1. Chlorophyll content detection using a low-cost hyperspectral imaging in Cannabis Sativa L.
- Author
-
Manggala, Braja, Chaichana, Chatchawan, Kornievsky, Mikhail, Fitriyah, Nursaadah Syahro, Wanison, Ramnarong, and Syahputra, Wahyu Nurkholis Hadi
- Subjects
- *
ARTIFICIAL neural networks , *CANNABIS (Genus) , *HYPERSPECTRAL imaging systems , *CHLOROPHYLL , *CROPS , *MEDICAL marijuana - Abstract
Cannabis Sativa L. has been clinically proven to be used for antioxidant, anti-inflammatory, antiemetic, parasitic, antibacterial, antiviral, antifungal, antipyretic, painkiller, etc. Consequently, this plant began to be widely produced as an agricultural crop with selling value. More than that, the production of medicinal cannabis products has been permitted in several countries across the world. Furthermore, detecting the chlorophyll content is very important to monitoring the quality of cannabis leaves since the chlorophyll content plays an essential role in determining photosynthetic capacity and plant growth. Chlorophyll content can be assessed using a bench-top method, which might give an accurate and stable result. However, the bench-top method is destructive, costly, and time-consuming. The hand-held SPAD chlorophyll meter is widely used nowadays as a non-destructive instrument. Nevertheless, the price of this device is not affordable, especially for small holder. This work aims to develop a relatively cheap, non-invasive, and accurate method for chlorophyll content detection. A low-cost Hyperspectral Imaging (HI) based on the mobile phone was used in this research. Low-cost HI can read reflected values at visible (400 - 700 nm) wavelengths by applying digital filters to photographs. The reflectance value at each wavelength was plotted and considered equal with the hyperspectral imaging system method afterward. Then, Artificial Neural Network (ANN) was used to analyze spectral data and generate a model prediction to predict unknown data. The best model predictions were built with 10- and 50-layer sizes. The coefficient determination (R2) exhibited acceptable model predictions of 0.785 and 0.8, with root-mean-squared deviation (RMSE) of 3.396 and 3.439, respectively. A low-cost HI device has great potential to detect chlorophyll content on the plant; hence pre-processing and resolution quality should be improved. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF