17 results about "Recursive convolution" patented technology

The application discloses a forest fire danger dynamic early warning system and method, and belongs to the technical field of forest fire early warning. The method continuously collects meteorological parameters, vegetation water content, terrain characteristics and historical lightning point distribution data through multiple source sensing nodes, inputs the multiple source data into a recursive convolution network after aligning the multiple source data according to the same time index, uses a time attention module in the recursive convolution network to dynamically allocate weights to the aligned data, obtains fire danger inducing feature sequences of each sensing node at different time steps, maps the fire danger inducing feature sequences to a pre-constructed fire danger grade atlas, determines an initial fire danger grade at the current moment by comparing the distance between the feature sequence and each grade node center vector, and generates a fire danger dynamic distribution grid covering the whole monitoring area according to the initial fire danger grade and a fire danger propagation speed parameter shared between adjacent sensing nodes, and the fire danger dynamic distribution grid is used as real-time early warning output.

The invention provides a medical image segmentation method and system based on an RU-Net model. The method comprises the following steps: S1, inputting an image into the RU-Net model; s2, capturing basic features in the image through a series of first convolutional layers in the encoder; s3, reducing the spatial dimension of the features through a pooling layer, and extracting abstract features; s4, the features are processed through a recursive convolutional layer, so that the model can accumulate and learn the features on a plurality of time steps, and the understanding of deep features of the image is enhanced; s5, recovering the spatial resolution of the feature map through an up-sampling layer in the decoder, and fusing the spatial resolution with the feature map from the encoder; s6, further processing the fused feature map through a series of second convolutional layers and RCL, and reconstructing a fine structure of the image; and S7, outputting a segmentation result through a final convolutional layer. According to the method, the abnormal areas in the medical image can be automatically detected and accurately classified, so that the diagnosis efficiency and accuracy are greatly improved.

The application discloses a kind of industrial grade 4D video generation system and method based on space-time anchoring and mixed expert parallel, belong to artificial intelligence video generation technical field.System includes eight big modules of global configuration, rigid space-time anchoring, 4D space-time coding and decoding, timing synchronization mixed expert, mass data flow loading, multistage distributed training, real-time inference scheduling, post-processing enhancement;Method is through building multidimensional rigid space-time anchor point tensor to eradicate iterative cumulative error, uses non-recursive 4D convolution to realize video global coding and decoding, uses multi-path mixed expert network to improve model capacity, combined with multistage distributed parallel to break through memory bottleneck, and realizes model hardware resident and multi-path real-time generation at inference end.The application completely solves the problem of timing drift and structure tearing of traditional autoregressive architecture, has the characteristics of high stability, high efficiency and low delay, and is suitable for film-level industrial video intelligent production scene.

The present invention belongs to the technical field of image denoising, and in particular relates to an efficient EBSD denoising and data restoration system and method. The system comprises: a variational autoencoder (VAE) for generating a noise mask corresponding to an input image, learning the data's latent representation through encoding and decoding processes, and optimizing the modeling of noise distribution; a foreground network (ReNet) for extracting background information from noisy images and restoring a clearer background image by learning the residual information from the denoising process. PReNet is designed as a recursive convolutional neural network, primarily used to effectively extract and restore background information from noisy EBSD images. Its unique iterative and residual learning structure improves detail preservation and noise removal capabilities; and a generative adversarial network (SAGANDiscriminator) employs a convolutional neural network architecture with a self-attention mechanism and spectral normalization to ensure stability and efficiency during adversarial training.

The invention provides an optimal scheduling method for improving the reliability of a pumped storage system under wind-solar multi-state convolution. The method comprises the following steps: firstly, forming a total load curve under preset time granularity; then establishing a time sequence probability distribution model of wind and light output based on output point prediction data and prediction error normal statistical characteristics of each wind power plant and photovoltaic power station under the preset time granularity; the method comprises the following steps: constructing an equivalent continuous load curve of a power system under preset time granularity by adopting a wind-light time sequence multi-state convolution and thermal power generating unit two-state recursive convolution algorithm, further calculating an expected value of insufficient electric quantity and a marginal reliability factor in each time period, and quantifying the marginal reliability influence of load growth; and finally, on the basis of the net load curve deducting the wind and light point output predicted value and the marginal reliability factor, a pumped storage optimal scheduling model considering multiple physical constraints is constructed and solved by taking the maximum improvement of the system reliability through pumped storage optimal scheduling as a target. The method can reduce the expected value of the total electric quantity of the system by optimizing the scheduling strategy.

Methods, systems, and apparatus, including medium-encoded computer program products, for processing data representing a graph through each of a plurality of layers of a graph neural network to generate an output. The plurality of layers include a graph recursive convolutional layer.

The invention relates to the field of power transmission of a power system, and provides a phase domain frequency change modeling method and device for a direct current transmission line, and the method comprises the steps: building a phase domain frequency change mathematical model based on the physical characteristics of the direct current transmission line; when the propagation time delay of the electromagnetic wave on the direct-current power transmission line and the simulation step length are in a non-integer multiple relationship, adaptively adjusting interpolation table parameters through the trained deep learning model, and performing interpolation reconstruction on the reflection current based on the interpolation table parameters; the deep learning model is used for representing a nonlinear mapping relation between interpolation table parameters and system dynamic response; and based on the reflection current after interpolation reconstruction and the phase domain frequency change mathematical model, adopting a recursive convolution algorithm to carry out time domain simulation under a non-linear sudden change working condition. According to the DC power transmission line phase domain frequency change modeling method and device provided by the invention, the modeling precision of the DC power transmission line in a high-frequency transient process is remarkably improved, and waveform distortion and simulation errors are reduced.

The application discloses a learning-based lightweight adaptive multi-stage dense reconstruction method and device, which comprises the following steps: obtaining multiple original images; generating a feature map of multiple scales of each original image through a preset feature pyramid based on a depth separable convolution; determining a feature body of multiple scales of multiple source images corresponding to each reference image; using an intra-view and inter-view double-flow adaptive aggregation mode to aggregate the cost volume with weights; at the top scale, generating a cost volume of the reference image, and using a hierarchical recursive convolution network to generate a depth map; at other scales, generating a residual cost volume of the reference image by referring to the depth map of the last scale, using the hierarchical recursive convolution network to generate a residual depth map, and adding the residual depth map and the up-sampling of the depth map of the last scale to obtain the depth map at the current scale; and fusing the depth maps generated at the multiple bottom scales to obtain a dense point cloud. The method can solve the problem of weak texture extraction.

The invention discloses a non-linear hydrodynamic calculation method for a shallow water floating body based on a finite water depth Green function, which realizes double breakthrough in precision and efficiency. A finite water depth Green function strictly meets a water bottom boundary condition, instantaneous wet surface dynamic iteration is combined, and shallow water nonlinear wave body interaction is accurately captured; a far-field Green function and a recursive convolution dimension reduction technology are analyzed by adopting a stable-phase method, so that the calculation efficiency is greatly improved; the influence of shallow water on diffracting force, radiation force and motion attitude of the floating body is researched, a floating body motion differential equation is solved according to nonlinear hydrodynamic force borne by the floating body to obtain the position, motion speed and motion attitude of the floating body, balance of precision and efficiency is achieved, the shallow water nonlinear calculation bottleneck is overcome, and high-precision calculation is achieved.

The application belongs to the field of electromagnetism, and particularly relates to a kind of time domain finite difference method open field simulation high-order perfectly matched layer implementation method and system, aims at solving how to avoid the problem of cascade solution and parameter coupling in PML under the same algorithm consumption as existing high-order PML. The method comprises: constructing PML tensor, the inverse of which is the weighted sum of the inverses of multiple complex frequency shift PML sub-tensors; constructing PML auxiliary variables based on PML tensor and Maxwell equations, including electric field component coupling auxiliary variables and magnetic field component coupling auxiliary variables; updating PML auxiliary variables using recursive convolution method or auxiliary differential equation method, and updating electric field intensity components and magnetic field intensity components according to the updated PML auxiliary variables. The time domain finite difference method open field simulation high-order PML implementation method and system provided in the application effectively overcome the defects of existing high-order PML cascade solution complexity and parameter coupling under the same algorithm consumption.

The invention relates to a direct current power transmission system-oriented fault initial wave process analysis method and equipment. The method comprises the following steps of: calculating time delay of a propagation function minimum phase system; processing the rational fraction of the propagation function and the rational fraction of the wave impedance by adopting a discrete recursion convolution method in combination with the time delay of the minimum phase system of the propagation function to obtain a low-order dispersion equivalent voltage source recursion, and integrating into a high-order dispersion equivalent voltage source recursion of the propagation function according to a linear time-invariant system superposition theorem; a wave impedance equivalent voltage source and transmission line equivalent resistance are obtained by combining the recursion law of the RC parallel network; constructing a direct-current transmission broadband Bergeron equivalent calculation circuit; and analyzing the fault initial wave process by using a direct-current transmission broadband Bergeron equivalent calculation circuit. The method can be suitable for any sampling rate, transition resistance, fault position and fault type, and good precision is kept within 10-20 ms after disturbance occurs.

The application discloses a Chinese sentence semantic similarity calculation method based on a recursive convolutional neural network. It belongs to the field of data cleaning in data mining. The method is calculated based on a semantic similarity network model, comprising: obtaining a sentence pair data set, extracting the sentence pair in the data set, and obtaining the text sentence pair after word segmentation; extracting the preliminary semantic features and deep semantic features of the sentence pair; using an Attention layer to generate the weight of the sentence pair feature vector according to the deep semantic features of the sentence pair; inputting the weight of the sentence pair feature vector into a full connection layer to generate the similarity of the sentence pair. The application enriches the output connotation of the word vector, better reflects the original sentence pair input by using an adaptive pooling layer, improves the model performance, designs the Attention layer to assign corresponding weights to the sentence pair features, makes the model focus on the key information in the data, extracts more direct semantic dependency relationships in the sentence, and improves the accuracy of the sentence pair semantic similarity judgment.

The present application relates to new energy automobile energy consumption management and artificial intelligence prediction technical field, specifically relates to a kind of energy consumption prediction method based on multi-source physical characteristics and multi-scale recursive convolution, comprising: step S1: multi-source physical characteristics acquisition and joint embedding;Step S2: multi-scale down-sampling and recursive trend injection;Step S3: local slice convolution and heterogeneous feature fusion;Step S4: double-path deep feature fitting and residual compensation;Step S5: result reconstruction and energy consumption prediction output.The present application has the technical effect that by constructing multidimensional collaborative time feature extraction system, the precision and robustness of energy efficiency prediction under complex physical working conditions are significantly improved.

The application discloses a remote sensing image target detection method fusing multi-scale context features and channel enhancement, and the steps include: 1, pre-processing remote sensing image dataset; 2, constructing a neural network model based on cascade recursive convolution and attention mechanism; 3, using the model of step 2, constructing a multi-scale context feature enhancement network; 4, using sub-pixel convolution and adaptive sampling factor to design a spatial pyramid channel enhancement network to perform multi-scale feature fusion and detection. The application can solve the problem of various target scales in remote sensing target detection, improve the detection effect of multi-scale targets, and complete efficient detection of unique features of remote sensing image targets in complex backgrounds.

This invention belongs to the field of computer vision and deep learning technology, and provides a method and system for small target detection in UAV aerial photography based on frequency domain-wavelet fusion. A small target detection head branch is added after the second layer output of the backbone network: after downsampling to a preset pixel resolution, it is adaptively weighted and fused with the deep features output by the detection head through a structure. A decoupled head structure is used to output classification prediction, bounding box regression prediction, and confidence prediction respectively, for detecting small targets with a resolution smaller than the preset pixels. The feature map output by the backbone network is input into a LOWTC-based feature extraction module, which performs two-level wavelet decomposition, recursively convolving the low-frequency subband to capture long-distance context, and preserving edge details in the high-frequency subband. After reconstruction by inverse wavelet transform, it is fused with the deep semantic feature structure at multiple scales. This solves the technical problems of low detection accuracy and susceptibility to noise interference in small target detection in UAV aerial images.

This application discloses a computing system to identify a path within a physical layout design of an integrated circuit. The path can have a cell configured to drive a current through a plurality of stages to a sink pins in the integrated circuit. The computing system can generate a timing for the path using non-recursive convolution to determine responses at each of the stages based on the current and loads associated with the stages, and determine a delay for signal propagation on the path based on the timing associated with stages of the path. The computing system can identify at least one timing violation for the path based on the delay for signal propagation on the path, generate a timing report including the path, the delay for signal propagation on the path, and the timing violation for the path, and modify the physical layout design based on the timing violation.

The invention relates to the field of wireless communication, in particular to a mountain complex environment sensing integrated intelligent anti-interference method, which comprises the following steps of: modeling a dynamic anti-interference problem into a Markov decision model; wherein the spectrum waterfall plot at the current moment is taken as an environment state, the decision of selecting the bandwidth and the frequency band at the current moment and whether to communicate is taken as actions, and a reward function is constructed according to the communication rate, the power consumption and the cost of switching the communication equipment. According to the method, the interference identification degree is enhanced through the OTFS waveform, the spectrum waterfall plot is constructed to realize high-dimensional state characterization, the deep recursion convolutional neural network is designed to optimize communication resource allocation, and finally the communication rate and the sensing stability are improved.