Automated Detection of Culture Vessel Contamination in Plant Tissue Culture Using Machine Learning

The rising demand for reliable plant tissue cultures in agriculture underscores the critical need for early detection of microbial contamination in culture vessels. Conventional detection methods often lack sensitivity in identifying contamination at initial stages, which can result in compromised cultures and substantial economic losses. This research focuses on developing an automated detection system for microbial contamination in plant tissue cultures. The study involved the design and implementation of a macroscopic image acquisition prototype of culture vessel contamination of Oryza sativa. The acquired images underwent preprocessing, including normalization and enhancement techniques, followed by feature engineering for effective data extraction of color and textural features. The extracted features were subsequently analyzed using supervised and unsupervised machine learning models. Supervised machine learning models such as support vector machine, K-nearest neighbors and Discriminant analysis showed high accuracy with Discriminant analysis scoring 99.6%, which is the highest. K-means clustering accurately identified the grouping by distinguishing contaminated from non-contaminated samples, while the Isolation forest indicated the presence of microbial contamination. Overall, the approach presented in this research proves the capability of detecting microbial contamination with macroscopic image processing and machine learning.