Machine Vision

AOI Machine Components and Vision System Architecture

A component-focused guide for automated optical inspection systems.

An AOI machine is a complete inspection architecture, not a camera mounted above a conveyor. Reliable inspection depends on how mechanical presentation, optics, lighting, software, controls, and data handling work together. If one part of the architecture is weak, the system may produce false rejects, miss defects, or become difficult to maintain during production changes.

The most important design principle is repeatability. The same product must appear in the same position, under the same lighting, with the same camera view, at the same inspection timing. Machine vision software is powerful, but it cannot fully compensate for unstable part location, vibration, glare, poor focus, or uncontrolled ambient light.

Image Acquisition Components

The camera captures the inspection image. Selection depends on required resolution, inspection speed, field of view, sensor type, interface, and trigger method. A small defect requires enough pixels across the defect to measure it reliably. A fast line requires a camera and lighting trigger that can freeze motion without blur.

The lens determines magnification, working distance, depth of field, and distortion. A wide field of view may inspect more area but reduce detail. A telecentric lens may be needed for precise measurement because it reduces perspective error. Lens selection should be based on the inspection feature, not only camera resolution.

Lighting and Fixture Design

Lighting is often the difference between a stable AOI machine and a frustrating one. Ring lights, bar lights, coaxial lights, dome lights, backlights, and low-angle dark-field lights all reveal different features. A scratch, solder bridge, embossed mark, or transparent edge may require a specific light angle before the software can detect it.

Fixtures are equally important. A part that shifts by a few millimetres may move the feature outside the expected inspection window. For flexible products, reflective parts, or assemblies with height variation, fixture design must support the optical plan.

Software, Controls, and Reject Handling

Vision software uses tools such as pattern matching, measurement, edge finding, blob analysis, OCR, colour inspection, and classification. The result is normally sent to a PLC or controller. If the result fails, the machine may reject the part, stop the line, trigger an alarm, or request operator review.

Reject handling must be designed carefully. A good reject signal is useless if the wrong part is removed from the conveyor. Timing, encoder tracking, part spacing, and reject confirmation all matter. The main automated optical inspection guide explains how these components support the full AOI workflow.

Maintenance and Changeover

AOI machines must remain stable after cleaning, product changeover, lighting replacement, camera adjustment, or software edits. Good systems include calibration procedures, access control, recipe version management, and image review tools. Operators should know which settings they can adjust and which require engineering approval.

For Malaysian manufacturers running multiple product models, architecture should consider future recipes and fixture changes. A machine built only for one product may become difficult to adapt when the production mix changes.

Choosing the Architecture for the Defect

The architecture should be selected after the defect is defined. A surface scratch on glossy plastic needs a different optical setup from a missing screw, wrong label, solder bridge, or connector alignment issue. If the defect is three-dimensional, the system may need multiple camera angles, structured light, or height measurement. If the defect is colour-based, lighting colour and camera colour response become important.

Cycle time should also be reviewed early. A system that works during offline trials may fail to keep up with production if image capture, processing, communication, or reject timing is too slow. The architecture must match both inspection accuracy and line speed.

Service access should be included in the architecture review. Cameras, lenses, light modules, filters, and fixture surfaces may need cleaning or replacement. If these parts are difficult to access, maintenance teams may disturb alignment during routine work, creating inspection drift after the machine is returned to production.

Control and Data Boundaries

A practical AOI design also defines which decisions belong to the vision system and which belong to the line controller. The vision system may classify the defect, but the PLC may decide whether to reject, stop, or request manual review. Result data should be structured so quality teams can separate product failures from machine faults, recipe errors, and communication failures. Clear boundaries make troubleshooting faster when production pressure is high.