What is foreign object detection and how does it work?

Foreign object detection evaluates the visible surface of products on the conveyor from camera images and flags objects that should not be there, such as plastic, glass, metal shavings, threads or insects. The result appears in the dashboard and a reject signal goes to line control; every decision is stored with its own image.

Does foreign object detection need extra hardware?

Yes, at the inspection point. It needs a fixed camera and controlled lighting; on transparent packaging and glossy surfaces, lighting directly determines the result. Video is processed either on an Algi Edge device on site or on ALGI AI servers, and the reject signal reaches line control through a dry contact. Your existing metal detector and X-ray line stay in place.

Are my existing IP cameras compatible?

IP cameras that provide an RTSP stream can be used, but image detail and lighting are what decide it: to catch small objects, the product surface has to fill the frame. If you share the product type, the smallest object size you want detected and the conveyor speed, we assess whether your current camera is sufficient before installation.

How long does setup take, and when does it go live?

Sample collection starts once the camera and lighting are in place. A meaningful inspection needs enough images of clean products and of products carrying objects from your own line, so the time depends on your production speed. After the model is adapted, the system runs first in reporting mode and then in reject mode.

Is foreign object detection GDPR and KVKK compliant?

Inspection looks at the product, not at people. The system does not recognize or identify operators and keeps no per-person records; the stored image is the product inspection frame. With an on-site Algi Edge device, live video stays on your premises; with server processing, video is processed on ALGI AI servers. If operators enter the frame, your existing duty to inform them about cameras still applies.

QUALITY CONTROL

Foreign Object Detection in Food Production

Foreign object detection flags a piece of plastic, a glass shard, metal shavings, a thread or an insect left on the surface of a product on the conveyor, and the item is rejected from the line. The check runs on every product before packing, using camera images. The camera sees the surface: it does not replace the X-ray system or metal detector that looks inside the product, but complements them by catching objects in the visible area that they miss. Every decision is stored with its image and written to the batch report.

Get a Quote Back to Quality Control

What is foreign object detection?

Foreign object detection is an inspection layer that uses camera images to single out objects that should not be on the visible surface of a product or its packaging. Plastic particles, glass shards, metal shavings, wood splinters, threads, hair and insects are defined as separate classes.

The limit of the method should be clear from the start: the camera sees the surface, not the inside of the product. Metal inside the product body needs a metal detector, and density differences need X-ray. The camera's strength begins where those are not enough. Plastic and glass that a metal detector does not register, and colour and texture differences that X-ray cannot distinguish, are caught on the visible surface. The three methods work side by side, not as replacements for one another, and your existing metal detector and X-ray line stay in place.

The system does not work from a generic list but from samples collected on your own line. It learns what your product normally looks like and flags deviations from that. Because the reference is your own product, its colour, texture and surface pattern are part of what the model treats as normal, and conveyor speed is taken into account as well.

Each flagged item keeps its inspection image, so a quality team can see exactly what was rejected and why, rather than relying on a count alone.

How does foreign object detection work on the line?

Setup follows three steps.

The inspection point is set up. The camera is placed at the pre-packing point of the conveyor with fixed lighting. For transparent packaging and glossy surfaces, angled lighting that breaks up reflections is used; glass and clear plastic can only be distinguished with the right lighting.

Samples are collected. Images of clean products and of products carrying foreign objects are collected from your line, and the object types are labelled. With this data the model is adapted to your product's colour, texture and conveyor speed.

The decision is connected to the line. Video is processed either on an Algi Edge device installed on site or on ALGI AI servers. The inspection result appears in the dashboard, and at the same time a reject signal goes to line control through a dry contact, so a pusher removes the product from the conveyor. If preferred, decisions are only reported and rejection is left to the operator.

Every decision is stored with the product image and a timestamp. Reported by shift, line and object type, this record goes into your HACCP file and serves as evidence in customer audits. A recurring object type also points to the station it comes from.

Which industries use it?

Food & Beverage → Industrial Manufacturing →

Frequently asked questions

What is foreign object detection and how does it work?: Foreign object detection evaluates the visible surface of products on the conveyor from camera images and flags objects that should not be there, such as plastic, glass, metal shavings, threads or insects. The result appears in the dashboard and a reject signal goes to line control; every decision is stored with its own image.
Does foreign object detection need extra hardware?: Yes, at the inspection point. It needs a fixed camera and controlled lighting; on transparent packaging and glossy surfaces, lighting directly determines the result. Video is processed either on an Algi Edge device on site or on ALGI AI servers, and the reject signal reaches line control through a dry contact. Your existing metal detector and X-ray line stay in place.
Are my existing IP cameras compatible?: IP cameras that provide an RTSP stream can be used, but image detail and lighting are what decide it: to catch small objects, the product surface has to fill the frame. If you share the product type, the smallest object size you want detected and the conveyor speed, we assess whether your current camera is sufficient before installation.
How long does setup take, and when does it go live?: Sample collection starts once the camera and lighting are in place. A meaningful inspection needs enough images of clean products and of products carrying objects from your own line, so the time depends on your production speed. After the model is adapted, the system runs first in reporting mode and then in reject mode.
Is foreign object detection GDPR and KVKK compliant?: Inspection looks at the product, not at people. The system does not recognize or identify operators and keeps no per-person records; the stored image is the product inspection frame. With an on-site Algi Edge device, live video stays on your premises; with server processing, video is processed on ALGI AI servers. If operators enter the frame, your existing duty to inform them about cameras still applies.

Related rules from the library

ALL →

PPE_DETECTION

Dijital Atölye listens to the problem in the field, prototypes, validates and productizes. Some of the products running in the field today were born on this line.

Bring the problem, let's talk prototype →