35 results about "Hierarchical coding" patented technology

The present application relates to the technical field of aquaculture, in particular to a multi-source aquaculture data low-latency anti-interference communication method and system for unmanned feeding ship, comprising the following steps: multi-source aquaculture data acquisition; edge data fusion and instruction layered coding encapsulation, real-time fusion processing through a lightweight neural network at the unmanned feeding ship end to generate decision instructions corresponding to the aquaculture event; multi-node anti-interference networking; data transmission scheduling; data reception and state feedback. By deploying a MobileNet-SSD lightweight neural network at the unmanned feeding ship end, the present application realizes edge real-time cross-modal fusion of aquaculture water environment data and feeding state data, combines layered coding design driven by aquaculture events and channel quality change rate adaptive transmission strategy, avoids data redundancy, guarantees complete transmission of key instruction information when channel quality deteriorates, and effectively solves the problem of insufficient data fusion and transmission protocol adaptability.

The application discloses a time capsule expression structure reservation method based on deep learning and a knowledge graph, and aims at solving the problems that the expression and organization structure of a user's sealed multi-modal content in a digital time capsule service is difficult to be identified and recorded, and the display according to the uploading time or the fixed template of the media type in the opening stage is easy to cause the distortion of the expression intention and the fragmentation of the receiving experience. The multi-modal content input in the sealing stage is normalized and time-aligned, the content representation is obtained by using multi-modal hierarchical coding, the segmentation boundary and hierarchical relationship are determined and the key weight and object information are extracted by combining span boundary detection and a pointer network, the narrative rhythm parameters are obtained by using a neural time point process modeling based on a media switching event sequence, a timing knowledge graph is fused and constructed, and the arrangement sequence and the presentation control parameters are generated under the sequence constraint and the segmentation continuous constraint by graph representation learning, and the presentation data for the display in the opening stage is generated, so that the technical effects of reserving the user's expression structure and rhythm and differentially arranging and presenting in the future are realized.

This invention provides a lightweight knowledge base construction method based on hierarchical coding. The method includes processing documents in a knowledge document library using a lightweight semantic coding model on a cloud server to generate corresponding high-dimensional semantic vectors. Each high-dimensional semantic vector is mapped to a fixed-length binary fingerprint using a pre-defined Locality Sensitive Hash (LSH) function cluster, constructing a global fingerprint index library. On a mobile terminal, fingerprint index data from the cloud server is received and stored. Responding to user-inputted query text, a query fingerprint is generated using the same lightweight semantic coding model and LSH function cluster as on the cloud. By calculating the similarity between the query fingerprint and the fingerprints in the locally stored fingerprint index data, relevant document information is retrieved, and the results are returned. The beneficial effects of this invention are that it can reduce the knowledge base storage overhead by more than an order of magnitude while keeping the accuracy loss of semantic retrieval within 5%.

This invention relates to a beamforming and resource allocation method based on communication-aware joint optimization, applied to hierarchical coding communication scenarios. The method includes the following steps: Step 1, initializing the beamforming matrix and rate allocation vector; Step 2, solving the perception optimization subproblem based on the current beamforming matrix and rate allocation vector to obtain a new beamforming matrix; Step 3, solving the communication optimization subproblem based on the current beamforming matrix and rate allocation vector to obtain a new rate allocation vector, and determining whether convergence has occurred. If not, proceed to Step 2; if yes, proceed to Step 4; Step 4, solving the communication-aware joint optimization problem based on the current beamforming matrix and rate allocation vector to obtain the final beamforming matrix and the final rate allocation vector, thus realizing beamforming and resource allocation. Compared with existing technologies, this invention has advantages such as large capacity, high speed, and low complexity.

The present application relates to the technical field of digital circuit control, and particularly relates to a contact state monitoring and control method of a direct current contactor. The method comprises the following steps: monitoring the contact state of the direct current contactor and collecting the driving code signal of the direct current contactor, and extracting the characteristic code after filtering; performing spiral hierarchical identification on the characteristic code, analyzing the hierarchical information to distinguish the driving instruction, and obtaining the characteristic code identification result; distributing the characteristic code identification result to the corresponding driving circuit; outputting the rated driving voltage to the corresponding direct current contactor coil according to the characteristic code identification result, driving the coil to attract the armature, and closing the contact of the direct current contactor. The present application realizes fine management and control and abnormal processing of the direct current contactor by combining filtering and spiral hierarchical coding identification, and improves the instruction execution accuracy and the control whole-process stability.

The application discloses a double-branch Raio gating graph aggregation multi-modal meningioma segmentation method, belongs to the cross technical field of computer vision and medical image processing, is used for meningioma segmentation, and comprises the following steps: preparing a data set, constructing a deep neural network model, and performing neural network training; inputting the data set into the trained deep neural network model; first inputting a double-branch multi-modal input module to output a multi-modal fusion feature tensor; then inputting the multi-modal fusion feature tensor into a hierarchical Swin Transformer encoder to output three-dimensional segmentation masks of enhanced tumors, tumor cores and whole tumor regions of meningioma. Through double-branch multi-modal input fusion, hierarchical Swin Transformer coding, edge perception modulation and Laplace gating graph aggregation, high-precision three-dimensional segmentation of enhanced tumors, tumor cores and whole tumor regions of meningioma is realized.

A method of signalling a signal adjustment when encoding an input signal using a hierarchical coding scheme to manage a display aspect ratio. The hierarchical coding scheme comprises encoding a downsampled version of the input signal using a first encoding method to produce a first encoded signal. When the downsampling operation of the hierarchical coding scheme is a non-square downsampling operation, the method signals an adjustment to cause a pixel aspect ratio of the first encoded signal to be adjusted by a scaling factor, wherein the scaling factor is determined from the non-square downsampling operation. Additionally, correction is applied in an enhancement level of the hierarchical coding scheme to ensure that the signalled change of pixel aspect ratio does not lead to unintended consequences in the enhanced reproduction at a decoder-side.

The application relates to the technical field of artificial intelligence, and discloses a multi-modal large language model intelligent decision-making method and system fusing cross-domain information. The method comprises the following steps: performing hierarchical coding processing on cross-domain multi-modal input data corresponding to a target decision event, generating observation evidence data and conditional constraint data, and performing modal confidence estimation processing, cross-domain conflict relationship construction, gating fusion processing and historical decision feedback updating on the basis of the observation evidence data and the conditional constraint data. The method can distinguish the category tendency of the corresponding mode from the reliability of the evidence, identify real conflicts and pseudo conflicts, reduce the interference of low-quality modes and invalid differences on the fusion decision result, and simultaneously perform closed-loop correction on the modal confidence estimation parameters and the gating fusion parameters based on the historical decision feedback data, thereby improving the accuracy, stability and self-adaptive ability of the multi-modal large language model intelligent decision-making fusing cross-domain information in a time-varying environment.

Methods and systems for executing a Gaussian online splatting model for simultaneously localizing and mapping a surrounding 3D space are disclosed. The model is configured to receive an image-based data sample representing a first field of view of the 3D space and, using a Gaussian 3D map of the model, to render a new image-based data sample representing a new field of view distinct from the first, as well as to render the corresponding speech features. By incorporating a hierarchical encoder and a contrastive speech-image pre-training model (CLIP model) into the architecture of the Gaussian online splatting model, the overall architecture is configured to operate in near real-time.

The application discloses a smart shelf goods location health state evaluation method and system, and relates to the technical field of smart warehouse equipment state monitoring.The application realizes complete reservation and cross-goods location correlation of goods loading dynamic characteristics by collecting dynamic differential waveforms of goods in the migration process between goods locations and splicing the waveforms into data stream fragments along the migration path.On this basis, layered coding of the shape and time constant of each waveform can accurately depict the bearing response characteristics of different goods locations, and then through the step-by-step consistency comparison along the path, the interference of goods differences and environmental noise is effectively eliminated, and the sensitivity and specificity of goods location health state discrimination are significantly improved.

The present application relates to the field of physical exploration technology, in particular to a pseudo-random code emission and inter-well multi-parameter tomography exploration method and system, comprising the following steps: according to the spatial position, hole depth structure and target layer distribution of a plurality of boreholes in the inter-well exploration area; dividing the boreholes into a plurality of emission well sections and receiving well sections corresponding to the target layer, and configuring different emission well sections with mutually distinguishable pseudo-random code section identifiers, respectively, to generate layered coding observation sequences corresponding to each emission well section, receiving well section and code section identifier. The present application binds the emission well section and the orthogonal code section identifier one by one, realizes the distinguishable superposition and efficient collection of multi-well section signals in a continuous emission process, and realizes the accurate separation of mixed responses through matched filtering, significantly improving the collection efficiency and data availability, and solving the problems of low efficiency and easy signal aliasing of traditional step-by-step excitation.

The application provides a kind of multi-modal adaptation device for low-resource multilingual document analysis, comprising: visual encoder, by progressive block merging, image is hierarchically encoded, a set of feature maps is generated, and projection and spatial alignment are carried out, to generate the visual feature map of uniform space grid, while extracting page-level description vector through global average pooling;Adaptive visual-text mapping module, the visual feature map is decomposed into horizontal, vertical, diagonal and global four kinds of anisotropic representation, the gate weight of four direction branches is dynamically calculated according to the description vector and language embedding, the dynamic weighting of fusion features is realized, and the visual feature embedding is obtained by mapping the fusion features to the decoder embedding space;Autoregressive multilingual decoder, by adding separable two-dimensional position coding in visual feature embedding and flattening as sequence as visual prefix, combined with task prompt word containing language identifier, task type, output format and optional spatial prompt, to generate output.

This invention relates to the field of multimedia data processing technology, and more particularly to a semantically aware compression and AI-powered intelligent processing method for multimodal digital media. This method extracts visual and textual features and constructs a cross-modal semantic association graph. It then labels the semantic hierarchy of data blocks based on semantic centrality and performs hierarchical encoding and storage. During processing, target nodes in the association graph are retrieved according to task requests, data from the corresponding bitstream channels is selectively decoded, and multimodal information is fused based on node connectivity. This method achieves efficient storage of compressed data and accurate, rapid reconstruction processing for specific AI tasks.

A method for extracting crop planting structure and calculating area in irrigated areas includes: acquiring a temporally continuous and spatially registered remote sensing dataset; selecting key phenological period images of key temporal phases of target crop growth changes from the remote sensing dataset to form a key phenological period image set; constructing a labeled plot-level sample feature library by associating prior information related to spectral and index features and prior information of irrigation zones with labels; constructing a multi-temporal crop semantic segmentation network based on SegFormer, including a hierarchical Transformer encoder and a lightweight multilayer perceptron decoder, and training the constructed multi-temporal crop semantic segmentation network using the labeled plot-level sample feature library; overlaying pixel-level crop category probability maps with cultivated land plot boundary vector data, and obtaining plot-level crop types using post-processing methods; and based on the plot-level classification results, calculating the planting area of ​​each major crop in different irrigation zones, and outputting crop planting structure maps and area statistics.

The application discloses a kind of multi-contract interactive transaction network modeling method and system based on hypergraph, belong to information security technical field, the application first abstracts four kinds of heterogeneous nodes of user, contract, token and application program, generates feature vector using the layered encoder of fusion static attribute and dynamic timing;Second, based on atomic operation, price influence and other semantic rules, combined with space-time clustering algorithm dynamically generates hyperedge, and the weight is calculated by fusing the multi-factor function of time attenuation, fund size and attack mode;Finally, sliding window is used to construct timing sub-hypergraph, and the global network is updated and matrix representation is generated by union operation.The application accurately describes cross-contract collaborative behavior such as lightning loan by hyperedge, and improves the real-time and accuracy of transaction modeling by space-time feature fusion, providing high-dimensional and accurate data basis for attack detection and vulnerability positioning in the field of decentralized finance.

The application discloses a kind of super-resolution high-definition image generation methods based on optimization generative neural network, constructs super network model: based on hierarchical encoding-decoding structure, utilize Mamba residual block to construct an end-to-end super network model;Shallow feature extraction module is converted into high-dimensional feature space by single-layer convolution to input, then multi-scale feature processing is carried out by hierarchical Mamba processing module, finally, the final super-resolution image is obtained by upsampling module, multi-scale feature fusion module and image reconstruction module.The application is used for the processing of river channel satellite remote sensing image, and high-quality image super-resolution processing can be realized efficiently.Can the existing 10m resolution river remote sensing image be converted into 2.5m resolution (4 times super-resolution), effectively improve the river detail recognition ability.

The application discloses a kind of electric vehicle charging load prediction method and system based on space-time frequency multi-domain state space model.The method first uses discrete wavelet transform to decompose non-stationary load data into orthogonal low-frequency trend component and high-frequency characteristic component, the frequency domain component is fused with time embedding feature and space embedding feature, realize the multi-dimensional feature extraction of original charging load sequence;Subsequently, the extracted features are input into the hierarchical coding structure, and the core uses parallel bidirectional Mamba processing flow to synchronously capture the forward and backward long-term time dependence of the sequence.Thirdly, the encoded latent variable is input into the symmetric multi-layer decoder, the multi-scale information from the encoder is integrated using the skip connection, and the recursive feature recovery and refinement are performed through the bidirectional Mamba flow.Finally, the high-precision prediction result is output through the linear mapping layer.The application uses a linear computational complexity state space model to solve the problem of excessive computational overhead when the traditional Transformer model processes long sequences, as well as the non-linear, volatility and long-distance dependence problems that traditional models cannot handle.Experiments show that the method improves the prediction accuracy while greatly reducing the inference power consumption, and the system is suitable for deployment on the power grid scheduling terminal with lightweight requirements, helping to balance and schedule the load in the power network.

The embodiment of the application discloses a resource allocation method, device and equipment of a satellite-ground network and a storage medium, and specifically discloses the following: constructing a three-dimensional decision matrix according to a link existence state between a satellite and a ground station, a channel allocation identifier and a resource allocation weight; using a preset hierarchical coding scheme to code the three-dimensional decision matrix into a chromosome individual; the chromosome individual includes a link layer gene, a channel layer gene and a control layer gene; generating an initial population based on a preset communication constraint condition and the chromosome individual; the communication constraint condition is generated based on at least one of a communication elevation angle constraint, a propagation loss constraint and a channel conflict constraint; optimizing the initial population based on a fitness function to obtain a target resource allocation scheme; the fitness function is determined based on total interference, channel utilization and communication quality loss; and the target resource allocation scheme is used for allocating communication resources.

The present document relates to a method of layered encoding of a compressed sound representation of a sound or sound field. The compressed sound representation comprises a basic compressed sound representation comprising a plurality of components, basic side information for decoding the basic compressed sound representation to a basic reconstructed sound representation of the sound or sound field, and enhancement side information including parameters for improving the basic reconstructed sound representation. The method comprises sub-dividing the plurality of components into a plurality of groups of components and assigning each of the plurality of groups to a respective one of a plurality of hierarchical layers, the number of groups corresponding to the number of layers, and the plurality of layers including a base layer and one or more hierarchical enhancement layers, adding the basic side information to the base layer, and determining a plurality of portions of enhancement side information from the enhancement side information and assigning each of the plurality of portions of enhancement side information to a respective one of the plurality of layers, wherein each portion of enhancement side information includes parameters for improving a reconstructed sound representation obtainable from data included in the respective layer and any layers lower than the respective layer. The document further relates to a method of decoding a compressed sound representation of a sound or sound field, wherein the compressed sound representation is encoded in a plurality of hierarchical layers that include a base layer and one or more hierarchical enhancement layers, as well as to an encoder and a decoder for layered coding of a compressed sound representation.

The present application relates to a semantic chain feedback and reinforcement learning driven image fusion method, belonging to the technical field of computer vision and image processing. It includes: inputting the preprocessed input image into the frozen dual-path hierarchical encoder, extracting multi-scale visual features and shallow prior features in the prior injection branch; on each intermediate scale, cross-modal collaborative fusion is performed on the infrared features, visible light features, the last layer propagation features and shallow prior features in the multi-scale visual features; chain semantic enhancement is performed on the multi-scale fusion features: obtaining several candidate fusion images; optimizing the unfrozen model parameters; inputting the infrared image and visible light image to be processed into the optimized network, and selecting the noise-free reference candidate as the final fusion image output. This method cooperatively optimizes through chain semantic progressive modulation and difference driven reinforcement learning, improving the visual quality and information expression ability of the fusion image.

This invention relates to an anti-interference data transmission method for a 10kV faulted high-voltage line, belonging to the field of signal transmission technology. The method includes: acquiring fault data and subcarriers; inputting the fault data into a fault-channel cooperative sensing model, outputting fault type and channel state information; performing hierarchical coding based on the fault type and channel state information to obtain hierarchical error correction coding; generating an anti-phase interference cancellation signal using a phase conjugate cooperative arc interference cancellation unit, and aggregating subcarriers to obtain aggregated subcarriers; mapping the hierarchical error correction coding to the aggregated subcarriers to obtain a modulated signal; superimposing the anti-phase interference cancellation signal and the modulated signal to achieve initial interference cancellation, obtaining a mixed signal; performing signal compensation using a multi-source power supply relay unit; performing secondary interference cancellation on the compensated mixed signal; and demodulating and decoding the signal after secondary interference cancellation to restore the fault data, thus completing the fault data transmission.

This invention discloses a log encoding / decoding method and system, relating to the field of data processing technology. The method includes: when encoding log text, extracting metadata and content data from the original log text; after hierarchically encoding the content data according to a global dictionary and data type, concatenating the metadata and encoded content data into a binary log block and storing it; when parsing the binary log block, extracting metadata from the stored binary log block, loading a field dictionary based on the metadata, and decoding the field names and values ​​respectively to output a structured log. This method adapts to the massive log processing needs of networked device scenarios, comprehensively optimizing log storage efficiency, parsing speed, and cross-system compatibility through innovative log encoding rules.

The application discloses a glycosylated RNA detection kit based on hierarchical coding chain reaction and application thereof, and relates to the technical field of biological detection. The kit comprises: (1) a metabolic marker module comprising: Ac4ManNAz; (2) a double-recognizing proximity assembly module comprising: 1) a glycosyl recognition probe (T SA ): used for recognizing a sugar chain of glycosylated RNA, and the nucleic acid sequence is shown as SEQ ID NO. 1; 2) an RNA recognition probe (T RNA ): used for recognizing five kinds of glycosylated RNA, namely U1, U3, U35a, Y5 and U8, and the nucleic acid sequences are shown as SEQ ID NO. 2 to SEQ ID NO. 6; (3) a connecting chain: the nucleic acid sequence is shown as SEQ ID NO. 7; and (4) a hierarchical coding chain reaction module. The application can in-situ "code" detection of up to five glycosylated RNA species in the same reaction system, and rich molecular pedigree information can be obtained in one experiment, thereby breaking through the technical bottleneck that most of existing methods can only detect single or limited targets. The whole reaction system is simple to operate and easy to popularize.

This invention discloses a glacier image segmentation method and system based on hybrid self-attention and contextual aggregation, comprising: acquiring and preprocessing remote sensing images; extracting multi-scale features using a hierarchical Transformer encoder; performing hybrid self-attention and cross-scale cross-attention at 1 / 16 and 1 / 32 scales to achieve semantic alignment and long-range dependency modeling; employing bottom-up multi-scale fusion decoding and progressive upsampling; introducing global priors through pyramid pooling, and combining channel variance selection and efficient channel attention for feature enhancement; introducing residual boundary refinement branches to correct contours, and outputting glacier segmentation results. This invention improves segmentation accuracy in complex backgrounds, blurred boundaries, and small target scenes while reducing the number of parameters and computational overhead, making it suitable for large-format, high-resolution remote sensing scenes.

This invention discloses a self-supervised image denoising method based on adaptive noise density region segmentation, comprising the following steps: performing local statistical analysis on the input noisy image to generate a noise density response map; adaptively segmenting the noisy image based on the noise density response map to generate a mask that distinguishes between high and low noise regions; obtaining corresponding region masks by using different mask rate sampling strategies for high and low noise regions, obtaining a training mask through element-wise combination, and performing blind spot processing on the occlusion positions of the training mask; constructing a denoising network architecture, using a hierarchical encoder-decoder structure as the backbone, with each input channel performing initial feature extraction through convolution operations, mapping noise density information into high-dimensional features and fusing them with image features; performing self-supervised network training on the denoising network architecture based on the training mask, and outputting a denoised image based on the trained denoising network. This invention achieves a balance between efficient noise reduction and detail preservation.

A method of signalling a signal adjustment when encoding an input signal using a hierarchical coding scheme to manage a display aspect ratio. The hierarchical coding scheme comprises encoding a downsa

The application discloses a kind of to BIM model and GIS model is carried out coordinate registration method and device.Method includes: the naming rule of design BIM model and the hierarchical coding rule of BIM model coding;Establish the database of the hierarchical coordinate conversion parameter of storage BIM model;From database, the first to N sub-coding of BIM model to be designed and its corresponding hierarchical coordinate conversion parameter value are obtained;Based on the first cumulative conversion matrix, the modeling coordinate system is created for the BIM model to be designed, and after design is completed, the BIM model designed is guided to engineering project coordinate system;Second platform for integrated BIM model determines the second cumulative conversion matrix of the BIM model in geodetic coordinate system relative to GIS model coordinate origin;Based on the second cumulative conversion matrix, the BIM model is registered into GIS model using geodetic coordinate system.The application provides unified coordinate identification and application scheme across software, cross-platform.

This application provides a full-chain digital scheduling method for laboratory testing tasks, including: decomposing the attack chain stages of laboratory threat assessment testing task requirements information to generate a testing task scenario feature vector table with attack chain stage labels; constructing a multi-dimensional profile of the technical capabilities of existing laboratory testing resources based on the testing task scenario feature vector table to generate a testing resource technical capability matrix with attack chain stage correlation coefficients; generating a task-resource scheduling configuration table with risk level identifiers based on the testing resource technical capability matrix; performing digital instruction hierarchical encoding on the task-resource scheduling configuration table to generate a scheduling execution and anomaly handling record table with instruction time-series trajectories; and calling a two-dimensional performance-delay model of scheduling strategies to generate strategy scheduling parameters adapted to the full-chain scheduling requirements for full-chain digital scheduling of laboratory testing tasks.

The application discloses a kind of material coding automatic generation method based on knowledge graph, comprising the following steps: obtaining the material main data of target material, executes preprocessing;Based on standardized material main data, material coding knowledge graph is constructed;Standardized material main data is executed hierarchical path construction and parameter slot processing;Based on slot parameter vector, construct hierarchical TransE embedding model;Utilize historical coded material sample to train hierarchical TransE embedding model;Hierarchical triple sequence and slot parameter vector are input target embedding model and execute embedding reasoning;In material coding knowledge graph, each level is executed coding segment candidate search and constraint filtering;According to hierarchical coding segment length table, candidate coding segment combination sequence is spliced, and material coding knowledge graph is updated.The application adopts hierarchical TransE embedding model, realizes material coding automatic generation.

This application discloses a method and system for scheduling sports replay data based on time-sliced ​​prefetching. By dividing the sports replay source data into time slices and performing hierarchical encoding, a mapping relationship between time slices and hierarchical data is constructed, and a basic playback state and hierarchical loading state are established. The prefetch window range is determined based on the target playback position, and the prefetch order is determined by combining time distance and hierarchical prefetching scheduling is executed. A resource allocation model is constructed by combining network bandwidth status, cache occupancy status, and playback rate information to generate scheduling priority parameters and dynamically adjust the hierarchical data loading ratio. Enhanced layer data is delayed and supplemented, and merged and updated with core layer data. Simultaneously, the scheduling process is adjusted in real time based on user playback behavior, enabling the sports replay data to achieve progressive loading of hierarchical data and coordinated optimization of resource utilization while ensuring playback continuity.