Vision-guided robots that can locate, inspect and sort.
Machine vision can let a robot find products that are not presented in a fixed position, verify attributes before handling and route parts according to inspection or identification results.
Define what the camera must decide before selecting hardware.
A vision project starts with the decision the system must make: where is the part, which way is it facing, is a feature present, or which destination should it reach? Samples must represent normal variation, defects and surface conditions.
Camera resolution alone does not determine performance. Lens selection, field of view, working distance, lighting, motion blur, product contrast and calibration all matter.
- 2D part location and orientation
- Vision-guided conveyor picking
- Presence and feature inspection
- Barcode, QR and character reading
- Colour or variant classification
- Robot-to-camera calibration
Lighting and product presentation are often the decisive factors.
Reflective, transparent, dark or highly variable products can require controlled illumination and a carefully designed imaging station. The cell should shield the inspection from ambient-light changes where those changes could affect the result.
For moving-conveyor applications, encoder tracking and timing must coordinate the image, position calculation and robot interception point.
- Lighting trials with representative products
- Fixed or robot-mounted camera arrangement
- Background and contrast control
- Calibration and coordinate transforms
- Conveyor tracking
- Reject routing and result logging
Test the complete range of good and bad product conditions.
A successful demonstration with one ideal sample is not enough. Validation should cover the expected product range, presentation variation, defects, contamination and environmental conditions.
Acceptance criteria should define detection performance, false-reject expectations, cycle time and what happens when the system cannot reach a confident decision.
- Representative sample set
- Defined pass/fail criteria
- Uncertain-result handling
- Cycle-time verification
- Access for cleaning and adjustment
- Backup of jobs, calibration and parameters
Frequently asked questions
These answers provide an initial planning framework. Final requirements are confirmed against the product, process, site and applicable safety obligations.
Can a robot pick randomly positioned products?
Often, if the camera can reliably locate each product, identify a valid gripping point and provide coordinates quickly enough for the required cycle. Overlap and occlusion can limit performance.
Is 2D or 3D vision required?
2D vision is effective for many flat or controlled-height applications. 3D may be needed when height varies, products overlap, surfaces are complex or a depth map is required.
Can vision inspect a part while guiding the robot?
Yes. One system can sometimes locate a product and check features, although the imaging geometry and resolution must support both tasks within the cycle time.
What samples are needed for a vision trial?
Provide the full range of good products, known defects, colours, finishes, packaging variations and any contamination or presentation conditions expected in production.
Continue planning your automation project
Use these pages to compare approaches, define the application and prepare the information needed for a useful feasibility review.
How to choose an industrial robot
A practical selection framework covering payload, work envelope, cycle, safety and interfaces.
Read more →Robot automation cost
Understand whole-cell cost drivers before comparing project quotations.
Read more →Discuss your application
Send product, process and output information for an initial engineering review.
Read more →3D vision bin picking
Combine 3D sensing, grasp selection and collision-aware motion.
Explore →Robotic quality inspection
Move sensors or products through controlled inspection views.
Explore →Discuss the production task with an automation engineer.
Send the product details, required output, available space and a photo or short video of the current process. We will review the application and identify the next engineering step.