Deep Learning-Based Approach for Efficient 3D Object Detection in Autonomous Vehicles
D Gnanakumar, G V S S Mouli, E. Reddy, Gaurav Yadav, G. Sharmila · 2024
The YOLOv3 deep learning architecture is utilized in this project to develop an advanced framework for 3D object detection in vehicle automation. This technology ensures precise and rapid detection of essential objects like cars, pedestrians, and barriers, crucial for safe autonomous driving. By integrating 2D bounding boxes with 3D coordinate estimations, the system analyzes input from multiple sensors to identify and classify various elements in the driving environment. This capability is vital for navigating complex scenarios that require swift decision-making, particularly in urban settings, highways, and unpredictable traffic. A primary objective is to enhance computational efficiency for real-time applications, achieving an impressive 89.55% detection accuracy. Performance evaluations demonstrate the system’s adaptability across diverse conditions, including varying lighting, weather, and road types. The combination of 3D object recognition and rapid processing positions this framework as a key advancement in developing reliable and secure automated vehicles.