DSOR: A Scalable Statistical Filter for Removing Falling Snow from LiDAR Point Clouds in Severe Winter Weather

Abstract

For autonomous vehicles to viably replace human drivers they must contend with inclement weather. Falling rain and snow introduce noise in LiDAR returns resulting in both false positive and false negative object detections. In this article we introduce the Winter Adverse Driving dataSet (WADS) collected in the snow belt region of Michigan's Upper Peninsula. WADS is the first multi-modal dataset featuring dense point-wise labeled sequential LiDAR scans collected in severe winter weather; weather that would cause an experienced driver to alter their driving behavior. We have labelled and will make available over 7 GB or 3.6 billion labelled LiDAR points out of over 26 TB of total LiDAR and camera data collected. We also present the Dynamic Statistical Outlier Removal (DSOR) filter, a statistical PCL-based filter capable or removing snow with a higher recall than the state of the art snow de-noising filter while being 28% faster. Further, the DSOR filter is shown to have a lower time complexity compared to the state of the art resulting in an improved scalability.Our labeled dataset and DSOR filter will be made available at https://bitbucket.org/autonomymtu/dsor filter.

3D ToF LiDAR in Mobile Robotics: A Review

“…JI-IL et al [63] removed snow particles based on the intensity between laser points to improve speed and accuracy of existing distancebased filters. Kurup et al [64] proposed a PCL-based filter for snow conditions by calculating the outlier threshold with k-nearest neighbor searching. 4) Datasets: Some well-known datasets, such as KITTI [16], Waymo [119], and nuScenes [20], have played a recognized role in promoting the development of outdoor robotics and autonomous driving.…”

“…Our previous work in this regard includes the EU long-term dataset [6], that was collected with four LiDARs, a front-facing and a rear-facing stereo camera, two side-pointing fisheye cameras, a long range radar, a GNSS-RTK and an IMU, across different seasons and weather conditions. Winter Adverse Driving dataSet (WADS) [64] was collected in the snow belt region in Michigan, USA. Point-wise segmentation labels of 22 classes were provided, including the noises of point clouds in severe winter weather.…”

Filter Methods for Removing Falling Snow from Light Detection and Ranging Point Clouds in Snowy Weather

“…On the basis of LIOR, Zhong et al (10) optimized the ROR to the dynamic radius outlier removal (DROR), which further improved the effectiveness of the filtering method. Kurup et al (11) proposed the dynamic statistical outlier removal (DSOR), which optimizes the SOR method by adaptively adjusting the standard deviation threshold in the SOR method as the LiDAR detection distance increases, which further improves the filtering accuracy compared with the DROR. Because it has low time complexity, there is space for further optimization in the future.…”

4DenoiseNet: Adverse Weather Denoising From Adjacent Point Clouds

“…Digital Object Identifier 10.1109/LRA.2022.3227863 localization, object detection, and navigation. Previous work uses either a classical approach [16], [17], [18], [19], [20] or a learned approach [21]. Unlike previous work, our work utilizes spatial-temporal data as an input to a neural network.…”

“…The problem of removing noise caused by adverse weather is not trivial because of the nature of the LiDAR sensor, the sparsity of the point cloud, occlusions caused by the noise, and the varying density of the noise. This leads to statistical or hard-coded filters [16], [17], [18], [19], [20] to remove valid points and preserve the clutter. To address this problem, we use a deep learning architecture that learns an equivariance function.…”