// Key takeaways
- The camera is one part of nine; most station problems start in the links around it.
- Lighting sets the contrast the software works with, so it is chosen for the defect.
- Where the processing runs (vision sensor, smart camera, controller or PC) sets capacity, not the principle.
- A reject without a confirmation sensor cannot prove that a failed product left the line.
What parts does an inspection station need?
Every station, from a single vision sensor to a multi-camera cabinet, has the same chain of parts. The table gives each one’s job in one line; the inspection technology overview goes deeper into how each is chosen.
| Part | Its job | What decides the choice |
|---|---|---|
| Trigger | Tells the station a product is in position, once per product | Product spacing, speed changes, transparent or unstable products |
| Camera, vision sensor or smart camera | Captures the image | Smallest detail to be seen, field of view, line speed, colour or monochrome, number of views |
| Lens | Projects the scene onto the sensor at the right size and sharpness | Field of view and working distance; a telecentric lens for some measurement |
| Lighting | Makes the defect stand out from the good surface, the same way every time | The surface (matte, shiny, clear) and the defect |
| Processing | Runs the inspection in the time between products | Number of cameras and tools, deep learning, cycle time |
| Software | Turns the image into a pass or fail and data, per product recipe | The tools the checks need |
| I/O and PLC link | Hands the result to the line and receives the product to run | PLC make and model, signals and protocols on the chosen device |
| Reject and confirmation | Removes failed products and proves they left | Product weight and stability, gap between products, air supply |
| Enclosure and mounting | Holds alignment and keeps out daylight, dust and washdown water | Cleaning regime, protection rating, space on the line |
How is the image captured?
The trigger is usually a photoelectric sensor just before the camera, or an encoder on the conveyor where the speed changes. The rule is one trigger per product: a missed or double trigger upsets everything after it.
The camera is an area scan camera for separate products, or a line scan camera for continuous material such as film or board. Moving products need a global shutter and a short exposure, so they do not blur. Resolution follows the smallest defect: it must cover several pixels across the field of view. The lens sets that field of view at the working distance, and its focus and aperture are locked once set. The camera and lens selection guide works through the choice.
The light decides what the camera can see at all. A backlight shows outlines and fill levels, low-angle light shows scratches, a dome hides glare on shiny packs. Chapter 5 explains why.
Vision sensor, smart camera or PC: where does the processing run?
- Vision sensor: camera, light and a small tool set in one housing, for one simple check from one view. HIKROBOT lists its SC1000, SC2000 and SC3000 series as vision sensors (HIKROBOT catalogue 2025 Q4).
- Smart camera: camera, lighting control and inspection software in one housing. HIKROBOT’s SC5000X and SC6000 smart cameras run its VisionMaster platform, including deep-learning modules (HIKROBOT catalogue 2025 Q4).
- Vision controller or PC: separate industrial cameras connected to an industrial computer. This suits several views, heavy processing or many products; HIKROBOT positions its VC-series vision controllers for multi-camera work such as positioning, inspection, measurement and recognition (HIKROBOT catalogue 2025 Q4).
The inspection logic is much the same in all three; what changes is how many tools, views and products the hardware can carry. Choosing a HIKROBOT smart camera covers the trade-off.
What does the inspection software do?
The software holds a recipe for each product: where to look, which tools to run, the limits, and what to send to the PLC. We build stations on HIKROBOT VisionMaster, which HIKROBOT describes as software for visual positioning, size measurement, defect detection and information recognition, and which can control cameras, I/O and light sources from within the platform (VisionMaster product page). HIKROBOT cameras are first connected and set up in its MVS client software (HIKROBOT camera user manual); the inspection then runs in VisionMaster.
addaScan works with VisionMaster as an independent integrator; we do not claim official HIKROBOT partner or distributor status. Chapter 3 explains how the tools reach a decision.
How do the PLC link, reject and enclosure fit in?
The station decides; the line PLC acts. Results reach the PLC over digital outputs, typically pass or fail, result valid, ready and alarm, or over an industrial network that can also carry codes and recipe numbers. Output type must match the PLC input card: HIKROBOT’s VC3000 manual, for example, sets NPN outputs for source-type PLC inputs and PNP outputs for sink-type (VC3000 user manual).
The reject, an air jet, pusher, diverter, flap or line stop, and its confirmation sensor complete the chain; chapter 4 covers them. The enclosure is easy to underrate: a shroud keeps daylight and overhead lights out so the station’s own light is the only one that counts, rigid mounts stop the camera moving after set-up, and housings or lens caps with a suitable IP rating protect it where the line is washed down.
Record interfaces as well as the equipment list
The equipment list identifies what was bought; the interface record explains how it works together. Record which device owns the trigger, which signal means ready, how a recipe is acknowledged and how missing results are handled. Include the light controller and its settings: changing a lamp can alter the inspection even when the camera and software files are unchanged.
| Requirement or condition | Meaning or risk | Practical check |
|---|---|---|
| Camera and lens | Model suffix, sensor mode, focus and aperture | Known target image after replacement |
| Lighting | Geometry, current setting and trigger timing | Reference product under locked exposure |
| Processor | Software build, licence, recipe and calibration | Restore test on the approved configuration |
| PLC and reject | Signals, timeouts, item identity and feedback | End-to-end fault challenge with physical product |
Worked example — hypothetical: A hypothetical spare camera matches the original interface but has a different sensor format. Reusing the same lens may change the field of view, making the old regions and calibration invalid. The spare must be restored and validated as a configured measurement or inspection system, not accepted merely because a live image appears.
EMVA’s GenICam standard provides a common camera programming interface. It does not make every feature, lens, proprietary recipe or electrical connection interchangeable. JCGM 100 sections 4–6 are relevant when calibration and measurement uncertainty form part of the station’s output. (EMVA: GenICam) (JCGM 100:2008, Guide to the expression of uncertainty in measurement)
Keep the original image together with the active recipe and the reference judgement when investigating a disagreement. Change one controlled setting at a time and re-run the relevant challenge samples. Once a correction works, verify the neighbouring requirements too: improving contrast for one feature can hide another. Record the revised configuration and the evidence for release so maintenance does not have to reconstruct the decision from memory.
Frequently asked questions
What is the difference between a vision sensor and a smart camera?
Both put the camera, light and processing in one housing. A vision sensor has a small tool set for simple checks such as presence, orientation or a count. A smart camera runs a wider tool set, in HIKROBOT’s case a VisionMaster build with deep-learning modules, for labels, codes, text and defects on one view.
Does every vision station need a PC?
No. One simple check fits a vision sensor, and one or two checks on a single view often fit a smart camera. Several cameras, heavy deep-learning work, an operator screen or long image storage usually point to a vision controller or PC.
Is the reject mechanism part of the vision system?
It is part of the station, whether it comes with it or is already on the line. Each proposal lists the reject, the confirmation sensor and any PLC changes as included or excluded, so nothing is assumed.
Sources
- HIKROBOT VisionMaster product page (vendor page, checked 2026-10-03; application areas and control of cameras, I/O and light sources)
- HIKROBOT Machine Vision Smart Product Catalog, 2025 Q4 (PDF) (vendor catalogue, checked 2026-10-03; vision sensor, smart camera and vision controller series)
- HIKROBOT GigE Line Scan Camera User Manual (MVS client) (vendor manual, checked 2026-10-03)
- HIKROBOT VC3000 Series Vision Controller User Manual (PDF) (vendor manual, checked 2026-10-03; NPN or PNP output to match the PLC input)
- EMVA: GenICam (standard camera programming interface; read 2026-10-27)
- JCGM 100:2008, Guide to the expression of uncertainty in measurement (sections 4–6: uncertainty evaluation and combination; read 2026-10-27)