Efficient surveying and mapping system based on point cloud data

Abstract

The invention relates to the technical field of artificial intelligence, in particular to an efficient surveying and mapping system based on point cloud data. The system includes a basic point cloud acquisition unit for collecting initial point cloud data, and inputting the initial point cloud data into the basic point cloud extraction network to obtain basic point cloud data; a noise-adding point cloud acquisition unit for inputting the basic point cloud data The trained noise point cloud generation network obtains the noise point cloud data; the denoising network construction unit is used to train the denoising neural network using the noise point cloud data and the basic point cloud data; the collected target point cloud data input The trained denoising neural network obtains denoised point cloud data, and performs surveying and mapping based on the denoised point cloud data. The generalization ability of the denoising neural network under different working conditions is enhanced, and there is no need to manually label data, which reduces costs and improves surveying and mapping efficiency.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to an efficient surveying and mapping system based on point cloud data. Background technique [0002] The original point cloud data often contains a large number of hash points and isolated points. When obtaining point cloud data, due to the influence of equipment accuracy, operator experience and environmental factors, diffraction of electromagnetic waves, changes in the surface properties of the measured object, and data splicing Due to the influence of the quasi-operation process, some noise will inevitably appear in the point cloud data, resulting in different sampling resolutions. [0003] In order to achieve efficient surveying and mapping of point cloud data, the original point cloud data obtained directly is denoised and all kinds of noise points are eliminated. The existing commonly used processing methods are usually point cloud filtering operations, includ...

AI Technical Summary

Problems solved by technology

[0002] The original point cloud data often contains a large number of hash points and isolated points. When obtaining point cloud data, due to the influence of equipment accuracy, operator experience and environmental factors, diffraction of electromagnetic waves, changes in the surface properties of the measured object, and data splicing Due to the impact of the quasi-operation process, some noise will inevitably appear in the point cloud data, resulting in different sampling resolutions

[0003] In order to achieve efficient surveying and mapping of point cloud data, the original point cloud data obtained directly is denoised and all kinds of noise points are eliminated. The existing commonly used processing methods are usually point cloud filtering operations, including bilateral filtering, Gaussian filtering, and straight-through filtering. Filtering, random sampling consistency filtering and other filtering operations, but the problem with filtering operations is that it is necessary to select an appropriate filtering method according to the working conditions, which greatly increases the complexity of the point cloud denoising process

[0004] Those skilled in the art find that the existing technology has the following defects: the generalization ability of the neural network is insufficient, if there are multiple types of noise, it is difficult to achieve a good denoising effect; and the data labeling of the neural network is difficult, and the results obtained by using different filtering operations The filtering results are different, so it is difficult to have a definite label

Images