Mark học AI

Video #5.3 · AI's three eyes · Part 4/8 ·

LiDAR, radar or camera: which is best?

The three eyes compete in six events: distance, speed, color and text, night, fog and rain, and price. Each one is good at something.

Watch videoVideo in Vietnamese

Watch on

Three eyes take the field

LiDAR uses laser light, radar uses radio waves, and a camera uses the light that's already there. Today the three eyes compete in six events to see who is good at what.

EVENTS 1-2

Distance and speed

For distance, LiDAR is the most precise, down to a few centimeters. For speed, radar wins thanks to Doppler. A camera has to compare several frames to estimate either, so it comes last in both.

EVENTS 3-4

Color, text and night

For reading color and text, the camera has no rival. But at night the camera gets much weaker, while LiDAR and radar send out their own signal, so dark or light makes no difference.

EVENTS 5-6

Fog, rain and price

In fog, radar holds up best. On price, the camera is cheapest, radar sits in the middle, and LiDAR costs the most, though its price is falling fast.

TWO SCHOOLS

Cameras only, or the full set

In the real world there are two schools. In 2021, Tesla dropped radar from new cars sold in North America to rely on cameras only. Waymo's sixth-generation robotaxi carries 13 cameras, 4 LiDARs and 6 radars.

HOW IT WORKS

Why use several kinds

Several kinds are used because they fail in different ways. When the sun blinds the camera, radar is fine; when radar mistakes a can for an obstacle, the camera sees clearly. Both failing at once is much rarer.

This article is based on the video LiDAR, radar or camera: which is best? from the Mark học AI channel. Watch the video (in Vietnamese) to see the animations.

Found this useful? Send it to a friend

Facebook