69 results about "Mixed noise" patented technology

The invention provides a three-dimensional seismic data mixed noise suppression method based on an MSAT-Unet. The method comprises the following specific steps: constructing an MSAT-Unet network comprising a multi-scale expansion convolution residual module, a channel-space attention mechanism and a Transform convolution module; the encoder is improved into multi-scale expansion residual convolution, so that the receptive field is expanded, and the capability of capturing local details and global semantic information is enhanced; a channel-space attention mechanism is integrated behind the decoder, a direction sensitive context is extracted through multi-dimensional adaptive pooling, and details and edge recovery are enhanced; meanwhile, an improved decoder is a Transform convolution module, and the feature reconstruction and complex structure recovery capability is improved; the optimized model carries out training and reasoning on the three-dimensional seismic data, mixed noise can be effectively suppressed, a clear data basis is provided for subsequent interpretation, and the method is high in generalization and robustness and good in performance in the aspect of three-dimensional seismic data denoising.

The invention discloses a joint blind denoising method and system based on self-heuristic learning and Bayesian reasoning, belongs to the field of computational imaging, and solves the problems that in the prior art, the mixed noise modeling capability is insufficient, the performance is degraded under the condition of low signal-to-noise ratio, the combination of uncertainty quantization and regularization is lacked, and the generalization capability is limited due to data dependence. Comprising the following steps: collecting an original image and preprocessing; generating a noise data pair; an enhanced residual attention U-Net model is constructed; a noise estimation sub-network is adopted to extract noise features, the noise features are fused with original image features, and the model is trained; adopting the trained model to carry out multiple times of forward propagation on the same input image to obtain multiple groups of denoising results; calculating a mean value and a standard deviation to obtain a de-noising prediction and uncertainty heat map; and training the trained model again based on the uncertainty heat map, optimizing network parameters, and obtaining a final denoising prediction result and uncertainty estimation thereof. The method is suitable for complex noise distribution processing scenes.

The invention discloses a near-field electromagnetic wave imaging multi-mode noise cooperative suppression method, relates to the technical field of electromagnetic wave imaging, and aims to solve the problems that the suppression effect on speckle, stripe and Gaussian mixture noise is poor and a target structure is easy to lose in the prior art. According to the method, a complex field speckle suppression-multidirectional fringe separation-cross-channel Gaussian suppression three-stage cooperation scheme is adopted, and firstly, the amplitude and phase of a complex field image are processed through an adaptive threshold value to remove speckle noise; respectively executing horizontal / vertical ADOM filtering on the real part and the imaginary part of the complex field to eliminate stripe noise; and finally, Gaussian noise is suppressed by combining BM3D filtering and three-dimensional transform domain optimization, and robustness is improved through multi-frequency point fusion. Experiments show that the SSIM of the method is improved by 21.7% compared with that of a traditional method, the GSSIM and the PSNR are optimal, the structural features of the target can be reserved in a strong noise environment, and the method is suitable for scenes such as defect detection of near-field electromagnetic wave synthetic aperture imaging.

The invention discloses a wavelength modulation spectral signal denoising method based on an unsupervised auto-encoder. The method comprises the following steps: step 1, constructing a WMS harmonic signal data set; step 2, training an HA-CAE denoising network model; 3, the HA-CAE denoising network model is evaluated and optimized, and an optimal HA-CAE denoising network model is obtained; and step 4, integrating the optimal HA-CAE denoising network model to a sensor system to realize real-time denoising processing of the signal. According to the method, a targeted data set and an improved HA-CAE denoising network model are constructed, three attention mechanisms are fused to accurately capture signal local details, time sequence association and global channel characteristics, non-stationary mixed noise characteristics are adapted, the denoising effect is improved, the generalization ability and the real-time processing ability of the model are guaranteed, and the method is suitable for popularization and application. The method is suitable for wavelength modulation spectrum signal processing in various complex scenes such as industrial leakage monitoring and atmospheric environment detection.

The application discloses a near-field electromagnetic wave imaging multi-modal noise cooperative suppression method, relates to the technical field of electromagnetic wave imaging, and aims to solve the problems of poor suppression effect of spot noise, stripe noise and Gaussian mixed noise and easy loss of target structure in the prior art. The method adopts a three-stage cooperative scheme of "complex domain spot suppression-multi-direction stripe separation-cross-channel Gaussian suppression", first removes the spot noise by adaptively processing the amplitude and phase of the complex domain image through a threshold value; then eliminates the stripe noise by performing horizontal / vertical ADOM filtering on the real part and the imaginary part of the complex domain, respectively; finally, combines BM3D filtering and three-dimensional transform domain optimization to suppress Gaussian noise, and improves the robustness through multi-frequency point fusion. Experiments show that the SSIM of the method is improved by 21.7% compared with the traditional method, the GSSIM and PSNR are optimal, the target structure features can be reserved in a strong noise environment, and the method is suitable for defect detection and other scenes of near-field electromagnetic wave synthetic aperture imaging.

This application belongs to the field of coherent diffraction imaging technology, specifically disclosing an ultra-low dose coherent diffraction imaging method and apparatus. Based on a blind source separation strategy, this application uses principal component analysis to process the acquired diffraction signals, performing noise separation and updating in reciprocal space. This effectively separates mixed noise energy into noise components, thus avoiding crosstalk to the reconstruction process and significantly improving the convergence stability and robustness of coherent diffraction imaging when reconstructing diffraction signals with extremely low signal-to-noise ratios under ultra-low exposure doses. Simultaneously, this application constructs noise components separately for each scanning position and correlates noise components at different positions through low-dimensional spatial projection, achieving non-stationary noise separation. This enables more effective handling of random noise caused by the low quantum efficiency of ultra-short band detectors, thus maintaining extremely high noise robustness and reconstruction accuracy even under ultra-low exposure doses, achieving an effective improvement in resolution.

This invention discloses a real-time rendering method and system for physically simulated volumetric clouds. The method includes the following steps: constructing a dynamic noise mixing model, and constructing complex cloud layers by mixing multiple noises based on the dynamic noise mixing model; constructing a multi-scattering illumination model, and using the multi-scattering illumination model to perform single-scattering calculations and multi-scattering approximations on the complex cloud layers to simulate illumination; based on the completed illumination simulation, dividing the cloud density field of the complex cloud layers into several voxel blocks, and performing adaptive light travel processing to complete the real-time rendering of volumetric clouds. This invention, without sacrificing image quality and rendering efficiency, integrates advanced cloud modeling theory with efficient GPU acceleration technology, and uses a mixed noise model to generate diverse cloud structures, accurately reproducing various typical cloud types such as high-altitude cirrus clouds, cumulus clouds, and stratus clouds.

The application provides an active noise reduction method and system for road noise and wind noise of an automobile, and the active noise reduction method comprises the following steps: S1, collecting a dynamic parameter signal and a total mixed noise signal during automobile driving; S2, constructing a dynamic calculation model of a Strouhal number according to a Reynolds number, deducing a real-time calculation model of a characteristic frequency of each noise source i according to the dynamic calculation model of the Strouhal number, and outputting a real-time characteristic frequency sequence corresponding to each noise source i; S3, constructing a global signal dictionary D, using an orthogonal matching pursuit sparse representation algorithm, and separating a single noise source signal in the total mixed noise signal by iteratively screening a dictionary atom with the highest matching degree with the characteristic frequency; S4, using a frequency tracking type fast adaptive filtering algorithm to generate a corresponding anti-phase cancellation signal; and S5, outputting the anti-phase cancellation signal in the automobile cabin, so that noise reduction can be performed on each noise source i.

The invention discloses an aircraft fuel quantity measurement self-adaptive filtering method, system and device based on multi-dimensional statistic dynamic adaptation and a medium. The method comprises the following steps that S01, signals are collected and preprocessed; dividing a fuel quantity digital signal output by a sensor into continuous time blocks according to a fixed time window; s02, calculating a multi-dimensional statistical magnitude; comprising the mean value, variance and kurtosis of the current time block; s03, performing multi-stage filtering processing; s04, performing data fusion and conversion; and keeping the original value of the data pulse point, and combining the original value with the fine filtering result of the non-pulse point to obtain a complete filtered signal. By monitoring the multi-dimensional statistical characteristics (mean value, variance and peak value) of a fuel quantity signal in real time and combining a coarse filtering-pulse detection-fine filtering multi-stage structure, filtering parameters are dynamically adjusted, efficient suppression of mixed noise (low-frequency shaking, Gaussian noise and pulse noise) is achieved, and the precision of fuel measurement under complex working conditions is improved.

The invention relates to the technical field of railway vehicle noise analysis and processing, and particularly discloses a railway vehicle noise distinguishing and extracting device and method.The device comprises an acquisition module, a processing module and an output module, and the acquisition module is used for acquiring carriage mixed noise in the running process of a railway vehicle; the processing module is used for performing frequency decomposition, noise classification and independent loudness analysis of various types of classified noise on the collected mixed noise of the carriage; the output module is used for outputting and / or storing the loudness values and the frequency characteristics of various types of noise after classification; according to the method, wheel track noise, passenger voice and train station reporting voice can be accurately separated and subjected to loudness analysis in real time, the problem that the passenger voice and the train station reporting voice are difficult to separate and distinguish is solved, and a scientific noise monitoring and management means is provided for rail transit operation enterprises; the passenger comfort level is improved, the station reporting volume and the operation service quality are optimized, and complaint and transformation cost caused by noise is reduced.

The invention relates to the technical field of image style migration, and provides a diffusion style migration method and system based on latent distribution re-anchoring dynamic injection, and the method comprises the steps: carrying out the second-order statistical alignment of a content hidden variable and a style hidden variable through a potential covariance re-coloring method; automatically positioning an injection starting point of the style according to the edge intensity of the content image, and injecting hidden distribution reanchoring noise into the initialized mixed noise hidden variable to realize hidden distribution reanchoring; the content self-attention features and the style self-attention features are fused through a soft + dynamic style injection module, in the reverse sampling process of the calibrated hidden variables, the fused attention features are gradually injected into an attention layer in the diffusion process at the injection starting point, diffusion denoising is carried out, and a style migration image is generated. According to the method, the image with rich structure details is protected, and the excellent content structure maintaining capability and the accurate style reappearance are realized at the same time.

The invention relates to the technical field of computer vision and video image processing, and discloses a passive night vision full-color video image enhancement method and system based on AI learning, and the method comprises the steps: constructing a Poisson to Gaussian mixed noise model through sensor metadata, generating a signal-to-noise ratio confidence map, and quantifying the physical reliability of pixels; constructing an optical flow energy functional through a signal-to-noise ratio weighted data item and a semantic boundary constraint regular item in combination with a semantic label graph, and calculating a high-robustness optical flow field; in a time-space fusion stage, taking the signal-to-noise ratio confidence map as a gating condition, and dynamically adjusting a time domain recursion fusion proportion of historical features and current features; and finally, driving an adaptive normalization unit by using the semantic tag graph, and retrieving a semantic chrominance priori library to restore the inherent physical color of the object. According to the method, physical noise and scene semantic priori are utilized, the problems of random noise interference, motion ghosting and color loss under low illumination can be solved, and clear passive full-color night vision imaging with real color is achieved.

The invention discloses a self-supervised laser speckle denoising method based on a transform domain, and relates to the field of laser measurement. Laser speckle noise is modeled into a Poisson-Gaussian mixture model, statistical characteristics of the laser speckle noise are represented, and efficient training of a denoising model is realized on the premise that a clean image does not need to be labeled by adopting a parameter estimation network fusing a residual attention strategy with a U-Net architecture and a self-supervised denoising network containing an asymmetric convolution Inception structure. A mixed noise domain is converted into a Gaussian noise domain through generalized Ansu transform, multi-loss function optimization is combined, the algorithm denoising performance is improved, the peak signal-to-noise ratio of a light spot image is remarkably improved, and then the precision of a laser measurement system is improved. The method is simple and efficient, adapts to different laser measurement devices, can effectively improve the measurement precision in the fields of laser radar, industrial detection, medical imaging and the like, and provides core technical support for the industries of intelligent manufacturing, automatic driving and the like.

The invention discloses a self-heuristic learning blind denoising method and system based on consistency difference guidance, belongs to the technical field of computer imaging, and solves the problems that the blind denoising performance of an SN2N method is insufficient under the condition of mixed noise distribution, the generalization ability of noise distribution is limited, and denoising of data with an extremely low signal-to-noise ratio is insufficient. The method comprises the steps that original picture data are collected and preprocessed, and an image data set is generated; generating a noise data pair by adopting an SN2N self-supervision method; performing consistency difference evaluation on the noise data pairs, and calculating a pixel-level difference chart; establishing a U-Net network based on a residual attention mechanism, wherein the network comprises an up-sampling module, a down-sampling module and a bottleneck layer; training the network by using a pixel-level difference graph, and constraining a loss function of the noise estimation branch according to the pixel-level difference graph; and adopting the trained network to predict and obtain a de-noising result. The method is suitable for living cell super-resolution microscope imaging and three-dimensional volume data denoising scenes.

Various examples are provided related to reconstructing images such as, e.g., medical images from low-dose image scans. Adversarial learning such as, e.g., a Cyclic Simulation and Denoising (CSD) framework can be used to address challenges of complicated mixed noise in real low-dose scans. The CSD framework can include a simulator model that can extract low-dose noise and features (e.g., tissue features) from separate image spaces into a unified feature space and a denoiser model that can learn how to remove noise and restore features, simultaneously. Both the simulator model and the denoiser model can regularize each other in a cyclic manner to optimize network learning effectively. The CSD framework in combination with phantom scans can embrace the realistic low-dose noise and features into a unified learning environment to address the challenge of real low-dose image restoration.

The invention aims to provide a voice training noise adding system and method based on a mixed noise generation model. The system comprises an input module, a noise environment enhancement module, a voice noise enhancement module and an output module. Wherein the input module is used for acquiring noise environment simple description information and clean voice data to be enhanced; the noise environment enhancement module converts the noise environment simple description information into a structured noise event sequence with time sequence characteristics; a voice noise enhancement module generates multi-source mixed noise according to the noise event sequence, and adds the multi-source mixed noise to the clean voice data according to a preset rule to obtain noisy voice data; and the output module is used for outputting the noisy voice data for voice model training. According to the invention, interaction characteristics of superposition, offset, interference and the like of different noise sources in the time dimension are fully joined, and the technical bottleneck that only linear superposition can be realized in a traditional mixing mode is solved.

The invention discloses a 4D wireless spectrum map reconstruction method in a dynamic environment, and the method comprises the steps: carrying out the modeling of missing spectrum data and mixed noise into a four-dimensional tensor model according to the four dimensions of two-dimensional space, frequency and time; in order to reduce the degree of freedom of the tensor model, the model is decoupled into an incomplete spectrum map corresponding to a single radiation source through a non-negative matrix factorization algorithm. Further considering the problem of mixed noise influence and calculation complexity, proposing an optimization function based on a dynamic threshold value and a fast algorithm combining random singular value decomposition and Nesterov accelerated gradient, and solving under a non-negative matrix factorization algorithm framework; and finally, reconstructing the frequency spectrum map corresponding to each radiation source by adopting parallel computing to obtain a complete 4D frequency spectrum map. According to the method, an accurate 4D spectrum map can be obtained based on spatial sparse sampling in a dynamic time-varying environment, and the high-precision spectrum situation sensing requirement in a complex noise environment is met.

This invention discloses a load forecasting method, which includes the following steps: decomposing target load power data with mixed noise components using empirical mode decomposition (EMD) to obtain a detrended subsequence of the target quantity; determining the K-value parameter range of variational mode decomposition (VMD) based on the target quantity; determining the optimal decomposition parameters of VMD by combining each detrended subsequence and the K-value parameter range; and predicting the load value corresponding to the target load power data based on the optimal decomposition parameters of VMD. Applying the load forecasting method provided by this invention achieves effective noise reduction of load power data with mixed noise components, significantly improving the accuracy of load forecasting. This invention also discloses a load forecasting device, equipment, and storage medium, which have corresponding technical effects.

The invention provides an image mixed noise self-adaptive suppression method and device of an ultrasonic endoscope, which are applied to electronic equipment in the ultrasonic endoscope, and the ultrasonic endoscope comprises the electronic equipment and a probe. The method comprises: controlling a probe to collect an ultrasonic image of an object lung area; performing noise feature extraction on each local image in the ultrasonic image to obtain a first feature, a second feature and a third feature of each local image; for each local image, correcting the first feature and the second feature by using a depth value of each pixel point in the local image, a depth gain compensation relationship and a lung ultrasonic signal attenuation relationship; determining a noise classification result of the target local image with noise based on the third feature of each local image and the corrected feature; determining a noise reduction scheme of each target local image by using the noise classification result of each target local image; and performing noise reduction on each target local image by using each noise reduction scheme. The denoising effect of the ultrasonic image can be improved.

This invention discloses an accurate detection system for noise and synchronization signals in A / D converters, specifically relating to the field of signal processing technology. It comprises four main modules: co-source timing allocation, multi-dimensional signal injection, dual-channel synchronous acquisition, and dynamic decoupling analysis. The co-source timing allocation module fans out a high-precision clock into multiple phase-correlated signals. The multi-dimensional signal injection module injects a composite excitation vector into the A / D converter under test. The dual-channel synchronous acquisition module, under clock constraints, synchronously captures the mixed noise data of the main channel and the jittered physical waveform of the reference channel clock. The dynamic decoupling analysis module uses a feedback calibration signal for time-domain registration, extracts intrinsic noise parameters through noise source stripping operations, and outputs timing compensation. This invention solves the problem of dynamic coupling and difficulty in separating synchronization jitter and intrinsic noise under high sampling rates, significantly improving the accuracy and robustness of A / D converter detection.

The invention discloses a long-sequence electrocardiosignal noise reduction method based on frequency domain features and a conditional diffusion model. The method comprises the following steps that S1, a clean electrocardiosignal and an electrocardiosignal noise signal e are obtained through preprocessing; s2, adding the clean electrocardiosignals and the mixed noise signals of different times, and normalizing the maximum value and the minimum value to generate noisy electrocardiosignals of different noise intensities; s3, constructing an electrocardiosignal noise reduction model, inputting the data set into the electrocardiosignal noise reduction model for optimization verification, and outputting a noise-reduced electrocardiosignal subjected to noise reduction processing; and S4, quantitatively evaluating the noise reduction effect by taking the distance sum of squares, the absolute maximum distance, the percentage root-mean-square difference, the cosine similarity and the signal-to-noise ratio difference value as evaluation indexes for the noise-reduced electrocardiosignals. The method has the advantages of being suitable for long-sequence electrocardiosignals, good in noise reduction processing effect, capable of reducing misdiagnosis risks, suitable for complex scenes, beneficial to model generalization and the like.

The invention discloses a photovoltaic module subfissure detection method, system and device and a medium, and relates to the technical field of photovoltaic module defect detection, and the technical scheme is characterized in that a first image set is acquired; training a deep convolutional generative adversarial network composed of a generator and a discriminator according to the first image set and a pre-configured training strategy, dynamically adjusting network parameters of the deep convolutional generative adversarial network in combination with real-time monitored training indexes in the training process until training stops, and outputting the trained generator; inputting the configured mixed noise vector and the category self-adaptive embedding vector into a trained generator to generate a second image set, and merging the first image set and the second image set to obtain a third image set; training a neural network according to the third image set to obtain a subfissure detection model; and performing defect detection on the to-be-detected electroluminescent image based on the hidden crack detection model, and outputting a hidden crack detection result for indicating that the photovoltaic module has hidden cracks or does not have hidden cracks.

The application belongs to the field of laminated diffraction imaging, and specifically discloses a high dynamic range image fusion laminated diffraction imaging method and system, the imaging method being: initializing an illumination probe and a sample to be measured, constructing a joint distribution containing a diffraction light field, a dark field and an exposure time sequence, solving a maximum likelihood estimation of absolute irradiance based on a joint probability density function, fusing a diffraction field high dynamic range absolute irradiance from directly measured diffraction light field, dark field and exposure time parameters, replacing an analog diffraction light field amplitude and keeping its phase unchanged, and simultaneously reconstructing the illumination probe and the sample to be measured according to the analog diffraction light field before and after updating. The application considers the influence of mixed noise, effectively extracts high-frequency overlapping correlation signals without losing or reducing the signal-to-noise ratio of low-frequency region diffraction signals, significantly reduces the requirement for the dynamic range of a detector, and avoids the increase of imaging system uncertainty without additional complex parameter calibration and modulation equipment in the imaging process.

The application provides a measurement preprocessing method based on correlation entropy and attention mechanism. The dynamic weight is calculated through the Gaussian kernel correlation entropy of the historical innovation and the current innovation in the sliding window, and the abnormal measurement is adaptively identified and suppressed. The method adopts a feedforward inhibition mechanism to block noise propagation, while retaining the optimality of the KF framework, and realizes efficient anomaly detection with lower complexity. Simulation experiments show that in the Gaussian mixed noise and impulse noise scene, the application significantly reduces the mean square error of KF, the dynamic window mechanism is faster than the noise statistical modeling algorithm in response speed, the tracking accuracy is equivalent to Huber-KF, and the parameter adaptability is stronger, which verifies the robustness and tracking performance of the method in non-Gaussian noise, especially in the impulse noise environment, and provides theoretical support for subsequent expansion to the maneuvering target scene and deep learning fusion.

The application discloses a PET detector signal simulation generation method based on artificial intelligence, comprising the following steps: generating an initial two-photon signal pair meeting initial energy and time difference physical constraints through a two-photon signal generator fusing a Transformer and a U-Net; decoupling mixed noise into multiple independent physical noise components according to a PET noise physical prior library by using a multi-branch decoupling network based on an attention mechanism, and outputting a pure signal and a noise component spectrum; performing linkage adjustment on the pure signal and the noise component according to target parameters set by a user through a space-time-noise collaborative regulator; and finally, outputting a customized signal data format according to a downstream task type. The application realizes high-fidelity, interpretable and controllable coincidence event level signal simulation, effectively solves the core problems of complex modeling, noise distortion, lack of correlation and poor scene adaptability of traditional methods, and significantly improves the efficiency and accuracy of PET detector research and testing.

The application discloses an underwater image enhancement method based on a channel attention and a deformation generative adversarial network, and comprises the following steps: acquiring an underwater image to construct a data set, and dividing the data set into a training set and a test set; constructing an adaptive channel attention module with a multi-scale receptive field for recalibrating channel weights; constructing a deformation convolution module for feature extraction and facing a convolution kernel offset; fusing the adaptive channel attention module and the deformation convolution module to generate a generative adversarial network; training the generative adversarial network based on the training set data to obtain a trained generative adversarial network; inputting the test set data into the trained generative adversarial network to obtain an enhanced underwater image; and constructing an adaptive channel attention module with different receptive fields by using a single hidden layer neural network and a global average pooling technology, which is helpful to reduce the influence of mixed noise on a feature layer and improve the enhancement consistency of an object of interest under different scene depths.

The application provides a method and system for removing random mixed noise in dynamic electrocardiogram, and relates to the field of digital signal processing of bioelectric signals. A noise signal is used to perform noise adding processing on an original noise-free electrocardiogram signal to obtain a noise-added dynamic electrocardiogram signal, and a training data set is formed. A denoising model is trained until the distortion degree of the noise-free electrocardiogram signal output by the denoising model and the original noise-free electrocardiogram signal is minimum in amplitude and angle. The dynamic electrocardiogram signal to be denoised is input into the trained denoising model to generate and output a noise-free dynamic electrocardiogram signal. The application uses a two-step method of frequency domain filtering and time domain filtering to remove random noise or random mixed noise in the electrocardiogram signal. The frequency domain filtering effectively removes noise outside a specific frequency range, and the effective frequency band range of the dynamic electrocardiogram signal is retained. The time domain filtering uses a deep learning algorithm for denoising, learns the characteristic matrix of the useful signal and the noise, and thus obtains a reconstructed signal with less distortion.

This invention discloses a radar super-resolution imaging method for mixed noise environments, belonging to the field of radar detection and imaging. By analyzing the noise characteristics of the actual environment, a new radar echo model is constructed, and the adaptive noise resistance capability of the modified loss function is utilized to achieve high-resolution target reconstruction in mixed noise environments. First, the traditional echo model is optimized to obtain an echo mathematical model closer to the actual environment. Then, the mathematical properties of the modified loss function are used to achieve good suppression of mixed noise. Finally, a generalized ridge regression estimation algorithm based on the minimization criterion is used to achieve a closed-loop solution of the objective function, thereby completing accurate reconstruction of the target scene in mixed noise environments. Compared with traditional super-resolution methods, the method of this invention can achieve forward-looking super-resolution radar imaging in mixed noise environments.

The invention discloses a mixed noise removal method combining low-rank decomposition and kernel adaptive filtering, and relates to the field of signal processing and image reconstruction. The method comprises the following steps of: separating a potential clean image structure by utilizing transform domain projection and approximate rank constraint; providing a sparse residual suppression strategy, coding abnormal pixels into a sparse matrix, and approaching an optimal solution by using an iterative soft threshold algorithm; a kernel minimum mean square filter structure is introduced, adaptive filtering is performed on a residual image, the overall structure supports end-to-end implementation, and the method can be applied to an online image preprocessing module in a near-field millimeter wave imaging system. Experiments prove that the method disclosed by the invention shows excellent denoising performance on a plurality of near-field image data sets, and each special noise evaluation index is obviously superior to that of the existing method. According to the invention, low-rank matrix decomposition and a kernel adaptive filtering mechanism are combined, the problem of Gaussian-salt and pepper mixed noise interference in the near-field millimeter wave image is solved, and the imaging definition and the structural fidelity are improved.

The invention provides a method for recognizing liquid hydrogen C-type ball valve inner leakage based on acoustic emission frequency characteristics, and belongs to the technical field of liquid hydrogen C-type ball valve inner leakage detection. After self-adaptive multi-scale wavelet packet decomposition, mixed noise reduction and coherent accumulation enhancement are carried out, a material frequency dispersion compensation model is established, parameters are extracted and input into a leakage feature recognition model to obtain a leakage probability value, and when the probability value exceeds a threshold value, rapid independent component analysis and sparse reconstruction are adopted to estimate the number of leakage sources; the space angle of a leakage source is positioned by using a multi-signal classification algorithm, the leakage aperture is calculated through thermodynamics-driven inversion, and fusion judgment is carried out by combining turbulence noise features extracted by a fractal theory. The technical problem that it is difficult to accurately recognize a tiny leakage source and achieve quantitative evaluation of the leakage degree in the strong background noise environment during inner leakage detection of the liquid hydrogen C-type ball valve is solved.