Researchers have found a new cost-efficient way for self-driving cars to “see” 3D objects along their path with high accuracy.
Autonomous Vehicles (AV) are automobiles that can guide itself without human conduction. It is also known as the driverless car, robot car or self-driving car. One thing that allows AVs to function is the detection of objects along its path. LiDAR is one of its sensors.
The Light Detection and Ranging (LiDAR) sensors are laser sensors used to detect 3D objects in the path of AVs. Though highly accurate, LiDAR is bulky, ugly, expensive and energy-inefficient. Regardless of their disadvantages, most experts have chosen LiDAR as the top sensor for AV to safely perceive pedestrians, cars and other hazards along its path.
“One of the essential problems in self-driving cars is to identify objects around them – obviously that’s crucial for a car to navigate its environment,” Kilian Weinberger, associate professor of computer science and senior author of the paper, said in a statement.
This led Cornell Researchers to seek simpler methods that mimic the function of LiDAR. They have discovered that by using two low-cost cameras on either side of the windshield, AVs can detect 3D objects with nearly LiDAR’s accuracy but at a portion of its cost.
Findings show that analyzing images from a bird’s eye view rather than the traditional frontal view is three times more accurate, making stereo cameras a simple and cost-efficient alternative for LiDAR.
“The common belief is that you couldn’t make self-driving cars without LiDARs,” Weinberger said. “We’ve shown, at least in principle, that it’s possible.”
Eventually, Weinberger said stereo cameras could possibly be utilized as the primary way of identifying objects in lower-end cars or as a backup method in higher-end cars that are also equipped with LiDAR.
Yan Wang, the first author of the paper, is a doctoral student in computer science. Wang, along with his colleagues, will present their paper titled, “Pseudo-LiDAR from Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving,” to the 2019 Conference on Computer Vision and Pattern Recognition, June 15 to 21 in Long Beach, california.