Self-Driving Car Navigation Boosted by 3D Object Detection System

A new IoT-enabled, 3D object detection system for autonomous vehicles (AVs) has been developed by a team of engineers from Korea.

The team of researchers from the Department of Embedded Systems Engineering at Incheon National University (INU) designed the system to give self-driving vehicles better navigation and obstacle avoidance capabilities even in low visibility.

The team called the system a “significant step” in AV object detection technology.

“Our proposed system operates in real time, enhancing the object detection capabilities of autonomous vehicles, making navigation through traffic smoother and safer,” said lead researcher Professor Gwanggil Jeon.

Currently, AVs use an array of sensors for navigation and depth perception, including lidar for obstacle detection, radar for navigation at night and in cloudy weather and cameras for a 360-degree view of a car’s surroundings. However, the team said these sensors often have reduced capabilities due to adverse weather, unstructured roads or occlusion.

To meet this challenge, the INU’s system is built on deep learning object detection tech You Only Look Once (YOLOv3) which uses both point cloud data and RGB images to “generate bounding boxes with confidence scores and labels for visible obstacles as output.”

Related:Self-Driving Car Company Chases Next-Level Autonomous Driving

The team tested the system’s efficacy using the Lyft dataset, which consists of road information from 20 AVs traveling on a predetermined route in Palo Alto, California, over a four-month period. 

Results from the tests showed “high accuracy,” with overall accuracy for 2D and 3D object detection at 96% and 97%, respectively.

"By improving detection capabilities, this system could propel autonomous vehicles into the mainstream,” said Jeon. “The introduction of autonomous vehicles has the potential to transform the transportation and logistics industry, offering economic benefits through reduced dependence on human drivers and the introduction of more efficient transportation methods."

Next, the team said it aims to explore additional deep learning algorithms for 3D object detection.