159 results about "Random noise" patented technology

A laser ranging method and device using the autocorrelation characteristics of light intensity random noise, the method comprising: step S1, a wide spectrum light source emits laser light, which is divided into a first light beam and a second light beam by a ranging probe, and the first light beam is defined as a reference light beam; step S2, the second light beam is emitted to a target object to be measured, and a light beam reflected by the target object to be measured is obtained and defined as a signal light beam; step S3, the reference light beam and the signal light beam are coupled to form a coupled light beam; and step S4, the coupled light beam is processed to obtain absolute distance information of the target object to be measured to the light emitting end surface of the probe. The device obtains absolute distance information of the target object to be measured based on light source intensity noise. The present application has the advantages of high time resolution and high precision measurement.

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.

The application discloses a kind of Φ-OTDR system coherence fading inhibition method and device based on spatial resolution internal redundant data, specifically: first, simulate random one-dimensional continuous signal and noise waveform, and carry out channel replication expansion, add different random noise to the waveform of each channel, obtain multi-channel signal and noise data set;Then set the structure, parameter, activation function, loss function and optimizer of time sequence neural network model, data set is sent into model and is trained;Then judge whether the model error meets the requirements, if it does not meet the requirements, adjust the structure, parameter, activation function, loss function and optimizer of model, retrain the model;If it meets the requirements, the actual data detected by Φ-OTDR system is processed and packed, and is sent into the trained model for prediction, to obtain the signal characteristics distributed along the whole optical fiber.The application has low complexity, reduces the influence of coherence fading noise on Φ-OTDR system, and improves the transmission ability of the system.

The application provides a low-light tip defect recognition method based on an attention-enhanced network, and belongs to the technical field of offshore wind turbine fault diagnosis, and comprises the following steps: S1: using a UAV to shoot a plurality of offshore wind turbine pictures containing a tip to form an original data set; S2: preprocessing the original data set, filtering pictures that do not meet the requirements, and arranging and sorting suitable pictures to form a structured original data set; S3: using a noise adding algorithm to add random noise to a single picture in the data set to enrich sample categories under different light conditions; S4: using the data set after adding noise for model training of the attention-enhanced network until a model meeting the accuracy and speed requirements is obtained; and S5: applying the mature model to engineering practice to finally form a complete low-light tip defect recognition method based on the attention-enhanced network.

This invention belongs to the field of protein fluorescence intensity prediction, specifically relating to a method, system, device, and storage medium for predicting the fluorescence intensity of protein sequences. The method includes: extracting features from a training dataset, training a generator and discriminator by incorporating random noise to obtain a generative adversarial model (GAP); supplementing data using the GAP to obtain a prediction dataset; extracting semantic features from the prediction dataset and performing local pattern enhancement on the hidden layer features; fusing these features to obtain the final features; using the final features as input to a prediction network to train the prediction network to obtain an optimal prediction network; and inputting the actual protein sequence into the optimal prediction network to obtain the fluorescence intensity corresponding to the actual protein sequence. This application improves the accuracy of predicting the fluorescence intensity of protein sequences.

This invention discloses a thin-layer weak signal enhancement method based on cyclic spectrum enhancement technology. The method first preprocesses the original seismic data using a Gaussian smoothing operator to suppress random noise while preserving the main signal features. Then, even-order derivative operations are introduced to enhance the high-frequency components and detailed information of the seismic signal, highlighting the reflection characteristics of the thin-layer interface. To address the issue of high-order derivatives easily introducing high-frequency noise, a Butterworth low-pass filter is used for directional noise reduction. Finally, an adaptive cyclic termination condition is used to determine the optimal derivative order, and the derivative seismic traces are superimposed onto the original seismic traces to achieve thin-layer weak signal enhancement. Numerical simulations and tests with actual seismic data show that this method can effectively enhance the weak signals of 10m (λ / 4) and 5m (λ / 8) thin layers, improve the continuity and clarity of the phase axis of thin-layer reflected waves, and thus enhance the identification capability of thin-layer reservoirs under complex geological conditions.

The application discloses a kind of underground space optical cable high fidelity sound restoration method based on DAS system, comprising: the real-time signal collected to optical cable is carried out data demodulation, identify whether it contains energy impact signal generated by the optical cable after being padded and being hit, if it contains, the energy abnormal distribution position in energy impact signal is positioned, obtains positioning point;Continuous analysis time domain signal of positioning point, after filtering low-frequency interference signal, phase anomaly signal is rejected;The sound signals of a plurality of equivalent sensing nodes adjacent to the positioning point are superimposed and averaged to suppress additive noise, and then the sound signal is restored.The application can realize rapid positioning according to the spatial distribution mapping relationship of optical cable, and after filtering background noise from the collected signal, the highly correlated signals of multiple spatial sensing nodes near the positioning position are linearly superimposed to suppress additive random noise, and high-fidelity sound signal restoration is realized.

This invention relates to the field of sensors and discloses a sensor drift correction method and system based on transfer learning, comprising the following steps: S1, collecting sensor data, including triaxial acceleration data, temperature, humidity, and air pressure environmental parameters; S2, calculating the magnitude of the triaxial composite vector. In this invention, a zero-point mapping model is constructed by using Gaussian process regression. The combined kernel function of periodic kernel, linear kernel, and RBF kernel can simultaneously capture the periodic fluctuations of environmental factors, the long-term trend of sensor aging, and short-term random noise. The maximum mean difference loss function is used to achieve adaptive distribution alignment between the source domain and the target domain, enabling the model to quickly adapt to the target domain. At the same time, a slope intercept joint judgment matrix is ​​introduced to achieve accurate differentiation between zero-point drift and real motion and tilt events, with an accuracy of over 96%. This effectively avoids measurement distortion caused by miscompensation and significantly improves the reliability and data credibility of the system.

This invention discloses a peak retrieval and feature extraction method suitable for environments with strong electromagnetic interference, belonging to the field of signal processing and high-speed data acquisition technology. Due to the presence of a large amount of broadband random noise and high-frequency interference in signals under strong electromagnetic environments, the peak features of signals are difficult to extract accurately. This invention proposes a peak detection and feature extraction method based on continuous comparison. This method constructs a sliding comparison model to quickly locate candidate peaks by comparing the amplitudes of adjacent sampling points, and combines a threshold decision mechanism to remove false peaks caused by noise. Based on this, feature parameters such as peak amplitude, peak position, and pulse width are extracted. Simultaneously, a dual-RAM ping-pong cache structure is introduced to achieve alternating storage and parallel output of feature data, avoiding data processing blockage and improving system real-time performance and stability. This invention has a simple structure, is easy to implement in FPGA, and can achieve high-precision peak detection and continuous data output under high-speed sampling conditions. It has good anti-interference performance and engineering application value.

This invention discloses a lightweight method for identifying pavement cracks based on a dynamic sliding window, specifically relating to the field of intelligent pavement defect detection technology. It addresses the technical problems of easy missed and false detections of cracks in complex pavement scenarios, and the difficulty in achieving continuous and stable crack identification. By performing denoising processing on the pavement image and extracting structural fidelity residuals, the linear structural features of small cracks are preserved while suppressing random noise and pavement texture interference. This ensures that the coverage ratio of the structural residual map in the neighborhood of the crack reference line is no less than 85%, and the missed detection rate of small cracks is no higher than 8%. Furthermore, by using a crack candidate prior field to drive the direction of the irregular adaptive window, combined with reliability-constrained cascaded forest discrimination, fracture compensation backtracking, and pseudo-line removal backtracking, the false detection rate is no higher than 12%, and the single-frame processing latency is no higher than 50 milliseconds. This method can output continuous, stable, and traceable crack identification results.

In an embodiment, input data (1) associated with a geological area and (2) including a plurality of types of input data is received. The input data is input into a multi-mode embedding encoder configured to generate an output (1) based on the input data, (2) representing a compressed version of the input data and (3) that is in a format compatible with a generative network. First random noise is generated. The output and the first random noise are input to the generative network to generate a first geological model. Second random noise different than the first random noise is generated. The second random noise, the output, and the first geological model is input to the generative network to generate a second geological model different than the first geological model.

This invention discloses a high-fidelity quantum state discrimination method and device based on automatic ion localization. The invention constructs a dynamic closed-loop compensation system of ion position update, effective discrimination region reconstruction, and adaptive adjustment of discrimination parameters. This system addresses the degradation of discrimination fidelity over time due to imaging drift in multi-ion quantum state discrimination, achieving long-term stability and high-fidelity quantum state discrimination. When constructing the effective discrimination region for ions, a physical constraint model based on the ion trap potential field distribution is introduced, utilizing bright spot features and the physical property of ions aligning along the ion trap axis to form linear ion chains, establishing a correspondence between ions and bright spot regions. By extracting the feature parameters of the point spread function in real-time from the average bright state image, and adaptively constructing local pixel regions highly correlated with the actual fluorescence distribution of ions, random noise interference from background pixels is suppressed to the maximum extent, fundamentally improving the signal-to-noise ratio of multi-ion quantum state discrimination.

Embodiments of the present disclosure provide a video texture migration method and device, electronic equipment and storage medium, which obtain an original video, perform feature extraction on a target frame in the original video to generate first feature information of the target frame, wherein the target frame is a video frame after an Nth video frame in the original video, and the first feature information represents an image structure contour of the target frame; perform feature fusion on the first feature information of the target frame and reference feature information to obtain second feature information of the target frame, wherein the reference feature information is used to represent an image structure contour of a reference frame, the reference frame is a video frame before the target frame, and the second feature information is the first feature information without random noise; and generate a texture migration video according to the second feature information of the target frame and corresponding texture feature information, wherein the texture feature information represents image texture details of a reference image, thereby avoiding frame flickering and improving the display effect of the texture migration video.

This application discloses a method and system for identifying 2D engineering drawing elements based on dynamic query-guided consistent projection, proposing a dynamic query-guided consistent projection decoder. This decoder achieves a single forward propagation projection from a random noise vector to an accurate element state vector. The method first initializes a set of vectors following a predetermined noise distribution as initial dynamic queries. These queries are fed into the dynamic query-guided consistent projection decoder (DQ-CPD). At each layer of the decoder, the current dynamic query vector not only interacts with other queries through a self-attention mechanism, but it is also input as a controller into a "contextualized hypernetwork." This hypernetwork generates a set of dedicated weight parameters in real time based on the current inference information carried by the query vector. This method organically combines dynamic queries, dynamic context, and one-step generation theory to construct a novel, efficient, accurate, and adaptive element identification method.

This invention discloses a signal Doppler tolerance and sidelobe optimization method based on a weighted ambiguity function template. The method includes determining a discrete ambiguity function based on a two-dimensional ambiguity function of a continuous signal and a radar transmitted signal; obtaining matched filtering results for signals with different Doppler frequency shifts and determining the main lobe and sidelobe positions based on the Doppler frequency shift results; for the Doppler frequency shift case, establishing a cost function for Doppler tolerance optimization and a weighted cost function for joint optimization of Doppler tolerance and sidelobe based on the main lobe and sidelobe positions, respectively; setting the weight coefficients of the weighted cost function according to signal characteristic requirements to obtain the target cost function; and using simulation parameters for iterative updating to obtain the optimization result. This invention can improve signal Doppler tolerance while reducing signal sidelobes, significantly improving the practicality of random noise signals; and it allows for setting weight coefficients according to signal characteristic requirements to adjust the optimization ratio of signal Doppler tolerance and sidelobe, providing strong flexibility.

This invention pertains to data processing technology for seismic exploration of sandstone-type uranium deposits, specifically an adaptive denoising method for sandstone-type uranium deposit seismic data based on sparse transform. The method includes: inputting noisy raw two-dimensional seismic data y; initializing the maximum number of iterations I, setting the outer iterative accumulator i=1, and designing a search range J for the denoising coefficient truncation percentage; for each i, calculating the curvelet transform coefficients s of y; for each j within the search range, applying hard thresholding for denoising, and performing an inverse curvelet transform on the denoised coefficients to obtain the denoised seismic data y. j ; Calculate the noise removal ε corresponding to each j j The entropy; compare the calculated noise ε removed for each j. j The entropy is calculated, and the percentage corresponding to the maximum entropy within the search range is taken as the optimal denoising effect, and the corresponding seismic data is output. If i reaches the maximum number of iterations, the loop is exited, and the denoised seismic data corresponding to the current maximum entropy value is used as the output result. This invention can protect the effective signal while suppressing random noise.

The application discloses a 12-lead electrocardiosignal denoising method and system based on a dual generative adversarial network, which comprises the following steps: obtaining time-adjacent clean electrocardiosignal segments and noisy electrocardiosignal segments; constructing a noise generative adversarial network, wherein the generator takes a random noise vector as input and simulates noise as output, the discriminator takes the noisy electrocardiosignal segment as a positive sample and a synthetic signal composed of the clean electrocardiosignal segment and the simulated noise as a negative sample for adversarial training, and a noise generator snapshot library is generated; performing dynamic noise injection on the clean electrocardiosignal segment to obtain 'noisy-clean' paired data; constructing a denoising network comprising a denoiser; obtaining the 12-lead electrocardiosignal containing noise and inputting it into the denoiser to output the denoised 12-lead electrocardiosignal. The application aims to overcome the defects in the prior art, such as the lack of real 12-lead electrocardiosignal denoising paired data and the poor model generalization ability caused by excessive dependence on the Gaussian noise assumption.

The application discloses an image repairing model training method, an image repairing method, an image repairing device and a storage medium. The image repairing model training method comprises the following steps: inputting a first image to be repaired and an i-th random noise in N random noises into a generating network of an image repairing model to generate an i-th second image; determining a first loss value according to a relative entropy of a first parameter and a second parameter; the first parameter is determined according to normalized results of the i-th random noise and the N random noises; the second parameter is determined according to normalized results of the i-th second image and N second images corresponding to the N random noises; the N is an integer greater than 1; and the weight parameters of the image repairing model are updated according to the determined first loss value.

This disclosure provides an image noise processing method, apparatus, electronic device, and storage medium. The method includes: acquiring at least two first sampled images of the same scene; determining pixel values ​​of pixels in the at least two first sampled images; wherein the pixel values ​​include random values ​​indicating the influence of random noise and non-uniform values ​​indicating the influence of non-uniform noise; determining residual noise information based on the difference between the pixel values ​​of corresponding pixels in two second sampled images; wherein the first sampled images include the second sampled images; and acquiring noise parameters for noise reduction based on the residual noise information.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining whether to generate sub-domain noise. One of the methods includes, for a message for a downstream system, determining whether to generate noise data that has the same sub-domain as a true value, the true value from a domain that has a plurality of different sub-domains including the sub-domain; using a result of the determination whether to generate noise data that has the same sub-domain as the true value, generating the message for the downstream system; and transmitting, to the downstream system, the message.

This invention provides a robot motion planning method, apparatus, device, and medium based on a diffusion model. Addressing the shortcomings of existing motion planning methods, such as high dependence on initial solutions, difficulty in balancing planning efficiency and smoothness, and poor generalization ability of learning models, this method proposes an end-to-end framework integrating prior learning and posterior guidance. After acquiring state data and distributing it according to function, the method uses a diffusion model for multi-step iterative denoising starting from random noise. The core of this invention lies in the following: in each denoising time step of trajectory diffusion, the control points of the current intermediate noisy trajectory are extracted and represented by B-spline low-dimensional parameterization; multi-objective comprehensive cost and gradient are constructed in real time based on multi-source states; and the cost gradient is incorporated as a guiding signal into the posterior sampling to guide the denoising update direction of the network. This invention deeply nests dimensionality reduction and cost guidance within a single iteration, enabling the rapid generation of multimodal smooth motion trajectories for adaptive dynamic obstacles without the need for post-processing smoothing.

The application relates to a two-stage time difference positioning enhancement method and system based on data fusion, and belongs to the field of radio signal processing and information fusion. The method aims to solve the problems of performance deterioration of traditional data fusion methods caused by coexistence of random noise and unknown system error in TDOA positioning and sensitivity of self-calibration technology to initial values. The technical scheme comprises the following steps: firstly, a first-stage fusion is performed on the biased measurement by minimizing the convex optimization of Stein loss, so that a high-quality initial positioning estimation after random noise suppression is obtained; and then, taking the initial value as a starting point, a joint observation model is constructed and iteratively solved, and the real position of the target and the system error are accurately estimated. The application can effectively improve the fusion precision and robustness of multiple sets of biased TDOA positioning data in a complex environment without the need of an external cooperative calibration source.

The application discloses an industrial defect detection method based on cross-layer collaborative multi-level attention and progressive feature enhancement, and belongs to the field of computer vision. The method first constructs a cross-layer collaborative multi-level attention feature extraction network, weights different level feature maps through slight, medium and severe three-level defect degree attention modules, and divides defect samples into three levels of simple, medium and difficult based on training loss value and confidence. Secondly, a conditional generative adversarial network is constructed, defect feature vectors, defect category labels, difficulty levels and random noise are taken as generator inputs, an enhanced defect feature vector of a specified category and difficulty is generated in a feature space, and an adversary is trained through a discriminator to optimize the generation quality. Finally, a progressive weighted training strategy is adopted to retrain the defect classifier, training data is introduced in stages according to the sample difficulty level, the loss weights of various samples are dynamically adjusted, and progressive model optimization from easy to difficult is realized.

This invention discloses a high-resistivity fault location method for distribution networks based on feeder segment model mismatch identification. First, the zero-sequence voltage of the bus and the zero-sequence current at each measuring point are collected, and the derivative of the zero-sequence voltage and the segment zero-sequence current difference of each feeder segment are calculated. Then, the data from the first power frequency cycle after the fault is selected, and the effective value of a selected low-frequency band is used as a scaling factor to normalize the segment zero-sequence current difference and the derivative of the zero-sequence voltage, achieving accurate alignment of the effective signals. Next, a first-order accumulation generator is used to enhance the signal, suppressing random noise and amplifying fault characteristics. The model mismatch degree of each feeder segment is calculated and multiplied with the fault intensity index to construct a comprehensive mismatch index as the location criterion. Finally, the fault segment is determined based on the maximum value of the comprehensive mismatch index. This invention fully utilizes the essential differences between healthy and faulty segments in the zero-sequence equivalent model, and combines the fault current intensity to construct a comprehensive criterion, possessing advantages such as strong anti-interference capability, good physical interpretability, and wide engineering applicability.

The application provides a seismic data noise attenuation method and device, the method comprising: obtaining seismic data to be denoised; inputting the seismic data into a pre-constructed artificial neural network to output noise intensity data; the artificial neural network is a seismic data noise intensity prediction model obtained by training seismic data samples based on a machine learning algorithm; limiting noisy seismic data based on Casado filtering setting space parameters, determining the size of the Hankel matrix in Casado filtering; obtaining noise intensity data limited by Casado filtering setting space parameters; performing Casado filtering on the noisy seismic data limited by the setting space parameters with the number of ranks to complete random noise attenuation of the seismic data; and performing iterative cycles with the setting space parameters to complete random noise attenuation of all seismic data. The application converts the noise intensity of the seismic data into the number of ranks, so that the rank of the Casado filtering changes with the signal-to-noise ratio of the seismic data, and the random noise attenuation effect is good.

The present application relates to the technical field of computers, and provides a waveform processing method and apparatus, a device, and a storage medium. The method comprises: acquiring a seismic record, wherein the seismic record comprises a plurality of seismic waves received by a plurality of receiving devices, and the seismic waves are transmitted by a transmitting device; performing spatial partitioning processing on the seismic record to obtain a plurality of sampled wavelets and wavelet information of the sampled wavelets; for any sampled wavelet, determining, on the basis of the wavelet information of the sampled wavelet, a transform operator and a deghosting operator corresponding to the sampled wavelet, and determining, on the basis of the transform operator and the deghosting operator, a target wavelet corresponding to the sampled wavelet, wherein the target wavelet is a deghosted wavelet; and concatenating the target wavelets corresponding to the plurality of sampled wavelets to obtain a ghost suppressed wave. Random noise is effectively attenuated, the problem that effective waves and ghost waves cannot be accurately separated during ghost suppression is further avoided, and the accuracy of ghost suppression is improved.

The application relates to the field of artificial intelligence, and discloses a regularization method for enhancing local robustness of a neural network, which comprises the following steps: reading original training data according to a preset batch size; receiving clean input samples, calculating and superimposing disturbance components according to a preset disturbance generation strategy to generate disturbance input samples constrained in a local neighborhood range; inputting the clean and disturbance input samples into a deep neural network model to perform forward propagation and obtain a prediction result; constructing a sample pair set by using the clean and disturbance input samples, calculating the difference norm of the sample pair in an output space and the distance norm in an input space and a ratio of the difference norm to the distance norm to obtain a regularization term; and constructing a joint loss function and updating network weights. The application constrains local smoothness of a model by regularization, and combines a diversified disturbance strategy to improve the robustness and generalization ability of the model under adversarial attacks and random noise.

This invention relates to a commercial cryptographic signature method resistant to side-channel attacks, comprising the following steps: Step 1: Network configuration initialization: Key generation and transaction signing in the system both occur within a designated network; Step 2: Private key conversion and public key generation: Incoming data is converted to an integer using the private key sk and a private key sk object is constructed. The corresponding public key is derived based on elliptic curve cryptography rules; Step 3: Public key and address generation: A hash operation is performed on the public key to generate a public key hash value. Combined with the address format generated by the version, a locking script is constructed based on the public key hash; Step 4: Noise injection operation: A delayed random noise is added during the commercial signature process. By injecting random noise into the signature algorithm, side-channel information is disrupted, preventing attackers from inferring the user's secret data based on the time information leaked during the transaction process, thus improving the system's ability to resist side-channel attacks.

The application belongs to the technical field of intelligent warehouses, and provides a power material RFID tag positioning method and system, which comprises the following steps: acquiring RFID passive electronic tag data; pre-processing the electronic tag data by using Gaussian filtering; and obtaining the positioning result of the power material according to the pre-processed electronic tag data and a preset full connection neural network. The electronic tag data is pre-processed by using Gaussian filtering, the interference of metal materials on the electronic tag data in the power warehouse is realized, the generated Gaussian template is subjected to a sliding convolution operation on the signal data sequence according to the weighted window of smooth transition, the random noise and peak interference caused by metal reflection are effectively filtered out through weighted averaging, the problem of metal material interference in the power warehouse is solved, and the positioning precision is improved.