“…In this paper, we present a simple yet effective framework for improving the robustness of object detectors in rainy conditions. Our framework uses a state-of-the-art adverse weather detection network [1] to identify and remove spray points in a LiDAR point cloud. The filtered data is then used as input to an object detector trained only on good weather data.…”
Section: Ground Truth Predictionsmentioning
confidence: 99%
“…As many autonomous driving application systems include both LiDAR and radar sensors, we explore the use of radar targets as an additional post-processing step to filter out ghost detections. We test our framework on the SemanticSpray dataset [1], which contains highwaylike scenarios in rainy conditions, and show that it improves the robustness of the evaluated object detectors to spray. Furthermore, since we do not require the re-training of the object detectors, performance in good weather conditions remains unchanged.…”
Section: Ground Truth Predictionsmentioning
confidence: 99%
“…Kurup et al [17] improve on this concept by proposing the DSOR filter, which incorporates the mean distance between neighbors during filtering. Piroli et al [3] detected vehicle exhaust by identifying the possible emission area for each vehicle in the scene and then finding regions in the point cloud where exhaust clouds are likely to be present. Heinzler et al [18] use a weather chamber to generate artificial fog and rain, and then use a lightweight CNN network to classify the LiDAR points associated with adverse weather.…”
Section: B Lidar Perception In Adverse Weathermentioning
confidence: 99%
“…LiDAR-based perception systems like semantic segmentation and 3D object detection perform extremely well in good weather conditions. However, their performance is seen to degrade when testing these models in adverse weather like snow, fog, and rain [1], [2], [3], [4], [5], [6], [7]. In this paper, we will focus on the effect of vehicle road spray on 3D object detection.…”
Section: Introductionmentioning
confidence: 99%
“…1. In the literature, 1 Institute of Measurement, Control, and Microtechnology, Ulm University, Germany {firstname.lastname}@uni-ulm.de 2 BMW AG, Petuelring 130, 80809 Munich, Germany {vinzenz.dallabetta, marc.walessa}@bmw.de and daniel.da.meissner@bmwgroup.com…”
LiDAR sensors are used in autonomous driving applications to accurately perceive the environment. However, they are affected by adverse weather conditions such as snow, fog, and rain. These everyday phenomena introduce unwanted noise into the measurements, severely degrading the performance of LiDAR-based perception systems. In this work, we propose a framework for improving the robustness of LiDARbased 3D object detectors against road spray. Our approach uses a state-of-the-art adverse weather detection network to filter out spray from the LiDAR point cloud, which is then used as input for the object detector. In this way, the detected objects are less affected by the adverse weather in the scene, resulting in a more accurate perception of the environment. In addition to adverse weather filtering, we explore the use of radar targets to further filter false positive detections. Tests on real-world data show that our approach improves the robustness to road spray of several popular 3D object detectors.
“…In this paper, we present a simple yet effective framework for improving the robustness of object detectors in rainy conditions. Our framework uses a state-of-the-art adverse weather detection network [1] to identify and remove spray points in a LiDAR point cloud. The filtered data is then used as input to an object detector trained only on good weather data.…”
Section: Ground Truth Predictionsmentioning
confidence: 99%
“…As many autonomous driving application systems include both LiDAR and radar sensors, we explore the use of radar targets as an additional post-processing step to filter out ghost detections. We test our framework on the SemanticSpray dataset [1], which contains highwaylike scenarios in rainy conditions, and show that it improves the robustness of the evaluated object detectors to spray. Furthermore, since we do not require the re-training of the object detectors, performance in good weather conditions remains unchanged.…”
Section: Ground Truth Predictionsmentioning
confidence: 99%
“…Kurup et al [17] improve on this concept by proposing the DSOR filter, which incorporates the mean distance between neighbors during filtering. Piroli et al [3] detected vehicle exhaust by identifying the possible emission area for each vehicle in the scene and then finding regions in the point cloud where exhaust clouds are likely to be present. Heinzler et al [18] use a weather chamber to generate artificial fog and rain, and then use a lightweight CNN network to classify the LiDAR points associated with adverse weather.…”
Section: B Lidar Perception In Adverse Weathermentioning
confidence: 99%
“…LiDAR-based perception systems like semantic segmentation and 3D object detection perform extremely well in good weather conditions. However, their performance is seen to degrade when testing these models in adverse weather like snow, fog, and rain [1], [2], [3], [4], [5], [6], [7]. In this paper, we will focus on the effect of vehicle road spray on 3D object detection.…”
Section: Introductionmentioning
confidence: 99%
“…1. In the literature, 1 Institute of Measurement, Control, and Microtechnology, Ulm University, Germany {firstname.lastname}@uni-ulm.de 2 BMW AG, Petuelring 130, 80809 Munich, Germany {vinzenz.dallabetta, marc.walessa}@bmw.de and daniel.da.meissner@bmwgroup.com…”
LiDAR sensors are used in autonomous driving applications to accurately perceive the environment. However, they are affected by adverse weather conditions such as snow, fog, and rain. These everyday phenomena introduce unwanted noise into the measurements, severely degrading the performance of LiDAR-based perception systems. In this work, we propose a framework for improving the robustness of LiDARbased 3D object detectors against road spray. Our approach uses a state-of-the-art adverse weather detection network to filter out spray from the LiDAR point cloud, which is then used as input for the object detector. In this way, the detected objects are less affected by the adverse weather in the scene, resulting in a more accurate perception of the environment. In addition to adverse weather filtering, we explore the use of radar targets to further filter false positive detections. Tests on real-world data show that our approach improves the robustness to road spray of several popular 3D object detectors.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.