Analysis of Lidar-Based Autonomous Vehicle Detection Technologies for Recognizing Objects

Abstract

Autonomous vehicle perception technologies extensively use vision, radar, and lidar sensors. The detection and classification of objects is the main purpose of vision sensors. Therefore, it is crucial to grasp perception technology before going any further. Perception involves organizing and interpreting information from the senses to understand the world. The primary goal of this sort of technology is to identify and track objects using various systems, with the identification and detection of barriers in varied driving environments constituting the biggest challenge. In this work, various perceptual tasks, including 3-Dimensional mapping, localization, and object detection, are performed using a Lidar-based system. The processing of image-like outputs from modern lidar sensors is investigated using general-purpose neural networks in detection and segmentation using DL perception. In this study, lidar-based perception is used, with lidar data combined with data from various perception tasks such as, localization, and object detection, 3-dimension mapping. The object detection model’s precision values vary according to variations in the number of training samples for the object detection model. Object detection techniques have a significant role in the performance of the model. The highest precision values are achieved, 9S.45%, for the YOLOx technique. The 95.35% precision value of the object detection model is computed when lidar is used as a camera.