Machine Learning based Biospeckle Technique for Identification of Seed Viability using Spatio-temporal Analysis

Viability assessment is one of the most important parameters for ensuring high crop yield. Hence, in this work, a machine learning (ML) based automatic approach for detection of seed viability is developed by using laser biospeckle technique. Temporal (absolute value difference (AVD)), as well as spatial features (contrast, and the spatial absolute value difference (SAVD)) from the acquired speckle images were extracted to train and test several state-of-the-art ML models. Obtained results showed that artificial neural network (ANN) based predictive model possess better performance as compared to other models with overall accuracy of 97.65% for classifying the viable seeds.