An Embedded System for Plant Transpiration Rate Detection Based on Machine Learning and Flexible Sensing

Accurately measuring the transpiration rate of plants is of great significance for precision agriculture and water resource management. The objective of this study is to develop and validate an embedded system that combines flexible humidity sensors with machine learning algorithms for real-time monitoring of plant transpiration. We used Parylene as the base, combined with Cr/Au interdigitated electrodes, and prepared a graphene oxide (GO) sensitive membrane through chemical vapor deposition and spin coating methods to create a flexible humidity sensor. Subsequently, we conducted systematic tests on the sensor's performance under different bending angles (0 to 90), different temperatures and humidity conditions (30% to 80% relative humidity), and determined its optimal working frequency to be 100 Hz. In the experiment, we used Epipremnum aureum to monitor the water status of plants, including sensor consistency tests, correlation analysis between sensor capacitance and plant physiological indicators (such as net photosynthetic rate, stomatal conductance, and transpiration rate), and drought stress experiments. Based on this, we developed an embedded system: initially based on the Arduino UNO platform, later upgraded to the STM32F407VET6 chip, and introduced the random forest algorithm in the system for predictive modeling. The results show that this system can accurately predict the transpiration state of plants, and the random forest model demonstrates high accuracy in processing time series data. This study provides valuable ideas for combining flexible electronics with machine learning for plant phenotypic analysis and has practical application significance for intelligent agriculture.