LIDAR-BASED THREE-DIMENSIONAL OBJECT DETECTION FOR AUTONOMOUS TRAM PERCEPTION SYSTEM

• Main Author: Cipta Ramadhan, Ibnu
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/70799
• Institution: Institut Teknologi Bandung

Three-Dimensional (3D) object detection is an essential task for applications such as autonomous vehicles. Its ability to percept the surrounding environment in 3D space makes it more practical to implement in a real-world application. As the perception system advance, Light Detection and Ranging (LiDAR)-based 3D object detection gains its popularity and many approaches have been proposed to achieve state-of-the-art performance. Several common approaches in LiDAR-based 3D object detection are projection approach, voxel approach, and points approach. On the other hand, Graph Neural Network (GNN)-based approach for 3D object detection is still in early development as compared to other approaches. Point-GNN is one of the first architectures to implement GNN in the Lidar-based 3D object detection task. While Point-GNN achieved leading accuracy in the KITTI benchmark, it has low inference speed which makes it not sufficient to be implemented in real-time applications such as an autonomous vehicle. In this work, we propose a method to improve the inference speed of Point-GNN by observing the contribution of different parameters in the graph construction part of Point- GNN. Our approach improves the inference speed up to 2.56 times faster than the original Point-GNN while maintaining the degradation in overall Average Precision (AP) not exceeding 3.5%. In addition, we also observed the contribution of different features in the GNN layer to improve the overall AP by adding new features in the GNN layer. Our approach manages to improve the BEV AP by 0.22% and 3D AP by 0.35% as compared with the original Point-GNN. This thesis also implemented the 3D object detection method to embedded computer Nvidia Drive AGX Pegasus, an embedded computer designed specifically for an autonomous vehicle. Because of compatibility issues, this thesis implemented another 3D object detection method namely CenterPoint on the embedded computer Nvidia Drive AGX Pegasus, and achieved an inference speed of 20.10 FPS.