262results about How to "Solve the loss" patented technology

The invention discloses a multi-layer semantic understanding large model agent construction and application method. The method comprises the following steps: S1, preprocessing input data; s2, analyzing the preprocessed data by using a large language model and dialogue history, identifying user intentions and extracting deep semantic features; s3, key information is extracted from data intending to be recognized, structured representation is carried out, and model configuration is adjusted through dynamic hyper-parameter optimization; s4, querying the knowledge graph, associating and organizing the extracted key information with the existing data, and generating a structured knowledge representation; s5, vectorizing the structured knowledge representation, and storing the vectorized structured knowledge representation in a vector database to support rapid similarity search; and S6, in combination with the generated unified semantic representation and the large model, performing deep analysis on user input, generating coherent and logical personalized answers, and returning the personalized answers to the user. The method can support various application scenes such as intelligent question and answer, content generation, sentiment analysis and multi-modal search.

The invention belongs to but not limited to the technical field of emotion recognition, and discloses a multi-modal classroom emotion recognition method and system based on modal adaptive learning, and the method comprises the steps: building a multi-modal data collection system, collecting the multi-modal data, such as texts, audios and videos, in a classroom teaching process in real time, and carrying out the real-time collection of the multi-modal data; and preprocessing and feature extraction are carried out on data of different modes. For emotion expression feature differences of different classroom scenes, a multi-modal data fusion method based on adaptive modal scoring (AMS) is provided, and a time sequence modeling and emotion classification model is constructed. The multi-modal fusion and time sequence modeling mechanism simulates the multi-dimensional perception of the teacher on the emotional states of the students in the class, and enhances the interpretability of the emotion recognition system. According to the system, weighted fusion is carried out on data features from texts, audios and videos through fusion of multi-modal data, and complete emotion feature representation is formed.

The invention discloses a large model knowledge base construction method based on multi-granularity retrieval. The method comprises the steps of obtaining a target text and analyzing a document structure, splitting and reorganizing text data according to three levels of titles, fragments and sentences, and constructing three levels of knowledge of the titles, the fragments and the sentences; using a large language model to construct a generative index as question-level knowledge according to the sentence-level knowledge; using a vector model to generate a first vector representing index keys of fragment-level knowledge and / or sentence-level knowledge, and a second vector representing index keys of title-level knowledge and / or question-level knowledge; and respectively storing the first vector and the second vector into a vector database, and storing the index values and the index types corresponding to the first vector and the second vector into a relational database. According to the method, multi-granularity document knowledge base construction is achieved, the expression modes of knowledge under different granularities are expanded, the accuracy of user query and knowledge matching is improved, and the problem that information is lost or incomplete in an existing method is solved.

A three-dimensional target detection and tracking method fusing multi-sensor information belongs to the technical field of computer vision, and can make up for the defect of a single mode by collecting camera and radar image information and fusing camera images and laser radar point clouds. Multi-target tracking is carried out by using a multi-data coupling method of 2D + 3D data, a remote object is identified through a camera, when the object enters an induction range of the laser radar, the laser radar can carry out more accurate positioning, and position estimation of the target in a 2D image is corrected by using a more accurate 3D position provided by the laser radar. The method is used for solving the problems of inaccurate target identification, target shielding and the like under complex road conditions; a traditional single-mode three-dimensional target detection algorithm cannot effectively fuse the advantages of multi-sensor data and is difficult to accurately identify objects in a distance or under a complex road condition.

The invention provides a helmet standard wearing detection method based on improved YOLOv8, and belongs to the technical field of computer vision target detection. The technical problem that an existing detection algorithm cannot judge whether the safety helmet is worn normatively or not is solved. According to the technical scheme, the method comprises the following steps: S1, collecting sample data; s2, a variable convolution Deform ConvolutionV2 is adopted to replace a traditional convolution, and a variable convolution is adopted to replace the traditional convolution; s3, DCNV2 improvement based on MPCA is carried out; s4, an improved multi-scale feature fusion module; and S5, training the improved YOLOv8 network model by using the helmet standard wearing data set to obtain a trained model. The beneficial effects of the invention are that the method not only helps a manager to effectively supervise the wearing condition of the safety helmet on the working site, but also provides a new idea and example for the target detection technology based on deep learning in the field of safety helmet detection.

The invention discloses an image semantic segmentation method based on cross-dimension multispectral edge fusion. The method comprises the following steps: firstly, constructing a bilateral multi-modal fusion encoder based on hierarchical attention, and extracting low-level and high-level features in an RGB image and a thermal infrared image; secondly, introducing an edge detection branch (namely a jumping edge guide structure), and enhancing the utilization of global semantic information; and finally, designing a deep supervision mechanism oriented to a plurality of loss functions, and supervising the network with the plurality of loss functions. According to the method provided by the invention, the RGB image and the thermal infrared image are used as input, various attributes and interaction between low-level and high-level features in a dynamic traffic scene are comprehensively considered, and the problems of information loss and the like are solved by fusing semantic and edge features, so that accurate and robust semantic segmentation of the RGB-thermal infrared image under a complex illumination condition is realized.

The invention relates to the technical field of ramie seed vigor evaluation methods, in particular to a ramie seed vigor evaluation method and system based on multispectral image analysis, and the method comprises the following steps: collecting a multispectral image and a photoautotrophic fluorescence image of a ramie seed to be detected; constructing a feature matrix of the multispectral fluorescence image, extracting a feature clustering center based on a feature clustering decomposition algorithm, and forming spectral feature data of the ramie seeds; carrying out dimensionality reduction on the feature matrix based on singular value decomposition to generate spectral principal components of the ramie seeds; projecting a spectral principal component to a low-dimensional feature matrix based on nonlinear embedded mapping; generating a spectral index matrix based on the low-dimensional feature matrix; and inputting the spectral index matrix into a depth inversion prediction model to obtain the vigor index of the ramie seeds, and effectively solving the problems of information loss and overlapping in the dimension reduction process in the traditional technology through the feature clustering decomposition and singular value decomposition technology in the processing stage of the multispectral image data.

The invention discloses a diseased fish detection method and device, and belongs to the technical field of computer vision detection. In order to solve the problems that in the prior art, diseased fish disease detection efficiency is low, accuracy is not high, and real-time detection cannot be achieved, a diseased fish detection model is constructed, and the diseased fish detection model comprises a backbone network, a neck network and a detection head network; the method comprises the following steps: inputting a diseased fish image into a backbone network based on a spatial deep convolutional neural network for feature extraction to obtain a visual feature map, and then inputting the visual feature map into a contrast-enhanced attention transfer structure of a neck network for feature fusion processing to obtain a fused feature map; and the fusion feature map is input into a detection head network, and a sick fish target detection result is obtained through detection of a multi-head prediction small target detection algorithm, so that fish school disease detection is performed in real time, the detection efficiency and detection precision of fish diseases are effectively improved, and the method has relatively high practicability and application universality.

According to the electric meter intelligent dynamic complementary reading method and device based on big data analysis, the method comprises the steps that historical data are analyzed, the reading data items and the reading priority of an electric meter are intelligently adjusted, task data sets needing complementary reading are combined into a composite task according to requirements, automatic execution is executed in the non-busy time period, and the intelligent dynamic complementary reading efficiency of the electric meter is improved. And the data acquisition efficiency and accuracy are improved. A master station system is used for issuing a configuration scheme and a task to terminal equipment, data items of different electricity meters are sorted, and deep mining and analysis are carried out based on collected data within a past set time (flexible adjustment can be carried out according to specific field requirements); and a set of personalized complementary reading configuration strategy suitable for various ammeters is deeply combed and constructed, so that intelligent dynamic complementary reading is realized. The measure not only effectively guarantees the basic data support of big data processing, but also assists a decision application level to obtain more accurate and effective results.

The invention provides a layer focusing illustration single-view 3D reconstruction method, which comprises the following steps of: performing region segmentation on an original illustration single view in a data set through SAM to obtain a region text pair fine-tuning CLIP model comprising the original illustration single view and a layer channel of the original illustration single view; extracting an unnatural contour line, semantic information and layer information from each original illustration single view in the data set T; selecting a basic element model to initialize a 3D point cloud according to a role posture in a single view, segmenting the human body point cloud into different parts through point cloud segmentation, carrying out noise growth and color perturbation on corresponding point cloud regions according to position information of an unnatural contour line, respectively mapping the point cloud regions into a submerged space, and obtaining a final potential representation by adopting LDM; and finally, the final potential representation passes through a decoder to obtain a final 3D representation, and 3D assets are reconstructed. According to the invention, the corresponding 3D assets can be generated according to a given illustration single view.

The invention provides a self-supervised medical image segmentation method and system based on structural relation modeling, and the method comprises the steps: carrying out the modeling of a complex relation among nodes, sub-graphs and a whole graph, thereby comprehensively capturing the structural information of the graph; by constructing and optimizing the structural relation model of the graph, the performance of graph representation learning is improved, the method is particularly suitable for graph data with a complex topological structure and non-uniform distribution, and the segmentation effect of the medical image is improved by modeling the organization structure relation in the medical image.

The invention belongs to the technical field of data storage, and particularly relates to a storage method and system of nonvolatile data, which realize high efficiency and safety of dynamic data storage by distributing at least two Flash pages for data storage, writing a basic information block into the head of each Flash page and associating a key value information block for each data object. According to the method, the length of the data entry can be dynamically adjusted, the problem that data may be lost when power failure occurs in the read-write process is solved through the strategy of writing first and erasing second, and meanwhile the purpose of prolonging the service life of Flash is achieved through dynamic management of the data entry and reasonable scheduling between pages.

The invention discloses a lithium battery health state and residual life prediction method and system. The method comprises the following steps: preprocessing original charging data, eliminating noise and abnormal values, and obtaining normalized data; and performing structured processing on the normalized data in a differential mode to generate structured data. And in combination with a convolution trunk line module, a channel sparse attention residual module, a space sparse attention residual module and a global attention residual module, extracting a global time sequence feature map from the structured data. The feature maps are fused with historical data of the health state of the lithium battery to form a fused feature map, prediction is performed based on the fused feature map, the health state and the residual life of the lithium battery are obtained, through an optimized difference mode, the information amount of data can be enriched, the model can sensitively capture subtle state changes in the continuous cycle period, and meanwhile, the model can be used for predicting the health state of the lithium battery. Potential time sequence features in the charging data can be effectively extracted through the attention residual module, and the prediction accuracy is improved.

The invention discloses an embedded firmware audio data modification method and device, a medium and computer equipment, and relates to the technical field of audio processing, and the method comprises the steps: carrying out the correlation analysis of adjacent audio frames in a second-level cache, carrying out the coding processing, and storing the coded audio frames in a third-level cache; dividing the audio frames in the three-level cache into M frames and N frames according to the computing power and bandwidth condition of the embedded device, locally encoding the M frames by the embedded device, and transmitting the N frames to a rear-end device; the back-end equipment receives and processes the N frames of audio signals, and returns a processing result to the embedded equipment; the embedded device combines the locally coded M frames of audio signals and the back-end processed N frames of audio signals according to a time sequence, and outputs complete audio signal data; through timestamp synchronization, interpolation frame supplementation and frame skipping processing, the problems of audio frame loss and discontinuous time sequence are solved, the integrity and continuity of audio signals are ensured, and the audio processing efficiency and quality are remarkably improved.

The invention discloses a database construction method and device for nonlinear data based on a large model. The method comprises the steps that a node is selected in an initial graph structure to input request data of a user, a retrieval result of the request data is retrieved and generated through an RAG system, and reply data of the request data is generated through the large model; writing the request data, the reply data and the retrieval result into a new node, determining a nearest node associated with the new node in the existing nodes, and forming a branch path with topicality by taking the nearest node as a father node of the new node; storing the new node in a database; wherein all the branch paths form a tree-shaped structure. According to the method, each dialogue node has a traceable father-son relationship by constructing a database storage mode of a multi-branch tree structure, and a user can branch a new dialogue path from any historical node. And the user can freely select different topic paths for dialogue, so that the dialogue context content of the large model is more accurate.

The invention discloses a rock mineral content detection system and a rock mineral content detection method for geological survey, which are characterized in that after image features and spectral features of rocks and ores are extracted, internal generality of the image features and the spectral features of the rocks and ores is mined by using a data processing technology based on deep learning, and the internal generality of the image features and the spectral features of the rocks and ores is used as a mask; according to the method, unique information in rock and ore image features and spectral features is emphasized, and the general information and the unique information of the rock and ore image features and the spectral features are comprehensively utilized through compensation type interactive fusion, so that accurate identification of rock and ore categories is realized. In this way, the problem of information loss in the multi-modal feature fusion process can be effectively solved, and therefore the accuracy of rock and ore classification and the calculation precision of the mineral content are improved.

The invention provides a closed-loop nerve synchronous regulation and detection system, and belongs to the technical field of neuroscience. The system comprises an electrophysiological regulation subsystem which is used for implementing nerve stimulation regulation on a detected object, has a filtering and amplifying preprocessing function and can receive and send out nerve regulation signals; the signal processing subsystem is used for receiving the preprocessed neural signals, sending out control signals for the electrophysiological regulation and control subsystem and synchronously analyzing and processing the preprocessed neural signals; and the neural multi-electrode is used for detecting an original neural signal. According to the system, a neural signal analysis method based on the power spectrum entropy of wavelet packet decomposition node reconstruction signals is used for calculating and analyzing neural signals collected by multiple neural electrodes, and then closed-loop regulation and control are carried out by comparing the neural signals with a set threshold value. According to the system, the experiment operation in the nerve regulation and control work is simplified, and the experiment efficiency is improved.

The invention discloses a connecting piece welding quality detection method and system, a medium and electronic equipment, and relates to the technical field of industrial automation and machine vision. The method specifically comprises the following steps: controlling a moving mechanism carrying an imaging device to move to a detection position, and obtaining a to-be-detected image containing a connecting sheet; inputting the to-be-detected image into a pre-trained target detection model, and obtaining a prediction bounding box of a connection sheet region; performing size expansion based on the prediction bounding box, and generating a region of interest covering the deformation region of the connecting sheet; in the region of interest, extracting a target feature point coordinate of a connecting sheet and a centroid coordinate of a welding spot by using an image processing algorithm; and calculating a distance value between the coordinates of the target feature point and the coordinates of the mass center of the welding spot, and judging the welding quality of the connecting piece based on the distance value. The invention aims to realize high-precision and high-efficiency welding quality detection of the connecting piece under the conditions of complex background interference and uncertain target form.

The invention belongs to the technical field of mobile phone display screen detection, and particularly discloses a defect detection method based on a smart mobile phone display screen. Graying and denoising the collected surface defect image of the original display screen; performing feature labeling and expansion on the preprocessed image; on the basis of the U-Net network, a residual block is inserted, common convolution is replaced by depth separable convolution, and an attention module is added; performing batch training on the cross entropy loss function by using a stochastic gradient descent method, continuously updating parameters of the network model, and optimizing the parameters of the network model by using an Adam method; when the output value of the loss function is reduced and does not change any more, iteration is stopped; and inputting a to-be-detected mobile phone display screen image into the trained network model for detection, if the mobile phone display screen is detected to have a defect, outputting a defect binary image of the mobile phone display screen, and calculating the real size of the defect. According to the invention, surface defects such as scratches and black spots of the mobile phone display screen can be rapidly and accurately detected.

The invention belongs to the technical field of polyurethane foam materials, and particularly relates to a modified China-hemp core powder composite polyurethane foam material which comprises the following components in parts by mass: 30-50 parts of polymer polyol, 10-40 parts of modified China-hemp core powder, 20-30 parts of a flame retardant, 1-2 parts of a catalyst, 3-4 parts of a foaming agent, 1-2 parts of a foam stabilizer, 65-80 parts of organic polyisocyanate and 0.5-1 part of white oil. The modified China-hemp core powder adopts titanium oxide modified China-hemp core powder; the invention also provides a preparation method of the foaming material. The problem of antibacterial durability of the China-hemp modified polyurethane foaming body is solved, an organic-inorganic composite antibacterial and bacteriostatic system can be formed by the titanium oxide material and the antibacterial material in the China-hemp core by utilizing the wrapping property, antibacterial and bacteriostatic properties and restorability of the titanium oxide material, and the antibacterial material is wrapped by the silica material, so that the problem of loss of the antibacterial material is effectively solved.

The invention provides an RGB-D semantic segmentation method based on feature collaboration and multi-scale feature fusion, and belongs to the field of computer vision and computer graphics. According to the method, semantic segmentation prediction is performed on an image based on an encoder-decoder network structure, multi-modal feature correction and fusion are realized by using multiple attention mechanisms at each stage of a network, feature representation of input data is optimized and enhanced, and a top-down multi-scale feature fusion strategy is used in a decoder. Through multiple attention mechanisms, the model can effectively capture global features and local features, fusion of different modal information under the same scale is realized, and fine aggregation of features under different scales is promoted. According to the method, cooperation of multi-modal features and fusion of multi-scale features are realized by applying multiple attention mechanisms, and the problems of difference between RGB and depth information, noise interference in data and loss of detail information in a network extraction process are effectively solved.

The invention belongs to the technical field of energy storage, and provides a battery pack SOC value calculation method, a battery pack, energy storage equipment and a medium. The method comprises the following steps: acquiring parameter information related to the SOC of the battery pack; building a mathematical model y (t) of the SOC value of the battery pack by using a neural network model according to the parameter information related to the SOC of the battery pack, and calculating an initial SOC value of the battery pack according to the mathematical model y (t); the humidity value of the current moment in the battery pack is obtained, the initial SOC value of the battery pack is corrected according to the humidity value of the current moment in the battery pack and the temperature value T (t) of the current moment of the battery pack, and the effective SOC value of the battery pack is obtained. Therefore, the stability and the safety of SOC calculation in a special environment can be improved, and the control capability of a user on the health state of the battery is enhanced.

The invention discloses an unsupervised multi-source information domain self-adaption method and device for rotating machine migration diagnosis, a medium and a product, and relates to the field of mechanical fault diagnosis. The method comprises the following steps: acquiring multi-source information; multi-source information is fused and compressed into a single fusion sample; constructing an information fusion enhanced domain adaptive self-attention network transfer learning model; and inputting the single fusion sample into the information fusion enhanced domain adaptive self-attention network transfer learning model to obtain a transfer diagnosis result of the rotating machinery. According to the method, the problems of negative migration, information loss, high calculation burden and the like in a known multi-source information domain adaptive method can be solved.

The invention relates to a TransUNet liver parenchyma image segmentation method based on an attention mechanism, and the method comprises the steps: introducing AxialAttention and a residual connection mode in TransUNet, solving a problem of information loss in skip connection through the method, better capturing a long-range dependence relation between positions in a liver parenchyma feature map, and improving the segmentation accuracy of the liver parenchyma image. The accuracy and expression ability of the liver parenchyma segmentation model are improved.

The invention provides an NVMe permanent memory allocation method and an NVMe controller, and is applied to the technical field of data storage, and the method comprises the steps that a host end reads permanent memory configuration information of the NVMe controller of an equipment end of NVMe equipment, and the permanent memory configuration information is information of configuring a permanent memory function by the equipment end in a power-on initialization stage; configuring each protocol register of the NVMe controller, and enabling a permanent memory function of a device end; and when the waiting duration after the permanent memory function of the device end is enabled is greater than or equal to the preset waiting duration in the permanent memory configuration information, taking the double-rate synchronous dynamic random access memory in the NVMe solid state disk of the device end as the permanent memory of the host end, so that the host end performs read-write operation on the permanent memory. According to the scheme, the double-rate synchronous dynamic random access memory of the equipment end is used as the permanent memory of the host end, so that the problems of memory shortage of the host end, memory power failure and data loss are solved.

The invention belongs to the field of computer vision, and particularly relates to a small target detection method and system based on feature fusion, and the method comprises the steps: inputting an image into a backbone network for feature extraction, and obtaining a multi-scale feature map; inputting the multi-scale feature map into CIAM for feature information integration to obtain feature information containing a target context; connecting output features of the 37th layer and the 50th layer in the backbone network to a fusion network in a cross-layer manner to obtain a fusion feature map combined with the original feature information of the target; and inputting the fused feature map into a detection network to obtain a detection result. According to the method, the CIAM is integrated into the yolov7 feature extraction network to integrate the target features and related background information in the multi-scale feature map, and meanwhile, the cross-layer feature fusion network is used to reserve the original features of the target, so that the problem of low recognition precision of the small target caused by lack of feature information and sensitive positioning error is solved.

The invention discloses a method and system for realizing high-speed data transmission through storage depth adjustment, and the method comprises the steps: mapping a storage space of PCIE (Peripheral Component Interface Express) equipment into a storage region with a set number of blocks, and carrying out the coding; the PCIE device stores data needing to be transmitted in storage areas in sequence, and each storage area serves as a data block; dynamically adjusting the length of the data block to ensure that the computer can respond to the DMA data transmission request; and the computer responds to the DMA data transmission request, reads the codes of the data blocks and processes the codes. According to the method, the problem of data loss caused by the fact that interrupt signals are not collected is solved through storage depth adjustment and data block coding methods. And the size of the data block in the storage area can be adjusted to adapt to data traffic under different sampling rates and match the processing speed of a computer.

The invention relates to an acoustic metamaterial general database construction method, device and equipment. The method comprises the steps that acoustic metamaterial models are generated in batches through modeling software; for each acoustic metamaterial model, internal entity object recognition and file rewriting are firstly carried out, and each entity object in the model is exported as an independent grid model file to be stored; discretization processing, voxel surface reconstruction and material physical parameter information addition are sequentially carried out on each grid model file, and a point cloud file of each acoustic metamaterial model is constructed; and finally, on the basis of the grid model file and the point cloud file corresponding to each acoustic metamaterial model and an acoustic absorption coefficient table of the acoustic metamaterial model calculated by simulation modeling software, constructing an acoustic metamaterial universal database. By adopting the method, the problem of cross-configuration data loss can be solved, and three-dimensional space feature aggregation and comprehensive characterization of the acoustic metamaterial model are realized.

The invention relates to a multi-modal medical image fusion method based on image denoising and illumination enhancement. The method comprises the following steps: inputting a multi-modal medical image into an illumination and contrast enhancement network for illumination intensity correction to generate an enhanced image; adding random noise to the multi-modal medical image to obtain a noise image; inputting the noise image into a denoising and reconstruction network to carry out denoising and reconstruction tasks; using the enhanced image and the multi-modal medical image as labels to jointly supervise and train a denoising and reconstruction network to obtain a feature encoder and a feature decoder with denoising and illumination enhancement functions, using the trained feature encoder to respectively extract features of different modal images to obtain modal feature maps, and using the modal feature maps to obtain the multi-modal medical image; and carrying out feature fusion on the modal feature maps of the different modal images, and reconstructing the fused feature maps by adopting a trained feature decoder to obtain a multi-modal medical image fusion result. The problems that image fusion is interfered by noise, illumination is degraded and fusion efficiency is low are solved.

The present application provides a communication method, device and system, the method comprising: receiving a first notification message sent by a gateway according to the mechanism start time determined by the time of starting the WiFi service; sending a first confirmation message to the gateway, so that the gateway monitors the Zigbee service initiated by the Zigbee node within the time window of each time period after the mechanism start time; communicates with the Zigbee node for WiFi service at a time outside the time window of each time period; starts a negotiation working mechanism, and after the mechanism start time, if a trigger event occurs, communicates with the gateway for Zigbee service within the time window of the time period. The method can effectively solve the problem of sudden information loss.