Autonomous vehicles are on the verge of revolutionizing how physical commerce moves around us. Already, autonomous delivery bots are operating in populated urban areas, navigating sidewalks and delivering packages to destinations.
Large semi-trucks without drivers will likely be present on roadways within the next decade. However, as these vehicles increase in size, weight, and value, the accuracy, safety, and guidance systems that pilot them will become increasingly critical and complex.
Several methods are currently used for autonomous vehicle guidance systems; Lidar and cameras are the most commonly utilized technologies in this space. Here, we'll compare Lidar vs. camera technology in autonomous driving applications and showcase their groundbreaking use.
What Is Lidar (and radar)?
Lidar stands for Light Detecting and Ranging and is functionally similar to radar technology. Radar, which relies on a radio wave's emission, reflection, and reception, utilizes electromagnetic frequencies ranging from 30cm to 3mm.
These emissions are reflected off a medium and received near the emission point. The time it takes for that light to travel indicates how far away the reflecting object is.
Fundamentally, the only difference between radar and Lidar is the emission frequency of the electromagnetic waves utilized in the system. However, this difference in frequency impacts the capabilities of both wavelength ranges. For example, radar can only detect objects larger than a bowling ball (24cm), but it can detect objects up to 10,000m away.
On the other hand, Lidar can detect objects 10,000 times smaller than a bowling ball but can only detect objects 200m away.
Lidar systems in autonomous vehicles are responsible for detecting objects that are both big and small, and frequently, objects that are further away than 200 meters aren't as immediately important. This makes Lidar the natural electromagnetic frequency choice for autonomous driving applications.
Cameras Are a Critical Element
Cameras used in autonomous cars are specialized image sensors that detect the visible light spectrum reflected from objects. Given that the sun emits extraordinary amounts of UV and visible light, image sensors can detect many frequencies of visible light. This is similar to how human eyes see the light (i.e., light is emitted by the sun, reflects off objects at various frequencies—what we call colors—and is received by the image sensor).
The visible light spectrum ranges from 380-740nm, making it perfect for sensing items larger than a strand of hair cut into 23 pieces. Most cameras and image sensors can't see objects that small without special lensing, but their need to see objects smaller than a human is uncommon.
For cameras in autonomous driving applications, accurate image recognition can be achieved with image sensor resolution as low as 1.2 megapixels. But, for reference, the human eye is considered equivalent to a 576-megapixel camera.
Lidar vs Cameras
The topic of Lidar vs. Cameras in autonomous vehicles is widely debated. The irony of these arguments is that these technologies rely on the same principle of electromagnetic emission, reflection, and reception of the sensors that complete it.
However, the fundamental difference between Lidar and camera technology is that Lidar emits the light it sees, whereas cameras don't. This gives Lidar the ability to calculate incredibly accurate distances to many objects that are simultaneously detected.
Therefore, the argument may shift from asking, "Which technology is better?" to the question of, "Which technology is more advanced?" Modern CMOS image sensors were invented by NASA's JPL in 1993 and quickly grew to fame as the technology of choice for modern cameras. Engineer David Hall created Lidar in 2005.
Given that modern cameras have been heavily developed and tested, is their technology is superior? The short answer is that no one knows quite yet. Both technology methods have been heavily utilized by various companies, ranging from garage AV startups to the largest automaker in the world. Still, none have yet been able to achieve fully autonomous driving capabilities.
The Future of the Lidar vs. Camera Debate
Both Lidar and camera technology are wonderfully powerful and have distinct advantages. Lidar can map entire cityscapes with sub-millimeter accuracy, while image sensors and camera technology are more advanced and widely produced. Tesla, which relies on cameras, claims its vehicles will be able to drive entirely autonomously by 2023.
When asked which technology he prefers, Elon Musk stated, "Humans drive with eyes and biological neural nets, so [it] makes sense that cameras and silicon neural nets are [the] only way to achieve a generalized solution to self-driving." Perhaps the Tesla founder is correct, or maybe the future of Lidar will see faster adoption. The first technology that can fully support autonomous driving and then be widely developed at scale will win the race to autonomy.
