Improved Estimation of Leaf Wetness Duration Using Deep-Learning-Based Time-Resolution Technique

Plant diseases are one of the leading causes of the loss of crops. Monitoring the plant growth conditions could help prevent the spread of diseases and ensure healthy agricultural productivity. A leaf wetness sensor (LWS) is used to detect wetness on the surface of the leaves. Leaf wetness duration (LWD) is an essential parameter that governs the possibility of the occurrence of diseases in plants. The sensor system used in this work for monitoring the environmental conditions transmits data every 30 min due to power consumption considerations. To improve the time resolution of the signal and accurate estimation of LWD, a deep-learning-based architecture is proposed. The data from LWS has been used to train the architecture and the results obtained have been compared with the existing methods to analyze its performance. Using the proposed architecture, the signal from LWS was benchmarked with the commercially available Phythos-31. The time resolution of data has been improved from ±30 to ±4 min. The average signal-to-noise ratio (SNR) values for three test signals with improved time resolution have increased from 19.36 to 22.16 dB. Similarly, the average root means square error (RMSE) of the estimated LWD values after time resolution improved from 0.55 to 0.20. During the experimental analysis, it has been observed that the proposed architecture estimates the LWD values better compared to the other time-resolution techniques.