Development of Early Detection Model of Plant Growth Anomalies in Greenhouses Based on Autoencoder and Unsupervised Learning Modeling

Abstract

Accurate monitoring of plant growth is essential for improving productivity and enabling early detection of disturbances in modern agricultural systems, particularly in greenhouse environments. This study proposes the application of an Autoencoder, an unsupervised deep learning method, to detect anomalies in plant growth data based on five key parameters: plant height, width, leaf count, temperature, and humidity. The dataset was analyzed using reconstruction error to identify data points that deviate from normal growth patterns. Experimental results show that the model successfully identified 10.5% of the test data as anomalies, with the highest errors attributed to leaf count and plant height. This approach demonstrates advantages in detecting non-linear patterns without requiring labeled data and shows potential for integration into sensor-based automatic monitoring systems in greenhouses. The study contributes to the advancement of precision agriculture by introducing an efficient, data-driven anomaly detection method that is easily integrable with modern monitoring technologies.