18 results about "Gaussian density" patented technology

The invention relates to a monocular video three-dimensional human body high-quality reconstruction method based on Gaussian splashing and normal perception, and belongs to the field of computer aided design, graphics and computer vision. The method comprises the following steps: firstly, based on a monocular video frame in an input three-dimensional human body data set, guiding posture deformation through a time sequence, and carrying out whole body correlation deformation on a standard space posture to obtain an observation space human body posture; secondly, through normal perception Gaussian optimization, adaptive Gaussian density control and human body normal mapping supervision are carried out on the human body posture in the observation space, and a human body model of Gaussian representation is obtained; then, carrying out human body shadow feature learning on the input human body normal map by adopting an illumination model to obtain human body shadow features; and finally, light and shadow enhancement human body rendering is performed on the Gaussian representation human body model in combination with the human body light and shadow features and human body color information input into the monocular video, a human body rendering effect is output, and the quality and reality sense of three-dimensional human body reconstruction of the monocular video are improved.

The invention provides a scene modeling method and device for electric power facilities, equipment and a medium. The method comprises the following steps: acquiring a plurality of frames of real images shot on the electric power facility at a plurality of visual angles; acquiring point cloud data of the electric power facility by adopting an SFM technology, and obtaining a plurality of frames of rendering images by adopting a 3DGS algorithm; for each frame of rendered image, calculating the gradient of each Gaussian body in the rendered image through back propagation based on the difference between the rendered image and the real image under the corresponding view angle; when the gradient value of any Gaussian body is greater than a preset gradient threshold value, adjusting the Gaussian density of a space region corresponding to the Gaussian body according to a covariance matrix after projection of the Gaussian body and a texture structure type at a position corresponding to the Gaussian body in a real image corresponding to the rendered image; and obtaining a scene model of the electric power facility based on the adjusted Gaussian density. The method is used for achieving the effect of improving the modeling precision of the electric power facility.

A deep learning-based environmentally adaptive outlier detection method according to an embodiment of the present invention is a method by which a computing system comprising a memory and a processor performs deep learning-based environmentally adaptive outlier detection and comprises the steps of: acquiring a corresponding fair quality image dataset for each predetermined material; acquiring corresponding Gaussian application information for each material, on the basis of Gaussian density estimation based on the fair quality image dataset; constructing a database comprising the Gaussian application information; determining whether a predetermined inspection target image is suspected of being defective, on the basis of the database; and providing a result of whether the predetermined inspection target image is suspected of being defective, wherein the Gaussian application information is information comprising mean (μ) and covariance (Σ) data of feature vectors extracted from N (N > 1) fair quality images included in the fair quality image dataset.

The invention discloses a drug-induced autoimmune response prediction method based on oversampling and classification modeling. The method comprises the steps that a drug-induced autoimmune response data set is acquired and preprocessed; the SNN-DPC is adopted to cluster the preprocessed data set; according to the JS divergence, selecting and marking clusters with class distribution similar to global distribution, and calculating the number of samples needing to be generated for the marked clusters based on a Kulczynski index and Gaussian density; synthesizing an oversampling sample in the mark cluster through density weighted three-point interpolation; performing LOF noise removal on the synthetic samples and rejecting boundary samples to form a balanced data set; and training and selecting an optimal classifier to construct a prediction model, and outputting a prediction category and probability. According to the method, data distribution is balanced through generated high-quality minority samples, the recognition capability of a classification model on drug-induced autoimmune positive samples is effectively enhanced, the accuracy of drug immunotoxicity prediction is greatly improved, and immunotoxicity evaluation in the drug research and development process can be assisted.

The application discloses a kind of based on Gaussian splashing air-ground scene three-dimensional reconstruction method, comprising: S1, obtains the multi-source air-ground scene image containing large pitch angle aerial photograph and ground view and camera pose;S2, based on the scene sparse point cloud generated by motion recovery structure, instantiation generates initial three-dimensional space element set;S3, introduce multi-measure adaptive density control mechanism to realize the adaptive matching of Gaussian density in different scale space;S4, by camera Jacobian matrix, three-dimensional space element is projected to two-dimensional view plane, after depth ordering, along the direction of ray extraction color and opacity, by Alpha mixing formula integration output predicted two-dimensional rendering image;S5, introduce multi-dimensional geometric constraint to guarantee the consistency of physical surface;S6, after S2-S5 closed loop circulation to convergence, extract and output high-precision three-dimensional reconstruction point cloud and new view two-dimensional rendering image.

The application discloses a method for processing image degradation problems of Gaussian rendering based on gradient control, comprising the following steps: obtaining a degraded image and performing pretreatment, using a pre-trained 2D image recovery model to recover the pretreated degraded image to obtain a preliminary recovery image; using a sparse point cloud generated by a motion recovery structure to initialize a 3D Gaussian model; using the preliminary recovery image to train the 3D Gaussian model, introducing a gradient control strategy in the training process, and iteratively optimizing Gaussian parameters to solve the view inconsistency problem; using the optimized 3D Gaussian model to process, completing the processing of the degraded image and three-dimensional scene reconstruction. The application converts various types of input image degradation problems into view inconsistency, introduces a gradient control strategy to alleviate the inaccuracy of Gaussian densification and optimization caused by view inconsistency, and well adapts to various degradation scenes, while maintaining high-fidelity rendering and real-time rendering capability.

This invention discloses a robust 3D hair reconstruction method with adaptive Gaussian ellipsoids and complex lighting, relating to the fields of 3D vision and computer graphics. Addressing the problems of sparse Gaussian ellipsoid distribution, unreliable orientation supervision signals, and geometric and texture coupling interference in existing multi-view hair reconstruction methods under non-ideal lighting, this invention employs a two-stage approach. It utilizes anisotropic Gaussian units to implicitly encode hair orientation and directly transfers the photometric constraint gradient to hair nodes through hair-aligned Gaussian dual representation, achieving efficient and high-precision hair-level reconstruction. Simultaneously, an adaptive lighting preprocessing flow is constructed, using a visual language model to locate low-quality viewpoints and dynamically setting the upper limit of Gaussian density. A three-stage adaptive density strategy is used to fill geometric gaps. Decoupled two-stage optimization is employed to eliminate lighting bias and restore true hair color. Furthermore, confidence-weighted orientation loss and 3D interpolation-guided upsampling are used to improve the orientation field quality and constraint coverage.

The invention provides a pattern-layout decoupling-based stylized handwritten mathematical formula generation method, which comprises the following steps of: by taking a handwritten formula sample of a reference style and a target print body formula as input, encoding category information of mathematical symbols in the print body formula and position information corresponding to the mathematical symbols, extracting category features and overall layout features of local mathematical symbols of the standard mathematical formula; extracting style features of a reference style handwritten formula by using a style feature coding module, performing feature fusion on the style features of a reference sample and overall layout features of a standard formula through adaptive global feature fusion, and estimating layout information of the handwritten formula by using a layout generator; in combination with mathematical symbol category information and style characteristics of a standard formula, generating a handwritten track of each symbol based on a Gaussian density mixture model; integrating the position and the handwriting track of each symbol to generate a target handwriting mathematical formula; according to the method and the device, the stylized complex layout handwritten formula is generated by decoupling the character image generation and the geometric layout generation of the handwritten formula, the generation difficulty of the handwritten formula is effectively reduced, and the accuracy and the authenticity of the handwritten formula generation are improved.

The invention provides a transformer substation 3D Gaussian high-precision modeling method and device based on RTK real-time fusion and multi-scale anti-aliasing, and belongs to the technical field of transformer substation digital twinning. According to the method, through multi-source data synchronous acquisition, RTK-vision-IMU real-time pose fusion, structure perception Gaussian density distribution, multi-scale anti-aliasing optimization and GIM format output, the problems of modeling dislocation, edge aliasing, storage redundancy, poor compatibility and the like in the prior art are solved. According to the method, the modeling precision reaches 0.5 cm, the edge SSIM is improved by 43%, the storage capacity is reduced by 65%, the method can be directly connected with an electric power PMS, 0.2 mm-level defect recognition and intelligent operation and maintenance are supported, and the method has remarkable engineering practical value.

This invention relates to the field of underwater positioning technology, specifically to an AUV cooperative positioning method based on generalized maximum correlation entropy and capacitive Kalman filtering (CKF) that can effectively improve positioning accuracy. This invention employs GMCC and introduces a generalized Gaussian density kernel function to measure the error of the nonlinear filtering algorithm using generalized correlation entropy. The generalized Gaussian kernel function has the advantages of having many parameters, flexible variation, and adaptability. It can achieve the maximum correlation entropy when the error is minimized. Because the generalized Gaussian kernel contains even-order higher moments of the errors of two variables, it can better handle heavy-tailed noise and large outliers. Experimental verification shows that the technical solution of this invention has better robustness and reliability, and can better handle problems such as outliers and heavy-tailed noise interference. Compared with the traditional CKF, this invention can improve positioning accuracy.

The invention is applicable to the field of computer vision, and provides a point cloud shape editing method based on position coding, which comprises the following steps of: in a point cloud shape editing process, mapping a three-dimensional coordinate of a point cloud to a high-dimensional space to generate a high-dimensional position code; dynamically adjusting the high-dimensional position code by using a deformation network formed by a lightweight standard multi-layer perceptron (MLP); an image obtained through rendering after position optimization is subjected to image consistency loss based on a pre-training model and a loss function based on sparse point cloud geometric prior, and the geometric structure in the deformation process of the image is effectively restrained; a geometric perception Gaussian density control strategy is introduced, and the shape editing precision is improved. According to the method, when geometric shape changes of a complex point cloud editing task are processed, high accuracy and stability are shown, and an efficient and convenient solution is provided for creation and editing of digital content.

The invention discloses a ship course and speed collision risk heat map generation method and system based on multi-source fusion and application, and the method comprises the steps: firstly fusing AIS, radar and electronic chart data, obtaining dynamic and static obstacle information, and calculating a multi-dimensional collision risk parameter set including the minimum encounter distance, the minimum encounter time, the prow passing distance and the prow passing time; then, constructing a danger course set of each target based on a safety threshold value, and fusing to obtain a total space danger course set; thirdly, introducing parameters representing ship maneuvering performance to construct a dynamic risk field model based on an annular Gaussian density function, calculating a cumulative probability on a danger course interval as a risk index, and expanding the risk index to a speed dimension to generate a space-time risk coefficient set; and finally, generating an omni-directional dynamic risk thermodynamic diagram taking the course and the speed as coordinate axes. According to the invention, visual, comprehensive and dynamic navigation risk situation assessment and decision support is provided for the ship.

This invention provides a SAR image aircraft detection method and system based on fuzzy loss, relating to the field of radar detection technology. The fuzzy loss-based SAR image aircraft detection method includes: defining the aircraft target using a two-dimensional non-normalized Gaussian density function; using an SFP structure designed based on FPN, firstly using DSM and AEM to suppress background interference and employing an attention mechanism for feature enhancement; then, fusing shallow detail information with deep, rich semantic features; finally, fusing features at different scales to obtain a comprehensive feature; and based on the extracted comprehensive feature, combining ADL and Floss loss functions to obtain the detection result through a detection head. This fuzzy loss-based SAR image aircraft detection method effectively solves the problem of insufficient aircraft detection accuracy in existing related technologies.

The invention relates to the technical field of anomaly detection, in particular to an anomaly detection method based on semi-nonnegative matrix factorization and Gaussian estimation, and the method comprises the steps: S10, dividing a normal data set X into N sub-data sets Xi (1 < = i < = N), inputting the N sub-data sets Xi into a joint anomaly detection network W for training, and outputting an anomaly detection model M; s20, the joint anomaly detection network W comprises a semi-nonnegative matrix factorization network and a Gaussian density estimation network; s30, the semi-nonnegative matrix factorization network extracts a feature data set Ci and decomposes the feature data set Ci into a basis matrix F and a coefficient matrix G; s40, reconstructing the Ci, and outputting a reconstructed feature data set Ci '; s50, calculating a mean value parameter and a covariance parameter by the Gaussian density estimation network; s60, outputting a logarithm probability density value p; s70, based on S10-S60, finding out n quantiles in p, and taking the n quantiles as a threshold value delta; and S80, inputting to-be-detected data y into the anomaly detection model M, and outputting an anomaly detection result R. According to the invention, Gaussian density estimation can be carried out more accurately, and the anomaly detection effect is improved.

The invention provides a data anomaly detection method and device and electronic equipment, and the method comprises the steps: constructing a local Gaussian density function of a first node based on a local first data set of the first node, the first node being any node in a plurality of nodes; sending the local Gaussian density function of the first node to a block chain, wherein the block chain is used for storing the local Gaussian density functions of the plurality of nodes; the contract address of the local Gaussian density function of the first node in the block chain is sent to a center node, a global Gaussian density function sent by the center node is received, and the mean value parameter of the global Gaussian density function is the mean value of the mean value parameters of the local Gaussian density functions of the multiple nodes obtained from the block chain; the covariance parameter of the global Gaussian density function is the average value of the covariance parameters of the local Gaussian density functions of the plurality of nodes; and performing anomaly detection on the first data set based on the global Gaussian density function to obtain a detection result of the first data set so as to improve data anomaly detection accuracy.

This invention discloses a method, system, and application for generating collision risk heatmaps for ship course and speed based on multi-source fusion. The method first fuses AIS, radar, and electronic chart data to acquire dynamic and static obstacle information and calculates a multi-dimensional collision risk parameter set, including minimum encounter distance, minimum encounter time, distance past the bow, and time past the bow. Then, based on safety thresholds, a set of dangerous course profiles for each target is constructed and fused to obtain a total set of dangerous course profiles across all space. Next, parameters characterizing ship maneuverability are introduced to construct a dynamic risk field model based on a ring Gaussian density function. The cumulative probability across dangerous course intervals is calculated as a risk index and extended to the speed dimension to generate a set of spatiotemporal risk coefficients. Finally, an omnidirectional dynamic risk heatmap with course and speed as coordinate axes is generated. This invention provides ships with intuitive, comprehensive, and dynamic navigation risk situation assessment and decision support.

The invention discloses a gradient control-based Gaussian rendering method for processing an image degradation problem, which comprises the following steps of: acquiring a degraded image and preprocessing the degraded image, and restoring the preprocessed degraded image by using a pre-trained 2D image restoration model to obtain a preliminary restored image; initializing a 3D Gaussian model by using a sparse point cloud generated by the motion recovery structure; training a 3D Gaussian model by adopting the preliminary recovery image, introducing a gradient control strategy in the training process, and iteratively optimizing Gaussian parameters to solve the problem of view inconsistency; and the optimized 3D Gaussian model is adopted to process the degraded image, and processing of the degraded image and reconstruction of a three-dimensional scene are completed. According to the method, various types of input image degradation problems are converted into view inconsistency, and a gradient control strategy is introduced to relieve inaccuracy of Gaussian densification and optimization caused by view inconsistency, so that the method is well adapted to various degradation scenes, and the real-time rendering capability is also maintained while high-fidelity rendering is maintained.