2026-05-24T09:29:24Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-1952912024-12-23T18:01:48Zoai_pmh:alloai_pmh:repo_items

195291 https://hdl.handle.net/2286/R.2.N.195291 http://rightsstatements.org/vocab/InC/1.0/ All Rights Reserved 2024 77 pages Masters Thesis Academic theses Text eng Wang, Shun-Yen Zhao, Junfeng JZ Wishart, Jeffrey JW Suo, Dajiang DS Chen, Yan YC Arizona State University Partial requirement for: M.S., Arizona State University, 2024 Field of study: Systems Engineering Simultaneous Localization and Mapping (SLAM) algorithms play a crucial role in Automated Vehicle (AV). These vehicles utilize sensors such as cameras, Light Detection and Ranging (LIDAR), and Radio Detection and Ranging (RADAR) to perceive their surroundings and use other sensor data such as Global Navigation Satellite System (GNSS) data to determine their location. High Definition (HD) maps enable automated vehicles to precisely pinpoint their location and navigate with lane-level accuracy. Additionally, HD maps provide information such as traffic light locations, lane placement, crosswalks, and more. However, manually labeling objects is a time-consuming process. Existing LIDAR SLAM algorithms struggle to visualize lane markings and road boundaries, making it difficult to accurately label lanes. Cameras have been widely used for lane marking detection, but they are sensitive to weather conditions and environmental lighting. To address these challenges, this thesis enhances the existing hdl-graph-slam algorithm by incorporating LIDAR intensity data into the point cloud map. To achieve this, the RANSAC method has been introduced to excute ground plane extraction. Next, a road marking detector based on the Otsu thresholding method is employed, which separates LIDAR point clouds into two segments: the road surface and road markings. The Otsu thresholding method has been modified for better accuracy, improving the determination of vehicle's pose and enhancing the efficiency of the algorithm. The results were compared to the result of the original hdl-graph-slam. As a result of using the modified SLAM algorithm, the road markings are visualized with an error of only approximately 0.5 meters. These enhancements contribute to the robustness and accuracy of AVs, particularly in scenarios involving lane detection and road boundary recognition. Robotics Automotive Engineering Lidar LIDAR intensity Otsu thresholding method Road Marking SLAM Enhancing HD Map Creation Using LIDAR Intensity Data for Road Marking Detection with a Simultaneous localization and Mapping