CNN-Based Lidar Point Cloud De-Noising in Adverse Weather

Abstract

Lidar sensors are frequently used in environment perception for autonomous vehicles and mobile robotics to complement camera, radar, and ultrasonic sensors. Adverse weather conditions are significantly impacting the performance of lidarbased scene understanding by causing undesired measurement points that in turn effect missing detections and false positives. In heavy rain or dense fog, water drops could be misinterpreted as objects in front of the vehicle which brings a mobile robot to a full stop.In this paper, we present the first CNN-based approach to understand and filter out such adverse weather effects in point cloud data. Using a large data set obtained in controlled weather environments, we demonstrate a significant performance improvement of our method over stateof-the-art involving geometric filtering. Data is available at https://github.com/rheinzler/PointCloudDeNoising.• The first CNN-based approach to lidar point cloud

Fast and Accurate Desnowing Algorithm for LiDAR Point Clouds

“…WeatherNet can be optimized for denoising purposes by reducing the depth, inserting a dropout layer and adding a dilated convolution to the base block of the network. While WeatherNet achieves better performance than RangeNet and DROR [7], it is somewhat inappropriate to compare DROR to WeatherNet in dense fog situations because DROR was developed expressly for desnowing purposes. In dense fog situations, the performance of DROR deteriorates because of the differences in sparsity between snow and fog.…”

“…Fortunately, as the LiDAR technology continues to develop and becomes affordable, interest in using LiDAR sensors in autonomous driving applications is increasing, and research to address the shortcomings of LiDAR is also becoming more active [2] [3] [4] [5] [6]. Recently, research on snow removal has been carried out by applying deep learning [7] or improving an existing filter [8] [9]. These studies significantly improved the accuracy and speed of snow removal.…”

Simulation-Based Data Augmentation Using Physical Priors for Noise Filtering Deep Neural Network

“…There are multiple other data augmentation techniques based on class weights for imbalanced class distribution or random duplication of points proposed by (Griffiths and Boehm, 2019b) and (Qi et al, 2017). Research work for autonomous driving vehicles that deals with noise due to adverse weather conditions proposed augmentation model for fog and rain (Heinzler et al, 2019). It utilizes distance matrix, intensity matrix, extinction coefficient, and point scattering rate for the augmentation of rain and fog.…”

“…Existing sophisticated deep learning methods such as PointCleanNet (Rakotosaona et al, 2020) have dealt with sparse noise points for meshes. There are two significant studies for noise filtering due to adverse weather conditions for autonomous driving systems, which do not deal with sensor noise (Heinzler et al, 2019) and (Stanislas et al, 2019). Developing a deep neural network for noise filtering requires a thorough investigation of the diverse annotated dataset.…”

Deep Learning Method on Target Echo Signal Recognition for Obscurant Penetrating Lidar Detection in Degraded Visual Environments

“…Light detection and ranging (Lidar) sensors are widely used in detection applications, such as autonomous driving [ 1 , 2 , 3 ] and unmanned ground vehicles (UGV) [ 4 ], due to the characteristics of high resolution [ 5 ] and precision [ 6 ]. However, when they are applied in degraded visual environments (DVE), such as smoke, fog, dust, and rain, the laser signals of Lidar sensors are attenuated and absorbed by scattering effect and extinction effect of aerosols, thus, it is difficult to extract effective beat signals and detect the targets further [ 7 ].…”