A Comprehensive Review on LiDAR Based 3D Deep Learning Object Detection Algorithms

Recent advancements in autonomous vehicles and robotics have underscored the importance of robust 3D object detection(3DOD) systems. Among various sensor modalities, 3D LiDAR stands out for its ability to provide dense and accurate point cloud data(cloud of points signifying the reflected rays from the surface of objects present in the environment of the sensor), crucial for detecting and localizing objects in 3D space. This paper surveys the state-of-the-art (SOTA) deep learning models specifically tailored for 3D object detection for 3D LiDAR data. We classify these models based on their foundational architectural principles and methodologies, including voxel-based, point-based, and range-image-based techniques.Popular Key datasets, evaluation metrics and considerations for real-time application are discussed, highlighting advancements and ongoing challenges in the field. Finally,this review seeks to guide future research towards enhancing the accuracy, speed and robustness of 3D object detection systems utilizing 3D LiDAR technology.