15results about How to "Resolve blur" patented technology

The invention provides a clothes defect detection method and system based on multi-spectrum characteristic harmonic complex, and relates to the technical field of image detection. The method comprises the following steps: establishing a multi-spectrum characteristic harmonic aggregation complex defect detection model and carrying out pre-training, wherein the multi-spectrum characteristic harmonic aggregation complex defect detection model comprises a spectral domain frequency division network, a spectrum harmonic aggregation network, a boundary spectrum complex network and a target detection output layer; inputting a to-be-detected image in the original clothing image data into the multi-spectrum feature harmonic congelation complex defect detection model, and processing the to-be-detected image by using a spectral domain frequency division network, a spectrum harmonic congelation network and a boundary spectrum complex network in sequence to obtain feature representation with enhanced edge perception capability, and performing target detection and classification on the feature representation through a target detection layer, and outputting a defect detection result of the to-be-detected image. The problems that in the actual production process, the garment flaws are difficult to recognize, and the manual detection efficiency is low are solved, and efficient and accurate detection of the garment flaws under the actual engineering application scene is achieved.

The application relates to the technical field of intelligent transportation and carbon governance, in particular to a passenger car carbon emission monitoring and regulation system and method based on multi-source spatio-temporal data. The application solves the problem of scattered and inconsistent data by preprocessing and uniformly mapping multi-source spatio-temporal data. A three-dimensional carbon emission calculation model of vehicle-space unit-time period is constructed to realize accurate calculation of single-trip carbon emission of a single vehicle and positioning of high-carbon emission hotspots. A clustering algorithm is used to complete differentiated division of user groups. A dynamic early warning threshold is generated based on a warning threshold prediction model, regional carbon emission targets and historical data to reduce false positives and false negatives. A warning type-policy tool-adaptation object mapping rule library is established, and a regulation strategy is generated by combining early warning signals and real-time data to realize automatic and accurate generation of regulation strategies from problem identification. The model parameter, early warning threshold and mapping rule are optimized by using policy implementation effect data to form a monitoring, early warning, regulation, optimization closed-loop management and realize accurate regulation.

The invention discloses a wake detection model construction method based on frequency domain multi-scale perception, which relates to the technical field of ocean observation, is used for capturing weak wake features to carry out sea surface internal wave wake detection, and comprises the steps of constructing a wake image data set, constructing a wake detection model by taking an RT-DETR model as a baseline model, and constructing a multi-scale texture information selection network. A backbone network of the baseline model is replaced, a frequency domain multi-scale attention perception module is constructed, an AIFI module of the baseline model is replaced, a golden cudgel convolution down-sampling module is adopted to replace a standard convolution down-sampling module of the baseline model, training and optimization are performed on the wake detection model based on the wake image data set, and a target detection result is obtained. According to the method, the wake detection model AUVRT-DETR is constructed, the detection precision is improved while the advantage of light weight is kept, the optimal balance is achieved between the detection precision and the calculation efficiency, and the model parameter quantity and the calculation overhead are remarkably reduced while the highest detection precision is achieved.

The invention relates to a method and system for reconstructing a hyperspectral image based on an RGB image, and the method comprises the steps: S1, selecting an RGB image in a space; s2, processing the RGB image to obtain end member information of the RGB image; s3, generating end member spectral representation based on the end member information of the RGB image; s4, encoding, decoding and attention modeling are carried out on the RGB image, and abundance information of the RGB image is obtained; and S5, constructing a rough hyperspectral image based on the end member information of the RGB image and the abundance information of the RGB image, performing feature enhancement on the rough hyperspectral image and the RGB image, and performing fusion to obtain a final hyperspectral image. According to the invention, the hyperspectral image can be effectively constructed based on the RGB image.

The embodiment discloses an automatic potential inspection system for a subsea pipeline, which is characterized in that a magnetic detector and a cesium optical pump are arranged on an underwater load platform, position signals between the subsea pipeline and the underwater load platform are respectively acquired, and an automatic inspection path signal of the underwater load platform is acquired based on the signals acquired by the magnetic detector and the cesium optical pump; the problem that in the prior art, image information is fuzzy when automatic routing inspection path signals are obtained depending on images is solved, the problems that the pipeline detection capacity and precision are limited by independently adopting magnetic detection equipment are solved, the obtained automatic routing inspection path signals are high in precision, and meanwhile the potential collection precision is not affected.

The invention relates to the technical field of fuel cell production, in particular to a fuel cell bipolar plate adhesive tape segmentation method, system, medium and equipment based on gradient attention and a multi-scale void residual structure, a MobileNetV3-U-Net fusion architecture is adopted, lightweight MobileNetV3 is used as an encoder, high-level semantic features of an image are extracted step by step through a multi-layer inverse residual block, and the high-level semantic features of the image are extracted through a multi-scale void residual block. Meanwhile, a DG Attention module is embedded in the bottommost layer of the encoding stage, the DG Attention module guides attention weight calculation by extracting gradient features and long-strip-shaped features and strengthens feature capture of the edge and the slender structure of the adhesive tape, a decoder reserves multi-layer jump connection of U-Net, meanwhile, a traditional convolution operation is replaced with a Res2DilatedConv module, and therefore the decoding efficiency is improved. Multi-scale feature extraction is realized through channel grouping, multi-voidage convolution and cross-group feature fusion, the structural difference of the adhesive tape under different widths and bending degrees is adapted, and efficient and accurate segmentation of the adhesive tape region in the fuel cell bipolar plate is realized.

This invention provides an intelligent inspection method, system, equipment, and medium for railway freight inspection, belonging to the field of railway freight inspection technology. The method includes: acquiring images of the freight cars using a line-scan camera array on both sides and the top of the track; performing spatiotemporal synchronization and adaptive parameter adjustment based on an encoder and a precise clock protocol; performing geometric correction, flat-field correction, motion compensation, and stitching on the images to generate a two-dimensional unfolded image of the freight car; using a deep learning-based freight car segmentation model to segment and number freight car instances, and extracting images of individual freight cars; finally, using a damage detection model integrating severity assessment to identify damage, outputting a structured result including damage category, location, confidence level, and severity score. This invention achieves non-contact, full-coverage, and high-precision intelligent inspection of freight cars, effectively improving inspection efficiency and automation levels.

This invention discloses a method for constructing a disease knowledge graph of the musk deer based on the GPLinker model, belonging to the field of natural language processing technology. The method includes: inputting the labeled disease corpus into the GPLinker joint extraction model; an encoding module extracting multi-scale semantic features from the disease corpus; a boundary awareness module obtaining the probability distribution of the boundary label corresponding to each word segment based on the multi-scale semantic features, obtaining the sum of the probabilities of the first label and the second label based on the probability distribution, and obtaining a gating weight signal; fusing the gating weight signal and the multi-scale semantic features to obtain enhanced features; a GPLinker module obtaining the target entity and the target relationship between the target entities based on the enhanced features, obtaining triples based on the target entities and entity relationships, and obtaining the musk deer disease knowledge graph based on the triples. This solves the problem that existing musk deer disease knowledge graph construction methods cannot accurately identify entity boundaries, resulting in low knowledge extraction quality.

The application discloses a dynamic power quality disturbance signal detection method, device and medium, and the method steps comprise the following steps: S01, acquiring a dynamic power quality disturbance signal to be detected; S02, dividing a signal frequency band, and performing time-frequency analysis on the dynamic power quality disturbance signal by using an improved variable resolution S transform method; the improved variable resolution S transform method is an S transform based on a Gaussian window, and a Gaussian window scale factor is used to optimize the Gaussian window function; S03, performing a synchronous squeezing rearrangement operation on the improved variable resolution S transform result, so that energy components scattered in adjacent frequency intervals are aggregated to real frequency axis positions, and a time-frequency matrix after synchronous squeezing rearrangement is obtained; and S04, detecting signal components of the time-frequency matrix after synchronous squeezing rearrangement. The application can reduce energy diffusion and realize high-precision detection of non-stationary multi-scale complex dynamic power quality disturbance signals.

The invention relates to the technical field of image enhancement, in particular to an image enhancement generation method based on face key point self-adaptive deformation. Comprising the steps of receiving a popular science museum face image data stream, performing topology reconstruction through a three-dimensional parameterized model, outputting a source face grid, mapping the source face grid to a UV parameter space, and decoupling and outputting an albedo texture tensor and a geometric texture tensor; calculating vertex displacement through a preset topology corresponding relation, constructing a differential homeomorphic deformation field, and constructing a continuous manifold coordinate field through nonlinear space adaptation; constructing an implicit neural texture representation network, performing overfitting training, inputting coordinate points of the continuous manifold coordinate field into the implicit neural texture representation network point by point, performing frequency modulation, calculating texture color values, and generating an enhanced facial image. According to the method, the continuous manifold coordinate field, the implicit neural texture representation network and the frequency modulation mechanism are deeply coupled, so that reconstruction from a low-resolution feature to a high-resolution visual signal is realized.

This invention belongs to the field of laser micromachining and relates to a coaxial monitoring system for ultrafast laser micromachining. The system includes an ultrafast laser, a beam-splitting plate, a laser focusing lens, an intermediate lens group, a filter, a fixed-magnification telecentric lens, and a high dynamic range camera. The ultrafast laser emits ultrafast laser light. The beam-splitting plate and the laser focusing lens are arranged sequentially from front to back along the optical path of the ultrafast laser. The laser focusing lens focuses the ultrafast laser light onto the workpiece, generating visible monitoring light on the workpiece. The visible monitoring light passes through the laser focusing lens and is reflected by the beam-splitting plate to form reflected visible monitoring light. The intermediate lens group, filter, fixed-magnification telecentric lens, and high dynamic range camera are arranged sequentially from front to back along the optical path of the reflected visible monitoring light. This invention provides a coaxial monitoring system for ultrafast laser micromachining that improves imaging clarity and resolution.

The invention discloses a sonar image super-resolution method and system based on a diffusion generation model, and belongs to the technical field of underwater image processing. The method comprises the following steps: generating an initial diffusion state based on an input image and Gaussian noise obeying standard normal distribution; inputting the current diffusion state and the current iteration time step into a diffusion generation model to obtain total noise contained in the image in the current diffusion state; combining the current diffusion state, separating the total noise into structural noise and random noise, and fusing the structural noise and the random noise to obtain optimized noise; calculating a preliminary estimation image corresponding to the current iteration time step based on the optimized noise and the current diffusion state; sampling noise obeying standard normal distribution is generated, and the diffusion state of the next iteration time step is generated based on the preliminary estimation image, the optimization noise and the sampling noise; and obtaining a super-resolution image until the inverse diffusion iteration process is finished. According to the invention, a high-resolution image which is high in fidelity, globally consistent in signal-to-noise ratio and adaptively deployed can be obtained.

The long-distance synchronous acquisition method and system of multiple normal and dynamic visual sensors provided by the embodiments of the present application comprise multiple normal visual sensors, multiple dynamic visual sensors and a control processor. The method is applied to the control processor and comprises the following steps: determining a target normal visual sensor and determining a target dynamic visual sensor; acquiring a first synchronization signal of the target normal visual sensor and sending the first synchronization signal to the remaining normal visual sensors after first delay compensation processing; sending the first synchronization signal to each dynamic visual sensor after second delay compensation processing; and sending a second synchronization signal emitted by the target dynamic visual sensor to the remaining dynamic visual sensors after third delay compensation processing, so as to perform data synchronous acquisition and improve the accuracy and synchronism of image data acquisition of multiple visual sensors.

The application discloses a sonar image super-resolution method and system based on diffusion generative model, and belongs to the technical field of underwater image processing. The method comprises the following steps: generating an initial diffusion state based on an input image and Gaussian noise conforming to a standard normal distribution; inputting a current diffusion state and a current iteration time step into a diffusion generative model to obtain total noise contained in an image under the current diffusion state; combining the current diffusion state, separating the total noise into structural noise and random noise, and then performing fusion to obtain optimized noise; calculating a preliminary estimated image corresponding to the current iteration time step based on the optimized noise and the current diffusion state; generating sampling noise conforming to the standard normal distribution, and generating a diffusion state of a next iteration time step based on the preliminary estimated image, the optimized noise and the sampling noise; and obtaining a super-resolution image until the inverse diffusion iteration process ends. The application can obtain a high-resolution image with high fidelity, a high signal-to-noise ratio, global consistency and adaptive deployment.

This utility model discloses a plant disease and pest detection device, including a self-propelled vehicle, a rotating seat, an L-shaped support column, a rotating cylinder, a visual inspection instrument, and a leaf-picking mechanism. The self-propelled vehicle has a rotating seat mounted on its top, connected to the L-shaped support column. A rotatable rotating cylinder is located at the end of the support column, and a turntable is located at the bottom of the rotating cylinder. A visual inspection instrument and supplementary lighting are installed below the turntable. The turntable has a leaf-picking mechanism that drives a two-way toothed plate to mesh with an incomplete gear via a threaded rod, thereby actuating two leaf-picking plates to separate branches and leaves, improving image clarity. This device has autonomous mobility and can perform multi-angle, high-efficiency detection in complex planting environments, making it suitable for automated monitoring of agricultural diseases and pests.