Finding the Non-destructive Monitoring System of Papaya by Machine Learning Algorithm

The economic significance of the tropical fruit papaya (Carica papaya) stems from its excellent nutritional and therapeutic value. A crucial activity in the fruit sector is packing papaya fruit according to its maturity stage. Manually grading papaya fruit using human eyes is laborious and harmful. This research aims to provide a new method for non-destructively determining the maturity condition of papaya fruits. The research proposed two methods for classifying papaya maturity state, both based on transfer learning (TL) and machine learning (ML). Additionally, several transfer learning and machine learning methods are compared. Two hundred photographs of papaya fruits, 50 from each of the three phases of ripeness, were used for the experiments. The machine learning technique comprises three classifiers, each with its unique kernel function and three distinct sets of features. Machine learning methods make use of features and classifiers such as support vector machine (SVM), k-nearest neighbor (k-NN), histogram of oriented gradients (HOG), local binary pattern (LBP), and gray-level co-occurrence matrix (GLCM). Using ResNet50, ResNet101, ResNet18, VGG16, VGG19, Google Net, and Alex Net are the seven previously trained models that make up the transfer learning method. With a training time of just 0.0996 s and a perfect accuracy rate, the weighted k-NN with the HOG feature is the best machine learning-based classification model. With 100% accuracy and 1 min and 52 s of training time (including early stop training), VGG19 outperforms all other transfer learning technique-based classification models. Using a transfer learning methodology, the suggested method (VGG19) for papaya fruit maturity classification obtained 100% accuracy, an improvement of 5% over the current method.