AI-based foreign-matter (NTRM) detection and visual analysis on the raw-tobacco line
The raw-tobacco bed passes through a camera and lighting cabin at belt speed. An AI model finds and classifies non-tobacco material such as plastic, paper or string in the image; a signal goes to the line control and the reject gate opens.
Class, position and time of every detection go to the line control; the gate diverts the item into the bin, and the class counters accumulate on the operator display and in the shift report.
- 01 Flow
- 02 Detect
- 03 Signal
- 04 Reject
The tobacco bed passes through the cabin at belt speed.
- Class
- —
- Size
- —
- Detections
- —
Scenes are illustrative simulations; they are not customer references or performance guarantees.
Need
The tobacco bed is irregular, multi-layered and variable in brown tones; string, paper and plastic fragments stay partly buried among the leaves and their shapes never repeat. A classic threshold or colour rule either misses them or raises constant false alarms.
Solution approach
The belt passes through a cabin that blocks ambient light; bar lights from both sides illuminate the bed without shadows and a downward-looking camera shoots at a period matched to belt speed. An AI object-detection model boxes and classifies the non-tobacco item in the image; position and time are passed to the line control and the reject gate opens when the item reaches the end of the belt.
Process steps
- 01
Flow
The tobacco bed passes through the cabin at belt speed.
- 02
Detect
The AI model finds the non-tobacco item in the image and boxes it.
- 03
Signal
The item is classified; its position and time are sent to the line control.
- 04
Reject
At the end of the belt the reject gate opens and the item drops into the bin.
Related applications
Let us assess your application together
We test your sample in our laboratory and report on resolution, optics and lighting selection.