We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
KI und Qualität - Detektion und Entfernung Pyrrolizidinalkaloidhaltiger Beikräuter im Erntegut mittels sensorgestützter Sortierung.
- Authors
Kronenwett, Felix; Maier, Georg; Schulte, Henning; Tron, Nanina; Krähmer, Andrea
- Abstract
Pyrrolizidine alkaloids (PAs) are secondary plant substances and primarily serve to protect the plant against predators. When PA-containing weeds are co-harvested in crop production, PAs enter food (salads, herbs, teas) or herbal medicines as contaminants. Due to their liver-toxic and genotoxic mode of action, they represent a potential health hazard. PA plants are a particular problem in the cultivation of medicinal and aromatic plants, as just a few plants per hectare are enough to make the harvest unusable for trade. Control or reduction of weeds already in the field is usually only possible mechanically and with very high labour costs. Increasingly, this is no longer economically viable for growers. In addition, imported, dried goods are often contaminated with higher contents of PAs than permitted. There is currently no way to remove plant parts containing PAs from the goods. The joint research project "Detection and removal of weeds containing pyrrolizidine alkaloids from cultivated plants after harvest - PA-NIRSort", funded by the Federal Ministry of Food and Agriculture (BMEL) and the Agency for Renewable Resources (FNR) (FKZ 220132165), should investigate possible solutions for quality control. With the development of an automated post-harvest detection and removal of contaminating, toxic PA weeds, an efficient improvement of quality control can be created. This was to be developed on the basis of hyperspectral near-infrared spectroscopy (hyperspectral-NIRS), in combination with a compressed air sorting unit. Therefore, a method for the detection, classification and physical ejection of PA-containing plant parts in a material stream of medicinal and spice plants (nettle, lemon balm and mint) was developed and implemented. The detection is based on hyperspectral imaging sensors in the short-wave infrared range. The obtained hyperspectral data form the basis for training an AI-based classification model. Classification using NIR spectroscopy achieves a detection rate of over 90 %. With the help of the detection model, a sensor-based sorting system was developed with the aim of physically ejecting PA-containing plant parts and purifying the stream of plant material. The focus was on optimising the material transport on the conveyor belt. For this purpose, an air flow generated from compressed air was used to calm the material. Thus, a sensor-based sorting process was successfully implemented by means of pneumatic fast-switching valves.
- Subjects
LEMON balm; CULTIVATED plants; PYRROLIZIDINES; NEAR infrared spectroscopy; WEED control; HYPERSPECTRAL imaging systems; WEEDS; AROMATIC plants
- Publication
Julius-Kühn-Archiv, 2023, Issue 476, p21
- ISSN
1868-9892
- Publication type
Article
- DOI
10.5073/20230821-134542-0