test set through the optimized improved YOLOv8 model, and outputting a fish identification result. According to the method, the feature extraction and recognition precision is effectively improved, the computing resource demand is reduced, and the robustness and adaptability of the model are enhanced; according to the method, the learning efficiency and the recognition precision of the model are remarkably improved, data dependence is reduced, and the model has high generalization ability.

The invention relates to the technical field of system testing, and provides a test case filtering method and related equipment based on an accessibility prediction model, and the method comprises the steps: generating a function call graph and a plurality of control flow graphs of a target system, and carrying out the integration to obtain an interprocess control flow graph; calculating the reachability representation of the test case based on the interprocess control flow diagram; performing feature extraction on the test cases to obtain case features, and constructing an association diagram of the test cases; carrying out reachability prediction by utilizing a reachability prediction model based on the use case characteristics and the association graph to obtain a prediction reachability representation; optimizing the reachability prediction model by using all the prediction reachability representations and all the reachability representations, and performing reachability prediction on the target test case by using the optimized reachability prediction model to obtain a final reachability representation; and filtering the target test case according to the final accessibility representation to obtain a filtering result. According to the method, the universality of test case filtering can be improved.

The application provides an industrial equipment fault diagnosis model construction method and related equipment, and relates to the technical field of industrial equipment fault diagnosis. The method comprises the following steps: obtaining a physical mechanism model of the industrial equipment; the physical mechanism model comprises a physical constraint equation; constructing a deep learning network; the deep learning network comprises a composite loss function constructed according to the physical constraint equation; collecting full-time series state data of the industrial equipment during operation; the full-time series state data comprises specified time series state data collected when the industrial equipment is in a healthy state; based on the full-time series state data, the deep learning network is trained for parameter optimization to obtain a final available industrial equipment fault diagnosis model. The method of the application aims to solve the problems of weak early fault identification ability, poor interpretability, lack of physical meaning and strong data dependence in the existing industrial equipment fault diagnosis technology, and effectively improves the intelligent level and operation and maintenance efficiency of the industrial equipment fault diagnosis.

The application discloses a boiler combustion optimization parameter optimization method and system based on an autoencoder. The method collects real-time operation data from a boiler DCS system through OPC communication and carries out preprocessing, including cleaning, feature engineering and normalization and the like. An autoencoder model is constructed to carry out unsupervised feature learning, essential features of a combustion process are directly extracted from a large amount of unlabeled historical data, and low-dimensional hidden space features are generated. A multi-objective function with the object of maximizing boiler efficiency and minimizing pollutant emission is constructed, a Pareto optimal solution set is searched in the hidden space, and the final executable optimization parameter is calculated through an aggregation strategy. The optimization parameter can be directly output to a combustion control system, so that the optimization effect is continuously effective.

The invention provides an electroencephalogram signal cross-equipment decoding classification model training method, a decoding classification method and equipment. The training method comprises the following steps: pre-training a hidden space deep neural network corresponding to a classification model based on an adversarial generative network based on a pre-training feature vector and a decoding type label corresponding to each electroencephalogram signal segment; using the network as a discriminator in a classification model, and inputting the feature matrix of the electroencephalogram signal fragment into an auto-encoder based on the convolutional neural network in the classification model to train the classification model; and training the classification model into an electroencephalogram signal cross-device decoding classification model for outputting a decoding type of electroencephalogram signal fragments and an electroencephalogram signal acquisition device type. According to the method, the problem that a decoding classification model trained by electroencephalogram signals of cross-manufacturer equipment is poor in reliability and effectiveness can be solved, and the data utilization rate of the electroencephalogram signals for training is increased.

This invention relates to the interdisciplinary fields of biomedical engineering, artificial intelligence, and neuromodulation, and provides a cognitive fatigue assessment and electrical stimulation intervention system based on a combined EEG-behavioral AI model. The system includes: a data acquisition and preprocessing module for simultaneously acquiring and preprocessing the user's EEG signals and behavioral data; a real-time fatigue assessment module for extracting and fusing features from the preprocessed EEG signals and behavioral data using the combined AI model, outputting the user's real-time fatigue assessment results; the combined AI model is constructed based on the Transduction Information Maximization (TIM) algorithm to achieve cognitive fatigue assessment in scenarios with few samples; and a closed-loop electrical stimulation intervention module for dynamically adjusting transcranial electrical stimulation parameters and performing targeted intervention based on the real-time fatigue assessment results. This invention achieves real-time, accurate detection and personalized targeted intervention of cognitive fatigue states through multimodal data fusion and closed-loop intervention.

This invention discloses a remote sensing image target detection method based on single-sample contrast feature changes. Belonging to the field of target detection technology, the method includes: acquiring an original remote sensing image set; preprocessing the original remote sensing image set to obtain an augmented image set; establishing a multi-head contrastive learning (MBR) training feature encoding network; inputting the augmented image set into the MBR training feature encoding network to obtain a trained feature extraction network; acquiring a remote sensing image to be tested and a conditional image; constructing a conditional target detection framework, wherein the conditional target detection framework includes the trained feature extraction network; and inputting the remote sensing image to be tested and the conditional image into the conditional target detection framework to obtain several targets in the remote sensing image that share the same interest category as the conditional image. This invention can extract same-category and rotation-invariant features from the remote sensing image and the conditional image, reducing the model's data dependency and improving its transferability.

The invention provides a non-vision-field imaging method and device based on quasi-Fresnel transformation, and belongs to the technical field of optical non-vision-field imaging. The method comprises the following steps: collecting photon echo data when non-vision-field imaging is carried out on a non-vision-field target as a detection signal; calculating a convolution size parameter of the detection signal based on quasi-Fresnel transform; calculating a two-dimensional complex representation of the detection signal based on the convolution size parameter; calculating a two-dimensional complex representation of the non-view target based on the two-dimensional complex representation of the detection signal; and calculating reflectivity and depth information of the non-vision target based on the two-dimensional complex representation of the non-vision target so as to realize imaging of the non-vision target. According to the method, the two-dimensional features of the surface of the non-vision-field target are fully utilized, the non-vision-field target can be accurately and efficiently reconstructed, and the method has the advantages of being high in calculation speed, small in used memory, low in data dependence and accurate in imaging and positioning.

This invention relates to a method for calculating the residual magnetism of a converter transformer based on inrush current, belonging to the field of power systems. Based on obtaining the deep saturation excitation characteristic curve of the converter transformer, this method uses the per-unit value of the effective inrush current generated by no-load closing at the peak voltage of the converter transformer as a parameter to calculate the theoretical per-unit value of magnetic flux. Using the starting point for calculating the per-unit value of residual magnetism, the ratio of the rated voltages of the primary and secondary sides of the converter transformer, and a weighting constant, a superposition weighting formula for the residual magnetism of the converter transformer and the rated voltage magnetic flux is determined. Finally, the residual magnetism of the converter transformer core is calculated using the calculated per-unit value of the theoretical magnetic flux corresponding to the inrush current and the superposition weighting formula for the residual magnetism and the rated voltage magnetic flux. The method proposed in this invention is simple to implement, highly operable, and has minimal impact on the converter transformer, providing an effective verification strategy for calculating the residual magnetism of converter transformers and evaluating the effect of residual magnetism elimination.