86 results about "Noise sensitivity" patented technology

The invention particularly relates to a binocular vision-based three-dimensional reconstruction method, which comprises the following steps of: calibrating a binocular camera based on an improved Zhang Zhengyou calibration method to obtain internal and external parameters and a distortion coefficient of the camera; performing stereo correction on the image by using the internal and external parameters of the camera and the distortion coefficient obtained by calibration, so that the binocular image meets an epipolar constraint condition; a multi-strategy optimized semi-global stereo matching algorithm is adopted to process the image after stereo correction, and a disparity map is generated; based on the generated disparity map, generating a three-dimensional point cloud through a triangulation principle; carrying out anti-interference processing and registration optimization on the three-dimensional point cloud; and performing global splicing on the three-dimensional point clouds subjected to anti-interference processing and registration optimization based on a sequential registration error sharing strategy to complete three-dimensional reconstruction. According to the method, the key problems of large calibration error, weak texture matching failure, point cloud noise sensitivity and registration accumulative error in a traditional method are solved, and the reconstruction precision and stability are remarkably improved.

The invention relates to the technical field of artificial intelligence and data analysis, in particular to an unsupervised anomaly detection method and system based on comparative potential fusion. The method aims at solving the problems that in the prior art, an unsupervised anomaly detection method is limited in feature expression ability, sensitive in noise, insufficient in potential feature discrimination and lack of statistical interpretability in detection results. According to the method, the global potential features generated by comparison learning and the self-encoder reconstruction residual error are fused, the statistical model is combined for self-adaptive threshold judgment, the problems of insufficient feature expression and high noise sensitivity in multi-source heterogeneous time series data anomaly detection are effectively solved, and the method has the advantages that the detection precision and robustness are improved, and the dependence on labeled data is reduced.

The invention discloses a human body posture estimation method based on radar point cloud imaging and multi-dimensional feature fusion, and relates to the cross technical field of computer vision and radar signal processing, and the method comprises the following three key technical links: firstly, improving the target resolution through spatial energy distribution estimation; reconstructing target three-dimensional space distribution by using the positive correlation between radar signal energy and a target reflection area and adopting a least square estimation algorithm; secondly, constructing a structured multi-dimensional point cloud matrix, and converting sparse radar point cloud into high-information-density imaging representation through a distance-speed hierarchical sorting strategy; and finally, designing a multi-dimensional feature fusion attitude estimation network, integrating three-dimensional convolution, a multi-head attention mechanism and a gating circulation unit, and realizing collaborative extraction of spatio-temporal features. According to the method, the problems of sparse target features, noise sensitivity and poor universality in traditional millimeter wave radar attitude estimation are solved.

The invention provides a video content enhancement method for a low-illumination environment, and the method comprises the steps: achieving the data preprocessing based on an original low-illumination video frame sequence through frame synchronization, color space conversion and local brightness analysis, generating a noise sensitivity thermodynamic diagram through multi-feature unsupervised learning, and constructing a noise perception gating mechanism through the combination of affine transformation. Dynamic modulation of the characteristic channel is realized; in the multi-scale network structure, a channel attention module is used for carrying out layer-by-layer self-adaptive adjustment on a noise sensitive area; a basic illumination image and an edge enhancement image are generated through double-branch decoding, and then weighted fusion is carried out in combination with a noise thermodynamic diagram, so that brightness balance and detail enhancement are realized; a noise smoothing regular term is introduced during end-to-end training, so that the network achieves dynamic balance between an enhancement effect and noise control.

The invention discloses a space-time density peak value clustering method and system based on shared neighbor weighting, and the method comprises the steps: generating a weight coefficient through quantifying the proximity of a space-time neighbor in time and space dimensions; fusing a weight coefficient, a sample density degree and a shared neighbor similarity, and innovatively calculating a shared neighbor weighted local density which can better reflect spatio-temporal data characteristics; a decision value is generated in combination with the relative distance to accurately select a class cluster center; class cluster division is carried out by adopting a dynamic priority ranking strategy based on a maximum heap priority queue, so that the boundary point distribution precision and efficiency are effectively optimized; and finally, abnormal points are identified through a self-adaptive dynamic threshold value based on statistical distribution characteristics. According to the method, the problems that traditional density peak value clustering is poor in spatio-temporal data adaptability and high in noise sensitivity are solved, and the clustering accuracy, robustness and efficiency in applications such as intelligent traffic flow analysis and regional hot spot monitoring are remarkably improved.

The invention relates to the technical field of infusion monitoring management based on central monitoring, and discloses an infusion monitoring management method based on central monitoring. Collecting the residual volume at regular time through a liquid level sensor, calculating the instantaneous flow velocity by combining the volume difference and time difference of adjacent samples, setting the order of a fractional derivative, and recursively solving a generalized binomial coefficient to obtain a derivative value; constructing a sliding window according to a preset time window, extracting a median and a median absolute deviation, and adaptively generating upper and lower limit thresholds; and comparing the derivative and the threshold value mark abnormity in real time, and regularly summarizing to generate a monitoring report. According to the integrated process, the problems of noise sensitivity, lag and missing detection of traditional monitoring are solved, fractional order derivatives strengthen tiny drift perception, robust statistics restrains interference, a self-adaptive threshold gives consideration to sensitivity and robustness, automatic high-frequency sampling and periodic reporting improve monitoring precision and response speed, manual burden is relieved, and management is optimized.

The invention discloses a thermal hydraulic flow field prediction method based on an evidence physical information neural network, and the method comprises the specific steps: collecting different physical quantity data sets in a thermal hydraulic system, and carrying out the preprocessing of the data; constructing data loss; constructing physical loss; constructing an evidence uncertainty modeling mechanism, and dynamically allocating evidence weights according to the reliability of different data and physical constraints; constructing a total loss function by integrating the data error, the physical residual error and the evidence reliability; and outputting the trained neural network model and prediction. According to the invention, by introducing an evidence learning mechanism, the uncertainty of the model is dynamically estimated in the network training process, and the adaptability of the model to data with inconsistent noise sensitivity degrees is significantly improved; the uncertainty of the evidence is mapped into a physical quantity weighting coefficient, a uniform normalized weighting loss function is constructed, and the subjectivity and robustness problems caused by manual setting of hyper-parameters for each physical quantity in a traditional physical information neural network are effectively avoided.

The invention discloses a high-speed rail line gale prediction system and method, and relates to the technical field of high-speed rail gale prediction. The system for predicting the gale along the high-speed rail comprises a multi-scale grouping representation module, a comprehensive prediction module and a data enhancement module, the multi-scale grouping representation module is used for generating a wind speed preliminary prediction value; the comprehensive prediction module is used for generating a final wind speed prediction value; the data enhancement module expands the diversity of the original wind speed data set. Based on the high-speed rail line gale prediction system, the high-speed rail line gale prediction method comprises the following steps: grouping multi-scale data; carrying out multi-dimensional attention feature fusion; calculating a preliminary predicted value; calculating a final predicted value; carrying out combined optimization training; expanding an original wind speed sample; and performing consensus enhancement training. The method has the advantages that through collaborative design of multi-scale feature fusion, dynamic adaptive weighting and unsupervised enhanced training, the three core problems of sudden wind speed report missing, noise sensitivity and marking dependence in high-speed rail gale prediction are solved, and the operation efficiency and safety of high-speed rail lines are improved.

The application relates to the technical field of system-in-package, and discloses a functional chip SIP (System in Package) system-in-package method and system. The method constructs a reduced-order discrete thermal state predictor through multi-order exponential attenuation fitting, generates a DVFS gear-noise hazard degree spectrum mapping table based on a synchronous switching noise spectrum and an ADC noise sensitivity curve, generates a multi-DVFS working condition robust grounding network topology by optimizing a narrow bridge connection structure parameter by using a sequential quadratic programming, quantizes equivalent noise interference amounts of each DVFS gear on an ADC chip, and jointly optimizes DVFS gear selection and power consumption upper limit distribution under a rolling time domain mixed integer quadratic programming framework, simultaneously performs online model correction through exponential weighted moving average and Kalman filtering, and realizes cooperative satisfaction of thermal constraints and noise constraints.

A kind of atomic gyroscope pumping rate stable composite control method based on phase reshaping extended state observer and power approaching sliding mode control.Aiming at the problems that standard linear extended state observer has steady-state estimation error under the action of ramp disturbance, disturbance estimation phase lag and it is difficult to balance between observation bandwidth improvement and noise sensitivity, a composite corrector composed of lead network and lag network is cascaded at the output end of disturbance estimation of extended state observer, the disturbance estimation channel is shaped in frequency domain, phase lead is provided in mid-frequency band to reduce compensation delay, amplitude lag characteristic is used to suppress high-frequency noise, and parameter constraint design is used to realize zero-error estimation of ramp disturbance;At the same time, a power approaching sliding mode controller is constructed to improve the adaptability to the nonlinear, internal and external coupling and parameter uncertainty of atomic gyroscope system, and the phase reshaping extended state observer is combined to effectively estimate and feed forward compensate the lumped disturbance, so as to improve the stability and accuracy of pumping rate control.The present application can balance dynamic anti-disturbance performance, steady-state control accuracy, convergence speed and noise robustness without significantly improving the observation bandwidth, and is suitable for atomic gyroscope precision control system affected by ultra-low frequency slowly varying disturbance.

The invention discloses a knowledge tracking method and system based on cognitive decoupling, and belongs to the technical field of knowledge tracking, and the method comprises the steps: carrying out the coding of historical learning interaction data of students, and generating a question feature vector and an answer feature vector; respectively decomposing the question feature vector and the answer feature vector into a stable cognitive mode component and a random factor component; performing time sequence modeling on the stable cognitive mode component based on an attenuation attention mechanism, and simulating short-term disturbance of the random factor component to the knowledge state based on the attenuation attention mechanism to obtain dynamic knowledge state representation; and according to the dynamic knowledge state representation and the current question feature vector, predicting the correct answering probability of the student to the next question. The method solves the problems of noise sensitivity, insufficient dynamic modeling, single feature representation and the like in the prior art, and finally realizes higher prediction precision, stronger robustness and finer-grained knowledge state tracking.

The application discloses an improved extended state observer method based on frequency domain disturbance compression, which comprises the following steps: an extended state observer designed for an optoelectronic tracking system is reconstructed from a frequency domain perspective, a frequency domain disturbance compression algorithm is implemented, and a disturbance compression extended state observer is obtained; the disturbance compression extended state observer is converted into a frequency domain form in combination with a linear feedback controller; the stability of the disturbance compression extended state observer and the stability of a closed loop system are analyzed to obtain stability constraint conditions; a parameter setting method of the disturbance compression extended state observer is set; and on the basis of meeting the stability constraint conditions, the influence of the disturbance compression extended state observer on disturbance suppression and noise sensitivity is analyzed. The application effectively reduces the upper limit of the disturbance and its derivative by changing the disturbance characteristics perceived by the observer, significantly improves the estimation accuracy of the state and the disturbance, and improves the anti-interference ability and noise insensitivity of the system.

The invention provides a medical invoice edge detection model construction method and device, equipment and a medium, and the method comprises the steps: obtaining a medical invoice image set, and marking the invoice semantic anchor point of each medical invoice image in the medical invoice image set, so as to construct a sample set; the mask auto-encoder is improved, and an optimized mask auto-encoder is obtained; performing iterative training on the optimized mask auto-encoder by using the sample set to obtain a trained mask auto-encoder, and extracting a multi-layer residual network from the trained mask auto-encoder; constructing an initial edge detection model based on a multi-layer residual network and a double-branch attention fusion module; and training the initial edge detection model by using the sample set to construct a medical invoice edge detection model. By adopting the medical invoice edge detection model construction method and device, the equipment and the medium, the noise sensitivity is reduced, the edge feature extraction capability is improved, and the image detection precision is improved.

The invention discloses a silence cabin intelligent control method and system based on an AI digital matrix, relates to the field of silence cabin intelligent control, and constructs a high-dimensional silence cabin multi-modal feature tensor, namely the AI digital matrix, by obtaining multi-source heterogeneous data such as environment, biological features and equipment states and using time synchronization, matrix construction and normalization cleaning technologies. On the basis, performing user activity scene semantic analysis on the feature tensor to generate an activity weight vector reflecting the current demand of the user; and then prediction modeling of the acoustic-thermal coupling effect is carried out based on the weight vector, optimization is carried out in a prediction result set by using a multi-target Pareto optimal algorithm, and optimal control parameters giving consideration to noise sensitivity, thermal comfort and energy consumption are calculated. And finally, a PWM signal is generated through closed-loop feedback to drive a fan and light, so that the breathing effect of the fan is eliminated on the physical level, and active precise regulation and control of the environment of the mute cabin are realized.

The application provides a spectral submanifold-based fast and accurate control method for a cable-driven tensegrity robot, and belongs to the field of control of the cable-driven tensegrity robot. First, end position data is taken as sampling data, data in the direction of the z coordinate axis is removed from the truncated data to obtain training data after truncation. Second, according to the Takens delay embedding theorem, the training data is projected onto a spectral subspace, a spectral submanifold is parameterized, a reduced-order dynamics is obtained, and an optimal solution is found. Finally, the position data and control input of the sampling end center point are open-loop controlled, the data of the end center point is projected onto the spectral subspace, is converted into an optimal problem to learn a control matrix, and then an optimal control problem is constructed. The application has low dependence on high-dimensional data, weak noise sensitivity, strong robustness of the reduced-order model, can efficiently realize fast and accurate control of the end pose of the cable-driven tensegrity robot, and has strong adaptability and expansibility.

This invention relates to an efficient particle sphere generation method based on principal component analysis (PCA), belonging to the field of quantum computing. It solves the problem of low data representation efficiency caused by excessive segmentation and noise sensitivity in existing particle sphere algorithms. The technical solution includes: initializing the particle sphere set using a parallel sampling strategy; calculating the splitting gain and setting a threshold; calculating the optimal splitting direction based on the PCA core sample set; splitting the particle spheres along the principal direction and iteratively updating the set. This method achieves efficient data compression, reduces computational complexity, enhances the representation ability of complex geometric structures, and has strong robustness.

This invention discloses a method for estimating the speed and position of a permanent magnet synchronous motor (PMSM) based on a third-order switched extended state observer. Specifically, based on the mathematical model of the PMSM, a back-EMF observer based on a sliding mode observer is constructed to obtain the estimated back-EMF. Further, the estimated back-EMF signal is processed using a low-pass filter and amplitude normalization. The processed back-EMF is then used as the input signal to a position estimation scheme based on the third-order switched extended state observer, achieving accurate estimation of the PMSM's speed and position under different operating conditions. This invention is simple in principle and easy to implement. It combines the advantages of both third-order linear and third-order nonlinear extended state observers, improving estimation accuracy and dynamic performance while reducing the adverse effects of noise on the estimation scheme. This solves the technical problems of low estimation accuracy, high noise sensitivity, and difficulty in parameter tuning in existing estimation schemes.

The invention discloses an ultrasonic flight time measurement method based on GAF image coding. The method comprises the following steps: acquiring an original ultrasonic echo signal and carrying out wavelet noise reduction processing; extracting an envelope signal from the original ultrasonic echo signal subjected to wavelet noise reduction processing; the envelope signal is coded into a two-dimensional image matrix by a Gramer angle field; inputting the Gramer angle field image subjected to image standardization into an improved Swin Transform network to carry out depth feature extraction; carrying out regression calculation on a flight time estimation value according to the depth features; and outputting a material thickness or defect positioning result according to the flight time estimation value. According to the method, by fusing signal image coding and depth feature learning, the problems of noise sensitivity and limited resolution in traditional ultrasonic thickness measurement are effectively solved, the thickness measurement precision and stability are remarkably improved, and the method is particularly suitable for the nondestructive testing requirements in complex industrial scenes such as high temperature and corrosion.

A clipper assembly includes a handle, a blade, a motor, and circuitry. The handle includes a first end, a second end, and a speaker positioned between the first end and the second end. The blade is coupled to the second end. The motor is positioned in the handle and is operable to actuate the blade. The circuitry is configured to control the speaker and the motor.

The invention discloses a noise-carrying unbalanced sample image regression network training method and application, and relates to the technical field of image processing. The problems of low training efficiency, high noise sensitivity and distribution imbalance caused by dependence on manual intervention weight adjustment and a complex preprocessing process in the prior art are solved. According to the method, an ordered candidate feature space is constructed through discretization annotation values, and the geometrical relationship of adjacent vectors is stabilized through cosine constraints; constructing a regression network to output a prediction feature vector, and dynamically dividing positive and negative candidate sets based on a relaxation factor; a geometric constraint collaborative optimization mechanism is adopted, and end-to-end training is carried out in combination with zoom-in loss and zoom-out loss; and normalizing the similarity, weighting and fusing the discrete candidate values, and outputting a continuous regression result. According to the method, labeling noise interference is shielded by loosening a safe area, sample distribution deflection is automatically corrected by using global gradient balance, and the robustness and prediction continuity of a model in a complex industrial scene are remarkably improved.

The invention discloses an energy storage converter virtual inertia control method based on a heterogeneous second-order generalized integrator-frequency locked loop (HSOGI-FLL), and aims to solve the problems that in virtual inertia control of an existing energy storage converter, power grid angular frequency operation often needs to be depended on to provide control input, and noise interference is easy to introduce in frequency differential operation, and implementation is complex. The invention discloses the energy storage converter virtual inertia control method based on the heterogeneous second-order generalized integrator-frequency locked loop (HSOGI-FLL). According to the method, on the basis of a typical SOGI-FLL control structure, different SOGI output signals are adopted to construct an FLL control loop, an HSOGI-FLL structure is formed, power grid angular frequency differential signals can be directly and accurately extracted, and power grid angular frequency differential operation does not need to be carried out. According to the method, noise sensitivity and implementation complexity caused by power grid angular frequency differential operation can be avoided, reliable input signals are provided for virtual inertia control of the energy storage converter, then the energy storage converter can effectively provide virtual inertia support for a power grid, and the method can be suitable for the field of grid-connected control of various energy storage converters.

The application discloses an electromagnetic array positioning method and device based on LSTM-EKF fusion and a readable medium, and relates to the electromagnetic positioning field.The method comprises the following steps: calculating the relative distance between each electromagnetic transmitting module and an electromagnetic receiving module at the current time through a magnetic dipole model; acquiring the position coordinates of each electromagnetic transmitting module, constructing input data together with the relative distance between each electromagnetic transmitting module and the electromagnetic receiving module at the current time, and inputting the input data into a trained position coordinate prediction model of the electromagnetic receiving module to output the prediction value of the position coordinates of the electromagnetic receiving module at the current time; initializing the estimated value of a state vector and a covariance matrix thereof to obtain an initial estimated value of the state vector and an initial covariance matrix, and inputting the initial estimated value of the state vector and the initial covariance matrix into an EKF algorithm to output the final estimated value of the state vector.The application solves the problems of the existing electromagnetic positioning technology, such as noise sensitivity, susceptibility to metal interference, significant decline in positioning accuracy and limited working range.

The invention discloses a crack three-dimensional accurate quantification method and system based on semantic point clouds, which are used for solving the problems of skeleton extraction distortion, tail end missing, noise sensitivity and insufficient multi-scale fusion in the existing crack quantification method, and the method comprises the steps: obtaining three-dimensional point cloud data with crack semantic tags, and carrying out the denoising and smoothing preprocessing; segmenting independent crack individuals through a density-based clustering algorithm; dynamically extracting skeleton points by adopting a curvature weighted L1 median algorithm, and adaptively adjusting the point density according to the curvature; skeleton end points are complemented from the original point cloud through main direction projection; calculating crack length, width, direction and type parameters based on the complete skeleton point set; the system correspondingly comprises a point cloud preprocessing module, a segmentation module, a skeleton extraction optimization module, a parameter calculation module and the like. According to the method, high-precision and automatic three-dimensional quantification of a crack structure with complex bending and strong noise interference is realized, and the accuracy, the integrity and the anti-noise capability of skeleton extraction are remarkably improved.

This application provides a hierarchical sensitivity-adaptive model smooth annealing quantization training method and apparatus, relating to the field of artificial intelligence technology. The method includes: determining quantization noise sensitivity based on the trace of the Hessian matrix; determining the annealing temperature based on the quantization noise sensitivity and the current training step number; determining multiple logical values ​​based on the distance between the weights to be quantized and multiple discrete quantization center points; converting the logical values ​​into a probability distribution based on the annealing temperature and a normalized exponential function; and obtaining the quantization weights by weighted summation of the multiple discrete quantization center points based on the probability distribution. The method and apparatus provided in this application determine the sensitivity of each layer through the trace of the Hessian matrix and adaptively adjust the annealing rate, solving the problems of low optimization efficiency and accuracy loss; by fitting the quantization function using a normalized exponential function, a smooth gradient path is constructed, solving the gradient mismatch problem; enabling large language models to have high model performance at extremely low bit compression rates.

The invention relates to the technical field of medical imaging, in particular to a self-supervised deep learning magnetic resonance image reconstruction method and device and electronic device.The method comprises the steps that multi-channel k-space data are acquired, the multi-channel k-space data comprise multiple pieces of parallel magnetic resonance acquired k-space data, and the k-space data comprise magnetic resonance images; and based on the multi-channel k space data, carrying out self-supervised training on a pre-established expanded neural network. In the training process, a single-channel image is generated, an under-sampling mask is applied to a discarded coil, the single-channel image and the under-sampling mask serve as network input, and a network is supervised and trained through an under-sampling matrix. And reconstructing a magnetic resonance image by using the trained expanded neural network. Therefore, the problems of complicated calculation, sensitivity to noise, dependence on a large amount of full sampling data and the like in related technologies are solved.

The application provides a knowledge management and expression method, relates to the technical field of data information management, and comprises the steps of judging whether a mode layer and knowledge data are matched, generating a graph structure framework, obtaining core features of the mode layer, and generating a knowledge graph containing knowledge point nodes, correlation relationship edges and visual elements, etc.The application aims at the weak explainability of the existing knowledge graph, converts abstract knowledge features into intuitive visual elements through multi-dimensional visual rules, records specific data by matching node feature forms, and ensures logical compliance through node text matching degree and knowledge correlation index checking; solves the noise sensitivity problem, filters invalid correlations through matching degree and correlation index threshold, eliminates isolated nodes, effectively reduces entity connection errors, relationship ambiguity and other noises, and guarantees the accuracy and reliability of the graph; and the graph is split into connected subgraphs to realize block management, greatly reduces query delay, and improves use efficiency.

The invention provides a Raman spectrum gas component quantitative analysis method based on an XGBoost algorithm and application of the Raman spectrum gas component quantitative analysis method, belongs to the technical field of spectrum analysis, and is particularly suitable for real-time gas monitoring in the olefin polymerization and cracking process. According to the method, Raman spectrum data of mixed gas and pure substances of all components are collected through a Raman spectrum enhancement cavity, after pretreatment, key characteristic peaks are extracted by adopting signal-to-noise ratio screening, correlation minimization or principal component analysis, and then an XGBoost model is trained with the intensity of the characteristic peaks and gas components are predicted. The problems that traditional Raman analysis is high in nonlinear interference and sensitive in noise and a traditional machine learning model is low in precision and poor in interference resistance are solved, the mean square error and the anti-interference capacity are remarkably optimized, and a high-precision and high-robustness gas component detection scheme is provided for a complex industrial environment.

The invention relates to a kidney tumor image segmentation method and system based on multi-scale feature extraction, and the method comprises the steps: extracting the multi-scale features of a kidney tumor image to be processed, dividing feature similarity groups, carrying out the optimization through combining the texture and shape, and obtaining similar feature groups; generating a complementary information flow, adaptively enhancing an interaction weight between groups by using an attention mechanism, and obtaining an enhanced complementary feature representation; evaluating and denoising noise, establishing association mapping among groups based on features after denoising, and performing multi-scale feature fusion and boundary refining by adopting a U-Net network to obtain tumor boundary representation; and the boundary fuzzy degree is evaluated, iterative optimization is carried out by adopting a graph convolutional network, optimized collaborative hierarchical representation is obtained, and finally a high-precision kidney tumor segmentation result and a pixel-level contour are output. According to the method, through multi-scale feature cooperation, attention interaction and boundary optimization, the problems of boundary blur and noise sensitivity in kidney tumor segmentation are effectively solved, and the accuracy and robustness of segmentation are remarkably improved.

The application discloses a SAR image shadow identification method with an attention mechanism, comprising the following steps: acquiring a SAR image and performing pretreatment; then inputting a parallel CNN encoder and a ViT encoder respectively, and extracting multi-scale features; using a bidirectional feature fusion module to perform deep fusion on the multi-scale features; then inputting a decoder, recovering the feature resolution through step-by-step upsampling, using an attention gate module and an attention residual module to optimize the features in the decoding process, and introducing a convolution block attention module in the fusion stage of the decoder path to enhance spatial context information; performing deep supervision on lateral outputs at different scales, using a weighted loss function to train, and generating shadow region identification results. The application can adaptively learn geometric context information from SAR image data, eliminate the dependence on high-precision DEMs, and effectively solve the problems of boundary blur, under-segmentation and noise sensitivity.

The invention relates to a Fiedler cobweb privacy protection noise mechanism construction method based on Transform, and belongs to the technical field of high-order networks and privacy protection noise. According to the method, from two dimensions of node attributes and local topology, bipartite representation of nodes and hyperedges is constructed, deep interactive modeling is performed on the nodes and the hyperedges through a Transform trans-attention mechanism, and an enhanced hypergraph incidence matrix is generated. And constructing a normalized hypergraph Laplacian matrix based on the enhanced hypergraph incidence matrix, and extracting Fiedler values and corresponding feature vectors through spectral decomposition so as to describe global structure features of the hypergraph. A Fiedler value is used as a global reference, the association influence of each hyperedge is calculated in combination with enhanced hypergraph association, the maximum influence is used as a noise sensitivity scale, a Fiedler noise disturbance mechanism is designed, differential protection is applied to a query result, effective defense of epsilon-level cobweb privacy protection is achieved, and the privacy protection effect of a cobweb is improved. Therefore, the high-efficiency availability of data query is kept while the risk of hypergraph sensitive information leakage is reduced.