Bottle Counting
to Catch Short Cases Before They Ship

Bottle counting with vision looks at each pack from above and detects every cap or container top. The model is trained on your packs, so it recognises containers through glare, shrink film and different cap colours. The count is compared with the target for the running product, and any empty slots are marked with their position. The pack passes or is rejected in the same cycle, and its image is stored. Counts for every product are also logged, which supports shift and batch records.

Beverage crates, open cases, trays, shrink-wrapped multipacks and cartons before closing are common formats. Containers can be glass or plastic bottles, cans, jars or tubs. The camera needs to see the top of each container, so cases are usually checked before the flaps close. Mixed formats on one line are handled by selecting the right target and layout for each product. Pack layouts are defined once per product, so changeovers load the right target automatically.

A checkweigher works well when container weights are consistent and light compared with the pack. When bottles vary in fill or glass weight, or the case itself is heavy, one missing bottle can fall inside the weight tolerance. Vision sees the empty slot directly, and many lines use both: weight for overall fill and vision for count and slot checks. Catching the gap at the packer keeps short-shipped cases away from customers.

Yes. Returnable crates often come back with missing bottles, wrong bottle types or foreign objects in the slots. The camera checks every slot before refilling or after packing, so crates with gaps or wrong items are rejected. Crate colour, wear and stacking marks are included in training so they do not affect the count. Separate classes for gaps, wrong bottles and foreign objects show which problem is most common.

Directly after the case packer or crate filler is usually best, so a short case is rejected before it is closed, wrapped and palletised. A top-down view over the outfeed conveyor sees every container top. Where cases are closed immediately, the check moves to the packer infeed or uses a view before the flaps fold. The position also needs space for a reject before the case reaches the palletiser.

Double-stacked bottles, tilted containers and items lying on their side can confuse simple counting. The model is trained on these conditions so it flags them instead of counting them as normal. Where height matters, such as a container sitting on top of another, a 3D sensor can be added to confirm one container per slot. Separating these conditions from true gaps keeps the reject rate honest.

Yes. Conveyor counting of loose bottles or containers tracks each item as it passes the camera and adds it to the production count. That supports shift records and filler output monitoring. Each item is counted once even when containers touch or bunch together, because tracking follows it across several images. The same camera can flag containers missing a cap or label while it counts.

Send images or samples of full and short packs, the pack formats and targets for each product, and the line speed. Tell us where the camera can sit and where rejects can go. With that, we can recommend camera position, lighting and integration for case count verification and container counting on your line. Short packs from real production are the most useful examples for training.