Point cloud semantic segmentation method for large part of large aircraft

Abstract

The invention discloses a large-scale aircraft large component point cloud semantic segmentation method, which comprises the following steps: obtaining whole machine actual measurement point cloud data of a large-scale aircraft, and carrying out k neighbor graph modeling to obtain a k neighbor graph; constructing a pyramid dynamic graph convolution network, performing multi-scale second-order random walk sampling on each walk kernel convolution module in the pyramid dynamic graph convolution network through a k neighbor graph, and performing Fisher vector coding in combination with a Gaussian mixture model to obtain unified feature coding representation of a multi-scale graph; and inputting actually measured point cloud data into the pyramid dynamic graph convolutional network, performing multi-scale point cloud feature extraction, sequentially inserting high-scale point cloud features back into low-scale point cloud features, then performing point cloud feature conversion, and converting the point cloud features into a semantic segmentation result of the large aircraft component. According to the method, precise large-scale point cloud semantic segmentation is realized.

Description

technical field [0001] The invention relates to the field of three-dimensional point cloud model detection, in particular to a point cloud semantic segmentation method for large parts of a large aircraft. Background technique [0002] Shape analysis is an indispensable part of large aircraft processing, assembly and entry maintenance. Due to the weak rigidity of the large-scale grading panel itself and the design of the conformal tooling, when the large-sized thin-walled parts are processed on the conformal tooling, Distortion occurs at the butt edges. Excessive deformation will cause irregular fluctuations in the skin of large aircraft, that is, ripples appear on the surface of the body. If the excessive deformation is near the nose pitot tube, it will directly affect the flow field near the static pressure probe, resulting in deviations in the height, pressure and other data measured by the static pressure probe. When the ripple is serious, it will affect the accuracy of ...

AI Technical Summary

Problems solved by technology

If the excessive deformation is near the pitot tube of the machine head, it will directly affect the flow field near the static pressure probe, causing deviations in the data such as height and pressure measured by the static pressure probe. When the ripple is serious, it will affect the accuracy of the data characteristics, causing misjudgment by pilots and affecting flight safety. Therefore, it is necessary to analyze the shape deformation of large aircraft

[0003] Large-scale aircraft point cloud data can reach hundreds of millions of levels, and has the characteristics of disorder, which makes it difficult to use regular methods such as voxel or projection to represent and learn

The geometric features of large aircraft and large parts have a large range of scale changes in the point cloud representation. The existing point cloud semantic segmentation network is limited by the representation of neighborhood features, and it is difficult to capture large-scale and long-distance semantic dependencies.

Images