Area scan vs line scan cameras

The core difference: matrix vs line sensor

An area scan camera carries a two-dimensional matrix of pixels and captures a complete rectangular image in a single exposure, exactly like a conventional camera. Each trigger yields one frame, so the part can be presented statically or frozen in motion with a short exposure or strobe. This makes area scan the natural fit for discrete objects that arrive one at a time and can be imaged within a fixed field of view.

A line scan camera has a sensor only one (or a few) pixel rows tall but very wide, often 2k, 4k, 8k or 16k pixels across. It does not produce a usable image on its own; instead it captures one line at a time while the object moves underneath, and software stitches thousands of lines into a tall, continuous image. The relative motion between camera and part is therefore not a nuisance but the mechanism by which the second image dimension is built.

Put simply: area scan freezes a scene, line scan scans a moving surface. That single distinction drives every downstream decision about triggering, lighting, optics and interface bandwidth.

When area scan is the right choice

Area scan is the default for the large majority of inspection tasks and should be your starting assumption unless something in the application rules it out. It excels when parts are discrete and indexed: bottles on a star wheel, electronic assemblies on a pallet, stamped components on a conveyor stopped or triggered at a station. A single proximity sensor or PLC signal triggers the grab, and a strobed light freezes motion so even fast-moving discrete parts are captured sharp.

It is also the simpler system to specify, mount and maintain. You select a resolution that resolves your smallest feature across the whole field of view (typically allowing 3-5 px on the smallest defect), pick a matching lens, and you have a complete imaging chain. Brands such as Basler, iRAYPLE and Omron / Sentech offer broad area scan ranges over GigE Vision, USB3 Vision and CoaXPress, covering most resolutions and frame rates a discrete-part line needs.

• Discrete, separable parts that can be triggered individually
• A bounded field of view where one frame covers the whole region of interest
• Stop-and-go or indexed motion, or motion that can be strobe-frozen
• 2D feature checks: presence/absence, dimensional gauging, OCR/OCV, surface defects on flat parts
• Simpler integration, lower lighting demands and easier setup are priorities

When line scan is the right choice

Line scan becomes necessary, not merely preferable, when the material is continuous or its geometry defeats a single matrix frame. Web materials such as paper, film, foil, textile, glass and metal strip run endlessly past the camera, so there is no discrete part to trigger and no natural frame boundary; a line sensor images them as one seamless, arbitrarily long strip. Cylindrical and rotating objects are the other classic case: rotating the part under a line sensor unwraps its curved surface into a flat image without the perspective distortion a matrix camera would suffer at the edges.

The second driver is width combined with detail. To inspect a wide product at high pixel density with area scan you would need either an enormous, expensive matrix sensor or multiple cameras stitched together with overlap and calibration headaches. A single line scan camera with 8k or 16k pixels across the web delivers uniform, high cross-web resolution in one optical axis, and the along-travel resolution is set purely by line rate versus speed. This resolution-over-width advantage is often the decisive economic argument.

• Continuous web or strip material with no discrete part boundary
• Rotating or cylindrical parts whose surface must be unwrapped flat
• Very wide products needing high, uniform resolution across the width
• Effectively unlimited image length in the direction of travel
• Avoiding the cost or stitching complexity of a single very large matrix sensor or a multi-camera array

Encoder synchronisation and image geometry

Because a line scan image is assembled line by line as the material moves, the vertical (travel-direction) scale of the image depends directly on how far the material advances between line captures. If you clock the camera at a fixed line rate but the belt or web speed varies, the image stretches and compresses, distorting measurements and confusing defect classification. For this reason production line scan systems are almost always driven by a quadrature encoder coupled to the transport.

The encoder issues pulses proportional to actual material travel, and each pulse (or a divided multiple of them) triggers one line capture. This locks one image row to a fixed increment of real-world distance, so geometry stays correct regardless of acceleration, deceleration or speed drift, and the square-pixel calibration across and along the web can be made to match. Area scan, by contrast, usually needs only a simple presence trigger per part, which is one reason it is easier to commission.

When sizing a line scan system, verify that the required line rate at maximum line speed is achievable, because line rate, exposure time and available light are tightly coupled, and that the encoder resolution gives at least one pulse per desired image row at top speed.

Lighting: the line scan tax

Line scan is far more demanding on illumination than area scan, and underestimating this is a common project failure. Each line is exposed for only a very short window, often tens of microseconds at high line rates, and only the narrow strip directly under the sensor contributes signal. You therefore need to pour a large amount of light into a thin, extremely uniform line precisely aligned to the sensor, rather than flooding a whole field of view as you can with area scan.

In practice this means high-output line lights, frequently high-intensity LED bar or focused-line geometries, and careful attention to uniformity across the full sensor width. Lighting specialists such as DCM Sistemes supply the concentrated line and bar illumination these systems require. Expect lighting to be a larger share of the system cost and integration effort than on an equivalent area scan station, and budget for it early.

Area scan can usually tolerate more relaxed, general-purpose illumination and benefits readily from strobing, where a bright pulse synchronised to the trigger freezes motion and lets you reuse modest continuous-rated LEDs at high peak output.

Interface bandwidth and frame grabbers

The data rate from a fast, wide line scan camera is high and sustained: thousands of lines per second multiplied by thousands of pixels per line produces a continuous stream that can saturate slower interfaces. While modest line scan cameras run over GigE Vision (~1 Gbit/s, up to 100 m, PoE) or USB3 Vision (~350-400 MB/s but only 3-5 m), high-speed, high-width line scan commonly needs CoaXPress (CXP-6/CXP-12, multiple GB/s) with a dedicated frame grabber in the PC.

A frame grabber is doing more than moving data in line scan applications. It typically handles the encoder input, divides pulses to the desired line trigger rate, manages line buffering and reconstructs frames of a defined height from the continuous stream, offloading deterministic timing from the host CPU. Area scan systems can more often run grabber-free over GigE or USB3, which keeps the bill of materials and software stack simpler.

If your line scan requirement points to CoaXPress plus a frame grabber, confirm the grabber supports your camera's CXP link configuration and the encoder/IO features you need before committing to the camera.

Putting it together: a practical decision path

Start with the part and the motion. If you can present discrete items, trigger them individually and cover the region of interest in one frame, choose area scan; it is simpler, cheaper to light and faster to deploy. Move to line scan only when the application forces it: continuous web, rotating or cylindrical surfaces, or a width-and-detail combination that an affordable matrix sensor cannot meet.

Then validate the line scan-specific chain end to end. Confirm you can deliver enough light into the line at the required line rate, that an encoder will keep geometry stable across speed changes, and that the interface and frame grabber can sustain the data stream. Each of these is a place where line scan projects commonly stall, so size them before ordering hardware.

If the inspection also involves complex or variable surface defects, deep-learning classification on either sensor type can outperform rule-based methods; AIS Vision can integrate the full imaging, motion and software chain, with Rosepetal handling the AI defect analysis where conventional tooling reaches its limits.