The first real decision in specifying a vision camera is not brand or resolution but sensor architecture, because area-scan and line-scan cameras see the world in fundamentally different ways. An area-scan camera takes a whole rectangular photograph in one exposure, the way most cameras work. A line-scan camera has a sensor only one row of pixels tall and builds a picture by capturing thousands of those rows in sequence as the product streams past. This guide explains how each works, when discrete snapshots beat continuous scanning and when it is the other way around, and why line-scan systems live or die by a motion encoder.
Area-Scan vs Line-Scan Camera in one line: An area-scan camera captures a full two-dimensional frame in a single exposure, making it the natural choice for discrete parts that can be imaged as a snapshot. A line-scan camera has a one-row-tall linear sensor and assembles an image line by line as the product moves beneath it, which suits continuous webs, roll-to-roll material, cylindrical surfaces, and very high resolution inspection, and it needs a motion encoder to trigger each line so image height tracks the conveyor speed.
An area-scan camera is what most people picture when they think of a camera. Its sensor is a rectangular grid of pixels, and a single exposure captures the whole grid at once, producing a complete two-dimensional frame. This is a snapshot in the ordinary sense: point it at a part, trigger it, and you have a picture of that part with a fixed width and height. It is simple to set up, easy to reason about, and it works with the widest range of lighting and optics, which is why the great majority of inspection stations use area-scan cameras.
A line-scan camera works in a way that is stranger at first but powerful in the right setting. Its sensor is a single line of pixels, one row tall and many pixels wide. On its own that line captures only a thin strip of the scene, so a line-scan camera never takes a finished picture in one shot. Instead the part or the material moves steadily past the camera, and the camera captures a new line thousands of times a second. Software stacks those lines in order, and the accumulated strip becomes a full image. The vertical dimension of that image is not fixed by the sensor at all; it is built out of motion.
The consequence of that difference runs through everything else. An area-scan image has a fixed height set by the sensor, while a line-scan image has a height set by how many lines were captured, so a line-scan system can produce an image of essentially unlimited length as long as the material keeps moving. That is the key to why the two architectures suit such different jobs: one photographs objects, and the other continuously unrolls a surface into an image.
Area-scan cameras win for discrete parts that arrive one at a time and can be captured as a snapshot. A bottle, a casting, a circuit board, a moulded component, or a stamped label all fit naturally in a single rectangular frame, and the inspection is a matter of triggering the camera when the part is in position and analysing the resulting image. Because the whole field is captured at once, there is no need to coordinate the exposure with the part's motion beyond the trigger, and setup is straightforward. For the majority of station-based inspections, area-scan is simply the right and easier answer.
Line-scan cameras win where the product is continuous or where the geometry defeats a single frame. Web material such as paper, film, foil, textile, and sheet metal never stops and has no natural boundaries, so imaging it as a stream of lines is the only way to inspect every part of it without gaps or overlaps. Cylindrical parts are the other classic case: a can, a bottle, or a shaft can be rotated in front of a line-scan camera so its curved surface is unwrapped into a flat, evenly lit image, something an area-scan camera cannot do because the curvature distorts and darkens the edges of a single frame.
The third argument for line-scan is sheer resolution. Because the image is built line by line, a line-scan system can achieve extremely high across-web resolution and very long images without needing an impossibly large two-dimensional sensor. Inspecting a wide web for tiny defects at high detail is a job area-scan struggles with, because covering the same width and detail would need either many cameras or a sensor far larger than is practical. For very high resolution surface inspection, line-scan is often the only economical route.
The practical catch with line-scan is that the image is built out of motion, so the motion has to be measured, not assumed. If a line-scan camera captured lines on a fixed timer while the conveyor sped up and slowed down, the resulting image would stretch and squash, because the same number of lines would cover different amounts of real material. To keep the image faithful, each line must be triggered by actual movement of the product, and that is the job of a motion encoder, a device on the conveyor or roller that emits a pulse for each small increment of travel. The camera captures one line per pulse, so a fixed distance of material always maps to a fixed number of lines regardless of speed.
This encoder coupling is what makes line-scan images dimensionally trustworthy and is also what makes the setup more involved than area-scan. The encoder resolution has to match the desired pixel spacing, the camera has to be synchronised to the encoder signal, and the lighting has to be bright and steady enough for the very short per-line exposures that high line rates demand. Get those right and the system produces a geometrically stable image of arbitrary length; get them wrong and the image distorts in ways no software can fully undo. For area-scan the equivalent concern is simply triggering the single exposure at the right moment, which is far simpler.
Both architectures feed the same downstream picture once the inspection produces results. Whether a station is snapshotting discrete parts or scanning a continuous web, its defect counts, reject rates, and throughput end up as values a PLC and a SCADA layer can gather. On a web line especially, the line rate is tied to the conveyor speed through that same encoder, so the running speed, the line rate, and the defect rate are all related signals worth watching together. A cloud SCADA platform such as Merobix can collect those counts and rates centrally, so a supervisor sees defect density rising or a line slowing across sites from a control room or a phone, turning a stream of local inspection results into a live view of line health regardless of which camera architecture produced them.
Use line-scan when the product is a continuous web or roll, when you need to unwrap a cylindrical surface into a flat image, or when you need very high across-web resolution over a long or wide surface. Use area-scan for discrete parts that can be captured as a single snapshot, which covers most station-based inspections. Area-scan is simpler to set up, so prefer it unless the geometry or continuity of the product pushes you to line-scan.
A line-scan image is built by stacking captured lines as the product moves, so the image height comes from motion rather than the sensor. An encoder measures the actual travel and triggers one line per increment of movement, so the image stays proportional even when the conveyor speeds up or slows down. Without it, on a fixed timer, the image would stretch and compress with speed changes and become unusable for measurement.
Not well. An area-scan camera captures fixed-height frames, so covering a continuous web means capturing overlapping or gapped frames and stitching them, which is awkward and risks missing defects at the seams. A line-scan camera captures the web as one continuous, gap-free strip of arbitrary length, which is why web, film, foil, and sheet inspection almost always uses line-scan rather than area-scan.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.