The Guide to Machine Vision Cameras in 2026
The camera is the foundation of every vision system, and the choice has more moving parts than it used to. More sensor technologies, more interfaces, more camera types. This guide covers each camera type we supply, how it works, where it fits, and how to choose between them.
Choosing a camera used to be simple. You picked a resolution, an interface and a manufacturer, and placed an order. Today there are more ways to capture an image than at any point in the industry's history, and the right answer depends entirely on what you are inspecting and how it moves in front of the camera.
This page is about choosing the right type of camera. If you want to understand who makes what across the European market, our guide to machine vision suppliers in Europe covers that separately. Here we stay focused on the technology and the decision, and on the camera ranges we actually supply and support.
Camera types at a glance
Seven camera types cover the vast majority of industrial applications. Use this as a shortcut, then read the section that fits before you specify.
| Camera type | Best for | Indicative investment | Clearview range |
|---|---|---|---|
| Area scan | General 2D inspection, gauging, code reading, optical character verification, presence and absence. The default starting point for most applications | Lower to mid | Teledyne, LUCID, Zebra, CIS |
| Line scan | Continuous web and conveyor inspection, print and surface quality, wide or cylindrical parts unrolled into one seamless image | Mid to high | Teledyne, LUCID |
| 3D | Height, volume and shape measurement, bin picking and robot guidance, defects that are invisible in 2D | Mid to high | Teledyne, Zebra, LUCID, Photoneo |
| Smart camera | Self-contained pass and fail at the point of inspection with no external PC. Open platform models run your own code too | Mid | Zebra |
| Fixed industrial scanner | Reliable barcode and 2D code reading, track and trace, logistics and warehousing | Lower to mid | Zebra |
| Thermal | Temperature measurement, seal inspection, hot spot detection, battery and EV safety | Mid to high | FLIR, Teledyne Calibir |
| Hyperspectral and SWIR | Material and chemical discrimination, food sorting, contaminant detection, recycling and research. SWIR also images beyond visible light | High | Teledyne, LUCID, EVK |
Investment columns are indicative bands only. For realistic budget ranges by component, see our machine vision pricing guide.
Area scan cameras
Area scan cameras are the workhorse of machine vision. They capture a complete two dimensional image in a single exposure and are used in most inspection, measurement and identification tasks. If you are new to vision, this is almost certainly where you start.
How they work
The sensor is a rectangular grid of pixels. When triggered, the camera exposes the sensor and outputs a complete frame, then waits for the next trigger. This frame based approach is simple, well understood, and supported by every vision software platform.
What to get right
Match the resolution to the smallest feature you need to see across your field of view, rather than assuming more pixels are always better. Pick a global shutter for anything in motion, since a rolling shutter exposes rows in sequence and distorts moving objects. Watch the sensor format, because a larger sensor needs larger and more expensive lenses to cover it. And size the frame rate to your real line speed. Over specifying any of these adds cost and data for no benefit.
Best suited for
Station based inspection where the part is stationary or briefly paused, where the field of view fits a single frame, and the task is triggered. This covers the majority of vision applications.
Clearview supplies area scan cameras from Teledyne, LUCID, Zebra and CIS across the resolution, speed and interface range. Browse area scan cameras.
Line scan cameras
Line scan cameras capture a single row of pixels at a time and build a complete image as the object moves past. They are the standard for continuous motion inspection: webs, conveyors, print quality, and anything effectively endless in one direction.
How they work
The sensor is a single line, typically from 1K to 16K pixels wide. The camera fires continuously at a fixed line rate, synchronised to the motion of the object, usually through an encoder. Each line becomes one row of the final image, so the image can be as long as you need at a consistent resolution across the full width.
What to get right
Choose the line resolution from the object width and the smallest feature you must detect. Match the line rate to the speed of the material so the image is neither stretched nor compressed. Decide on colour early, since print and colour sorting usually need a trilinear sensor. And size the interface to the data rate, because line scan produces a continuous stream rather than discrete frames.
Best suited for
Continuous material that is longer than it is wide, or where you need more resolution across the width than a single area scan frame can provide. Paper, film, foil, textile, printed web, and round parts rotated under the camera.
Clearview supplies line scan cameras from Teledyne and LUCID across the resolution and interface range. Browse line scan cameras.
3D cameras
3D cameras capture depth as well as a 2D image, which makes them the tool for measurement, robot guidance and any inspection that is genuinely three dimensional. There are four main technologies, and they are not interchangeable. The right one depends on accuracy, speed, surface type and budget.
Laser triangulation
A laser line is projected onto the surface and a camera mounted at an angle measures how the line deforms to calculate height. This is the most accurate 3D technology and the standard choice for in line surface profiling and dimensional measurement. In our range this is Teledyne Z-Trak and Zebra AltiZ.
Structured light
A coded pattern of light is projected onto the object and the 3D shape is calculated from how the pattern deforms. It captures a full surface in a single snapshot, which makes it fast and well suited to bin picking and robot guidance. In our range this is Photoneo.
Time of flight
The camera measures the time light takes to travel to the object and back, producing a depth value at every pixel at once. Time of flight is less accurate than triangulation but fast and simple to integrate, which suits robotics, navigation and material handling. In our range this is the LUCID Helios family.
Stereo vision
Two cameras view the scene from slightly different positions and depth is calculated from the disparity between them, the same way human eyes judge distance. Passive stereo relies on existing surface texture, while active stereo projects a pattern so it can still work on plain surfaces. Stereo copes well with ambient and outdoor light and has no moving parts, with the trade off of lower accuracy than triangulation or structured light. In our range this is the Teledyne Bumblebee X.
Before you commit to 3D
Not every application that looks three dimensional needs a 3D camera. A 2D camera with the right lighting, such as a dark field or grazing angle, can reveal surface features that appear to need depth. 3D costs significantly more, so test both approaches on real samples first and only pay for depth when the task genuinely requires it.
Smart cameras
A smart camera combines sensor, processor and I/O in a single unit. It captures the image, processes it, makes the decision and outputs the result directly, with no external PC, frame grabber or separate software box on the line. The important distinction, and one that is easy to miss when comparing models, is open platform versus closed.
Open platform smart cameras
An open platform smart camera runs a full operating system on the device, typically Linux or embedded Windows. That means you are not limited to one toolset. You can run the manufacturer's own software, a third party SDK, or your own algorithms directly on the camera. The Zebra Iris GTX is the example in our range: it runs Zebra's Aurora software, but it can equally run your own code, so the platform grows with your application instead of boxing it in.
Closed smart cameras
A closed smart camera only ever runs the manufacturer's own software. That can be perfectly fine for a simple, fixed task, and it is often quicker to set up, but you are tied to that vendor's capabilities. If your requirement grows beyond what their tools support, your only route is to replace the camera.
Best suited for
Well defined inspection at a single point: presence and absence, basic code reading, simple measurement and pattern verification, especially where space is tight and there is no room for a PC. For complex, multi camera systems or heavy custom processing, a PC based system with an industrial camera gives more headroom.
Clearview supplies smart cameras from Zebra. Browse smart cameras.
Fixed industrial scanners
Fixed industrial scanners are purpose built for reading codes reliably at a fixed point on a line: barcodes and 2D codes for track and trace, sortation, and warehouse and distribution flows. Where a smart camera is a general inspection tool, a fixed scanner is optimised for decode performance and uptime on identification tasks.
In our range these are the Zebra FS series, managed through Zebra's Aurora software so a whole estate of scanners and cameras can be set up, deployed and run from one platform. Higher models in the range take C mount lenses and a choice of sensors, and machine vision capability can be added by licence, which blurs the line between a scanner and a smart camera when you need it to.
Best suited for
High reliability code reading and track and trace, logistics and warehousing, and any line where dependable decoding matters more than flexible image processing.
Clearview supplies fixed industrial scanners from Zebra. Browse fixed industrial scanners.
Thermal cameras
Thermal cameras detect infrared radiation rather than visible light, so every pixel represents a temperature value. That makes them the right tool whenever temperature is the thing you care about: seal inspection on food and drink packaging, process and temperature monitoring, hot spot detection, and battery and EV safety.
Watch out for
Thermal cameras do not see through glass. Glass is opaque at the infrared wavelengths most thermal cameras use, so the camera reads the glass surface, not what is behind it. The germanium optics they need are also far more expensive than standard lenses, which lifts the system cost.
Clearview supplies thermal cameras from FLIR and Teledyne Calibir. Browse thermal cameras.
Hyperspectral and SWIR cameras
Hyperspectral cameras capture many narrow wavelength bands rather than the three a colour camera sees. A single hyperspectral camera might record a hundred or more bands, revealing the chemical composition, moisture content and material properties of an object that are invisible to standard imaging. Push broom designs build the spectral image line by line as the product moves, while snapshot designs capture it in one go.
Typical uses are food sorting, contaminant and foreign material detection, recycling and material classification, pharmaceutical analysis and research.
Watch out for
Hyperspectral generates large amounts of data and needs specialist processing, and it costs considerably more than standard imaging. For many tasks a mono camera with the right filters and lighting does the job. Reach for hyperspectral when the spectral information is genuinely essential and nothing simpler will resolve it.
SWIR is not hyperspectral
You will also find SWIR cameras grouped in this part of the range, and they are a different thing worth keeping separate when you specify. SWIR stands for short wave infrared, roughly 900 to 1700 nanometres. A SWIR camera is normally a monochrome camera that images in that band, not a camera that splits light into many spectral bands the way hyperspectral does. It earns its place on tasks visible light cannot do, such as seeing moisture and fill levels, inspecting through certain coatings and materials, silicon wafer inspection, and sorting by material. So both sit close together in the range, but one resolves spectral signatures across many bands and the other simply sees in a waveband beyond the visible.
Clearview supplies hyperspectral and SWIR cameras from Teledyne, LUCID and EVK. Browse hyperspectral and SWIR cameras.
Embedded and board level vision
Embedded vision builds a compact, board level camera into a product rather than a factory enclosure: a medical device, a mobile robot, a sorting machine. It uses compact interfaces and runs processing on embedded platforms such as ARM or NVIDIA Jetson, which suits volume OEM products where size, weight, power and per unit cost are the constraints.
A word from experience
We regularly see companies move from a GenICam compliant industrial camera to a MIPI based embedded camera to shave unit cost, then find the development effort dwarfs the saving. If you are weighing that switch, budget the engineering time realistically before you commit.
Clearview supplies embedded vision computers from Neousys alongside board level cameras. Browse embedded systems.
A quick word on sensors
More than any other single factor, the sensor drives a camera's price and performance. A few fundamentals are worth understanding before you specify.
Modern CMOS sensors now dominate and are excellent across the range, so the most expensive camera is rarely the one you need. Shutter type matters most for motion: a global shutter freezes movement, a rolling shutter is cheaper but distorts moving objects. Mono or colour is a real decision, not a default, because a mono sensor gives you more usable resolution and sensitivity, so only choose colour when the application needs it. And spectral response into the near infrared can solve problems that visible light cannot, such as seeing through certain coatings or improving contrast under IR lighting. Sony's Pregius and Starvis families turn up across many of the cameras in our range, with Pregius the common choice for global shutter machine vision.
A deeper guide to sensors, sensor generations and how to read a sensor datasheet is coming as a companion piece.
Camera interfaces
The interface decides how the camera connects to the processing unit and how much data it can move. Get it wrong and it bottlenecks the whole system, so choosing it is as important as choosing the camera.
| Interface | Frame grabber | Power over cable | Best for |
|---|---|---|---|
| GigE Vision | No | PoE available | The default for most applications. Long cable runs, no grabber, wide support. Available in 1, 2.5, 5, 10 and 25 Gigabit grades |
| USB3 Vision | No | Yes | Short cable runs, bench and lab setups, cost effective single camera systems |
| CoaXPress | Yes | Yes | Highest bandwidth. High resolution, high speed area and line scan |
| Camera Link | Yes | Yes | Established installed base. Being succeeded by CoaXPress in new designs |
| Camera Link HS | Yes | Fibre option | Very high bandwidth over long fibre runs, large line scan systems |
For most new projects, GigE Vision in the right speed grade is the starting point. Move to CoaXPress when you need bandwidth that GigE cannot provide, and use USB3 for bench and short cable setups.
How to choose, in five questions
With this many options the decision can feel heavy. These five questions get you to the right camera type in order.
What are you inspecting?
Define the task before the hardware. What feature must be detected, what are the pass and fail criteria, what tolerances apply.
How does the object move?
Stationary or triggered points to area scan. Continuous motion points to line scan. Depth points to 3D. Temperature points to thermal. Material or chemistry points to hyperspectral. Code reading points to a fixed scanner.
What resolution do you need?
Work it out from the field of view and the smallest feature. A camera that exceeds your requirement is not better, it is more expensive and generates more data.
What speed do you need?
Match frame rate or line rate to your real production speed. Over specifying speed costs money on both the camera and the interface.
What is the budget?
Use our pricing guide for realistic ranges and the budget planning guide to build a full project budget including lenses, lighting, software and integration.
If the answers are clear, you can specify with confidence. If they are not, that is what testing is for. Our Insights lab in Thame is equipped with every camera type in this guide, and the engineering team can evaluate your samples under real conditions before you buy.
Frequently asked questions
What is the difference between area scan and line scan?
An area scan camera captures a full 2D frame in one exposure and suits stationary or triggered parts. A line scan camera captures one row at a time and builds an image as the object moves past, which suits continuous material and wide or endless surfaces.
Do I really need a 3D camera?
Only if the task genuinely needs depth, such as height, volume or shape measurement, or robot guidance. Many features that look three dimensional can be revealed by a 2D camera with the right lighting, at a fraction of the cost. Test both on real samples before committing.
What makes a smart camera open platform?
An open platform smart camera runs a full operating system, so it can run the manufacturer's software, a third party SDK or your own algorithms on the device. A closed smart camera only runs the manufacturer's own software, which limits how far the application can grow.
Which interface should I choose?
GigE Vision is the default for most applications because of long cable runs and wide support. Move to CoaXPress for the highest bandwidth, and use USB3 for short cable bench setups.
How much does a machine vision camera cost?
It depends heavily on resolution, speed, interface and type. Our pricing guide gives transparent ranges across the whole system, not just the camera.
Not sure which camera fits?
Our engineers have been specifying machine vision cameras for years. We would rather spend half an hour helping you get the specification right than sell you something that does not fit, and we can prove the approach on your own samples in our Insights lab before you buy.
info@clearview-imaging.com | +44 (0)1844 217270