Method for generating large batch of point cloud data sets

Abstract

The invention discloses a method for generating large batch of point cloud data sets. The method comprises the following steps: establishing a three-dimensional model for an object in a scene; re-sampling points for each vertex and each patch in the three-dimensional model to form a surface point cloud of the three-dimensional model; combining the surface point clouds of components in the scene togenerate scene point cloud information; processing the scene point cloud information by adopting a simulated camera model to form simulated scene point cloud information acquired by a simulated structured light system; generating a large batch of point cloud data sets by repeating for many times. According to the method, the training data volume of a neural network for point cloud processing is improved, and meanwhile, the method has pertinence for specific scenes and processing tasks, so that the performance of the neural network is improved.

Description

technical field [0001] The invention relates to the research field of machine vision, in particular to a method for generating large batches of point cloud data sets. Background technique [0002] At present, point cloud data processing is a relatively popular field in the field of machine vision. As a kind of three-dimensional data, point cloud can provide detailed three-dimensional information for objects and environments. Its data processing is an important aspect of three-dimensional vision technology. There are many applications in the field. [0003] Inspired by deep learning and the ability of neural networks to learn features in 2D images, neural network-based point cloud processing methods are currently being developed. But the difference is that the training data sets of neural networks currently used to process two-dimensional images come from a wide range of sources, while the acquisition of three-dimensional point clouds is much more difficult than that of two-...

AI Technical Summary

Problems solved by technology

But the difference is that the training data sets of neural networks currently used to process two-dimensional images come from a wide range of sources, while the acquisition of three-dimensional point clouds is much more difficult than that of two-dimensional images. The point cloud data sets used to train neural networks Currently even more scarce

For some special application scenarios, it is more difficult to obtain point clouds

Images