Machine Vision & Industrial Imaging

An area scan camera captures a complete two-dimensional image in a single exposure. This type of camera is ideal for discrete parts that can be presented to it while remaining still or under strobe lighting. In contrast, a line scan camera captures one row of pixels at a time, gradually building a complete image as the object moves past it. This camera is suitable for continuous materials such as webs, sheets, and rolls, as well as cylindrical parts that rotate beneath it. Line scan operation requires precise control of the motion, typically synchronized with an encoder, and lighting that is concentrated into a bright line rather than distributed over a wide area.

The choice of camera connection is typically determined by factors such as cable length, the number of cameras, and data rate, rather than by maximum speed alone. GigE Vision operates using standard Ethernet cables over distances of up to 100 meters; it supports multiple cameras through typical network switches and does not require a frame grabber, making it a common starting point for many applications. USB3 Vision is a straightforward and cost-effective option for connecting a single camera at short ranges, although the cable length is generally limited to a few meters unless an active extension is used. CoaXPress is a high-bandwidth, low-latency, point-to-point interface commonly used where demanding data rates and deterministic triggering justify a frame grabber. Camera Link and Camera Link HS are still in use, particularly for line scan systems, but they also require a frame grabber.

When selecting a lens, three key factors must be considered: sensor size, the working distance from the lens to the object, and the desired field of view. The focal length can be roughly determined by multiplying the sensor size by the working distance and then dividing by the field of view. Additionally, the lens must meet two important conditions: its image circle must completely cover the sensor to prevent dark corners in the image, and it must have sufficient resolving power to match the sensor's pixel size. It is also essential that the lens mount type is compatible with the camera.

A telecentric lens maintains a consistent magnification across its designed working range, meaning that an object does not appear larger as it gets closer to the lens. In contrast, conventional lenses can introduce perspective errors; for instance, when a part is positioned slightly higher on a fixture, it may measure as larger than it actually is. This distinction is important when measuring dimensions, rather than simply detecting presence. Telecentric lenses tend to be larger and more expensive, and to have a fixed field of view, which is why they are primarily specified for measurement and metrology rather than for general applications.

The requirement for a frame grabber depends on the type of interface being used. GigE Vision and USB3 Vision connect to standard ports commonly found on most computers, so they do not require a frame grabber. In contrast, interfaces such as CoaXPress, Camera Link, and Camera Link HS do require a frame grabber, which is a dedicated card that receives image data. Although a frame grabber adds cost and occupies a slot on the host machine, it offloads data handling from the CPU and provides precise triggering, which is especially important at high frame rates.

A global shutter exposes every pixel on the sensor simultaneously, capturing the entire image at the same instant. In contrast, a rolling shutter exposes the sensor one row at a time, so different parts of the image are captured at slightly different times. This timing difference is insignificant for stationary objects, but with moving objects, a rolling shutter can cause geometric distortion because the object may have moved between the first and last rows captured. Therefore, applications that involve imaging moving parts typically require a global shutter.

EMVA 1288 is a standard established by the European Machine Vision Association, and it is also recognized as an ISO standard. This standard defines a consistent method for measuring and reporting the performance of cameras and sensors. It includes various parameters such as quantum efficiency, temporal dark noise, saturation capacity, and dynamic range. The main advantage of EMVA 1288 is its ability to enable direct comparisons between cameras; two cameras evaluated using EMVA 1288 can be compared accurately, unlike marketing figures from manufacturers, which may not follow a consistent method. It is important to note that this standard characterizes only the camera and sensor, not the entire system, which includes the lens.

Lighting determines whether the feature you need to inspect is visible to the camera at all. No amount of resolution or software recovers contrast that was never captured. The geometry matters more than the brightness: backlighting produces a silhouette for edge and dimensional measurement, dark field lighting strikes the surface at a low angle so scratches and texture scatter light back while the flat surface stays dark, diffuse dome lighting removes reflections from curved shiny parts, and coaxial lighting suits flat specular surfaces. Lighting is frequently specified last and is a common cause of vision projects underperforming.

Resolution follows from the smallest feature you must detect and the field of view you must cover, not from a megapixel target. Work out how many pixels are needed to resolve the smallest feature for your inspection to be reliable, then combine that with the field of view to determine the required sensor resolution. Specifying more resolution than the task needs increases data rate, cost, and processing load, and can force a more expensive interface. It also demands a lens capable of resolving that detail, since a high-resolution sensor behind an inadequate lens gains nothing.