975 results about "Deconvolution" patented technology

The invention provides an unmanned aerial vehicle identification method and system for low-altitude security and protection. An optical compensation parameter set for unmanned aerial vehicle imaging optimization is generated in real time through a joint optimization model of an ambient light intensity change rate and a background motion vector field, and an original optical sequence in a target capture window is processed by using the parameter set. And constructing a time domain deconvolution kernel in combination with the motion characteristics of the unmanned aerial vehicle to generate an enhanced optical image resistant to motion blur. Non-linear weighted fusion is carried out through a disturbance intensity evaluation function, and a confrontation disturbance feature mask is formed. The enhanced optical image and the confrontation disturbance feature mask are subjected to airspace superposition operation, a multi-scale residual network is adopted to carry out target confidence estimation on the superposed image, an unmanned aerial vehicle recognition result is generated, and the unmanned aerial vehicle recognition accuracy and the anti-interference capacity in the low-altitude complex environment are remarkably improved through the technical scheme provided by the invention.

The invention discloses a metal surface quality detection method and system, and relates to the technical field of metal surface quality detection. The method is used for solving the problems of low microdefect detection precision, weak technological parameter relevance and closed-loop control deficiency of the high-reflection surface. The metal surface is irradiated through multi-angle coherent light field serialization, the phase offset of interference fringes is analyzed to generate three-dimensional shape data, and reflection noise interference is restrained. Defect depth gradient is extracted based on dynamic segmentation of process parameter constraint, deposition temperature and pressure deviation are quantified through deconvolution calculation, and process deviation feature distribution is constructed. Finite element simulation is utilized to generate a process-morphology mapping atlas library, cross-domain invariance features are extracted through depth constraint manifold alignment and comparative learning, and a causal correlation model of defect types and process parameters is established. And dynamically adjusting process parameters according to the weight gradient, and reflowing data to update the manifold rule. And high-precision three-dimensional defect detection, process deviation traceability and adaptive parameter optimization are realized.

The invention provides a river water quality parameter supervision method and system based on deep learning. The method comprises the steps of self-calibration multi-source data acquisition, diffusive water quality prediction, extreme water quality parameter simulation, reverse diffusion pollution positioning and water quality parameter intelligent supervision. The invention relates to the technical field of river water quality supervision, in particular to a river water quality parameter supervision method and system based on deep learning. By introducing a graph convolutional neural network and a physical diffusion constraint model, space-time diffusion trend modeling of pollutants in a river channel is realized; constructing an extreme pollution event simulation and attribution mechanism by combining a generative adversarial network and physical verification; further adopting a multi-modal Bayesian inversion model and a graph deconvolution structure to realize accurate source tracing of the pollution source; the system can dynamically sense hydrological changes, construct an adaptive threshold judgment mechanism, realize prediction, tracking and response to pollution risks, and provide efficient and intelligent technical support for river ecological safety management.

The invention relates to the technical field of fabric detection, and discloses a fabric defect intelligent detection method and system based on AI visual identification. According to the method, motion blur is quantized through motion state data, optical blur caused by fabric motion is eliminated through deconvolution solution, so that motion interference in the fabric transmission process is processed in a targeted mode, self-adaptive balance of the deblurring capacity and the feature retention capacity is achieved, and then based on the optical interference principle, the deblurring capacity and the feature retention capacity are improved. Through a dynamic calibration system combining hardware-level real-time compensation and multi-dimensional optical parameter calibration, dynamic optical parameter calibration of primary correction data is realized, then fabric defect characterization data is extracted to accurately obtain defect features, and finally, a detection-production line control closed loop is constructed through a quality quantitative index and a comprehensive risk value, so that fabric defect detection is realized. The fabric defect detection precision can be improved, so that the problem of high defect missing detection and false detection rate caused by optical data distortion due to movement and environment interference in a traditional method is effectively solved.

The invention discloses an image restoration method, an image restoration device and a storage medium, which are used for improving the structure restoration precision and the detail restoration capability of image restoration. The method comprises the following steps: acquiring multi-dimensional inertial data in real time; performing frequency domain analysis on the multi-dimensional inertial data by adopting sliding window short-time Fourier transform to obtain vibration intensity; if the vibration intensity does not exceed the preset threshold value, acquiring an image; calculating a definition index of the image; determining whether the image is a blurred image according to a preset definition standard and the definition index of the image; if the image is judged to be a blurred image, dividing the blurred image into a motion blurred image and a focusing blurred image; respectively constructing point spread function models of the motion blurred image and the focusing blurred image; performing deconvolution processing or depth reconstruction on the blurred image through a point spread function model to obtain a clear image; and recalculating the definition index of the clear image, and if the definition index does not exceed the definition threshold, triggering reacquisition or switching the repair model to execute secondary repair.

The invention discloses a deconvolution-based full-focusing super-resolution imaging method, and relates to the technical field of full-focusing super-resolution imaging, and the method specifically comprises the steps: 1, carrying out the data collection based on a full-matrix collection mode, and carrying out the initial image reconstruction through a full-focusing method, and 2, carrying out the reconstruction of an initial image through the analysis of a point spread function of an imaging system, the method comprises the following steps: step 1, establishing a physical convolution model to describe a fuzzy effect in an imaging process, and step 2, by taking minimization of an imaging error as a target and introducing sparse constraint, promoting generation of non-zero reaction only at a position where a defect actually exists in a reconstructed image and inhibiting background noise and artifacts. A point spread function modeling imaging process is introduced, sparse deconvolution solution is carried out on defect distribution by utilizing a fast iteration threshold method, and a super-resolution imaging scheme which has a physical basis and is high in robustness and calculation efficiency is provided for guided wave ultrasonic imaging.

The invention discloses a tunnel back break detection method, system and device and a medium, and particularly relates to the technical field of tunnel detection, and is characterized in that a point cloud feature vector and an image feature vector are fused by using a space-channel attention mechanism, and deconvolution processing is performed on a multi-dimensional fusion feature vector and tunnel point cloud data; concave points are detected according to the two-dimensional edge point set of the tunnel point cloud contour model, and a local concave envelope contour is constructed based on a detection result; fusing the pre-generated convex hull contour and the local concave hull contour to obtain a closed actual measurement section line; performing concave region division on the tunnel point cloud contour model based on the constraint parameter set to obtain a constraint effective feature point set; constructing a constraint actual measurement contour line by utilizing the geometric parameters of the design section and the constraint effective feature point set; fusing the closed actually-measured section line and the constrained actually-measured contour line to obtain a final actually-measured section line; and carrying out back break detection on the tunnel by utilizing the tunnel design section and the final actually measured section line to obtain a back break detection result.

The invention discloses a long video role scene decoupling generation method and system based on multi-modal diffusion, and relates to the technical field of image processing, and the method comprises the steps: S1, synthesizing the advanced features of a role and a scene through a SigLIP encoder and a DINOv2 encoder; s2, performing cross-modal feature fusion on the advanced features to obtain joint features, and compressing the joint features to obtain compact vectors; s3, generating text features according to the text prompt; s4, potential codes are generated from an input video through a causal 3D convolution encoder, the potential codes pass through a linear projection matrix and then are spliced with a memory state for dimension reduction, and a segmented potential vector sequence is obtained; s5, performing decoupling perception generation on the segmented potential vector sequence through an improved 3D-UNet, and performing deconvolution up-sampling reconstruction after deterministic sampling to obtain an RGB video segmented sequence; according to the method, the key problems of rough dynamic control, limited generation length and over-high resource consumption in long video generation are solved, and the quality and efficiency of the generated video are remarkably improved.

The invention relates to the technical field of voice processing, in particular to a mixed voice processing method, electronic equipment and a computer readable medium. The method comprises the following steps: collecting mixed voice and environment influence parameters; carrying out bionic frequency domain analysis on the mixed voice to obtain low-frequency attenuation compensation characteristic data; performing multipath effect propagation analysis on the low-frequency attenuation compensation characteristic data through the environmental influence parameters to generate channel distortion data; performing time domain-frequency domain joint deconvolution processing on the mixed voice by using the channel distortion data to generate a direct sound component and a reflected sound component; performing adversarial training based on the direct sound component and the reflected sound component to generate anti-multipath speech enhancement data; and constructing a dynamic frequency compensation filter based on preset environmental acoustic characteristics. Through the multi-stage signal processing, frequency compensation and real-time optimization technology, the output quality of the mixed voice is improved.

The invention relates to the technical field of image fusion, and discloses a sparse aperture optical system polarization image fusion method based on deep learning, and the method comprises the steps: obtaining a linear polarization degree image, a polarization angle image, and a polarization intensity image through a sparse aperture optical system, and constructing a polarization image fusion model comprising an encoder, a multi-mode fusion module, and a decoder; the encoder comprises two branches for respectively extracting three polarization image features, the multi-modal fusion module comprises an edge gradient compensation module for extracting multi-stage edge features, a polarization attention mechanism for adaptively weighting fusion features and a residual aggregation module for reserving original features, and the decoder uses multi-core deconvolution to decode aggregation features; and constructing a loss function and training a model in combination with the characteristics of the polarization images, and inputting the three polarization images to be fused into the trained model to obtain a polarization fusion image. According to the invention, effective fusion of polarization imaging and sparse aperture imaging can be realized, noise can be effectively suppressed, and the contrast and resolution of imaging can be improved.

The invention discloses a hyperspectral image and LiDAR data collaborative classification method based on double-domain mask and multi-scale local reconstruction, and belongs to the field of remote sensing image classification. According to the method, the problems of scarcity of annotation data and insufficient multi-source feature fusion precision in cross-modal classification of a traditional method are solved. According to the invention, feature learning is carried out through mask random image blocks and channels; a hierarchical multi-scale reconstruction architecture is designed, a lower-layer encoder learns fine-grained features, an upper-layer encoder recovers macroscopic semantic information, and multi-level feature space alignment is realized in combination with deconvolution up-sampling and adaptive pooling. According to the method, a multi-modal feature interaction mechanism and a cross-modal attention module are constructed by fusing the local feature extraction advantages of a convolutional neural network (CNN) and the global modeling capability of Transform, and the complementarity of heterogeneous data is enhanced. According to the method, through multi-level feature dynamic fusion and adaptive weight distribution, the collaborative classification precision of the hyperspectral image and the LiDAR data is improved. The method can be applied to remote sensing image classification.

The invention discloses a method for constructing a remote sensing image defogging network based on wavelet frequency domain heterogeneous enhancement, the network adopts a U-shaped architecture as a basic framework, the network input is a foggy image, and firstly, shallow layer features are extracted through a convolution block; then, a symmetric codec structure is adopted to learn layered representation, a codec comprises up and down sampling and a wavelet frequency domain heterogeneous enhancement module, and the resolution of up and down sampling is controlled through step convolution and deconvolution; the wavelet frequency domain heterogeneous enhancement module separates the high and low frequency features of the image through discrete wavelet transform, and performs heterogeneous enhancement on the separated high and low frequency features by combining the dynamic receptive field advantage of deformable convolution and the global perception capability of Fourier transform; therefore, the recovery of high-frequency local texture details and the removal of low-frequency global haze are effectively promoted. And finally, reconstructing a clear fogless image through the convolution block. According to the research algorithm, the texture features and natural colors of the scene can be precisely reduced.

The invention discloses an out-of-focus identification and restoration method for a microscopic system. The method comprises the following steps: 1, preparing a microscopic out-of-focus image data set and carrying out degradation modeling; 2, constructing an out-of-focus parameter prediction network based on multi-label parameter reasoning and microscopic system priori knowledge and an image restoration network based on a generative adversarial network, performing independent training, and then performing merging training through a circulation system formed by mutual connection based on predicted point spread function convolution and deconvolution operation; 3, model fine tuning based on a specific system and a new data set; and 4, performing model testing, performing large-view image sliding window detection and collage fusion, and explicitly outputting defocus parameters of the system during image shooting. The method can be applied to out-of-focus fuzzy recognition and restoration of microscopic imaging of various systems, and is beneficial to the accuracy of functions such as particle counting and particle size statistics of the systems, thereby further promoting the application of the full-depth-of-view microscopic system in biomedical detection.

The invention discloses an aberration correction and image quality enhancement method for a laminated structure image, and the method comprises the steps: carrying out the deconvolution preprocessing of a to-be-detected marked image according to an aberration priori set, and obtaining an aberration-free image; meanwhile, combining label data to obtain a data set; feature extraction is carried out based on shallow convolution according to the data set, and global feature information is generated through activation function operation; enhancing the feature data by adopting a frequency domain feature and spatial domain feature fusion strategy; the enhanced feature map realizes initial aberration restoration through an aberration correction module; the corrected feature map is processed by a double-channel attention mechanism, and the global context modeling capability of the self-attention mechanism and the spatial perception characteristic of the position attention unit are fused in parallel; and the image resolution is improved through a super-resolution reconstruction module comprising a sub-pixel convolution layer. By adopting the technical scheme of the invention, the accuracy of overlay error detection is improved.

The invention belongs to the technical field of urban traffic flow prediction, and particularly relates to a traffic flow prediction method and device based on dynamic contrast learning and multi-scale 3D convolution, and the method comprises the steps: obtaining a traffic feature matrix and an environment feature vector based on traffic track data and environment data, mapping the environment feature vector to the spatial dimension of the traffic grid to obtain an environment feature matrix, and splicing the traffic feature matrix and the environment feature matrix to obtain a target space-time matrix; constructing a traffic flow prediction model comprising a multi-scale 3D convolution module, a dynamic contrast learning module, an environmental feature gating fusion module and a time-space deconvolution prediction module, taking the target space-time matrix as input, and constructing a joint loss function based on contrast learning loss, gating fusion loss and prediction loss; therefore, the traffic flow prediction model is optimized. According to the method, the problems of a traditional prediction method in the aspects of capturing nonlinear space-time dependence, dynamic emergency response, environment factor collaborative modeling and the like are solved.

The invention discloses an ocean in-situ radioactivity monitoring method, which is characterized in that feature extraction is carried out based on a maximum likelihood expected value maximization (MLEM) algorithm, and spectrum unfolding operation is carried out by adopting a multi-layer perceptron. The method comprises the following steps: firstly, measuring FWHM related parameters of a CeBr3 detector by adopting a standard source, simulating the response of the detector to gamma rays in a marine environment based on a Monte Carlo method, and establishing a full-spectrum response matrix H in a range of 0-2048keV; based on the response matrix H, reconstructing a gamma energy spectrum y obtained by the detector by using an MLEM iterative algorithm; and the reconstructed spectrum is used as the input of a multi-layer perceptron (MLP), and qualitative and quantitative analysis of radionuclides in the energy spectrum is realized. According to the method, the discrete characteristic peak signal is convolved into the single path address corresponding to the characteristic energy, so that the accuracy of the MLP model in the spectrum unfolding process is effectively improved.

The invention discloses an image denoising and self-adaptive enhancement method based on pixel discretization and illumination guidance, and relates to the technical field of image processing, and the method comprises the steps: carrying out the down-sampling and noise reduction of an input image, and obtaining a preliminarily denoised image through the deconvolution operation, fuzzy pixel discretization and residual serialization; estimating an illumination component from the preliminarily denoised image by using an illumination estimation network; calculating a reflection image based on the estimated illumination component and the preliminarily denoised image according to an image enhancement theory; based on a dynamic splicing mechanism, splicing the reflection image and the illumination component in the channel dimension to obtain a spliced image; and with keeping of image details and brightness information as a criterion, denoising and enhancing are carried out on the spliced image through a reflection denoising network, and a denoised and enhanced image is output. According to the invention, high-quality recovery of low-illumination images is realized, and the intelligent level of an electric power inspection system is improved.

The invention provides a medical image segmentation method and system, computer equipment and a storage medium, and belongs to the field of image processing, and the method comprises the steps: extracting preliminary features of a medical image through depth separable convolution, and splicing the preliminary features with original image residuals to obtain a preliminary feature map; after an encoder performs average pooling dimension reduction, local details and global contour features of a dimension reduction feature map are extracted by using left and right branches of a lightweight convolution module LDB, then a downsampling feature map is obtained through channel attention CA weighted fusion, and attention is calculated in combination with a self-attention mechanism module EMHA to obtain a depth feature map and a bottleneck feature map; the decoder weights the depth feature map by means of a channel and space attention to obtain a CBAM enhanced feature map, upsamples the bottleneck feature map and then splices the bottleneck feature map with the CBAM enhanced feature map, features are extracted through an LDB module, and finally a pixel-level segmentation result is output through upsampling and deconvolution, so that image segmentation achieves the effects of high quality, low complexity and low operand.

The invention discloses a small target detection network based on sparse feature enhancement fusion. The method comprises the following steps: extracting multi-level features of different scales through a backbone network; the sparse feature enhancement module is used for performing deconvolution operation on shallow features to restore image details to the greatest extent, further improving the detection precision of the small target through feature integration from global to local, and enhancing the distinguishing capability of the small target in a complex background; the gradual gradient enhanced detection head is adopted, and the information transmission mode in the training process is dynamically adjusted, so that the network performance is effectively optimized through programmable information, and the detection effect is improved. The invention provides a small target detection network based on sparse feature enhancement fusion. A sparse feature enhancement module, a hierarchical feature fusion module and a progressive gradient enhancement detection head are provided, and a programmable gradient information mechanism is introduced to improve and optimize the detection head.

The invention provides a spatial domain identification method based on data interpolation and cell type deconvolution, and belongs to the technical field of bioinformatics. In order to solve the problems that gap information between adjacent points cannot be utilized in low-resolution spatial transcriptome data and prior information of cell types in a tissue space structure level cannot be fully integrated in a traditional method, the method comprises the following steps: acquiring a spatial transcriptome data set and a single-cell RNA sequencing data set, and performing data preprocessing on the acquired data sets; and carrying out data interpolation on the preprocessed spatial transcriptome data, and carrying out cell type deconvolution in combination with single-cell RNA sequencing data. And constructing a deep learning model based on the graph convolutional network. And training a deep learning model according to gene expression information, spatial position information and cell type information of the spatial transcriptome data after cell type deconvolution by using a self-supervised contrast learning strategy. And performing spatial domain identification on the to-be-detected data based on the trained model.

The invention discloses a motion blur removing method based on dynamic image interpolation, and the method comprises the steps: firstly correcting a blur track according to the dynamic change of time and direction; then, considering the stage speed change of the target, and carrying out weight updating on the initialized fuzzy kernel; then, under the condition that the target speed cannot be estimated, feature points are detected by using an SIFT algorithm, and the feature points are tracked by using an L-K optical flow method, so that a more accurate dynamic fuzzy kernel is constructed, and then deconvolution operation is performed by combining a neighbor interpolation technology, so that the definition of an original image is recovered; in order to further optimize the image quality, three image enhancement technologies, including histogram equalization, contrast enhancement and image sharpening, are combined to cope with visual interference in different environments, enhance the contrast and detail expressive force of the image, and further improve the image processing effect. Therefore, the unmanned aerial vehicle identification system can capture and identify the hostile target more accurately.

The invention discloses a fluorescence lifetime imaging phase analysis method based on deep learning, a terminal and a readable storage medium, and the method comprises the steps: obtaining counting data of time-dependent single photon counting of each pixel in a measured sample image, and measuring a response function of a used instrument, inputting the convolution attenuation data and the response function into a deconvolution neural network; the deconvolution neural network outputs a deconvolution signal of each pixel according to the response function and the counting data of each pixel; and performing phase transformation according to the deconvolution signal of each pixel, and obtaining the fluorescence lifetime of each pixel. According to the method, the time migration caused by the instrument response is regarded as the convolution caused by the instrument response function, and then the neural network is adopted to carry out deconvolution on the counting data, so that the time migration caused by the response function is calibrated and corrected, a traditional phase calibration process is not needed, and accurate analysis of the service life is realized.

The invention relates to a power load prediction system based on time sequence self-supervised representation learning, and the system comprises a data preprocessing module, an expansion time convolution network, a trend representation decoupler, a seasonal representation decoupler, a feature converter, and a predictor. The expansion time convolution network performs dynamic adaptive grouping on different variables in the same data processing unit, establishes an intra-group variable relationship by sharing a convolution kernel weight and stacking a plurality of expansion time convolution layers, and establishes an inter-group variable relationship through subsequent multi-group representation splicing operation and single-layer expansion time convolution; the trend representation decoupler is used for separating trend representation Z (T) from feature representation extracted from the expansion time convolution network based on a plurality of parallel one-dimensional causal convolution blocks with different scales; the seasonal representation decoupler adopts discrete Fourier transform to separate seasonal representation Z (S) from the feature representation extracted from the expansion time convolutional network; the feature converter maps a combined feature vector of trend representation and seasonal representation back to an original space from a potential space by stacking a plurality of deconvolution layers; and the predictor maps the feature representation of the original space into a power load prediction result by using a linear projection layer, and the adaptability, robustness and generalization ability of the model for modeling complex power load time series data can be improved.

The invention relates to the technical field of image processing, and particularly discloses a video monitoring image processing system which comprises a dynamic video acquisition module, a self-adaptive illumination enhancement and denoising module and an image segmentation module. The dynamic video acquisition module acquires video data through a fixed camera and a movable camera, and supports various resolutions and frame rates. The adaptive illumination enhancement and denoising module utilizes a multi-scale Retinex algorithm to realize brightness equalization, and combines bilateral filtering denoising to effectively improve the image quality under the low illumination condition. The image segmentation module is based on an improved SegNet network, and greatly improves segmentation precision and reasoning speed by introducing residual connection, a channel attention mechanism and a self-attention mechanism and adopting an up-sampling strategy combining bilinear interpolation and deconvolution. Experimental results show that the segmentation precision of the system is improved by 6.2%, and the reasoning speed is improved by 32%. The system is suitable for image processing and target segmentation in low illumination and complex scenes.

The invention provides a spectral data optimization method and device, computer equipment and a storage medium, and relates to the technical field of semiconductor detection.The method comprises the steps that original spectral data output by a spectrograph is obtained, and the central wavelength value of the original spectral data of each frame is calculated; querying from a preset database according to the central wavelength value to obtain a corresponding instrument linear function; and processing each frame of original spectral data and the corresponding instrument linear function by adopting a deconvolution algorithm to obtain optimized real spectral data. According to the scheme, algorithm compensation is carried out on each frame of spectral data, so that system distortion introduced by the spectrometer can be accurately compensated in a full spectrum range, the extraction precision of spectral features is greatly improved, and the magnitude order improvement of measurement precision is realized; in addition, the optimized spectral data more truly reflects the physical characteristics of the tested sample, the dependence of an optical model on empirical parameters is reduced, and the generalization ability and prediction reliability of the model are improved.

The invention provides a deep learning-based oral cavity image recognition method, a storage medium and a deep learning-based oral cavity image recognition system. The method comprises the steps of deploying a federated learning framework and collecting a multi-modal oral cavity image data set; extracting local features to obtain image features, and generating a modal adaptive weight map; a multi-head self-attention mechanism is used for fusing the cross-modal features to generate a fused feature map, and deconvolution up-sampling is carried out to form high-resolution multi-modal feature representation. A tooth segmentation mask is generated based on this representation, and an initial diagnostic report is generated. And aggregating the attention weight of each client through an encryption protocol, and generating interpretable decision support data. And finally, generating a structured clinical report by using a natural language. According to the method, the Grad-CAM thermodynamic diagram is combined with the encrypted and aggregated attention weight, so that the privacy security is guaranteed, the model interpretability is enhanced, the clinical credibility and the diagnosis decision efficiency are improved, and the problems of insufficient diagnosis precision of complex lesions and insufficient utilization of multi-modal information in the prior art are solved.

The invention relates to the technical field of biological microscopic imaging, and discloses a denoising and deconvolution super-resolution imaging method and device based on optical switch molecules, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring a plurality of fluorescence image sequences, wherein to-be-imaged structures of images in the fluorescence image sequences are marked by optical switch fluorescence molecules; each sequence is generated by utilizing the photoswitch characteristic of the photoswitch fluorescent molecules, and the sequence comprises at least one bright-state image and at least one dark-state image; performing de-noising processing on each bright-state image and each dark-state image in the plurality of fluorescence image sequences to obtain de-noised images; and carrying out deconvolution processing on the denoised image to obtain a corresponding super-resolution imaging image. The requirement for imaging equipment is low, the method can be used for various imaging technologies, the size of an imaging area is not limited, the requirement for large-view-field fixed biological sample imaging in an application scene can be met, and fixed cell and tissue large-view-field super-resolution imaging is achieved.

The invention discloses a composite fault signal separation method based on adaptive spectral kurtosis deconvolution. The method comprises the following steps: S1, carrying out noise reduction processing on an original composite fault signal; s2, separating the composite fault signal after noise reduction through Hilbert envelope demodulation; s3, optimizing a low-pass filter and a high-pass filter through a Bayesian objective function, and constructing a band-pass filter and a band elimination filter; and S4, carrying out iterative filtering on the composite fault signal after noise reduction through a band-pass filter and a band elimination filter, constructing a frequency-bandwidth fast spectral kurtosis graph according to a fast spectral kurtosis deconvolution method, obtaining a center frequency and bandwidth range corresponding to the maximum spectral kurtosis, and separating out the weak fault signal in the step S2. According to the method, a spectral kurtosis graph is constructed by combining a low-pass filter and a high-pass filter through a fast spectral kurtosis deconvolution method to determine the center frequency and the bandwidth range, targeted separation is carried out on composite fault signals, an initial filtering parameter range is provided for Bayesian optimization, and the fault separation efficiency and accuracy are improved.

The invention discloses a passive synthetic aperture method and device based on R-L deconvolution beam domain processing, a medium and equipment, and the method comprises the steps: constructing a formation distortion model of a towed array under disturbance, and processing a discretized formation distortion model through a Kalman filter, so as to obtain the absolute coordinates of each array element; determining a receiving signal time domain solution of each array element based on the sound pressure expression and the absolute coordinates; based on an ETAM algorithm, obtaining beam output of overlapped sub-arrays in adjacent time samples, and performing R-L iteration on the beam output to obtain a phase compensation factor; and performing phase compensation based on the wave beam output of the phase compensation factor to obtain each phase correction wave beam and the wave beam output of the synthetic aperture. The passive synthetic aperture algorithm and the R-L deconvolution wave beam forming technology are effectively combined, so that the accuracy of the phase compensation factor after formation distortion is generated is improved to the maximum extent, and the accuracy of the phase correction wave beam is improved. Therefore, the algorithm provided by the invention still provides high-precision target positioning under the condition that the ETAM algorithm fails.

The invention discloses a multi-scale adaptive lesion detection method based on mammary gland ultrasound. The method comprises the following steps: pre-processing an image; the input module is used for extracting features through convolution blocks to clearly display boundaries, and then deconvolution blocks are used for refining boundary features to complete image expression; a trunk module; in the neck network, the CA generates feature maps in two directions, the feature maps in the two directions are spliced, features are extracted through convolution operation, and attention weights in the two directions are further generated; and the detection head network performs positioning prediction, classification prediction and loss calculation, and finally outputs a result. According to the method, the precision and richness of feature extraction are improved, and the clear recognition capability of the model on different tissue boundaries is enhanced; the capacity of capturing multi-scale lesion features is improved, and the lesion features are better captured; important areas such as small calcification points, changes of cyst walls and boundaries of fibroadenoma can be highlighted, and the accuracy and specificity of detection are improved.