Line-scan vs area-scan cameras

An area-scan camera takes pictures. A line-scan camera takes one row of pixels at a time and relies on motion to build the image. That single difference drives every trade-off between them.

When line-scan wins

Continuous material - webs, foils, films, printed rolls. There is no natural "frame" on an endless material; line-scan turns the motion you already have into the second image dimension.
Very high resolution across the width - 8k and 16k pixel lines are routine. Getting the same cross-web resolution with area cameras means stitching several of them.
Rotating parts - unwrap a cylinder by imaging one line while the part spins. An area camera sees a curved, unevenly lit surface; a line-scan sees a perfect flat strip.
One line of light - illumination only has to be excellent along a single stripe. That is far easier (and cheaper at high intensity) than lighting a full field evenly.

When area-scan wins

Discrete parts presented one at a time.
Anything that requires a snapshot of a moving scene (strobe freeze).
Stop-and-go stations where there is no continuous motion to exploit.
Every project where integration simplicity matters - which is most projects.

The math that decides it

Line-scan couples camera, motion, and resolution into one equation. Everything follows from it:

pixel length along motion = speed / line rate

example: web at 2 m/s, line rate 80 kHz
         -> 2000 mm/s / 80,000 lines/s = 25 um per line

want 10 um pixels at the same speed?
         -> need 200 kHz line rate
         -> exposure per line: 5 us
         -> and ~5x the light intensity to keep the same gray level

That last line is the one that catches people: every doubling of speed or resolution halves the exposure time, and the light budget rises to match. Big line-scan systems are bright for a reason.

The integration trade-offs nobody warns you about

You are buying a motion-control project. Image quality is only as good as the speed measurement. A 1% speed error is a 1% geometric distortion. You will need an encoder, a good one, mechanically coupled to the actual material - not to a motor that hopes the material follows.
Vibration becomes image distortion. In an area image, vibration blurs. In a line-scan image, it draws visible wobble into straight edges - every mechanical disturbance is permanently recorded as geometry.
Aspect ratio depends on calibration. Cross-web resolution comes from the lens; along-web from speed and line rate. Get them mismatched and circles become ellipses. Plan a calibration procedure with a known grid from day one.
Lighting must be uniform along the line, always. A 5% brightness roll-off at the line ends becomes two permanent dark stripes along the whole image. Area cameras forgive lighting non-uniformity; line-scan archives it forever.
Debugging is different. There is no "live image" until things move. Commissioning needs the line running, which makes every iteration slower and every test more expensive than bench-testing an area camera.

My rule of thumb

If the part or material already moves continuously and the resolution demand is high, line-scan earns its complexity. If the parts are discrete and the motion would exist only to serve the camera, stay with area-scan - the simpler system you fully control beats the elegant one you fight.