646 results about "Structural similarity" patented technology

The invention discloses a panchromatic sharpening method based on multi-resolution panchromatic feature guidance, which comprises the following steps: firstly, designing a multi-resolution feature extraction network based on a spatial frequency Transform module, and respectively extracting multi-scale features from panchromatic and up-sampled multispectral images; secondly, in order to restrain and optimize the extracted detail features, a multi-head self-attention-based texture injection module is adopted to extract a dependency relationship between panchromatic and multispectral features, and more similar detail features are guided to be injected; thirdly, in order to solve information loss caused by multispectral image up-sampling, multi-level detail features are injected into a multispectral reconstruction network, the detail features are learned step by step and reconstructed to a panchromatic image size, and a spectrum fusion module based on adaptive channel attention is utilized to reconstruct a fused image in a high-quality mode; and finally, optimizing the model performance by combining reconstruction loss based on L1 norm, structural similarity constraint and transmission perception loss in training. According to the method, the spectral fidelity and the spatial resolution of the fused image are remarkably improved.

The invention provides an AI-based post competency and recruitment post matching method, and relates to the technical field of big data processing, and the method comprises the steps: obtaining multi-source heterogeneous data, and extracting dominant skill entities and implicit ability entities of candidates; constructing a candidate ability evolution graph; analyzing a post dynamic demand by combining a time sequence prediction model, and forming a dynamic post portrait including a current demand, a hidden demand and a future evolution demand; calculating the structural similarity between a candidate map and a post portrait through a map neural network matching algorithm, and generating and sorting comprehensive matching scores in combination with adaptive dimensions such as core capability difference, future integrating degree and team integration potential; team cooperative effect simulation is introduced, the influence of candidate addition on a team capability structure and a cooperative network is predicted, and the model is optimized through a closed-loop feedback mechanism; the talent recognition accuracy and coverage range are remarkably improved, prospective strategic talent matching is achieved, and the matching efficiency is improved.

The invention relates to the technical field of power plant management and control, and discloses an intelligent power plant management and control system based on the Internet of Things, and the system comprises the steps: when a plurality of labels related to the same equipment or parameter exist in different subsystems, according to a similarity index between the labels and an equipment association relationship; on the basis of the detected conflict label group, combining equipment historical operation and maintenance data and an upstream and downstream parameter flow relationship, constructing a label semantic evolution graph, and performing reasoning analysis on label equipment through a fusion rule engine and a graph neural network; through mapping knowledge domain fusion and semantic embedding comparison, matching and clustering among conflict labels are completed based on structural similarity and semantic relevancy, and a label alignment rule is constructed; according to a label coordination result, designing a mapping rule of a data field; and performing inter-system synchronous verification on a result after structure conversion and label standardization processing, and writing a standardized label into a unified semantic database. The method has the advantage of improving data semantic consistency.

The invention relates to the technical field of multi-image fusion recognition, in particular to a data fusion method and system for CCD visual detection, and the method comprises the following steps: obtaining a horizontal pixel row calculation gradient construction trend sequence, repairing an edge fracture to generate an integrity index, extracting a gray value to detect feature mutation, and distributing fusion weights to establish a mapping relation. And executing image fusion and balancing the contrast to generate a fusion matrix result. According to the method, the fracture edge region is identified, interpolation compensation is executed, the structural similarity index of the local gray sequence in the image overlapping region and feature direction mutation detection are combined, accurate identification of the edge matching result is guided, and fusion weight factor mapping corresponding to signal-to-noise ratio distribution is introduced; according to the method, the distribution relation between the pixels in the region and the credible weight is effectively established, the edge transition among the multi-source images is more natural through Poisson constraint and contrast balance adjustment of the fusion region, and the structural fidelity and the judgment stability of the fusion image are remarkably enhanced.

The invention discloses an image super-resolution reconstruction method based on deep learning, and belongs to the technical field of image signal processing. Comprising the following steps: S1, realizing multi-dimensional feature representation of a low-resolution input image by using a deep learning network; s2, realizing accurate alignment and lossless mapping of cross-scale features, and effectively solving the problems of feature dimension inconsistency and information distortion; s3, explicitly learning an increment mapping relation between low-resolution input and a reconstructed high-resolution image, and relieving the gradient disappearance problem of a deep learning network structure; s4, enhancing the naturalness and detail fidelity of the reconstructed image through adversarial training; s5, integrating pixel-level reconstruction loss, perception loss and structural similarity loss, and effectively balancing objective indexes and subjective visual quality of the reconstructed image; and S6, realizing joint optimization of deep learning network parameters through a unified gradient back propagation mechanism, and finally outputting a high-quality super-resolution reconstructed image.

The invention relates to the technical field of image feature extraction, in particular to a motion data analysis method and system based on image processing, and the method comprises the following steps: collecting skeleton and inertial data, marking an action boundary, extracting a direction sequence, detecting a turning frame, analyzing a structural similarity labeling posture, screening a sudden change frame, and extracting an edge direction. And fusing direction optimization attitude features. According to the method, the image skeleton coordinates and the three-axis inertial data are synchronously obtained, the inter-frame displacement and the acceleration amplitude are paired, the action boundary marking time sequence precision is improved, the bidirectional change sequence is constructed according to the joint relative position difference and the angular velocity integral, the action trend recognition continuity is enhanced, and the recognition precision is improved. A main axis direction and a joint arrangement vector are extracted by sliding a frame window, a trend sequence is constructed in combination with structural similarity, attitude state dynamic labeling is realized, an edge direction is introduced to analyze and identify a sudden change frame, and the local sensitivity of skeleton offset detection and the abnormal frame detection precision are improved.

The invention discloses a three-dimensional human body reconstruction method based on an implicit neural network and a diffusion model, and belongs to the field of three-dimensional reconstruction. According to the invention, a diffusible density volume representation is constructed, and identity-based rendering is realized by using a diffusion model. By adopting parameterized three-dimensional distribution representation, Pareto optimum, balanced realistic rendering and feasible prediction are realized in an unbalanced optimization framework. In order to increase the rendering training speed without sacrificing the quality, mixed multi-resolution hash grid features are adopted to enhance the directional feature representation capability. Meanwhile, a multi-scale structural similarity constraint (MSSC) is provided to maintain identity distinguishability and ensure geometric consistency, so that synthesis with high calculation efficiency and satisfactory visual effect is realized. According to the method, the three-dimensional human body model can be reconstructed from a single image based on limited geometric and appearance information, the method is excellent in topology credibility and visual fidelity, and a more flexible and efficient solution is provided for an actual application scene.

The invention discloses an automobile part defect detection method and system, and belongs to the technical field of image recognition, and the method comprises the steps: deploying at least one collection device at a detection line, transmitting first image data collected by the collection device to a data processing center, and generating second image data; establishing an illumination model, performing illumination analysis on the second image data, and identifying a first region and a second region; after the overexposure and shadow areas are identified, illumination correction and optimization are carried out on the image, and original details of the image are recovered; and performing quality evaluation on the repaired image, quantifying the image repairing effect by adopting a peak signal-to-noise ratio and a structural similarity index, and performing defect analysis after the evaluation is completed. In the implementation process of the technical scheme of the invention, by establishing the illumination model, performing illumination analysis on the second image data, and identifying an overexposure region and a shadow region, the image repairing efficiency is improved. And then different optimization methods are adopted for different areas, so that the image quality and the detection precision are improved.

The invention discloses a medical image generation method and device based on bimodal fusion, equipment and a medium, and the method comprises the steps: respectively extracting the visual features of a medical image and the semantic features of text description through an image encoder and a text encoder; mapping the two types of features to a shared semantic space by adopting comparative learning to realize cross-modal alignment; a first path captures hierarchical semantic information through a convolutional neural network, a second path retains local significant features through maximum pooling, and two paths of outputs are fused layer by layer to construct spatial context features; projecting the cross-modal features into a spatial feature map through a multi-layer perceptron, splicing the spatial feature map with coding features, and inputting the spliced spatial feature map into a decoder for up-sampling reconstruction; and performing joint optimization on the comparison loss and the structural similarity loss to realize end-to-end training. And finally, the SSIM index of the generated medical image is improved.

The invention relates to a multi-stage filtering road spilled object detection method based on dynamic difference analysis, which is suitable for automatic identification of unstructured foreign matters in video monitoring. The method comprises the following steps: firstly, extracting a reference image road mask, eliminating vehicle and pedestrian interference by using YOLOv8 detection, and extracting a motion candidate area through a frame difference method and background modeling; and then context expansion and super-resolution reconstruction are carried out on the candidate region, the candidate region is converted into an HSV space, multi-dimensional features such as color similarity, structural similarity and shadow determination are synthesized for screening, false detection is further removed in combination with inter-frame time sequence consistency, and finally a stable detection result is output. The method provided by the invention has the advantages of strong anti-interference capability, high adaptability, high detection precision and the like, and is suitable for the intelligent recognition task of the expressway thrown objects in a complex environment.

The invention relates to the technical field of software test automation, in particular to a self-repairing UI automatic detection method based on multi-modal fusion, which mainly comprises the following steps: extracting visual features to obtain feature vectors, and constructing a virtual DOM tree with semantic tags to carry out similarity calculation to obtain structural similarity; and carrying out weight dynamic adjustment on the feature vector and the structural similarity through a feature fusion decision maker, outputting a fusion weight proportion under a current confidence threshold, fusing the feature vector and the structural similarity through the weight proportion to form a decision combination, and generating a new locator. Through the method, the problem of test case vulnerability caused by a dynamic user interface is solved, the detection speed of real-time sensing of interface change is improved, and meanwhile, the positioning success rate during use of a new positioner automatically generated by a positioning strategy is also remarkably improved.

The invention relates to a pig image BDE reconstruction system and method based on differentiatable image rate filtering, and belongs to the field of image processing, and the system comprises an image preprocessing module which is configured to process an original low-bit image; the neural network module acts on the processing and reconstruction tasks of low-bit pig images, the characteristics of each image are collaboratively optimized in a task dividing and conquering mode, and finally a high-quality image is obtained; the loss function module is configured to optimize the network model performance by minimizing the difference between the network output image and the real image; the training module is configured to optimize network model parameters and a learning process by adopting a staged training strategy; and the test and evaluation module is configured to evaluate the image quality through the peak signal-to-noise ratio and the structural similarity index and measure the reconstruction effect of the model. According to the method, a two-stage training strategy is adopted, an image reconstruction task is decomposed into two sub-tasks which cooperate with each other, and spatial resolution enhancement and bit resolution improvement are achieved.

The invention relates to the technical field of knowledge processing, and particularly discloses a knowledge proposition error correction method and system based on knowledge graph optimization upgrading, and the method comprises the steps: extracting effective information from multi-source knowledge data through employing entity linking, relation extraction and attribute alignment methods, and constructing an initial knowledge graph. Through semantic analysis based on deep learning, grammar, semantic and logic analysis is carried out on knowledge propositions, and a proposition element triple is generated. A graph neural network is utilized to encode triples, proposition vectors and neighborhood feature vectors are obtained, structural similarity is calculated accordingly, and proposition errors are detected in combination with rules and a statistical method. An integrated learning algorithm is adopted to classify error traceability, and an error correction decision is made in combination with multi-source evidence, so that an accurate error correction result is obtained, and the accuracy and reliability of knowledge propositions are improved; the error correction process is more logical and reliable, blind error correction is avoided, and the quality and efficiency of knowledge proposition error correction are improved.

The invention discloses a radar echo extrapolation method and system based on frequency domain enhancement, and the method mainly comprises the following steps: obtaining and preprocessing a historical radar echo grayscale image sequence, generating a sequence sample through a sliding window, and dividing the sequence sample into a training set, a verification set and a test set; the method comprises the following steps: constructing a frequency domain enhanced U-Net network comprising an encoder-decoder structure, introducing a multi-scale deep convolution structure into an encoder and a decoder, and enhancing frequency domain features by using a frequency domain dynamic attention mechanism in jump connection; inputting the training set into the model for training by adopting a composite loss function comprising intensity weighted loss, frequency domain consistency loss and structural similarity loss; and inputting the test set into the trained model, and outputting a radar echo prediction result at a future moment. The method can be effectively applied to the fields of short-term and temporary weather forecast, severe convection monitoring and the like, and provides more accurate and reliable radar echo prediction support for meteorological disaster early warning.

The invention discloses a microscopic image seamless splicing and enhanced reconstruction method based on deep learning, and the method comprises the following steps: S1, collecting a plurality of original images with overlapped regions, and recording the spatial position information and imaging parameters of the original images; s2, preprocessing the original image to generate a standardized image sequence; s3, inputting the standardized image into a structure perception feature extraction network, and extracting a feature map fusing textures and structures; s4, inputting the feature image and the original image into an image registration module; s5, inputting the registration image into the boundary attention splicing network; s6, inputting the seamless image into the residual hierarchy reconstruction network, and enhancing image details through hole convolution and multi-scale branches; s7, image quality evaluation is executed, and the structural similarity, the signal-to-noise ratio and the edge retention rate are calculated; and S8, constructing a training set and carrying out end-to-end training optimization based on a joint loss function. According to the method, a multi-module deep network is fused, and seamless splicing and high-quality enhanced reconstruction of microscopic images are realized.

The invention discloses an MRI-PET mode conversion method and system based on a generative adversarial network, and belongs to the technical field of artificial intelligence medical image generation. And the multi-scale structure representation injection module injects multi-scale anatomical prior information at different stages of the encoder, and overcomes the limitations of insufficient utilization of prior information and single injection scale. And the adaptive semantic residual fusion module adopts semantic attention guidance and double-branch attention weighting, adaptively fuses fine-grained local features and global context information, harmonizes the difference between the fine-grained local features and the global context information in an abstract level and a semantic category, and solves the problems of feature conflict and semantic fuzziness in a bottleneck region. The direction sensing space-frequency discriminator realizes multi-dimensional and fine-grained adversarial supervision through a space, frequency and local image block multi-branch collaborative discrimination mechanism, and improves the structural fidelity and spectrum authenticity of a synthetic image. And the generated image is superior to the existing method in indexes such as structural similarity and peak signal-to-noise ratio, and has higher clinical practical value.

The invention belongs to the technical field of image processing, and particularly relates to a wood surface defect detection method and device based on multi-view anomaly detection, and the method comprises the steps: obtaining a target image based on an original wood image; constructing a wood surface defect detection model comprising an encoder, a decoder and a defect marking module, constructing positive and negative samples based on the target image, and training the encoder by using the positive and negative samples through contrast learning; the output of the encoder is used as the input of a decoder, a loss function is constructed through a mean square error and a structural similarity index to train the decoder, and the decoder outputs a reconstructed image; obtaining a residual image between the reconstructed image and the target image through a defect marking module to obtain a fused residual image, and marking a pixel region exceeding a dynamic threshold as a defect; and detecting the preprocessed to-be-detected wood image by using the trained wood surface defect detection model to realize wood defect detection and pixel-level positioning.

The invention discloses a three-dimensional dynamic scene graph construction method based on 3D Gauss, and belongs to the field of computer body intelligence. The implementation method comprises the following steps of: realizing object perception and semantic feature extraction of open vocabularies by utilizing a visual basic model; a 3D Gaussian scene with high fidelity and continuous object semantics is constructed through multi-view multi-dimension optimization of 3D Gaussian representation; constructing a multi-level three-dimensional scene graph, extracting spatial levels and semantic relationships among objects by using 3D spatial positions and semantic tags of instance objects existing in a semantic Gaussian graph, and constructing a multi-level spatial semantic topology to accurately represent an environment layout; according to the method for realizing local updating for the Gaussian scene graph based on the environmental structural similarity, environmental change detection is carried out through real-time RGB-D observation and the structural similarity between high-quality rendering views of the Gaussian scene graph, and corresponding local updating is carried out by using rapid training and differentiable rendering of 3D Gaussian representation. And the capability of adapting to a complex dynamic environment of the 3D Gaussian scene graph is improved.

The invention provides a sea surface height abnormal data downscaling method based on deep learning, and relates to the technical field of data processing, and the method specifically comprises the following steps: preprocessing data including AVISO data, SWOT data and key auxiliary variables; an enhanced super-resolution generative adversarial network ESRGAN is constructed, a high-resolution feature map is generated, and the ESRGAN comprises a generative network and a discrimination network; using a plurality of loss function combinations to optimize the performance of the generative network and the discriminant network, including adversarial loss, perception loss and pixel loss; and evaluating the pixel-level error and the spatial detail consistency output by the ESRGAN by adopting a root-mean-square error and a structural similarity index respectively. According to the technical scheme, the problems that in the prior art, a data downscaling method is limited in applicability and generalization ability, and is difficult to adapt to fusion requirements of different regions and multi-source data are solved.

The invention discloses a space target point cloud three-dimensional reconstruction and rendering method, belongs to the field of space target scene reconstruction and rendering, solves the problem of low accuracy of space target scene reconstruction and rendering, and comprises the following steps: generating an initial sparse point cloud and a camera pose of a space target by using an SFM; performing denoising and outlier processing on the initial sparse point cloud by using a bilateral filtering algorithm to obtain an initial point cloud; distributing an initial gray Gaussian volume for each point in the initial point cloud, and projecting the initial gray Gaussian volume to a two-dimensional plane by using a projection formula in combination with a camera pose; rendering the two-dimensional projection image by using a micro-rasterization rendering technology; fusing the luminosity loss function, the structural similarity loss function and the depth loss function to construct a total loss function, and optimizing an initial gray Gaussian body attribute in the rendered image by using a stochastic gradient descent method; a self-adaptive density control mechanism is introduced to adjust the density of the initial point cloud, and a space target point cloud three-dimensional reconstruction structure is generated; according to the invention, the reconstruction and rendering accuracy of the space target scene is improved.

The invention relates to the technical field of content recognition, in particular to an Al-based video content analysis method and system, and the method comprises the following steps: carrying out the image segmentation learning based on a video sequence through employing a U-Net convolutional neural network, analyzing scenes and elements in a video frame, recognizing and isolating key visual elements in a video through network learning, and carrying out the recognition of the key visual elements in the video. Comprising objects and figures. According to the method, image segmentation learning is carried out by adopting the U-Net convolutional neural network, key visual elements in the video can be identified and isolated more accurately, a clearer basis is provided for follow-up scene change and key event tracking, video content analysis is carried out by applying the graph neural network, the identification capability of dynamic scenes and events is enhanced, and the identification efficiency is improved. Deeper structured understanding is provided for video content indexing, the video quality is analyzed through a structural similarity index evaluation method, and the reason of visual quality reduction can be accurately recognized and improved.

The invention discloses a retrieval enhanced warehouse-level code completion method based on semantic topology and disambiguation redundancy pruning. Firstly, a key value code library for accurate mapping of source code fragments and induction sub-graphs is constructed through a specific slicing algorithm; in the retrieval stage, four-level layered optimization is adopted, structural similarity is evaluated by extracting a deep semantic relationship, trimming accurate duplicate items and using a new graph-based measurement (tradeoff is performed on editing according to topological importance), results are reordered to maximize correlation and diversity, candidate items are systematically refined, and the retrieval efficiency is improved. The problems of retrieval redundancy solidification and surface similarity misleading are solved, meanwhile, cross-module dependence is analyzed through an external perception identifier disambiguator, and the problem of cross-file symbol ambiguity is solved; and finally, fusing an optimization result to generate a prompt to drive the LLM to generate higher-quality output. According to the invention, through coordination of semantic and structural signals, strong performance can be obtained even in a large-scale and resource-limited code library. Meanwhile, the design allows it to be orthogonally complementary to other cross-file methods, providing collaborative improvements when used in combination.

The invention relates to the technical field of three-dimensional reconstruction, in particular to a near-infrared assisted low-light scene three-dimensional reconstruction method based on 3D Gaussian splashing, which is characterized in that generation of a Gaussian ellipsoid is dominated by a near-infrared image with a high signal-to-noise ratio, and a stable geometric basis is provided for a visible light information missing area; in the rendering stage, a near-infrared rendering image and a normal visible light rendering image are respectively generated through Gaussian ellipsoid shared geometric parameters and respective opacity and color attributes of two modes; through cross-modal structure similarity loss, a normal visible light image which is forcibly rendered is aligned with a near-infrared light image in structure, and clear near-infrared structure information is used to strictly constrain the recovery process of the color of the visible light image, so that the accuracy and authenticity of a recovery result are ensured.

The invention discloses a retraining-free pruning and recombination method for a sparse expert hybrid large model, and belongs to the technical field of large model compression and optimization. The method aims at solving the problems that due to the fact that an existing sparse expert hybrid (SMoE) model needs to load all expert parameters, memory occupation is too high, and deployment is difficult. According to the method, firstly, redundant experts are identified and pruned based on routing activation statistics; then, decomposing the pruned experts into neuron-level functional fragments, and redistributing the fragments to the reserved experts according to structural similarity; and finally, original fragments and newly distributed fragments are merged in the reserved experts through a weighted clustering algorithm, so that compact experts with fewer parameters and stronger expression ability are reconstructed. According to the method, fine-grained operation is carried out at the neuron level, the inherent representation conflict and dislocation problems among experts are effectively solved, the performance of the compressed model is remarkably improved, and reliable technical support is provided for deploying a large-scale SMoE model.

A bidirectional autoencoder learns or approximates an identity mapping as it trains a single network with a version of the new bidirectional backpropagation algorithm. Ordinary unidirectional autoencoders find many uses in image processing and in large language models. But they use separate networks for encoding and decoding. Bidirectional autoencoders use the same synaptic weights for encoding and decoding. The forward pass encodes while the backward pass decodes. Bidirectional autoencoders improved network performance and significantly reduced memory usage and used fewer parameters. Simulations compared unidirectional with bidirectional autoencoders for image compression and denoising. The models trained on the MNIST handwritten-digit and CIFAR-10 image datasets. The performance measures were the peak signal-to-noise ratio and the index of structural similarity. Bidirectional autoencoders outperformed unidirectional autoencoders and still reduced the number of trainable synaptic parameters by about 50%.

The invention discloses a multi-modal face recognition method and system based on deep learning, and relates to the technical field of face recognition, 3D structured light, near-infrared images and PPG blood flow signals are synchronously fused, occlusion completion is achieved through a double-branch generation network, PPG spectrum constraints generate blood vessel distribution, and biological feature authenticity is ensured; low-confidence sub-regions are divided based on a structural similarity algorithm, the weight of a loss function is dynamically adjusted, only problem regions are iteratively generated, and calculation redundancy is reduced; the depth feature of the complemented image and the blood vessel frequency spectrum main frequency band overlapping rate are extracted, the identity authenticity is doubly verified, and 3D mask and photo attacks are resisted; the performance bottleneck of traditional single-mode identification in a shielding scene is broken through, closed-loop optimization is driven through data, precision, efficiency and safety are considered, and the method is particularly suitable for high-requirement scenes such as medical treatment and finance.

The invention discloses a defogging method based on interlayer multi-scale sequence interaction and Fourier domain frequency-space enhancement, and the method comprises the following steps: S1, collecting a public foggy data set containing different concentrations of haze, and carrying out the preprocessing of an image in the public foggy data set, so as to improve the data quality and a model training effect; s2, constructing a defogging network framework based on interlayer multi-scale sequence interaction and Fourier domain frequency-space enhancement, wherein the framework comprises a U-shaped defogging network, an interlayer multi-scale sequence interaction module IMSIM and a Fourier domain frequency-space enhancement module FDFSEM; s3, training the network by using the marked image data set, defining a proper loss function, and training by using an optimizer; s4, evaluating a defogging effect by using two indexes, namely a peak signal-to-noise ratio (PSNR) and a structural similarity index (SSIM), optimizing the trained model, and deploying the model to practical application for online reasoning and real-time defogging; through the steps, defogging of the input feature map can be realized.

The invention provides a wind power data determination method and device based on geographical constraint attention. The wind power data determination method comprises the steps of obtaining topographic data, surface vegetation coverage data and wind power data; for the wind power data, combining the topographic data and the surface vegetation data, performing feature fusion through an adaptive attention mechanism to obtain a fusion feature containing multi-source information; performing shallow feature extraction, deep feature extraction and image up-sampling on the fusion features in sequence, wherein the deep feature extraction realizes multi-scale feature extraction and geographic constraint fusion through a multi-stage residual fusion module RFSTB, a wind field frequency domain decoupling module CFB and a geographic perception global attention module GAB; verifying the effectiveness of the model by using a peak signal-to-noise ratio (PSNR) and a structural similarity index (SSIM); and performing super-resolution reconstruction on the wind power data based on the trained model to generate wind power data with predetermined precision. The method provides support for fine evaluation and efficient development of wind energy resources in complex terrain areas.

The invention discloses a two-stage hyperspectral image band selection and target detection method, which comprises the following steps of: based on a two-stage band selection and background reconstruction network of a transformer, realizing band selection through a first-stage training transformer and realizing background reconstruction through a second-stage training transformer; a transform position coding module and a multi-head self-attention mechanism are used for learning similar features and difference features among wave bands, a full connection layer network is used as a clustering device, a structural similarity index and an Euclidean distance are added to serve as a loss function training model, wave band clustering is achieved, and finally a local variance is used for estimating the noise level of an image of each wave band. Obtaining a band training subset; the band subset is subjected to constraint energy minimization detection and then sent to a background feature extraction network composed of a transformer and a discriminator, background pixels are arranged through position coding, network learning background features are enhanced through a multi-head self-attention mechanism, an improved loss function is used for training, and therefore reconstruction of a background image is achieved. The invention provides a two-stage wave band selection and background reconstruction network based on transformer so as to realize hyperspectral target detection. The network comprises a first-stage wave band selection work and a second-stage background spectrum learning work. And difference detection is carried out on the final background reconstruction image and the original image, so that a final task is realized, and the detection precision is improved.

The invention discloses a weak supervision super-resolution method based on bimodal prior, and relates to the field of computer vision, and the method comprises the following steps: obtaining a scene text image, carrying out the feature extraction of the scene text image, and constructing an initial image super-resolution frame based on the extracted image features; the initial image super-resolution framework is optimized through a structural similarity loss function and a connection dominant time sequence classification loss function in combination with a back propagation mechanism, and a final image super-resolution framework is obtained; and inputting a pre-acquired scene text image into the final image super-resolution frame to obtain a reconstructed scene text image. According to the invention, through bimodal prior refinement processing, capture of a global structure and semantic information of a scene text image is enhanced, so that reconstruction quality of a super-resolution image is improved; and meanwhile, through global feature fusion processing, the overall quality of the generated high-resolution image is improved.