For 2026 industrial camera selection, the first decision is sensor class: CCD suits low-light scientific imaging, but CMOS dominates factory-automation deployments because each pixel integrates its own amplifier and ADC, enabling higher frame rates, lower power draw, and falling cost per megapixel [S3].
The second decision is shutter architecture: global shutter exposes all pixels simultaneously and is required for high-speed motion and machine-vision guidance, while rolling shutter exposes line-by-line and can show banding under strobe lighting. Slower-moving or stationary subjects on a tighter budget can run rolling shutter without issue [S3]. A third split, area-scan versus line-scan, follows from the application: area-scan handles 2D inspection, while line-scan covers continuous webs, rotating drums, and ultra-wide fields with strict motion-sync demands [S3].
Sensor Size, Color, and Resolution Sizing
Large-format sensors (1 inch and up) deliver higher per-pixel image quality and dominate low-light, high-precision, or wide-field-of-view systems, while small-format sensors (1/1.8 inch and below) keep system volume and cost down for cost-sensitive, space-constrained integration [S3]. The matching rule is non-negotiable: the lens's image circle must be at least equal to, and typically larger than, the sensor's target size, otherwise edge sharpness collapses.
Color versus monochrome is a direct trade against throughput and sensitivity: monochrome cameras capture only luminance, so they reach higher resolution, faster frame rates, and better near-infrared response, while color cameras add red, green, and blue channels for tasks such as food appearance grading or print color verification, with the cost of extra interpolation noise [S3]. Resolution itself is sized from field of view and required precision, not from catalog maxima: the worked example in [S3] is 1600 mm field width with 1 mm precision, giving a minimum 1600 pixels, a 2k (2048-pixel) sensor choice, and an actual precision of 0.8 mm at the chosen line rate.
Interface Bandwidth vs Cable Length
Interface choice is a function of three numbers: required bandwidth, cable run, and budget. Reference data from the 2026 selection guide [S3] shows USB 3.0 capping near 400 MB/s under 5 m for low-cost desktop inspection, GigE Vision at roughly 120 MB/s over 100 m for multi-camera factory synchronization, Camera Link at 850 MB/s over 10 m for high-resolution high-speed rigs, and CoaXPress at 6.25 GB/s over 100 m+ for 4K/8K imaging, with 5GigE and 10GigE sitting at 500 MB/s to 1 GB/s for balanced speed/distance needs [S3].
The comparison that matters for most plant buyers is GigE versus CoaXPress: GigE wins on cable length and multi-camera synchronization and is the default for distributed factory cells, while CoaXPress wins on raw bandwidth per channel for high-resolution line-scan and high-speed area-scan work where a single Camera Link channel would be too slow. The lower-tier interfaces are not interchangeable with this work, so specifying CoaXPress where only GigE cabling exists leads to scope creep, while under-speccing a 4K inspection line to USB 3.0 is a common cause of dropped frames.
Area-Scan, Line-Scan, and 3D Selection Logic
Area-scan cameras output rectangular frames and cover most stationary or low-speed 2D inspection: dimensional checks, defect detection, part presence/absence, and rough temperature mapping. Their limits are frame rate and per-row pixel count, so very wide or fast continuous surfaces push the spec out of the area-scan envelope [S3].
Line-scan cameras sweep one or a few pixel rows across a moving web, with the 2D image built up over time; they deliver high scan rates and high along-web resolution, but they require tight synchronization between the encoder, the trigger, and the line rate, and any speed fluctuation stretches or compresses the image. Sizing example from [S3]: a 1600 mm web with 1 mm precision and 22000 mm/s transport speed needs 27.5 kHz line rate, so a 2048-pixel 28 kHz line-scan camera is the minimum. 3D cameras (laser profiling, stereo, structured light) sit above this and are a separate buy driven by Z-axis accuracy, point density, and the work envelope, with premium imports (Keyence, Cognex) holding the high end and domestic suppliers (Hikvision, Huaran) gaining share in 2D while pushing into 3D.
Strobe Lighting, Shutter Choice, and Motion Blur
Rolling shutter is the most common failure mode in machine vision and is worth a paragraph of its own. Under pulsed LED or strobe illumination, a rolling shutter exposes rows at different times, so partially-illuminated frames show horizontal banding, which in practice ruins thresholding and edge detection. The 2026 guide's rule is direct: inspect fast-moving objects with global shutter, and only consider rolling shutter when the target is stationary or moving slowly enough that motion within one frame is below one pixel [S3].
Global shutter costs dynamic range and sensitivity at the same sensor size, because each pixel needs its own sample-and-hold circuit; that is why a global-shutter sensor of a given generation is often one or two stops behind its rolling-shutter sibling in low-light SNR. The choice therefore cascades into lighting: rolling-shutter systems usually need continuous high-power LED bars, while global-shutter systems can drive short, bright strobe pulses and freeze motion at lower continuous power. This is a small change on paper but it dictates the illumination budget and the heat load inside the vision cell.
Brand Landscape and Where to Get Specs
The 2026 brand map splits into two camps [S3]. Imported: Teledyne (Dalsa for area/line-scan and TDI, Lumenera for USB and board-level NIR, FLIR for area-scan plus the LadyBug 360° spherical array), Korea's Vieworks for area/line/TDI, and Japan's AVALDATA for infrared specials. Domestic: Hikvision Robotics for industrial IoT, smart logistics, and machine-vision integration, and 51camera for one-stop machine-vision parts including cameras, lights, lenses, frame grabbers, cables, and PCs, with mixed import/domestic tiers.
Cross-checking camera specs against the actual line is non-optional, and consolidated spec portals (the 51camera catalog, manufacturer datasheets, and standards such as GigE Vision, Camera Link, CoaXPress, USB3 Vision) are where a buyer should anchor the comparison, not vendor marketing pages. The selection process in [S3] is worth following as a checklist: define motion (line-scan if moving, area-scan if static), define color (color vs mono), define shutter (global for fast motion), compute sensor size and resolution from FOV and precision, then pick interface from bandwidth, distance, and budget.
Selection Walkthrough: 1600 mm Web at 22000 mm/s
The worked example in [S3] compresses the entire selection into one calculation. Given 1600 mm field width, 1 mm required precision, and 22000 mm/s transport speed: resolution is 1600 / 1 = 1600 pixels minimum, rounded up to a 2048-pixel (2k) sensor for an actual precision of 0.8 mm; line rate is 22000 / 0.8 = 27.5 kHz, so a 28 kHz line-scan camera is the minimum; this rules out USB 3.0 and GigE on bandwidth, makes Camera Link marginal, and points to CoaXPress or 10GigE for headroom.
The same logic applied to a static label-inspection cell: area-scan color or mono, global shutter if a conveyor moves the part during exposure, GigE Vision for cable runs up to 100 m, sensor sized from FOV and the smallest feature of interest, lens image circle at least equal to the sensor. For a precision CMM-style measurement, see the vision measuring machine selection guide; for a bridge or gantry CMM rather than a vision-only rig, the CMM buying guide covers the wider coordinate-measuring decision. For balance and roundness work that often sits next to a vision cell, the roundness tester selection guide is a relevant companion. Buyers who need a frame-and-motion overview before specifying the camera should also review the industrial camera encyclopedia entry.
Trackable signals for the next quarter: CoaXPress-over-Fiber and 10GigE module pricing, the rollout of 3D time-of-flight modules into mid-tier pricing, and any revision to the GigE Vision standard that affects multi-camera synchronization. These three data points will decide whether 2026 line-scan budgets shift from Camera Link to CoaXPress or to 10GigE, and whether 3D moves further down the price stack.
Spec-level background on the components involved: linear guide, and crossed roller guide.