Development of a digital nose system for early detection of plant stress

Early forest stress detection is essential for the preservation
of healthy forest ecosystems. The aim of this project is to develop an electronic
nose (e-nose) using metal oxide (MOx) gas sensors that can differentiate
between stressed and healthy trees by detecting volatile organic
compounds (VOCs). An Arduino microcontroller was used to collect data
from the gas sensors, while Python was implemented for data processing.
The system applied machine learning algorithms such as Linear Discriminant
Analysis (LDA), as a supervisded learning method, and Principal
Component Analysis (PCA), as an unsupervisded learning method, to
classify and perform dimensionality reduction on the sensor data. To
enhance portability and usability, a printed circuit board (PCB) was
designed, creating a compact and efficient e-nose for field testing. The
sensor array was tested with various materials found in stressed trees.
PCA was also applied to assess sensor sensitivity and evaluate sensor
configurations. Initial results demonstrated the e-nose’s ability to distinguish
between diseased and healthy trees with significant accuracy. PCA
showed good separation of VOC patterns but lower accuracy when detecting
multiple target gases. LDA provided clearer distinctions between
the two classes with minimal overlap. Although MOx sensors exhibited
high sensitivity, their low selectivity for specific gases affected classification
accuracy. The high sensitivity of MOx sensors often comes at the
expense of selectivity. Future research will focus on identifying specific
VOCs emitted by stressed trees using neural networks and improving the
e-nose’s ability to detect a wider range of compounds.