202 results about "Prior information" patented technology

The invention discloses a physical information neural network hyperspectral band selection method for target identification, which uses physical information SDI as prior information to be combined with a channel attention mechanism added in a deep neural network to guide band selection, and evaluates the advantages and disadvantages of band weights through a reconstruction module after the weights are generated. Wave band selection is carried out accordingly; according to the hyperspectral image target recognition waveband selection method based on the physical information neural network, compared with a traditional method, the process is simplified and the precision is improved through end-to-end learning, the utilization of spatial context information can be enhanced by adding an attention mechanism, and the model has interpretability by adding prior SDI information.

The invention relates to a circuit board defect identification method and system based on multi-dimensional image data, and belongs to the technical field of data identification processing, and the method comprises the following steps: obtaining synchronous image data of a circuit board to be detected in a plurality of imaging modes; performing space-spectrum joint registration on each modal image to generate a multi-dimensional image cube with a unified coordinate system and pixel alignment; inputting the multi-dimensional image cube into a pre-trained multi-branch heterogeneous fusion neural network; generating a pixel-level defect probability graph by utilizing a defect sensing context decoder, and performing geometric constraint optimization on the probability graph by combining prior information of a circuit board design layout; outputting defect types, positions and confidence coefficients, and establishing an interpretable defect fingerprint database according to the multi-dimensional response characteristics of the defects; the method has the beneficial effects that false defect signals generated by image noise and circuit board surface texture interference can be effectively inhibited, the omission ratio and the false detection ratio are greatly reduced, and pixel-level accurate defect positioning is realized.

The invention relates to the technical field of computer vision, and discloses a precise medicine delivery method and system, and the method effectively strips illumination artifacts through the construction of a local gradient structure tensor field and an anisotropic screening mechanism, remarkably improves the robustness of a visual front end in a light and dark alternating environment, and improves the accuracy of medicine delivery. Rigid body motion constraint and Lie algebra continuous modeling are utilized, high-precision space-time alignment between heterogeneous sensors is realized on the premise that scene depth does not need to be recovered, hardware delay errors are eliminated, a physical-driven probability weight dynamic allocation mechanism is constructed by introducing a multipath scattering index and a structural information entropy flux, and a dynamic space-time alignment algorithm is established. Environment degradation is sensed in real time, the observation weight is adjusted in a self-adaptive mode, the influence of the non-line-of-sight multipath effect and visual texture missing is effectively restrained, high-precision inertial dead reckoning can still be maintained in a blind area where a sensor is in full failure in combination with momentum prior information generated through Schel complement marginalization operation, and the method has the advantages of being high in precision and high in precision. And the navigation continuity and reliability of the distribution robot in a complex scene are ensured.

The invention discloses a road intelligent diagnosis method based on an AI large model, and relates to the technical field of road engineering detection, and the method comprises the following steps: building a unified time base line, constructing a phase reference field, injecting a calibration pulse, separating dynamic shadow time sequence texture and crack initial features in a road image sequence, and obtaining a road image sequence; generating prior information of dynamic shadows and cracks; under the constraint of prior information, polarization spectrum three-dimensional coding imaging is carried out, polarization angle distribution and spectrum gradient are extracted, a feature dictionary of dynamic shadows and cracks is constructed, and phase drift in the prior information is calibrated. According to the method, a unified time base line and a phase reference field are constructed, and accurate distinguishing of dynamic shadows and cracks is realized by combining polarization spectrum imaging and anti-fact distillation; the point cloud and acceleration data are fused for geometric reconstruction, and through phase conjugate extinguishing and closed-loop feedback, the continuity and accuracy of diagnosis are significantly improved, and maintenance error decisions caused by misjudgment are avoided.

The invention discloses a multi-modal speech enhancement method and device based on a deep learning model, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining the head posture data, binaural audio signals and visual context information of a user in a virtual reality environment, coding the binaural audio signals into three-dimensional space acoustic features, and carrying out the coding of the three-dimensional space acoustic features; and extracting virtual sound source position features and lip motion features from the visual context information, inputting the features into an immersive fusion enhancement network, selectively enhancing or inhibiting acoustic features from different spatial directions, generating an enhanced audio stream, and outputting the enhanced audio stream through a binaural rendering engine. According to the method, the technical problems that in the prior art, due to the fact that multi-modal prior information cannot be effectively fused, voice enhancement lacks spatial selectivity, and an interference sound source irrelevant to vision is difficult to restrain are solved, and head posture dynamic attention and lip motion cross-modal constraint are fused; the technical effects of natural binding of auditory attention and a visual focus and effective suppression of an interference sound source are achieved.

The invention discloses a geologic body boundary extraction method based on gravity total horizontal derivative anomaly, and belongs to the technical field of geophysical data processing and interpretation. Calculating gravity total horizontal derivative anomaly based on actually measured gravity anomaly data; in the sliding window, detecting whether the gravity total horizontal derivative abnormity meets a preset extreme value structure condition in multiple directions, and screening out boundary points based on an amplitude threshold value and a slope threshold value; according to the invention, a geologic body boundary extraction method based on gravity total horizontal derivative anomaly is established by using directional extremum features of gravity total horizontal derivative anomaly; compared with a traditional boundary recognition method, the method does not depend on prior information, is high in anti-noise performance, can quickly and accurately recognize the boundary position of the geologic body under a complex background, and has higher boundary resolution.

This invention discloses a clutter discrimination and suppression method based on multiple echo features, comprising: acquiring radar front-end clutter and target echo detection video data and dividing it into several connected regions; extracting multi-dimensional features and labeling prior information for each connected region; using the Relief feature selection algorithm to calculate and filter the weights of the multi-dimensional features, removing redundant features to obtain the optimal feature vector; using this feature vector to train a support vector machine classifier, and obtaining the optimal parameter model through cross-validation; acquiring measured clutter data, extracting corresponding features and inputting them into the model; and determining whether to remove clutter and retain targets based on the output. This invention effectively filters out dynamic clutter spots, avoids false deletion of weak targets, significantly reduces the false alarm rate of the system, and alleviates the computational burden of subsequent track processing.

The application discloses a variable working condition cross-component causal coupling modeling and fault diagnosis method and device, and relates to the technical field of rotating machinery fault diagnosis. The method comprises the following steps: collecting and preprocessing multi-component signals under variable rotating speed working conditions of a rotating machinery; based on multivariate Granger causal modeling, prior information and dynamic disturbance information are fused to construct a dynamic causal diagram; a double-view structure of a main view of the dynamic causal diagram and an auxiliary view of a correlation diagram is constructed, multi-scale coupling features are extracted through a multi-order spectrum filter and an attention mechanism, a symmetric InfoNCE contrast learning loss and a supervised classification loss are combined to optimize and train a model, and fault classification is realized. The device is correspondingly provided with signal acquisition, dynamic causal diagram construction, feature fusion, model training and fault prediction modules. The application improves the accuracy of fault feature representation and the robustness of diagnosis under variable working conditions, is suitable for complex industrial scenes with multi-component coupling and frequent rotating speed switching, and has strong engineering practicability.

The invention provides an aeronautical equipment task reliability assessment method based on prior information, and the method comprises the steps: determining the minimum sample demands of a task reliability test under different failure samples according to the technical demands of aeronautical equipment task reliability assessment and the characteristics that the data of the aeronautical equipment task reliability assessment are distributed in a Bernoulli test result; determining available test samples according to existing prior information of the aviation equipment, analyzing data characteristics, and setting beta distribution as prior distribution; according to the number of historical serious quality problem samples of the equipment, a supplementary test profile design is determined according to a task reliability test technology, and a subsequent supplementary test implementation scheme is designed; and according to a supplementary test result of subsequent supplementation, in combination with the Bayesian theorem, a posterior density function is determined, and a task reliability level under a corresponding confidence condition is given, so that the problems that a conventional task reliability evaluation model is not applicable and the demand quantity of supplementary test samples is large are solved.

This application relates to an encoding method, a decoding method, and an electronic device. An example method includes: performing inter prediction on a current frame, to obtain prediction information of the current frame; determining residual information of the current frame based on the prediction information of the current frame and an original picture of the current frame; determining first residual hyperprior information of the current frame based on the residual information of the current frame and prior information of the residual information of the current frame; encoding the first residual hyperprior information of the current frame, to obtain a first bitstream; and performing probability estimation based on second residual hyperprior information of the current frame and prior information of the residual information of the current frame, to obtain a residual probability distribution of the current frame.

This invention discloses a method and system for resisting slice reconstruction interference based on alternating frequency modulation slope polarity and time-series agility. It mainly addresses the problems of high complexity, poor real-time performance, and high dependence on prior information and training data in existing technologies. The scheme is as follows: At the transmitting end, a joint agile transmission signal is established based on the number of sub-pulses, sub-pulse duration, sub-pulse interval, sub-pulse bandwidth, and sub-pulse frequency modulation slope; at the receiving end, the received signal is segmented according to the recorded sub-pulse timing to obtain each segment signal, and pulse compression and delay compensation for both matched and mismatched channels are performed simultaneously; dynamic weights are generated based on the energy of the two compensated channels; an anomaly suppression factor is constructed based on the energy of the matched channels of each segment, and the output of each segment's matched channels is weighted and accumulated using this factor and the dynamic weights to obtain a one-dimensional range profile result. This invention does not require a large amount of training data or estimation of interference parameters, has low computational complexity, can suppress false targets generated by slice reconstruction interference while maintaining the energy of the real target, and can be used for target detection.

The invention relates to the technical field of artificial intelligence content generation, and discloses an AIGC video background generation control method based on hand-painted background structure priori. The method comprises the steps of obtaining a hand-drawn background image; performing digital processing to obtain a digital background image; segmenting the digital background image to generate a mask image of each background region, and assigning and generating a constraint type for each region, the constraint types at least comprising a static background constraint and a character / symbol stability constraint, and performing structured organization on the unique identifier of each region, the mask image and the corresponding constraint type to generate background prior information; in an AIGC video generation process, background prior information is used as a unified control framework, and differential constraints are applied to different regions in a reasoning stage. According to the method, the structured prior is provided through the hand-drawn background, the definition, stability and controllability of video background generation are effectively improved, and the unexpected linkage of the background and the foreground is reduced.

The application discloses a Tensor data processing method and device and related equipment. The method comprises the following steps: acquiring prior information of Tensor data, wherein the prior information comprises a dimension parameter and a storage address mapping rule of the Tensor data; determining each target element based on the dimension parameter of the Tensor data; converting the virtual storage address of each target element into a physical storage address by using the storage address mapping rule, wherein the physical storage addresses are continuous; storing each target element into the corresponding physical storage address; and performing a preset operation on each target element stored in the physical storage address by using a preset algorithm to obtain a processing result. Thus, the cache hit rate is improved.

The application discloses a deep learning magnetotelluric data inversion method and device based on sub-region coding, and the method comprises the following steps: dividing an inversion region of magnetotelluric data into multiple coding sub-regions; constructing a training data set according to prior information of the multiple coding sub-regions; training a deep learning model by using the training data set to generate coding with prior regions, and generating coding without prior regions by using direct mapping, so as to re-parameterize a geoelectric model and obtain a re-parameterization result; and optimizing the coding based on the re-parameterization result to realize magnetotelluric data inversion. The application can flexibly embed prior information with different complexities, degrees of certainty and spatial ranges; can adaptively improve inversion effect according to the quality of the prior information; can strengthen the reconstruction resolution in a specific region or direction; and has low complexity and low training calculation amount.

The present disclosure provides a 3D vehicle target detection method for fusing image texture features and prior information. Through pixel-level fusion of point clouds and images, the method achieves high-precision and real-time three-dimensional vehicle detection in combination with decision-level fusion of a feature distance and channel attention, improves the detection stability in a complex environment, and is suitable for autonomous driving and intelligent transportation systems.

The invention discloses an object-level semantic SLAM (Simultaneous Localization and Mapping) construction method and device based on NeRF. The method comprises the following steps: acquiring image data and camera parameter data; preprocessing the image data and the camera parameter data to obtain image preprocessing data and camera calibration parameter data; and utilizing an object-level semantic SLAM modeling model to process the image preprocessing data and the camera calibration parameter data to obtain camera pose data, object scene modeling data and scene object semantic segmentation data. According to the method, the NeRF technology is used for object modeling, and the problem that traditional object-level semantic SLAM object modeling precision is poor is solved; object semantic recognition is carried out by using an image segmentation large model SEEM, and the problem that a traditional object-level semantic SLAM cannot recognize objects except prior information is solved.

The invention relates to the technical field of security house management, in particular to a house management method, an electronic device and a storage medium, and the method comprises the steps: obtaining multi-source time sequence data which comprises the identity verification data, indoor behavior data and outdoor behavior data of a legal tenant renting a target house, extracting a digital portrait vector according to the identity verification data, the indoor behavior data and the outdoor behavior data, extracting a fusion feature vector according to the identity verification data and the indoor behavior data, and inputting the digital portrait vector and the fusion feature vector into a consistency discriminator to obtain a consistency score, and managing the target house according to the consistency score. According to the embodiment of the invention, the security house can be managed intelligently, the security house can be effectively prevented from being subleased, the security house management efficiency is improved, and the accuracy and robustness of sublease identification can be improved by constructing the consistent game network and using the personalized portrait of the tenant as the prior.

This invention discloses an intelligent tracking method for maneuvering swarm targets based on deep neural networks. This method incorporates deep learning into a Bayesian filtering framework, extracting the motion characteristics and process noise statistical characteristics of the swarm targets from measurement data using multiple deep neural networks. It then estimates the motion state transition matrix and process noise variance matrix of the swarm targets online, and uses Bayesian filtering to accurately estimate the motion state of the swarm targets' centroid. By modeling the swarm target contour as an ellipse centered at the centroid, the high-precision centroid motion state estimation results improve the contour estimation accuracy, ultimately achieving swarm target tracking. Compared to existing swarm target tracking methods, the proposed method does not require prior information to establish a target motion model, enabling high-precision tracking of non-cooperative maneuvering swarm targets in complex environments where prior information is lacking.

The application belongs to the technical field of signal processing, and particularly relates to a non-coherent scattering ionospheric parameter non-ambiguous estimation method based on Monte Carlo, aiming to solve the problems of inaccurate parameter estimation and poor robustness in ionospheric parameter inversion in the prior art. The method comprises the following steps: receiving measured echo data, estimating the signal noise level and determining the prior information combination mode; obtaining an expected index based on a signal noise threshold mapping library, and judging whether a preset decision threshold is met; if the preset decision threshold is met, starting an optimization selection strategy based on Monte Carlo, and obtaining a candidate solution set; performing cluster analysis on the candidate solution set, selecting a solution with the optimal fitting degree from the solution cluster containing the largest number of solutions, and taking the solution as the final non-ambiguous estimation value. Through systematic Monte Carlo simulation, the application can quantitatively evaluate the inversion reliability under the current data condition, effectively avoid invalid calculation under adverse conditions such as excessive noise or insufficient prior information, and improve the robustness and success rate of estimation.

This invention discloses a network path optimization method and system. The method includes: adjusting the adaptive weights of the heuristic function in the underlying model LPA* algorithm to intelligently change the magnitude of the heuristic function's influence with iteration; simultaneously, introducing a bias p in the K-value calculation to improve the algorithm's search efficiency; and performing unbalanced allocation of initial pheromones for the ant colony algorithm based on the initial path calculated by the underlying model. The upper-layer model applies the ACO algorithm to plan fiber optic network paths on a grid map to obtain the globally optimal network path. This invention, through the design of a hierarchical algorithm, provides the ant colony algorithm with prior information including pre-selected paths, enhancing the traditional ant colony algorithm's ability to guide pre-selected areas in path planning. This enables faster search for feasible paths, effectively solving the problems of slow search speed and susceptibility to local optima in the traditional ACO algorithm, and significantly improving the overall performance of the fiber optic network planning system.

This invention provides a data classification model training method, a data classification method, and an electronic device, relating to the field of image processing technology. The method applies weak and strong enhancement to each ultrasound image in the training dataset, obtaining corresponding weakly enhanced and strongly enhanced views. The weakly enhanced view is input into the primary task classifier of the student network to obtain the benign / malignant prediction probability distribution. The strongly enhanced view is input into the secondary task classifier of the teacher network to obtain the swelling prediction probability distribution. The swelling prediction probability distribution is mapped to the benign / malignant classification task space, and a cross-task consistency loss is calculated. Backpropagation optimization is performed on the student network based on the cross-task consistency loss. The teacher network parameters are updated using an exponential moving average based on the optimized student network parameters. Thus, structural prior information can be introduced as a supervisory signal through cross-task consistency constraints, improving the model's learning ability for key discriminative features.

This invention relates to the field of electromagnetic exploration technology, and discloses an efficient probabilistic inversion method, medium, and electronic equipment for electromagnetic resource exploration. The method employs a Fourier neural network from machine learning, training it with a large number of conductivity models to obtain the electromagnetic forward response. This network, through a cascaded structure of multi-layer neurons, automatically extracts features from a large amount of magnetotelluric response data, constructing a mapping from conductivity models to magnetotelluric responses. This avoids traditional forward calculations for conductivity models obtained from each sampling, accelerating the forward calculation time during sampling. Prior constraints related to the structure are constructed in advance and added to the inversion prior distribution, reconstructing the Metropolis-Hasting acceptance criterion. The prior information is used to quickly search for the posterior probability density distribution of the model, reducing the search time in the inversion space during sampling, and quantitatively evaluating the obtained inversion results.

The application relates to the field of image processing, and proposes a small sample semantic segmentation method and system fusing class label semantics, wherein a prior information generation module is designed, multi-modal data fusion of image data information and text data information serving as class labels is realized, a semantic segmentation model can more accurately understand image content, a multi-scale fusion module is further designed, original detail information of an image is further protected, the calculation performance and channel fusion capability of the semantic segmentation model are greatly improved, the parameter quantity of a decoder is greatly reduced while ensuring the speech segmentation precision, and the accuracy of target recognition and target positioning of semantic segmentation is greatly improved.

The invention discloses a reservoir parameter prediction method and device, electronic equipment and a medium. The method comprises the following steps: determining seismic attribute parameters capable of reflecting reservoir parameter essential characteristics, and predicting a seismic facies through a self-organizing competitive neural network algorithm; adding the seismic facies as transverse constraints into the low-frequency model, and constructing a phase-controlled low-frequency model; and carrying out reservoir parameter inversion through a self-adaptive pre-stack mixed domain inversion method. According to the method, low-frequency information containing seismic facies constraints is added in low-frequency model construction, in addition, in the inversion process, through self-adaption of prior information of different distributions, and meanwhile, mixed domain constraints are added in a likelihood function, the inversion result resolution and transverse continuity are improved, the reservoir description precision is greatly improved, and the method is suitable for large-scale popularization and application. And good support is provided for high-quality reservoir prediction.

The application relates to a power distribution network topology and parameter identification method under a partial node unobservable condition, and belongs to the technical field of intelligent power distribution networks; the method comprises the following steps: constructing a deep and shallow neural network model and introducing a virtual node to represent an unobservable area; dividing observable and partial observation subgraphs through sparse training; performing Bayesian randomization modeling on network weights, virtual voltages and topology variables, and introducing prior information; performing particle inference by adopting a Stein variational gradient descent algorithm, outputting edge existence probability and parameter confidence intervals; constructing a reinforcement learning search module based on a high-confidence edge set to optimize virtual node connection; finally, a connected topology and line parameters are generated, and a topology risk index and a parameter uncertainty index are output; the application can realize probabilistic identification of the topology and the parameters under the partial node unobservable condition, and improves the reliability, robustness and engineering applicability of the identification result.

The invention discloses a real-time abnormal sub-trajectory detection method based on multilevel self-supervised learning, and belongs to the technical field of spatio-temporal data mining and intelligent traffic systems. In order to solve the problems that in the prior art, fine-grained abnormal positioning is difficult and depends on source point-target point prior information, a multi-scale local encoder and a hierarchical fusion mechanism are adopted to capture hierarchical features of a track; further, random masking is carried out on the fusion embedded sequence, a context encoder and a decoder are trained to reconstruct masked fragments, and a normal behavior mode is learned under a self-supervised normal form; and finally, the reconstruction error of the track fragment is used as an abnormal score for judgment. According to the invention, the method can achieve the precise and real-time detection of local anomalies without the prior information of source points-target points (SD), and remarkably improves the detection accuracy and universality. Meanwhile, a set of novel fine granularity evaluation indexes are introduced, and a more scientific evaluation criterion is provided for field research.

The invention provides a priori knowledge-based cyclic redundancy check method, apparatus and device. The method comprises the steps of obtaining original to-be-checked data; performing period division on the original to-be-verified data, and determining the length of a to-be-complemented byte; performing preprocessing according to the original to-be-verified data and the to-be-complemented byte length to obtain intermediate data; obtaining a prior information value according to the length of the byte to be complemented and a preset prior information table; and performing cyclic redundancy check calculation on the intermediate data according to the prior information value to generate a target check result. According to the scheme, the cyclic redundancy check calculation is converted from multi-core to single-core, so that the chip area and logic resources can be saved, and the improvement of the cyclic redundancy check calculation rate and the reduction of power consumption and heat production are realized.

This invention discloses a frequency offset estimation method, apparatus, electronic device, and storage medium, relating to the field of communication technology. The method includes: receiving a baseband spread spectrum signal from a low-Earth orbit satellite to obtain a received signal; performing segmented frequency sweep compensation on the received signal to obtain a compensated measured signal; using prior information determined from the received signal to despread and descramble the compensated measured signal, and performing a coarse frequency offset estimation on the measured signal after the first despreading and descrambling; performing a second despreading and descrambling on the signal data after the coarse frequency offset estimation, and performing a fine frequency offset estimation on the signal data after the second despreading and descrambling. This invention introduces prior information to directly lock and track a specified target beam signal, significantly improving the directionality and efficiency of the measurement; by designing a low-complexity signal processing flow, it can significantly reduce the algorithm burden, ensuring rapid completion of neighboring satellite beam frequency offset estimation during service operation, meeting the real-time requirements of seamless handover.

The invention discloses a power distribution network measurement data restoration method and system considering edge enhancement and multi-scale detail restoration, and relates to the technical field of power system data processing and industrial time series data governance. Real working condition abnormity and dirty data are distinguished, and a repair mask and morphological enhancement prior information are generated; a model is generated through WGAN-GP training with mask consistency and boundary constraint, and a preliminary repair result is output through latent variable inversion; then, repairing details are optimized through edge-oriented form remodeling and multi-scale residual error weighting; and finally, performing back-check iteration by adopting a homologous statistical standard. According to the scheme, the problems of excessive smoothness and edge fault of long block repair are effectively solved, the authenticity and reliability of repair data are improved, and accurate operation and maintenance requirements of power distribution network state monitoring, scheduling optimization and the like can be met.

The invention relates to the technical field of container ship energy efficiency optimization, and discloses an ultra-large container ship energy efficiency optimization system based on intelligent navigation control, and the system comprises an environment sensing module which collects the multi-source data of the navigation environment and the self state in real time; the state prediction module is used for fusing data to construct a wave field and predicting a future waveform and ship response; the intelligent decision-making module is used for solving multi-objective optimization by using MPC based on the prediction information and generating a cooperative control instruction; the cooperative execution module analyzes the optimal instruction and converts the optimal instruction into a cooperative control signal of the propulsion and control system; and the efficiency feedback module is used for evaluating actual energy efficiency, constructing an efficiency map and feeding back and correcting the macroscopic route cost. According to the invention, a Kalman filtering framework in the state prediction module is utilized, and statistical prior information provided by a wide-area wave spectrum and observation input provided by a near-field high-precision sea surface elevation measurement value are deeply fused, so that high-precision and deterministic prediction of a local waveform sequence to be encountered by a ship is realized.