---
lang: en
title: "Automated visual quality inspection at line speed"
url: https://nabrio.com/use-cases/visual-quality-inspection
canonical: https://nabrio.com/use-cases/visual-quality-inspection
updated: 2026-10-27
publisher: Nabrio Co., Ltd. (nabrio.com)
---

# Automated visual quality inspection at line speed

Nara inspects every item at the inspection point for defects, colour, contaminants, size, presence, codes and printed text. Passing items are logged; failing items are rejected through your PLC, with an alert and an image for the record.

- Sectors: [Manufacturing](https://nabrio.com/industries/manufacturing.md), [More sectors](https://nabrio.com/industries/more-sectors.md)
- What it takes: Trained on your items. We train a model on your own products or defects in Reva, and Nara runs it on the line. Colour, code and text checks work out of the box. Defects specific to your product are trained in Reva.
- Products: [Nara](https://nabrio.com/nara.md), [Reva](https://nabrio.com/reva.md)

## The problem

Visual inspection by people is inconsistent and tiring, and it cannot keep up with line speed or cover every item.

## How it works

- **Input:** Camera at the inspection point, triggered by a part sensor
- **Process:** Checks for defects, colour, contaminants, size, presence, codes and text
- **Action:** Pass: log. Fail: reject through the PLC, switch the tower light, alert and log with an image

A part sensor triggers the camera as each item arrives. Nara runs the checks you configure: a custom defect or classification model trained in Reva, colour thresholds, code reading or OCR. If/switch logic decides pass or fail, writes the result to the PLC over Modbus, switches a tower light, and records the result and image in the database.

## What you need

- A Basler industrial or IP camera at the inspection point, with consistent lighting
- A part sensor to trigger capture
- A PLC or rejector reachable over Modbus or TCP/IP
- For product-specific defects: sample images to train a model in Reva

## Nara building blocks

- Basler and IP camera nodes
- Custom detection or classification model (Reva)
- Colour detection and thresholds
- Code reader
- OCR
- If / switch
- Modbus write to PLC
- Tower light
- Database table

AI models: [OCR (text reading)](https://nabrio.com/nara/ai-models.md), [Barcode and QR code reading](https://nabrio.com/nara/ai-models.md)

## What it improves

- Escape rate (defects reaching customers)
- Inspection staff needed
- Scrap and rework
- An audit trail for every item

## Where it runs

- [Chili powder graded by colour, with no training data](https://nabrio.com/customers/chili-powder-colour-grading.md) (A food producer)
- [AI vision added to an existing conveyor](https://nabrio.com/customers/conveyor-sorting-retrofit.md) (A conveyor system operator)

## Frequently asked questions

### Do we need training data?

Not always. Colour, code and text checks use Nara’s built-in tools with no training. Defects specific to your product need a model trained in Reva on your own images.

### Can it reject items automatically?

Yes. Nara writes the result to your PLC or rejector over Modbus or TCP/IP, so failing items are removed from the line.

## Related use cases

- [Defect detection and rejection](https://nabrio.com/use-cases/defect-detection-and-rejection.md)
- [Foreign object detection](https://nabrio.com/use-cases/foreign-object-detection.md)
- [Sorting and grading](https://nabrio.com/use-cases/sorting-and-grading.md)

Talk to us about visual quality inspection: sales@nabrio.com. [Download Nara](https://nabrio.com/download.md).