21results about How to "Reduce computational redundancy" patented technology

The invention discloses a road crack segmentation method and device based on deep learning, electronic equipment and a program product. According to the method, two collaborative and functional complementary processing paths are adopted to perform parallel processing on a to-be-segmented image so as to obtain corresponding feature information: on one hand, the to-be-segmented image is converted into a feature sequence containing spatial position information, and global semantic modeling is performed on the feature sequence; global semantic features representing the overall shape of the crack and long-range semantic dependence are obtained; and on the other hand, local texture and edge detail information of the crack is directly extracted from the to-be-segmented image, and detail enhancement features are obtained. Thirdly, fusing the global semantic features and the detail enhancement features to obtain target fusion features considering crack positioning, connectivity and boundary and texture expression; according to the segmentation result output based on the target fusion features, even tiny cracks can be effectively depicted, and then the continuity, boundary definition and robustness of tiny crack segmentation under the complex road background can be improved.

The present application relates to the field of digital image processing and information security, and discloses a document tampering detection method and system based on image data processing, comprising the following steps: first, extracting the noise residual and microscopic penetration characteristics of the document image, and constructing a physical potential energy field and a virtual viscous resistance field; then, using a Darcy law variant model for dynamic evolution, generating a virtual flow velocity vector field to simulate the sliding behavior of fluid in heterogeneous media; subsequently, constructing a heterogeneous graph based on the flow field divergence singular point and streamline trajectory, using a graph neural network to aggregate the node dynamics characteristics for deep reasoning, and finally generating a tampering positioning mask. The present application innovatively introduces fluid mechanics field theory, converts hidden static texture differences into significant dynamic flow field anomalies, solves the problem that the prior art is difficult to capture microscopic tampering traces, and significantly improves the detection accuracy and generalization ability in complex document scenarios.

A kind of flow video active interaction control method, equipment and medium for the scene of helping blind, construct the hierarchical task system for the scene of helping blind, train the semantic understanding and behavior alignment of multimodal large model to specific task of helping blind;Introduce the active interaction control mechanism for flow video, under the premise of not changing the bottom model parameter, realize the trigger output of environmental perception by the dynamic suffix switching of determination mode and generation mode;Finally, the incremental context updating strategy is used to optimize the inference process, while keeping the timing consistency and reducing the calculation redundancy.The invention solves the technical problems of lack of special data, passive interaction logic and calculation redundancy of multimodal large model in flow environment through the synergistic effect of data-driven task adaptation and active interaction mechanism.The general framework proposed in the invention provides an important theoretical basis and technical paradigm for the paradigm evolution of flow video understanding from "passive analysis" to "active interaction".

The present application relates to the field of data processing, and particularly relates to a user travel behavior label identification method and device, the method comprising: obtaining target user information and converting into a user feature vector; determining a travel behavior label corresponding to the user feature vector based on a plurality of preset travel behavior labels respectively corresponding to a plurality of determination conditions and a plurality of travel behavior label prediction models; if the user feature vector and any determination condition have a preset association relationship, then the travel behavior label corresponding to the determination condition is taken as the result; otherwise, the user feature vector is input into the plurality of travel behavior label prediction models, and the result is determined according to the model output. The present application sets different key feature subsets as determination conditions for different travel behaviors, and combines the two-stage architecture of rule priority and model bottoming, thereby ensuring the identification accuracy, reducing the calling frequency of complex models, and improving the overall identification efficiency.

The application relates to the technical field of image processing, in particular to an image super-resolution reconstruction method based on a multi-scale content perception mixer, which is realized by using an adaptive processing mechanism. The method comprises the following steps: shallow feature extraction is performed on a low-resolution image to be reconstructed, so as to obtain an initial shallow feature map; feature enhancement based on a feature pyramid and an attention mechanism is performed on the shallow feature map, so as to obtain a deep feature map; multi-scale content perception prediction is performed based on the deep feature map, so as to generate guide information for guiding calculation allocation, the guide information comprising a window classification binary mask and a window size; different image regions are allocated to different calculation paths for processing based on the guide information; the feature maps output by the calculation paths are recombined and fused, and then enlarged to a target resolution, so that a high-resolution image is finally obtained. The method realizes accurate classification of image regions and on-demand allocation of calculation resources, and significantly reduces the calculation complexity and the memory occupation.

This invention discloses an underwater electric field source localization method based on a determinant orthogonal propagation operator and adaptive hybrid-driven differential evolution, belonging to the field of electromagnetic positioning and detection. This invention acquires and receives array signals and calculates the covariance matrix. Based on the propagation operator, it constructs a projection operator matrix orthogonal to the array manifold to replace the noise subspace, thereby constructing a determinant spatial spectrum function and defining a cost function. Then, based on this cost function, it constructs an adaptive hybrid-driven differential evolution framework, fusing the alpha evolution algorithm and an adaptive differential evolution algorithm based on successful historical records to solve for the target position coordinates. This invention eliminates the need for eigenvalue decomposition, reducing computational complexity, and accelerates convergence through an adaptive hybrid-driven search strategy, enabling fast and high-precision electric field source localization in complex underwater environments.

The application provides an infrared and visible light image fusion method and system based on bidirectional semantic-space alignment and differential perception, and relates to the technical field of image processing. In order to solve the problem that the expression capacity is insufficient when the existing method processes complex cross-modal semantic interaction. The technical points of the application include: S1, collecting infrared images and visible light images; S2, fusing infrared and visible light images based on a semantic fusion network, wherein the semantic fusion network is based on a PSFusion network model, the SDFM module of the PSFusion network model is replaced by a differential perception complementary module DACM, the PSFM module is replaced by a deep semantic fusion block DSFB, and a space semantic alignment module SSAM is added after the deep semantic fusion block and the differential perception complementary module; S3, image reconstruction and supervision are performed on the aligned features. The method provided by the application shows superior potential compared with the current advanced image fusion algorithm.

The application discloses a fine-grained multi-modal video behavior recognition method and system guided by motion saliency, and the method comprises the following steps: selecting the first K image blocks with the most motion saliency from all image blocks corresponding to the DRDI image of each time segment as dynamic image blocks; selecting the corresponding K image blocks from the RGB image as static image blocks; splicing the dynamic and static image blocks under each time segment, and inputting the spliced image blocks into a visual encoder for fine-grained multi-modal interactive learning to obtain the feature representation corresponding to the current time segment; processing the feature representations of all time segments to obtain a video global feature representation; extracting a text feature representation from a text description of a behavior category; calculating the similarity between the video global feature representation and the text feature representation, and determining a video behavior recognition result according to the similarity. The application can improve the accuracy of video behavior recognition.

This invention discloses a high-efficiency video analysis method based on cloud-edge collaboration and adaptive MoE. The method includes: when an edge device performs a video analysis task, it first constructs a meta-request containing current task data and device resource constraints, and sends it to the cloud; the cloud uses a pre-trained basic model to extract features from the video data and clusters the data based on the feature space, dividing complex video scenes into multiple semantic subdomains; a lightweight expert model is trained in the cloud for each semantic subdomain; a routing decision model is further trained in the cloud; after deploying an expert model pool and routers on the edge device, for input data, the router first calculates the matching degree of each expert model and performs a comprehensive evaluation based on its computational cost. Through knowledge distillation, the large-scale basic model is decomposed into lightweight expert models, achieving a significant compression of the model size, enabling efficient deployment on resource-constrained edge devices.

The invention provides a disease and pest image recognition method and system. The disease and pest image of the surface of a target crop is acquired; segmenting the pest image into a plurality of super-pixel units based on the texture complexity of the target crop; determining color gamut differences of the superpixel units and neighborhood superpixel units under different color channels, generating color gamut difference vectors representing local anomalies of the superpixel units, and identifying candidate scab regions of the target crops; determining gradient histogram features of the candidate scab area under the red channel and the blue channel, and further determining mutual information entropy of a gradient histogram between the red channel and the blue channel; based on the color gamut difference vector and the mutual information entropy corresponding to the candidate scab area, cross identification of the target crop scab is carried out, and a disease and pest plaque on the surface of the target crop is obtained. According to the technical scheme provided by the invention, the scab area and the healthy area can be accurately distinguished under the crop surface with high texture complexity.

The invention discloses an operation quality evaluation method, device and equipment based on Beidou positioning, and a storage medium, and relates to the technical field of Beidou navigation and Internet of Things data fusion, and the method comprises the steps: carrying out the time sequence alignment processing of Beidou positioning terminal track data, the facility state data of Internet of Things sensor nodes, and mobile terminal image data, the aligned multi-modal data is obtained; extracting a space-time distance value between each track point in the track point position sequence data of the positioning terminal and each facility coordinate in the geographical coordinate set data of the facility from the aligned multi-modal data; determining associated facility data corresponding to each trajectory point according to the space-time distance value, and constructing a space-time constraint model of the trajectory and the facility point location according to the associated facility data; quantitative evaluation is carried out on the quality of the mobile supervision object based on the space-time constraint model, a space-time correlation evaluation result is obtained, and the space-time matching precision and correlation analysis efficiency of the Beidou positioning terminal and the Internet of Things facility in a complex urban environment are improved.

This application provides a resource scheduling method, apparatus, electronic device, medium, and program product, which can be applied to the fields of big data technology and artificial intelligence technology. The method includes: acquiring multi-dimensional feature data; constructing an input feature vector based on the multi-dimensional feature data, wherein the multi-dimensional feature data includes at least task information and resource information; inputting the input feature vector into an interpretable model, outputting a scheduling mapping function, and determining a target scheduling factor based on the scheduling mapping function; performing hierarchical classification of tasks and resources based on the target scheduling factor to obtain a multi-dimensional combined classification result; and generating a corresponding first resource scheduling result using the interpretable model based on the multi-dimensional combined classification result.

The invention belongs to the field of intelligent query, and particularly relates to an examination score and admission destination query and analysis method and system.The method comprises the steps that firstly, a query demand instruction is analyzed through a clustering algorithm, and a query cluster representing a query intention is generated; based on the initialized query concurrence tree, intelligently mapping query clusters to corresponding regional concurrence nodes according to examination number mapping or regional information, and establishing rapid indexes of examination numbers and nodes; analyzing a query intention condition field in the instruction, generating an effective query condition chain and evaluating query concurrency difficulty; and dynamically generating a query demand feedback information cluster by combining the real-time load state of each data storage node according to a comparison result of the concurrence difficulty and a preset threshold value. According to the method, intelligent routing of the query request and accurate scheduling of load awareness are realized, the processing pressure of a core node is effectively relieved, system overload and crash in a high-concurrency scene are avoided, and the throughput and service reliability of a score query system are remarkably improved.

The application discloses a multi-modal three-dimensional target detection method based on structural feature and semantic feature fusion, and belongs to the technical field of target detection.The application solves the problems of low precision, poor real-time performance and insufficient robustness of the existing method.The application constructs an explicit matching fusion mechanism of low-layer structure guided initialization, high-layer semantic optimization and time sequence expansion, projects 3D boundary boxes of a laser radar and a camera to multi-view image planes, calculates geometric similarity and category consistency constraints of 2D projection regions, and constructs an explicit matching graph in combination with low-layer structure information and high-layer semantic features.The application guides weighted aggregation of cross-modal target features with high confidence through a sparse matching graph, aligns historical frame targets to a current frame coordinate system to construct a space-time matching graph, aggregates historical information through expansion to a time sequence dimension, and realizes efficient and accurate three-dimensional target detection of the laser radar and the visual multi-view camera.The application method can be applied to multi-modal three-dimensional target detection.

The application discloses a kind of viscoelastic scalar P-wave equation construction and wave field numerical simulation method, belong to deep sea oil and gas field exploration development field.For the problems of insufficient physical completeness of existing scalar acoustic equation, high cost of multi-component elastic wave equation calculation, and difficulty of viscoelastic equation of constant Q model to fully reflect the wave field dynamics characteristics, the application establishes the dispersion relation based on constant Q model, constructs the viscoelastic multi-component wave equation containing fractional order space differential operator and time derivative coupling term, then introduces the dual projection operator to realize P-wave and S-wave mode decoupling, obtains the viscoelastic scalar equation describing only P-wave, and uses pseudo-spectral method and finite difference method for numerical simulation.The method can more accurately depict the propagation dynamics of seismic wave in strong attenuation medium, effectively decouples amplitude attenuation and phase dispersion effect, and provides reliable theory and calculation tool for high-precision seismic data processing and imaging.

This application discloses a method, apparatus, and device for detecting defects in printed circuit boards based on an improved YOLO, relating to the field of circuit board defect detection technology. The method includes: inputting the image to be detected into a backbone network, where the C3K2-IMCA module extracts features through cross-scale information interaction and spatial-channel co-modeling, outputting multiple first feature maps at different scales; inputting the first feature maps into a neck network, where the CCFM-Neck performs information interaction and fusion of feature maps at different levels to obtain multi-scale fused features; and inputting the multi-scale fused features into a detection head, where the LDCD-Head performs feature processing through detail enhancement convolution to obtain the defect detection result. This invention achieves lightweight modeling while suppressing background interference and enhancing small target detection capabilities, resulting in high detection accuracy, good real-time performance, and suitability for deployment on edge devices.

This invention discloses an air-ground collaborative stereoscopic electromagnetic spectrum mapping device and method. The device includes an airborne platform, a vehicle-mounted platform, and a ground station. First, it receives mapping parameters, plans a collaborative acquisition path between the airborne and ground platforms, and simultaneously collects, frames, and stores spectrum and spatiotemporal information. Second, after compensating the spectrum data from each platform, it performs three-dimensional fusion of air, time, and frequency. Based on the platform's real-time location and the fused data, it delineates dynamically complete areas and completes missing grids using an online optimization algorithm with a forgetting factor. Finally, it calculates the map coverage in real time; if the coverage exceeds a threshold, it replans the supplementary measurement path until a complete stereoscopic spectrum map is generated. This invention combines the advantages of airborne and ground platforms to achieve high-precision three-dimensional spectrum perception in large-area, complex scenes, improving completion accuracy and real-time performance.

The application relates to a state-aware dynamic asymmetric space-time Transformer long-term traffic speed prediction method and system, and belongs to the technical fields of intelligent traffic systems, traffic flow prediction and deep learning, which comprises the following steps: step 1: traffic road network directed graph construction and long-term prediction problem formalization; step 2: traffic speed time series data preprocessing; step 3: space-time embedding feature mapping; step 4: state-aware dynamic asymmetric space feature extraction; the macroscopic traffic flow theory is deeply integrated into an attention mechanism design, asymmetric space dependence of time-varying traffic flow is captured through four sub-steps of traffic state discrimination, dynamic mask generation, double-branch feature extraction and adaptive gate fusion, step 5: lightweight multi-scale time Transformer feature extraction; step 6: based on the deep space-time features of the output, the traffic speed prediction results of future time steps are directly generated through a linear mapping layer. The application realizes high-precision and high-stability long-term traffic speed prediction.

This invention relates to the field of audio processing and discloses a method, apparatus, and device for audio recognition of bullying events based on hyperbolic space. The method includes: acquiring real-time audio signals, extracting semantic features and acoustic event features, and embedding them into a hyperbolic space; obtaining the target state trajectory of the continuous time-dependent coupling evolution of semantic features and acoustic event features on a hyperbolic manifold within a preset observation window, and calling micro-operators or full operators based on phase transition states to monitor bullying events. This invention provides a high-capacity non-Euclidean geometric space through hyperbolic mapping, then characterizes the dynamic path of the causal evolution logic between semantic and acoustic events within this space, and finally performs resource scheduling based on the system's stable state. This allows for high-precision, real-time monitoring of complex and dynamically evolving school bullying behavior with low computational cost, significantly reducing frequent false wake-ups of devices due to environmental decibel fluctuations, while ensuring timely response to school bullying events.

The invention specifically relates to a fruit tree trunk identification and positioning method based on binocular machine vision, and relates to the technical field of autonomous operation of intelligent agricultural equipment, and the method comprises the steps: 1, creating and preprocessing a data set; step 2, carrying out optimization construction on a YOLOv7-tiny trunk identification model; 3, training a trunk identification model; 4, evaluating the trunk identification model; 5, detecting a trunk identification model; and 6, constructing a positioning model and solving the position of the trunk. According to the method, the tree trunk recognition performance is remarkably improved through multi-dimensional optimization, and the core advantage is reflected in the balance of recognition precision and operation efficiency; data set construction covers single independent, high and dwarf, complex illumination background and other multi-element scenes, the data robustness is high after HSV enhancement, overturning and other processing, and a foundation is laid for accurate recognition; in terms of model improvement, a PConv module replaces conventional convolution to reduce calculation redundancy, and a SimAM attention mechanism strengthens tree trunk key features, so that the improved YOLOv7-tiny model is optimal in an ablation experiment.

The invention belongs to an underground cable fault positioning method, and provides a traveling wave time delay estimation method and a related device in order to solve the problem that a traveling wave signal processing algorithm cannot completely meet high-precision and high-adaptability traveling wave time delay detection requirements in a complex electric power scene. Carrying out adaptive frequency band wavelet deconstruction and detail extraction on the first sensing signal and the second sensing signal respectively, and then carrying out dynamic parameter modal splitting based on variational problem solution on each reconstructed signal in the obtained multi-layer first reconstructed signal group and the second reconstructed signal group corresponding to the scale; the method comprises the following steps: obtaining a plurality of intrinsic frequency band modal components with specific sparsity, taking the first intrinsic frequency band modal component as a main mode, comprehensively considering the energy ratio and cross-correlation sharpness of two main modes in each layer, and determining a final traveling wave time delay estimation result according to the cross correlation coefficient of the two main modes under the optimal common decomposition layer. According to the method, the calculation redundancy is reduced while the precision is ensured, and the real-time performance is improved.