scholarly journals A Deep Learning-Based Perception Algorithm Using 3D LiDAR for Autonomous Driving: Simultaneous Segmentation and Detection Network (SSADNet)

2020 ◽  
Vol 10 (13) ◽  
pp. 4486 ◽  
Author(s):  
Yongbeom Lee ◽  
Seongkeun Park
Keyword(s):  
Deep Learning ◽  
Point Cloud ◽  
Autonomous Driving ◽  
Light Detection And Ranging ◽  
Point Cloud Data ◽  
Cloud Data ◽  
3D Lidar ◽  
Top View ◽  
Single Neural Network ◽  
Reflection Intensity

In this paper, we propose a deep learning-based perception method in autonomous driving systems using a Light Detection and Ranging(LiDAR) point cloud data, which is called a simultaneous segmentation and detection network (SSADNet). SSADNet can be used to recognize both drivable areas and obstacles, which is necessary for autonomous driving. Unlike the previous methods, where separate networks were needed for segmentation and detection, SSADNet can perform segmentation and detection simultaneously based on a single neural network. The proposed method uses point cloud data obtained from a 3D LiDAR for network input to generate a top view image consisting of three channels of distance, height, and reflection intensity. The structure of the proposed network includes a branch for segmentation and a branch for detection as well as a bridge connecting the two parts. The KITTI dataset, which is often used for experiments on autonomous driving, was used for training. The experimental results show that segmentation and detection can be performed simultaneously for drivable areas and vehicles at a quick inference speed, which is appropriate for autonomous driving systems.

scholarly journals Review: Deep Learning on 3D Point Clouds

2020 ◽  
Vol 12 (11) ◽  
pp. 1729 ◽  
Author(s):  
Saifullahi Aminu Bello ◽  
Shangshu Yu ◽  
Cheng Wang ◽  
Jibril Muhmmad Adam ◽  
Jonathan Li
Keyword(s):  
Deep Learning ◽  
Point Cloud ◽  
General Structure ◽  
Point Clouds ◽  
Autonomous Driving ◽  
Point Cloud Data ◽  
3D Vision ◽  
Cloud Data ◽  
3D Point Clouds ◽  
Learning Techniques

A point cloud is a set of points defined in a 3D metric space. Point clouds have become one of the most significant data formats for 3D representation and are gaining increased popularity as a result of the increased availability of acquisition devices, as well as seeing increased application in areas such as robotics, autonomous driving, and augmented and virtual reality. Deep learning is now the most powerful tool for data processing in computer vision and is becoming the most preferred technique for tasks such as classification, segmentation, and detection. While deep learning techniques are mainly applied to data with a structured grid, the point cloud, on the other hand, is unstructured. The unstructuredness of point clouds makes the use of deep learning for its direct processing very challenging. This paper contains a review of the recent state-of-the-art deep learning techniques, mainly focusing on raw point cloud data. The initial work on deep learning directly with raw point cloud data did not model local regions; therefore, subsequent approaches model local regions through sampling and grouping. More recently, several approaches have been proposed that not only model the local regions but also explore the correlation between points in the local regions. From the survey, we conclude that approaches that model local regions and take into account the correlation between points in the local regions perform better. Contrary to existing reviews, this paper provides a general structure for learning with raw point clouds, and various methods were compared based on the general structure. This work also introduces the popular 3D point cloud benchmark datasets and discusses the application of deep learning in popular 3D vision tasks, including classification, segmentation, and detection.

scholarly journals Multi-Dimensional Underwater Point Cloud Detection Based on Deep Learning

Sensors ◽  
2021 ◽  
Vol 21 (3) ◽  
pp. 884
Author(s):  
Chia-Ming Tsai ◽  
Yi-Horng Lai ◽  
Yung-Da Sun ◽  
Yu-Jen Chung ◽  
Jau-Woei Perng
Keyword(s):  
Deep Learning ◽  
Point Cloud ◽  
Three Dimensional ◽  
Point Clouds ◽  
Training Data ◽  
Network Architectures ◽  
Point Cloud Data ◽  
Data Types ◽  
Raw Data ◽  
Cloud Data

Numerous sensors can obtain images or point cloud data on land, however, the rapid attenuation of electromagnetic signals and the lack of light in water have been observed to restrict sensing functions. This study expands the utilization of two- and three-dimensional detection technologies in underwater applications to detect abandoned tires. A three-dimensional acoustic sensor, the BV5000, is used in this study to collect underwater point cloud data. Some pre-processing steps are proposed to remove noise and the seabed from raw data. Point clouds are then processed to obtain two data types: a 2D image and a 3D point cloud. Deep learning methods with different dimensions are used to train the models. In the two-dimensional method, the point cloud is transferred into a bird’s eye view image. The Faster R-CNN and YOLOv3 network architectures are used to detect tires. Meanwhile, in the three-dimensional method, the point cloud associated with a tire is cut out from the raw data and is used as training data. The PointNet and PointConv network architectures are then used for tire classification. The results show that both approaches provide good accuracy.

scholarly journals LiDAR Point Cloud Recognition and Visualization with Deep Learning for Overhead Contact Inspection

Sensors ◽  
2020 ◽  
Vol 20 (21) ◽  
pp. 6387 ◽  
Author(s):  
Xiaohan Tu ◽  
Cheng Xu ◽  
Siping Liu ◽  
Shuai Lin ◽  
Lipei Chen ◽  
...  
Keyword(s):  
Deep Learning ◽  
Computational Complexity ◽  
Point Cloud ◽  
High Speed ◽  
Point Cloud Data ◽  
Embedded Devices ◽  
Cloud Data ◽  
Manual Inspection ◽  
Order Of Magnitude ◽  
Supply Systems

As overhead contact (OC) is an essential part of power supply systems in high-speed railways, it is necessary to regularly inspect and repair abnormal OC components. Relative to manual inspection, applying LiDAR (light detection and ranging) to OC inspection can improve efficiency, accuracy, and safety, but it faces challenges to efficiently and effectively segment LiDAR point cloud data and identify catenary components. Recent deep learning-based recognition methods are rarely employed to recognize OC components, because they have high computational complexity, while their accuracy needs to be improved. To track these problems, we first propose a lightweight model, RobotNet, with depthwise and pointwise convolutions and an attention module to recognize the point cloud. Second, we optimize RobotNet to accelerate its recognition speed on embedded devices using an existing compilation tool. Third, we design software to facilitate the visualization of point cloud data. Our software can not only display a large amount of point cloud data, but also visualize the details of OC components. Extensive experiments demonstrate that RobotNet recognizes OC components more accurately and efficiently than others. The inference speed of the optimized RobotNet increases by an order of magnitude. RobotNet has lower computational complexity than other studies. The visualization results also show that our recognition method is effective.

Measuring Streetscape Features with High-Density Aerial Light Detection and Ranging

2020 ◽  
Vol 2674 (11) ◽  
pp. 192-206
Author(s):  
Yaneev Golombek ◽  
Wesley E. Marshall
Keyword(s):  
Point Cloud ◽  
High Density ◽  
Light Detection And Ranging ◽  
Quality Level ◽  
United States Geological Survey ◽  
Point Cloud Data ◽  
Street Trees ◽  
Light Detection ◽  
Cloud Data ◽  
Traffic Lights

This study investigates the feasibility of extracting streetscape features from high-density United States Geological Survey (USGS) quality level 1 (QL1) light detection and ranging (LiDAR) and quantifying the features into three-dimensional (3D) volumetric pixel (voxel) zones. As the USGS embarks on a national LiDAR database with the goal of collecting LiDAR across the continuous U.S.A., the USGS primarily requires QL2 or QL1 as a collection standard. The authors’ previous study thoroughly investigated the limits of extracting streetscape features with QL2 data, which was primarily limited to buildings and street trees. Recent studies published by other researchers that utilize advanced digital mapping techniques for streetscape measuring acknowledge that most features outside of buildings and street trees are too small to detect. QL1 data, however, is four times denser than QL2 data. This study divides streetscapes into Thiessen proximal polygons, sets voxel parameters, classifies QL1 LiDAR point cloud data, and computes quantitative statistics where classified point cloud data intersects voxels within the streetscape polygons. It demonstrates how most other common streetscape features are detectable in a standard urban QL1 dataset. In addition to street trees and buildings, one can also legitimately extract and statistically quantify walls, fences, landscape vegetation, light posts, traffic lights, power poles, power lines, street signs, and miscellaneous street furniture. Furthermore, as these features are processed into their appropriate voxel height zones, this study introduces a new methodology for obtaining comprehensive tabular descriptive statistics describing how streetscape features are truly represented in 3D.

scholarly journals TAILORED FEATURES FOR SEMANTIC SEGMENTATION WITH A DGCNN USING FREE TRAINING SAMPLES OF A COLORED AIRBORNE POINT CLOUD

2020 ◽  
Vol XLIII-B2-2020 ◽  
pp. 339-346
Author(s):  
E. Widyaningrum ◽  
M. K. Fajari ◽  
R. C. Lindenbergh ◽  
M. Hahn
Keyword(s):  
Deep Learning ◽  
Point Cloud ◽  
Semantic Segmentation ◽  
Blue Color ◽  
Point Cloud Data ◽  
Dynamic Graph ◽  
Cloud Data ◽  
Training Samples ◽  
Map Generation ◽  
Fast Development

Abstract. Automation of 3D LiDAR point cloud processing is expected to increase the production rate of many applications including automatic map generation. Fast development on high-end hardware has boosted the expansion of deep learning research for 3D classification and segmentation. However, deep learning requires large amount of high quality training samples. The generation of training samples for accurate classification results, especially for airborne point cloud data, is still problematic. Moreover, which customized features should be used best for segmenting airborne point cloud data is still unclear. This paper proposes semi-automatic point cloud labelling and examines the potential of combining different tailor-made features for pointwise semantic segmentation of an airborne point cloud. We implement a Dynamic Graph CNN (DGCNN) approach to classify airborne point cloud data into four land cover classes: bare-land, trees, buildings and roads. The DGCNN architecture is chosen as this network relates two approaches, PointNet and graph CNNs, to exploit the geometric relationships between points. For experiments, we train an airborne point cloud and co-aligned orthophoto of the Surabaya city area of Indonesia to DGCNN using three different tailor-made feature combinations: points with RGB (Red, Green, Blue) color, points with original LiDAR features (Intensity, Return number, Number of returns) so-called IRN, and points with two spectral colors and Intensity (Red, Green, Intensity) so-called RGI. The overall accuracy of the testing area indicates that using RGB information gives the best segmentation results of 81.05% while IRN and RGI gives accuracy values of 76.13%, and 79.81%, respectively.

scholarly journals CLASSIFICATION OF LIDAR DATA FOR GENERATING A HIGH-PRECISION ROADWAY MAP

2016 ◽  
Vol XLI-B3 ◽  
pp. 251-254 ◽  
Author(s):  
J. Jeong ◽  
I. Lee
Keyword(s):  
Machine Learning ◽  
Point Cloud ◽  
Three Dimensional ◽  
Autonomous Driving ◽  
Learning Objects ◽  
Support Vector ◽  
Feature Descriptor ◽  
Point Cloud Data ◽  
Cloud Data ◽  
On The Road

Generating of a highly precise map grows up with development of autonomous driving vehicles. The highly precise map includes a precision of centimetres level unlike an existing commercial map with the precision of meters level. It is important to understand road environments and make a decision for autonomous driving since a robust localization is one of the critical challenges for the autonomous driving car. The one of source data is from a Lidar because it provides highly dense point cloud data with three dimensional position, intensities and ranges from the sensor to target. In this paper, we focus on how to segment point cloud data from a Lidar on a vehicle and classify objects on the road for the highly precise map. In particular, we propose the combination with a feature descriptor and a classification algorithm in machine learning. Objects can be distinguish by geometrical features based on a surface normal of each point. To achieve correct classification using limited point cloud data sets, a Support Vector Machine algorithm in machine learning are used. Final step is to evaluate accuracies of obtained results by comparing them to reference data The results show sufficient accuracy and it will be utilized to generate a highly precise road map.

Creating a 3D Cuboid Map Using Multi-Layer 3D LIDAR with a Swing Mechanism

2018 ◽  
Vol 30 (4) ◽  
pp. 523-531 ◽  
Author(s):  
Yoshihiro Takita ◽  
Keyword(s):  
Coordinate Transformation ◽  
Point Cloud ◽  
Autonomous Robots ◽  
Point Cloud Data ◽  
High Definition ◽  
Cloud Data ◽  
Occupancy Grid ◽  
Occupancy Grid Maps ◽  
3D Lidar ◽  
Local Map

This paper proposes a method for creating 3D occupancy grid maps using multi-layer 3D LIDAR and a swing mechanism termed Swing-LIDAR. The method using Swing-LIDAR can acquire 10 times more data at a stopping position than a method that does not use Swing-LIDAR. High-definition and accurate terrain information is obtained by a coordinate transformation of the acquired data compensated for by the measured orientation of the system. In this study, we develop a method to create 3D grid maps for autonomous robots using Swing-LIDAR. To validate the method, AR Skipper is run on the created maps that are used to obtain point cloud data without a swing mechanism, and 11 sets of each local map are combined. The experimental results exhibit the differences among the maps.

Sign in / Sign up

Export Citation Format

Share Document