# Dust Filtering in LIDAR Point Clouds Using Deep Learning for Mining Applications

## TL;DR

This paper introduces a deep learning method to filter dust from LIDAR data in mining environments, improving sensor accuracy.

## Contribution

A novel neural network-based approach for real-time dust filtering in LIDAR point clouds is proposed and validated with real data.

## Key findings

- The proposed method achieves state-of-the-art results in dust filtering from LIDAR point clouds.
- A public database of LIDAR data from dusty environments is created to support future research.

## Abstract

In the domain of mining and mineral processing, LIDAR sensors are employed to obtain precise three-dimensional measurements of the surrounding environment. However, the functionality of these sensors is hindered by the dust produced by mining operations. In order to address this problem, a neural network-based method is proposed. This method is capable of filtering dust measurements in real time from point clouds obtained using LIDARs. The proposed method is trained and validated using real data, yielding results that are at the forefront of the field. Furthermore, a public database is constructed using LIDAR sensor data from diverse dusty environments. The database is made public for use in the training and benchmarking of dust filtering methods.