Denoising method for 3D point cloud data

Abstract

The invention relates to a denoising method for three-dimensional point cloud data, the method comprising: extracting depth data of three-dimensional point cloud data, establishing a two-dimensional point cloud matrix, elements in the two-dimensional point cloud matrix are depth data; extracting two-dimensional point cloud data K 3×3 sub-matrices N of the cloud matrix k ;Create submatrix N k The position index of the central element in the two-dimensional point cloud matrix; the sub-matrix N k The central element a within 2,2 respectively with a ij The absolute value of the difference is summed up and recorded as M1, where a ij is the sub-matrix N k elements within; if M1 is greater than the first threshold, determine the central element a 2,2 For the noise point, find the position of the noise point in the two-dimensional point cloud matrix according to the position index and discard the element corresponding to the noise point; if M1 is less than or equal to the first threshold value, keep the central element a 2,2 The element corresponding to the position in the two-dimensional point cloud matrix.

Description

technical field [0001] The invention relates to a data processing method, in particular to a denoising method for three-dimensional point cloud data. Background technique [0002] With the continuous development of 3D imaging technology, in recent years, a new generation of active sensors based on the time-of-flight measurement principle has been developed. The 3D point cloud data collected by the time-of-flight sensor mainly includes shot noise and dark noise, resulting in random Errors, mixed pixels, multipath reflections and scattering artifacts caused by the scene, errors caused by thermal fluctuations of the sensor system itself, lens distortion, ranging offset, and distance scale errors that have nothing to do with the scene. It is of far-reaching significance to denoise and preprocess the acquired 3D point cloud data before reconstruction. [0003] At present, the method for denoising processing of 3D point cloud data is mainly based on Gaussian noise and salt and pe...

AI Technical Summary

Problems solved by technology

[0003] At present, the method for denoising processing of 3D point cloud data is mainly based on Gaussian noise and salt and pepper noise of depth data, using Gaussian filtering and mean filtering for denoising processing. These two filtering methods are rich in processing details, including edge information. When there are more depth images, it is easy to cause distortion of image edge details, and these two filtering methods are aimed at Gaussian noise and salt and pepper noise, and the noise generated by the 3D point cloud data collected by the time-of-flight sensor is not limited to the above two , so only Gaussian filtering and mean filtering cannot filter other forms of noise well

Images