Related Experiment Video
Updated: May 11, 2025
08:18
Non-Aqueous Isolation and Enrichment of Glandular Capitate Stalked and Sessile Trichomes from Cannabis sativa
Published on: May 12, 2023
1.9K
Identification of non-glandular trichome hairs in cannabis using vision-based deep learning methods
Alon Zvirin1, Amitzur Shapira2, Emma Attal1
1Computer Science Department, Technion - Israel Institute of Technology, Haifa, Israel.
Journal of Forensic Sciences
|April 18, 2025
Summary
A new deep learning method accurately identifies cannabis using microscopic images of trichome hairs. This AI tool aids forensic labs in distinguishing real cannabis from illicit substitutes, improving drug trafficking investigations.
Area of Science:
- Forensic Science
- Computer Vision
- Botany
Background:
- Accurate identification of cannabis and its substitutes is crucial for forensic laboratories and law enforcement due to the harmful effects of these substances.
- Distinguishing genuine cannabis from non-cannabis plant material adulterated with synthetic cannabinoids is challenging using visual inspection alone.
- Current forensic identification methods, including colorimetric tests and expert microscopic analysis, are time-consuming and resource-intensive.
Purpose of the Study:
- To develop and validate a novel deep learning-based computer vision method for the accurate identification of non-glandular trichome hairs in cannabis.
- To provide an efficient and reliable tool for forensic departments and law enforcement agencies to combat illicit drug trafficking.
Main Methods:
- Collected a dataset of several thousand annotated microscope images of genuine cannabis and non-cannabis plant material.
- Utilized a deep learning-based computer vision approach for image analysis and classification.
- Established ground-truth labels using a combination of forensic tests, chemical assays, and expert microscopic analysis.
Main Results:
- The proposed deep learning method achieved an accuracy exceeding 97% in distinguishing cannabis from non-cannabis plant material.
- The AI model reliably identified non-glandular trichome hairs based on microscopic features.
- The method demonstrated potential to reduce reliance on traditional, labor-intensive forensic analysis techniques.
Conclusions:
- Deep learning offers a highly accurate and efficient solution for identifying cannabis through microscopic trichome analysis.
- This AI framework can significantly enhance the capabilities of forensic laboratories and law enforcement in detecting illicit cannabis products.
- The developed tool supports efforts to combat drug-related crimes and illicit drug trafficking.
