164results about How to "Reduce the amount of calculation" patented technology

This invention discloses a low-light scene analysis method based on multimodal feature fusion and clustering, belonging to artificial intelligence technology. It constructs a single-branch feature extraction network based on Transformer; a multimodal feature interaction and fusion module to achieve multimodal feature interaction and fusion; a multimodal fusion feature clustering module to cluster the features after multimodal interaction and fusion, utilizing the semantic and spatial distances between different features to achieve clustering while simultaneously downsampling the features, mitigating the problem of loss of detailed edge information in structured downsampling; and a multi-scale feature aggregation and decoding module to receive feature information from the encoding network and classify each feature pixel according to the semantic distance of the multi-scale features. This invention can fully utilize visible light and thermal image information and can be applied to scene analysis and navigation of unmanned systems in diverse and complex low-light scenes.

The application relates to a real-valued super-resolution direction estimation method of a high-order acoustic field sensor array, which uses a received signal covariance matrix to approximately calculate an estimated value of noise power, avoids the iterative operation process of noise power in the prior art, improves the calculation efficiency, and reduces the floating-point operation amount. By constructing an augmented matrix, the array receiving data matrix and the array manifold matrix with a multi-dimensional structure of an element are made into Hermitian matrices, so that the unitary transformation processing of the related parameters of the high-order acoustic field sensor array is realized. On the basis of obtaining the array receiving data matrix and the array manifold matrix in the real number domain, the array receiving data matrix and the array manifold matrix are applied to a sparse approximate minimum variance method with a variable exponential factor, and the completely real-valued sparse approximate minimum variance direction estimation with a variable exponential factor is realized.

The application discloses a visual perception method for assisted driving, comprising the following steps: S100, collecting assisted driving image data, and constructing input X after data set preprocessing; S200, constructing a multi-task neural network model F and substituting the input X into the model to obtain a plurality of branch task-specific outputs f i ; S300, calculating the final loss function L all of the multi-task neural network model F; S400, optimizing the minimum as the target of the loss function L all , and finally obtaining the optimal multi-task neural network model F * . A system using the above visual perception method for assisted driving is also disclosed, specifically comprising a data input module, a shared layer feature extraction module, a task layer structure search module and a model output module. The application has the beneficial effect that a multi-task model is constructed to effectively reduce the network parameter quantity and the calculation quantity, simultaneously reduce the online reasoning time, and is very beneficial to practical application.

The invention relates to a multi-task face sensing method and device, electronic equipment, a storage medium and a computer program product. The multi-task face sensing method comprises the steps that a to-be-detected face picture is acquired; a to-be-detected face picture is input into the multi-task face perception model, face features and positioning reference points are extracted from the face picture through the multi-task face perception model, and the positioning reference points are used for executing a positioning task about face key point detection; semantic reference points are calculated based on the positioning reference points through a multi-task face perception model, and the semantic reference points are used for executing semantic tasks about face state detection; and executing a positioning task and a semantic task based on the face features, the positioning reference points and the semantic reference points through a multi-task face perception model. Therefore, by explicitly utilizing the physical representation of the face key points, effective supervision for the semantic reference points is realized, the guiding effect of the face key points on the face perception task is enhanced, and the prediction effect of the face perception model can be improved.

The present application belongs to the technical field of data analysis, and particularly relates to a high-throughput biosensing system for early warning of algal blooms and pathogens, which comprises multiple sampling units, a microfluidic chip with a multi-channel structure, and a high-throughput surface-enhanced Raman detection unit with an integrated metal nanostructure surface, and obtains the original surface-enhanced Raman spectrum of the water sample through a Raman spectrum acquisition mechanism. The system uses compressed spectrum sampling and sparse reconstruction methods to reduce the data volume and maintain the integrity of the spectrum information, constructs a pathogen Raman fingerprint dictionary through a dictionary learning method, and realizes the identification of different algal metabolites, algal toxins and pathogen-related molecular components and the concentration interval judgment thereof in combination with a sparse representation classification mechanism. Based on the type and concentration interval of the pathogens, the system further generates algal bloom risk warning and pathogen risk warning, and realizes early discovery of abnormal changes in the water body. The present application has significant advantages in stability, sensitivity, throughput and real-time performance.

The application relates to the field of optical fiber sensing, and more particularly to an event detection method, system and device for optical cable detection and a storage medium. The method comprises the following steps: inputting optical fiber data of an optical cable into a first processing unit for processing to obtain space-time features; inputting the optical fiber data into a second processing unit for processing to obtain frequency domain features; inputting the space-time features and the frequency domain features into an attention processing unit after splicing for feature enhancement to obtain enhanced features; inputting the enhanced features into a bottleneck layer processing unit for processing to obtain bottleneck features; wherein the bottleneck layer processing unit comprises at least two bottleneck layer processing subunits, and the bottleneck layer processing subunits can perform feature transformation processing on the enhanced features; and inputting the bottleneck features into a backbone network for optical fiber event detection to obtain an optical fiber event detection result corresponding to the optical fiber data. The application can improve the accuracy of optical fiber event detection in an optical cable.

The invention discloses a composite material multi-scale image finite element analysis method considering manufacturing characteristics, and belongs to the technical field of special-shaped composite material structure multi-scale analysis. CT scanning is carried out on the special-shaped composite material structure, an internal image of the special-shaped composite material structure is obtained, statistical parameters of manufacturing characteristics are extracted through an image processing technology, and mesoscopic representative volume units are classified according to the statistical parameters; secondly, a self-consistent clustering analysis method is adopted for model unified order reduction, and an efficient solving database is constructed; mapping corresponding cell element categories according to defect characteristics of each region of the component, realizing scale solution through a concurrent multi-scale framework and a homogenization method, and predicting mechanical response and damage evolution of the structure; and finally, revealing a physical mechanism that macroscopic damage expands along a low-performance weak band based on spatial distribution of cell categories. The method has high calculation efficiency and prediction authenticity, and an effective tool is provided for performance evaluation and process optimization of the defect-containing composite material component.

This application discloses a robot map exploration method and chip based on reachable boundary points, including: Step A, the robot uses the boundary points in its current detection area to fit a preset boundary line, and selects the contour line used to delineate all connected detectable areas in the map to be optimized as the target contour line; Step B, the robot controls the two ends of the preset boundary line to expand respectively to obtain a preset boundary expansion line segment; if the expansion length of both ends of the preset boundary line is less than or equal to the preset perception radius, and both ends of the preset boundary expansion line segment intersect with the target contour line, then Step C is executed; Step C, based on the preset boundary expansion line segment, the robot segments the contour line to be tracked from the target contour line along the direction connected to the unknown area, and then searches for leading boundary points in the area enclosed by the contour line to be tracked and the preset boundary expansion line segment; Step D, based on the traversability of the leading boundary points searched in Step C, reaching boundary points are selected.

Embodiments of the present specification provide a method for question and answer reasoning of image-text content based on a target model, the target model comprising a plurality of decoding layers, and the image-text content comprising a plurality of images and a plurality of text segments; the method comprising: for a first image, performing first representation compression to obtain L visual compressed representations corresponding to the L decoding layers; for a first text segment, performing second representation compression to obtain L text compressed representations corresponding to the L decoding layers; replacing the images and the text segments in the image-text content with placeholders respectively, and inputting the plurality of decoding layers for decoding operation, comprising: in each decoding layer of the L decoding layers, superimposing the visual compressed representation and the text compressed representation corresponding to the decoding layer on the current hidden representation of the corresponding placeholder, and then performing fusion processing to obtain a decoding representation; and generating an answer text corresponding to the image-text content based on the decoding representation output by the last decoding layer.

This invention belongs to the field of intelligent detection technology for wind turbine blades, specifically relating to a feature detection method and system for multi-category surface damage on blades. The method includes: constructing a pyramid network structure EFPN to replace the baseline model's FPN+PAN neck network. The EFPN includes a pyramid feature enhancement (PFE) module, a multi-feature fusion (MFF) module, and a dynamic interpolation fusion (DIF) module, used to integrate multi-scale damage features and suppress background noise; designing a Focaler PIoU loss function, combined with a size-adaptive penalty factor and a dynamic sample weighting mechanism, to improve the model's ability to perceive difficult samples and the accuracy of bounding box regression; and using the DepGraph structured pruning method to compress the network, combined with an HDCFR hybrid knowledge distillation strategy, to reduce model complexity while maintaining detection accuracy. This invention achieves an mAP@0.5 of 84.5% on the WTBs GLE dataset, with reduced parameter and computational costs, combining high accuracy, lightweight design, and strong robustness, making it suitable for deployment on UAV edge devices.

A method for generating contraband images based on an improved DDPM (Discretionary Data Processing Model) is disclosed. The method includes: creating a contraband dataset; constructing a contraband image generation model based on the improved DDPM and generating feature maps of contraband in different poses; determining the weights of the contraband image generation model based on the improved DDPM; and generating an image of a specific contraband. The contraband image generation method based on the improved DDPM provided by this invention has the following advantages: it reduces the number of model parameters and improves model running speed; it improves model performance, perception ability, and efficiency, better handles multi-scale features, and reduces model parameters and computational load; it better captures global contextual information in the image, thereby improving the accuracy of image segmentation and allowing the model to focus more on important regions in the image.

This invention discloses a coal spill identification method based on information entropy, relating to the field of detection technology. The method includes the following steps: Step 1: For each frame image of the monitored area, convert it into a grayscale image and represent it using matrix A; Step 2: Perform edge detection on matrix A, and record the result as matrix C; Step 3: Calculate the information entropy of matrix C to obtain the frame entropy H of the grayscale image; Step 4: Process each frame of the real-time frame images generated sequentially in the monitored area through steps 1 to 3 to obtain a frame entropy sequence; Step 5: Calculate the frame entropy mean μ and variance σ of the frame entropy sequence. 2 Determine the threshold H for frame entropy alarm. gate Step 6: When the frame entropy of several consecutive frames exceeds the threshold H gate If a coal spill occurs, a shutdown command is sent. This invention utilizes a pure algorithm, avoiding the time-consuming and labor-intensive cleanup caused by coal blockage, and providing a reliable technical guarantee.

This invention provides an integrated method for pruning and controlling large-scale models of unmanned aerial vehicles (UAVs), belonging to the field of model lightweighting technology. The proposed method significantly reduces model size and computational load, greatly improves inference speed and control response performance, and enables the deployment of complex neural network controllers on resource-constrained airborne platforms. Simultaneously, the lightweight model maintains or approaches the control accuracy and stability of the original uncompressed model. This method has wide applicability and strong scalability, covering various scenarios from basic flight control to complex mission control, providing a feasible, efficient, and cost-effective solution for applying large models to airborne controllers of UAVs, and possesses high engineering practical value.

The invention discloses a high-dimensional sparse text-oriented classification method, system and device and a storage medium, and belongs to the technical field of natural language processing. The method comprises the steps of obtaining to-be-classified text data; based on the to-be-classified text data, constructing a candidate redundant feature set by calculating semantic similarity; joint mutual information of candidate redundant features in the candidate redundant feature set is calculated, and key features are obtained through screening according to the joint mutual information; inputting the key features into a pre-constructed text classification model to obtain a classification prediction result; wherein the text classification model comprises a dual-channel gating module, an attention module and a prediction module; the two-channel gating module extracts context features in the key features; the attention module weights the context features to obtain weighted features; and the prediction module obtains a classification prediction result according to the weighted features. According to the method, effective compression of feature dimensions and precise modeling of context semantics are realized.

This invention discloses a real-time rockfall detection method for open-pit mine slopes based on edge intelligence, belonging to the field of mine safety monitoring and early warning technology. The method includes: acquiring slope video to extract rockfall images, generating a dataset through manual annotation and region-adaptive dual-domain enhancement; constructing a DGN-YOLOv10 edge detection model, replacing the static convolutional layers of the backbone and neck network with dynamic convolutional modules DyConv, and replacing the C2f module with a lightweight composite module C2f_G_D; and using the Star Raven Optimization Algorithm (NOA) to dynamically and adaptively optimize the learning rate of the model to obtain the final model for rockfall detection. This invention suppresses environmental interference through region-adaptive dual-domain enhancement and the DyConv mechanism; it integrates the C2f_G_D module with the DFC attention mechanism to maintain the global receptive field while reducing the number of parameters and computational cost; and the NOA algorithm solves the local optimum problem under data scarcity.

The present application relates to a kind of pneumonia detection model training method, specifically to a kind of hierarchical topology pneumonia detection model based on collaborative learning system and method.It overcomes the difficulty of existing pneumonia detection model training system, low efficiency and the problem that local detection node cannot achieve optimal detection performance.First, each node uses local pneumonia data to pre-train respective pneumonia detection model locally, then, center node is based on the feedback data of local model to build hierarchical topology collaborative learning system;After that, each cluster node and child node uses local pneumonia data to train respective pneumonia detection model locally;Finally, collaborative training is carried out, including intra-cluster collaborative training and inter-cluster collaborative training;Intra-cluster collaborative training includes cluster node aggregation training and child node random interactive training.The present application has small data transmission, storage and calculation pressure, and is easy to implement;At the same time, different client is provided with different model parameters, so that each hospital detection node achieves local detection performance maximization.

This invention proposes a method and system for diagnosing open-circuit faults in three-phase inverter IGBTs, relating to the technical field of power electronic device fault diagnosis. The method involves lightweight processing of acquired historical data to obtain feature vectors; training a pre-defined three-phase inverter IGBT open-circuit fault diagnosis model based on these feature vectors; real-time acquisition of three-phase current data to be processed, followed by bounded normalization to obtain normalized current amplitudes; and generating IGBT open-circuit fault judgment results based on these normalized current amplitudes. Real-time data related to the three-phase inverter IGBT open-circuit fault is collected based on the IGBT open-circuit fault judgment results, and this real-time data is input into the trained three-phase inverter IGBT open-circuit fault diagnosis model to output the three-phase inverter IGBT open-circuit fault diagnosis results. This invention effectively achieves lightweight, efficient, and accurate diagnosis of three-phase inverter IGBT open-circuit faults.

This invention discloses a field weakening control method and system for a torque-enhanced permanent magnet synchronous motor for electric vehicles, relating to the field of torque-enhanced permanent magnet synchronous motors, including the following steps: Step S1, based on the actual speed of the torque-enhanced permanent magnet synchronous motor... ω m and given speed ω m * Step S2: Determine the current amplitude setpoint; Step S3: Calculate the current trajectory during MTPA control operation based on the inverter's limit voltage. U lim With stator voltage setpoint U s The difference Δ U Step S4: Determine whether field weakening control needs to be activated; Step S5: The field weakening area includes field weakening zone 1 and field weakening zone 2; Step S6: Generate a space vector pulse width modulation signal based on the current given value and the actual current value. This invention adopts the above-mentioned field weakening control method and system for torque-enhanced permanent magnet motors in electric vehicles to realize speed regulation control of torque-enhanced permanent magnet synchronous motors across the entire speed range.

This invention relates to the field of bid document management technology, and in particular to an automatic bid document review method and system based on natural language processing. The method includes: identifying target sections; determining historical performance and financial information; judging whether false performance exists; analyzing the complete sentences containing target keywords; calculating material scores; determining price deviation; and determining the review result. This invention achieves multi-dimensional and comprehensive review of bid documents by intelligently analyzing multiple sections such as qualifications, risks, materials, and pricing. This effectively improves the comprehensiveness and depth of the review. Utilizing natural language processing technology, it can accurately identify and extract key information from bid documents, overcoming the subjectivity and limitations of traditional manual review. Quantitative assessment based on risk level significantly improves the accuracy and relevance of risk identification. By comprehensively weighing material scores and price deviation, it significantly improves review efficiency.

The present application relates to the technical field of image recognition, in particular to a lightweight infrared small target real-time detection method, the detection method comprising the following steps: step one, constructing an infrared small target data set: obtaining the infrared small target and selecting and labeling it, step two, constructing an infrared small target detection network based on Mobilenetv3-Unet, step three, training the Mobilenetv3-Unet model using the infrared small target data set in step one to obtain the trained Mobilenetv3-Unet model, step four, using the trained Mobilenetv3-Unet model to detect the infrared small target of the input picture, the Mobilenetv3-Unet model in step two comprising a feature extraction backbone network and a feature aggregation network, wherein the feature extraction backbone network is an optimized Mobilenetv3 model, the present application not only improves the detection capability of the infrared small target in a complex scene, but also uses fewer computing resources, so that the small target detection algorithm can be easily deployed on resource-limited edge devices.

The application relates to a terahertz radar-based focusing imaging method, device, equipment and medium. The method comprises the following steps: obtaining echo signals of a uniformly rotating large-size target based on a terahertz radar, performing distance-direction pulse compression and azimuth-direction FFT on the echo signals to obtain a two-dimensional RD imaging result, estimating the rotation speed and longitudinal distance center offset of the target according to the spatial distribution law of the PSF inclination of each defocused point in the two-dimensional RD imaging result, constructing a final phase compensation function according to the finally estimated rotation speed and longitudinal distance center offset, performing second-order spatially-varying phase compensation and azimuth-direction FFT on a one-dimensional distance image after Keystone transformation according to the final phase compensation function, and obtaining a finally focused RD image. The method can reduce the calculation search amount of terahertz radar RD focusing imaging and improve the focusing accuracy.

The invention discloses a light-weight image fusion method based on a Mamb-KAN structure, and the method comprises the following steps: obtaining a registration source image, dividing the registration source image into overlapped image blocks, and constructing a Sheaf Laplacian matrix according to an overlapping relation and limited mapping; decomposing the Laplacian spectrum matrix, and projecting image block features into a Sheaf coefficient sequence; a Mama-KAN network is constructed, a Mama branch receives the coefficient sequence to extract fusion features, and a KAN branch calculates a block-level fusion weight according to the coefficient sequence and the fusion features; constructing a loss training network by using the Sheaf consistency energy and the reconstruction error; a lightweight network is obtained according to contribution pruning and quantification of the Mama channel and the KAN base unit to the consistent energy; in the fusion stage, a Sheaf coefficient sequence is generated and input into the lightweight network, and a source image is weighted according to a fusion weight to generate a fusion image. According to the invention, high-quality and low-overhead image fusion is realized.

This invention relates to a target localization and counting method based on point domain feature learning. The method includes: inputting the image to be detected into a backbone network to extract a three-level feature pyramid comprising top-level features, intermediate-level features, and bottom-level features; inputting the three-level feature pyramid into a multilayer perceptron module to capture the global long-distance dependence of its top-level features and obtain global information features; inputting the three-level feature pyramid into a visual centering mechanism module to aggregate local region features of its top-level features, obtain local information features, and connect them with global information features along the channel dimension to obtain mapped top-level features; then, combining these with intermediate-level and bottom-level features to form a mapped three-level feature pyramid; inputting this pyramid into a global centralization adjustment module; outputting an initial feature map to a multi-perceptron module; extracting channel features and spatial features and concatenating them to obtain a multi-perceptron feature map; then, passing through a regression head to output the predicted target count of the image to be detected; and finally, passing through a classification head to output the predicted target localization of the image to be detected.

The invention provides a surge management system of a fluid machine, an early warning method, a model training method, a computer readable storage medium and a computer program product. The surge management system comprises a computing device which is configured to operate a surge early warning model; the surge early warning model comprises a deep support vector convolutional neural network, and the deep support vector convolutional neural network has a deep separable convolutional network structure and is used for mapping input data reflecting the operation state of the fluid machine to a feature space to obtain corresponding feature points; in addition, the surge early warning model predefines a hypersphere in a feature space; and the calculation device is configured to trigger surge early warning when it is judged that the feature points are located outside the hypersphere. The surge early warning model adopted by the system can complete reliable development training under the conditions of scarcity of surge samples and imbalance of data sets, and light weight of a deep learning model is realized.

The application provides a lightweight steel surface defect detection method and device and processing equipment, which are based on a YOLOv11 model, deep adaptability optimization is performed, the steel surface defect detection task has better detection precision, the parameter quantity and the calculation quantity can be effectively reduced, the generalization ability is good, when the industrial terminal is deployed, the high performance and the lightweight can be considered, and thus the industrial application prospect is good.

The invention provides a polarization image defogging method based on a neural network, and belongs to the technical field of image processing, and the method comprises the steps: obtaining a plurality of fogged polarization images, and generating a polarization image package; constructing a neural network defogging model; and based on the constructed neural network defogging model, carrying out defogging processing on the foggy polarization image in the generated polarization image packet to obtain a fogless image. According to the image defogging method, the deep learning and the polarization imaging technology are combined, the problem that the calculation complexity of the polarization imaging technology is high can be effectively solved, the overall generalization ability and defogging level are improved, and better robustness and high adaptability can be shown in the face of different types of fog and different degrees of haze.

This invention provides an online inspection device for the dimensions of non-standard wood products, relating to the field of laser measuring instrument technology. It includes a machine body and rotating rollers. A conveying assembly is mounted on the machine body, comprising a conveyor belt and a motor. A detection assembly is also mounted on the machine body, including a laser measuring instrument and a sliding roller. A scribing assembly is mounted on the machine body, comprising a laser and a slider. A moving assembly is mounted on the machine body, including a lead screw and a motor. This online inspection device for the dimensions of non-standard wood products reduces the amount of measurement and computation required by the laser measuring instrument, lowering the performance requirements of the instrument. Simultaneously, one side of the non-standard wood product uses the sliding rollers on the machine body and movable plate as a reference line, preventing the non-standard wood product from tilting in the conveyor belt's transport direction, thereby ensuring the accuracy of the inspection structure. It is suitable for rapid inspection of the dimensions of non-standard wood products.

The invention discloses a ceramic membrane internal defect detection method based on parallel stripe convolution and topological convolution. The method comprises the following steps: 1) constructing a ceramic membrane plate internal defect data set; 2) in a feature extraction module, providing a parallel multi-branch strip convolution feature extraction module, and better extracting features while realizing light weight; in a detection head module, a self-adaptive topological convolution detection head is provided, and the module reduces calculation of redundant features; and 3) model training and optimization: carrying out multiple rounds of training on the self-built data set. According to the method, for the problem of YOLO model lightweight, parallel strip convolution is used in a feature extraction part, and features can be better extracted on the premise of reducing the network parameter quantity; in addition, adaptive topological convolution is used in a detection head part, so that redundant calculation of a large-area non-key background area can be effectively reduced.

The application discloses a speech recognition fine-tuning method based on subspace decomposition and reorganization. The low-rank space of the LoRA module is equivalently decomposed into multiple subspaces, and learnable weights are added to different subspaces, so that the reorganization of all subspaces is realized. The reorganized subspaces can be transformed into an efficient calculation mode during training and an inference mode during inference, which significantly improves the inference and storage efficiency, and solves the problem of high inference cost during high concurrency request while ensuring efficient fine-tuning of parameters.

The application discloses a centrifugal pump off-design condition diagnosis method and system based on a vibration signal, and the method comprises the following steps: collecting vibration signals of a centrifugal pump under multiple flow conditions, and constructing a vibration sample sequence with a working condition label; reconstructing the vibration signal by using a variational mode decomposition (VMD); performing a continuous wavelet transform (CWT) on the reconstructed signal, and mapping a one-dimensional vibration signal into a two-dimensional time-frequency graph; constructing a Light-CBAM-CNN diagnosis model, and training the same; inputting the vibration signal collected in real time into the trained diagnosis model after pretreatment, VMD and CWT, outputting a working condition category and a diagnosis confidence, and outputting an "uncertain" label and triggering a retest mechanism or an alarm when the confidence is lower than a threshold value. The application can realize online diagnosis of high-energy consumption off-design conditions of the centrifugal pump only by relying on the shell vibration signal, and has the advantages of small modification workload, low deployment cost, high identification reliability and strong energy-saving scheduling guiding significance.