Case Study

Multi-camera orange blemish detection & sorting.

An automated orange inspection and sorting system using image processing and multiple cameras to grade fruit by colour and surface characteristics. Oranges move individually along a conveyor while multiple cameras cover every side; a processor-based board detects blemishes and yellow/green/dark areas, classifies each orange (good, bad, unripe, over-ripe) and diverts it to the respective container.

Talk to deliveryAll case studies

Industry

Agriculture

Region

Agriculture / Image Processing / Automation

The project focuses on developing an automated system that detects the grade and quality of oranges based on their colour and visible surface characteristics. Image processing identifies blemishes, yellow areas, green areas and dark areas; based on the detected characteristics, oranges are separated by quality and sent to their respective containers. The solution is designed for faster automated processing and uses multiple cameras to cover all sides of an orange.

Business requirements & challenges

The system had to grade oranges by colour, on purpose-built hardware, and sort them automatically at speed — while seeing every side of the fruit.

Key requirements

Automatically detect the orange grade based on its colour using image processing
Design and develop hardware — with camera interface logic — to implement the algorithm
Automatically send each orange to its respective container based on detected quality
Use multiple cameras to cover all sides of an orange, so surface issues aren't missed

Major challenges

Process oranges fast enough — a challenge for traditional processors on the market
Ensure only one orange appears in the camera frame at a time

Solution

A conveyor-belt system moves oranges through the inspection area one by one, multiple cameras cover all sides of each fruit, and a processor-based board implements the algorithm and improves image-processing results — detecting blemishes and yellow, green and dark areas.

Conveyor-belt system presents oranges one at a time to the cameras
Multiple cameras cover all sides of the orange for complete surface coverage
Processor-based board implements the algorithm and improves image-processing results
Detects blemishes and yellow, green and dark areas to judge quality and ripeness
Throughput extensible via a high-end processor, an optimised algorithm and a better camera module

Stack

Technologies used

Image ProcessingProcessor-Based BoardCamera ModuleMultiple CamerasCamera Interface LogicConveyor Belt SystemAutomated SortingQuality DetectionColor-Based Analysis

How it works

Process workflow

01Orange on conveyor belt — oranges placed on a belt and moved toward the inspection system
02Detect orange using camera module — multiple cameras capture the orange, one at a time, from all sides
03Image processing — analyse the visible characteristics, looking for blemishes and yellow, green and dark regions
04Quality classification — separate into good quality, bad quality, unripe and over-ripe
05Automated sorting — send each orange to its respective container

What's new

Key innovations

Multi-camera inspection

Multiple cameras cover all sides of the orange, so surface issues aren't missed by relying on one view.

Colour-based analysis

Determines quality from the orange's colour and visible surface characteristics.

Blemish & area detection

Detects blemishes and yellow, green and dark areas across the fruit surface.

Quality & ripeness classification

Classifies each orange as good, bad, unripe or over-ripe.

Processor-based processing

A dedicated processor-based board implements the algorithm and improves image-processing performance.

Automated sorting

Each orange is automatically directed to its respective container.

Business benefits

Increased accuracy and efficiency of orange sorting — helping customers find more end users for their products
Increased revenue — supplying good-quality fruit to the juice-making industry
Reduced manual labour cost — automation cuts the manual effort in the sorting process

At a glance

Key highlights

Region	Spain
Primary Objective	Automated orange quality grading & sorting
Quality Parameter	Colour + visible surface characteristics
Detection	Blemish, yellow, green, dark areas
Classification	Good, bad, unripe, over-ripe
Input	Camera images
Camera Setup	Multiple cameras
Coverage	All sides of orange
Feeding System	Conveyor belt
Processing	Processor-based board
Output	Respective container
Future Improvement	High-end processor + optimised algorithm + better camera
Business Value	Accuracy, efficiency, revenue opportunity, reduced labour cost

More from agriculture.

See all case studies

Hyperspectral yield prediction for wheat & rice

Agriculture

Featured

Hyperspectral yield prediction for wheat & rice

Hyperspectral image processing at the Plant Phenomics Centre, IARI — enabling accurate, early-stage, non-destructive crop-yield estimation for wheat and rice by correlating spectral signatures (photosynthetic efficiency, chlorophyll, water stress) with final yield.

Plant Phenomics Centre, IARI

Agriculture

Multi-modal phenotyping — disease, nitrogen & biomass

A comprehensive AI-based image-processing framework at the Plant Phenomics Centre, IARI — integrating RGB, hyperspectral and NIR imaging across controlled and field environments to assess wheat and rice traits: biomass, near real-time disease detection and scoring, nitrogen and water-level classification, NDVI soil-vegetation segmentation and trait-specific spectral analysis, feeding the AgriX decision-support tools.

Plant Phenomics Centre, IARI

Agri & Manufacturing

FPGA-based cashew grading — 10 cashews/sec automated

An automated cashew-grading system that uses machine learning and image processing to grade cashews by size and identify bad-quality nuts. A conveyor belt positions one cashew per camera frame while a custom FPGA-based board processes images and grades at ~10 cashews/second, automatically diverting each nut to its respective container.

Indian Cashew Processing Industry

Make this kind of impact at your organisation.

Talk to us

// loading…

Multi-camera orange blemish detection & sorting.

Business requirements & challenges

The system had to grade oranges by colour, on purpose-built hardware, and sort them automatically at speed — while seeing every side of the fruit.

Key requirements

Automatically detect the orange grade based on its colour using image processing

Design and develop hardware — with camera interface logic — to implement the algorithm

Automatically send each orange to its respective container based on detected quality

Use multiple cameras to cover all sides of an orange, so surface issues aren't missed

Major challenges

Process oranges fast enough — a challenge for traditional processors on the market

Ensure only one orange appears in the camera frame at a time

Solution

Conveyor-belt system presents oranges one at a time to the cameras

Multiple cameras cover all sides of the orange for complete surface coverage

Processor-based board implements the algorithm and improves image-processing results

Detects blemishes and yellow, green and dark areas to judge quality and ripeness

Throughput extensible via a high-end processor, an optimised algorithm and a better camera module

Process workflow

01Orange on conveyor belt — oranges placed on a belt and moved toward the inspection system

02Detect orange using camera module — multiple cameras capture the orange, one at a time, from all sides

03Image processing — analyse the visible characteristics, looking for blemishes and yellow, green and dark regions

04Quality classification — separate into good quality, bad quality, unripe and over-ripe

05Automated sorting — send each orange to its respective container

Key innovations

Multi-camera inspection

Multiple cameras cover all sides of the orange, so surface issues aren't missed by relying on one view.

Colour-based analysis

Determines quality from the orange's colour and visible surface characteristics.

Blemish & area detection

Detects blemishes and yellow, green and dark areas across the fruit surface.

Quality & ripeness classification

Classifies each orange as good, bad, unripe or over-ripe.

Processor-based processing

A dedicated processor-based board implements the algorithm and improves image-processing performance.

Automated sorting

Each orange is automatically directed to its respective container.

Key highlights

Region	Spain
Primary Objective	Automated orange quality grading & sorting
Quality Parameter	Colour + visible surface characteristics
Detection	Blemish, yellow, green, dark areas
Classification	Good, bad, unripe, over-ripe
Input	Camera images
Camera Setup	Multiple cameras
Coverage	All sides of orange
Feeding System	Conveyor belt
Processing	Processor-based board
Output	Respective container
Future Improvement	High-end processor + optimised algorithm + better camera
Business Value	Accuracy, efficiency, revenue opportunity, reduced labour cost