6 results about "Point density" patented technology

The invention discloses a topology enhancement prototype and geometric self-adaption few-sample point cloud semantic segmentation method, which belongs to the technical field of three-dimensional vision, and comprises the following steps of: extracting geometric features of point cloud from a support set and a query set by using a feature encoder; a topology guiding prototype optimization module is introduced, persistent coherence analysis is carried out on point clouds of each category in a support set, multi-scale topology features are extracted, topology descriptors are generated, and the topology descriptors are used for guiding structural adjustment of prototype representation; a difficult sample correction module is designed, a difficult sample area in a query set is identified based on a topology instability index, and fine correction is performed on geometric features in combination with local point density and curvature information; and constructing a topology perception optimization module. According to the method, topology priori is systematically introduced into a few-sample point cloud segmentation task for the first time, and a more stable and fine semantic segmentation result is realized under the condition of not depending on a large amount of annotations.

The present application relates to the technical field of seismic data processing, and particularly relates to a seismic data sampling method, device, equipment, medium and product, wherein the method comprises: substituting shot point data of an initial bubble point set into a point density function to obtain a point density function value of a shot point; if the point density function value of the shot point is greater than a point density threshold value, deleting the shot point to obtain an intermediate shot point set; judging whether the point density function values of all shot points of the intermediate shot point set are all less than or equal to the point density threshold value; if the point density function values of all shot points of the intermediate shot point set are all less than or equal to the point density threshold value, taking the intermediate shot point set as a uniform shot point set, and randomly sampling shot points of the uniform shot point set; the uniformity of random sampling can be improved.

It is provided a method for encoding a point cloud within a volume, the method being performed in an encoder. The method comprises: recursively splitting the volume into sub-volumes until each sub-volume satisfies a completion criterion, in which case the sub-volume is an end sub-volume; estimating, for each end sub-volume comprising points, its density of points, by selecting a density value, from a set of pre-selected density values, that is closest to the density of points of the end sub-volume; encoding the density of the sub-volumes by encoding a general mode density value, being the most common density value of the sub-volumes, and encoding density values of any sub-volumes comprising points, where the density value of the sub-volume differs from the general mode density value; and providing indications of the sub-volumes and the encoded density values. Corresponding embodiments for the decoder side are also provided.

The invention provides a high-risk operation real-time identification and millisecond-level alarm method for an operation and maintenance session, and the method comprises the steps: calculating the angle deviation between vectors in a hidden malicious behavior segment through employing cosine similarity, adjusting the center coordinate and radius expansion of a hypersphere according to an angle deviation sequence, fusing the deviation trend and point density distribution, and obtaining a high-risk operation real-time identification and millisecond-level alarm. Obtaining an expanded hyper-spherical boundary parameter; reversely adjusting the center coordinate adjustment amount and the boundary adaptive expansion amplitude of the high-dimensional curved surface space through the expanded hyper-spherical boundary parameters, and updating curvature local deformation parameters to obtain a corrected curved surface boundary model; and updating the state of the operation and maintenance session through an alarm response mechanism, obtaining an alarm confirmation mark and a response state, and adjusting the session access permission level according to the alarm confirmation mark to obtain updated session context data.

The present disclosure is a kind of based on depth datum unified variable resolution seabed terrain grid generation method and device, the method comprises: data preprocessing;Point density and slope calculation;According to the point density, the function relationship between the point density and the slope correction amount is established, and the slope is corrected;Based on the inverse distance weighted method, the corrected slope is smoothed;According to the slope threshold, the point cloud is filtered;Variable resolution triangular irregular grid is constructed.This embodiment effectively eliminates redundancy and noise by filtering repeated points of multi-source point cloud, ensures the accuracy and consistency of input data;Combined with point density and slope information, the terrain features are corrected and optimized, so that the terrain model can adaptively reflect the detail difference of different regions;Adaptive threshold screening improves the retention rate of key terrain features;Variable resolution TIN construction effectively reduces the overall data volume while maintaining important terrain details, balances modeling efficiency and storage cost.

The invention discloses a point cloud registration method adaptive to point density change. The method comprises the following steps: step 1, collecting point cloud data; step 2, point feature extraction; 3, removing point pairs based on the relative height of the points; 4, carrying out point filtering and fine registration based on probability distribution; according to the invention, large-scale point cloud registration adapting to point density change can be realized in a large-range outdoor environment with communication bandwidth change.