Rauschunterdrückung für Laserscannermessungen auf der Basis eines Trainierten Adaptiven Filters

Details

Veröffentlichungs-Nr.

EP4783106

Anmeldetag

23. Januar 2025

Veröffentlichung

29. Juli 2026

Rechtsraum

EP

IPC

G06T5/70, // G01S17/89

Abstract

The invention pertains to a method of reducing noise in 3D data. The method comprises receiving 3D data that has been captured by a first laser scanner with a first resolution in a first environment. The method further comprises reducing noise in the 3D data by utilizing a noise reduction filter, wherein the noise reduction filter has filtering settings. The invention is characterized in that a trained machine learning model uses the filtering settings, wherein the filtering settings are learned by the machine learning model in a training process. The training process comprises multiple steps. Firstly, it comprises receiving training data from at least two laser scanners, the training data comprising a first set and a second set of 3D data that are structured as point-clouds. Secondly, it the training process comprises clustering the second set of 3D data to produce clustered data. Furthermore, the training process comprises training the machine learning model using the clustered data.

Anmelder

Firma

LEICA GEOSYSTEMS AG

Land

🇨🇭 Schweiz

Fachgebiete