With intelligent driving sweeping the automotive industry, "camera" is a term everyone knows. Marketing often implies that the more cameras a car carries, the smarter it is. In fact cameras have permeated every corner of our lives — phones, computers, community surveillance, smart doorbells — and it is becoming increasingly hard to find a scene without one.

Yet although cameras are so familiar, how they work remains a mystery to many. How does a camera actually "see" the world? What signal does it output? And what processing happens before the data reaches an AI model? This article takes the camera apart to show the mechanism behind it.

Inside the camera

Taking a typical automotive camera as an example, the internal structure consists mainly of an optical assembly and two core printed circuit boards. At the front are the lens and the image sensor — the optical sensing body itself; at the back is the ISP module responsible for signal processing.

The image sensor converts light into electrical signals and is the core sensing element; the example uses an onsemi CMOS sensor. The rear signal processor handles the task of seeing clearly, performing a series of processing and optimisation steps on the raw image. The example uses a Mobileye chip, which not only performs image signal processing but also integrates target recognition and tracking.