Artificial intelligence identification of perennial plant leaves using electrical properties

Artificial intelligence algorithms, including Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machine (SVM), k-Nearest Neighbors (KNN), Decision Tree (DT), Gaussian Naive Bayes (GNB), Linear Discriminant Analysis (LDA), Multilayer Perceptron (MLP), Random Forest (RF), Artificial Neural Network (ANN), and One-Dimensional Convolutional Neural Network (1D-CNN), were used to create plant classification models based on non-destructive Electrical Impedance Spectroscopy (EIS) measurements obtained from the leaves of perennial plants Vitis vinifera, Prunus laurocerasus, Magnolia grandiflora, Tilia cordata, Corylus avellana, and Olea europaea. Two datasets were created from the EIS data. One of them is the impedance, and the other combined data consisting of impedance, the imaginary part of impedance, phase angle. These datasets were augmented by 20% through random increases according to a normal distribution function within the standard deviation limits for each frequency, yielding two additional datasets. Models were then developed for each artificial intelligence approach using these datasets individually. Out of the 40 models created (four each for PLS-DA, SVM, KNN, DT, GNB, LDA, MLP, and RF), 32 underwent cross-validation, and measurement and evaluation were performed during training for the remaining 8. Subsequently, all models were tested with a standard dataset randomly created from the original data, and confusion matrices were generated. Using the data from these confusion matrices, testing accuracy, precision, recall, and F1 scores were calculated. The SVM, DT, LDA, RF, ANN, and 1D-CNN models showed particularly promising results in distinguishing each plant species. Among the trained artificial intelligence models, the 1D-CNN model trained with the augmented combined dataset achieved the best performance with an accuracy of 99%. The results obtained in this study indicate that, despite similarities in shape and color, perennial plants can be distinguished using EIS data, enabling the development of AI applications for plant classification.