1 / 5

AI & Computer Vision

Count it. Classify it. Keep a record of every batch.

LUKK AI Video Analytics

Counting and classification with fixed cameras and an on-site AI server, for poultry, agriculture and minerals.

Count and classify: Live or dead, good or rotten, one type or another, from the same camera view
Processed on site: The AI runs next to the cameras. Full video does not have to leave the facility
No line changes: Fixed cameras placed around your existing conveyor or belt
A record per batch: Time-stamped results linked to each consignment, batch or sample

Enquire Now WhatsApp

Local support in SG · MY · IN
Customised to your site
Hardware + AI + software

Overview

Meet the LUKK AI Video Analytics

Fixed cameras and an AI server on site count what passes the line and tell you what it is: live or dead birds, good or rotten fruit, one mineral type or another. Every result is time-stamped, tied to a batch and shown on a web dashboard.

The problem we solve

Counting and sorting by eye is slow, tiring and hard to check later. Totals are written on a sheet at the end of a load, quality is judged by whoever happens to be on shift, and there is little record to look back on if a figure is questioned.

Poultry: live and dead birds are counted by hand at unloading, so mortality figures depend on one person and one moment.
Agriculture: counting coconuts or oranges and picking out rotten fruit is repetitive work that varies from person to person and shift to shift.
Minerals: telling mineral types apart by eye needs experience, and results are hard to compare across shifts and sites.
Across all three: no time-stamped record to review, no easy comparison between batches, and problems found late.

How it works

Step 1: Assess: We visit the site and look at the line, camera positions, lighting, power and connectivity. Nothing is chosen before this.
Step 2: Install: Fixed cameras and the edge AI server go in, with an independent connection to the cloud. The conveyor or belt is not modified.
Step 3: Tune: We record video from your own line and tune the model on it. Counting zones and limits are set to your rules.
Step 4: Prove and run: Automated results are compared with manual counts in an operational trial. Once validated, the system runs and reports every batch.

Security, compliance and trust

Access control for the dashboard and system, with authentication for administrators.
Video is processed on site. Count events, not a continuous video stream, are what is sent.
Protected communication between the site and the cloud service.
Hardened configuration of the edge equipment, with monitoring of system and security events.
Controls on stored data and a defined process for incident reporting. Data retention and hosting location are set per project.
bizSAFE Level 3, Certificate No. E49709. Seagate Technology Partner, Partner Program.

Features

Built for real operations

Poultry: Live & Dead Counting

Counts live birds on the conveyor. Counts dead birds placed into the collection area, separately. Works out mortality per consignment (dead divided by live). Keeps a record for each lorry-load.

Agriculture: Counting & Quality

Counts fruit and produce as it passes the camera, for example coconuts or oranges. Marks each item as good or rotten from its appearance. Shows the good and rotten split for each batch. Flags batches above a limit you set.

Minerals: Type Detection

Identifies mineral types from the camera image. Counts pieces of each type. Keeps a record for each sample or batch.

On-site AI & Connectivity

Detection, tracking and counting run on a server next to the cameras. Count events are held locally if the link drops and sent on when it returns. Own connection, cellular or wired, so the facility network is not needed for core operation.

Dashboard & Reports

Live and recorded counts in a browser on a PC or tablet. Batch history and trends. Limits and alerts configured to your requirements. Export and reporting.

Accuracy & Access

Models tuned on video from your own line. Results compared with manual counts in an operational trial before sign-off. Controlled access for authorised users only.

At a glance

Count and classify: Live or dead, good or rotten, one type or another, from the same camera view
Processed on site: The AI runs next to the cameras. Full video does not have to leave the facility

No line changes: Fixed cameras placed around your existing conveyor or belt
A record per batch: Time-stamped results linked to each consignment, batch or sample

Need more details? Ask us →

Who it's for

Poultry receiving hubs and farms: Live and dead bird counts per lorry-load, mortality trends, a record that can be reviewed later.

Fruit and coconut growers and packers: Counting at intake or on the packing line, spotting rotten fruit, batch quality figures.

Food processors and distributors: Checking incoming produce and keeping a record per supplier delivery.

Mining, quarrying and mineral processing: Telling mineral types apart on a belt or sample table, counting pieces by type.

Operations, quality and IT teams: One system, one dashboard, controlled access and an audit record.

FAQ

Questions, answered

Do we have to change how we operate?+

No. The system is designed around your existing line. Cameras and equipment are placed so they do not obstruct people, products or other equipment, and the conveyor or belt is not modified.

Can it count and classify at the same time?+

Yes. The same camera view gives the count and the classification, for example live and dead birds, good and rotten fruit, or mineral type.

Which fruit and minerals can it handle?+

Coconuts and oranges are the examples on this page. Other produce and mineral types are assessed at the site assessment, using video from your own line.

How accurate is it?+

Accuracy depends on the camera view, lighting and how clearly items can be seen. It is measured on your own site against manual counts during the operational trial, not assumed in advance.

What if the connection drops?+

Counting carries on and events are stored locally, then synchronised when the connection returns.

Does it need our network?+

Not for core operation. The edge server uses its own connection to the cloud.

What could reduce accuracy?+

A moved camera, a changed angle, poor or changing light, or heavy obstruction of the counting area. Mounts are secure, the view is fixed and maintenance checks include the camera set-up.

How long does implementation take?+

A typical poultry programme runs sixteen weeks, including a four-week operational trial. Agriculture and minerals are scoped after the site assessment.

Enquire

Interested in the LUKK AI Video Analytics?

Ask for a quotation, more details or a demo. Tell us a little about your site and our engineers will come back with the right configuration.