Leveraging Deep Belief Networks and Multi-Class Support Vector Machine for Nutritional Deficiency Identification in Grape Plant Leaves

Modern fertilization includes identifying nutritional deficits in plant leaves since soils frequently fall short of providing growing crops with all the nutrients the earth needs. The creation of a fertilization program to deliver the desired micronutrients while not overproducing others will be made easier by determining the amount of vitamins and minerals that are required. The indications of agricultural malnutrition are nutritional insufficiency indicators in plant leaves, which may be physically examined from the plant morphological elements and so offer adequate information to identify the nutritional deficit. This study was intended to focus on nutritional deficiency identification in grape plant leaves. In this work, Deep Belief networks (DBNs) are employed for training the model. The output of the model is fed to a Multi-Class Support Vector Machine (SVM) to classify the grape plant leaf images as healthy, N-deficiency, P-Deficiency, K-Deficiency, Ca-Deficiency, Mg-Deficiency and S-Deficiency. Accuracy and Misclassification rate is used to analyze the performance of the model. The results indicated that the suggested model achieved an accuracy of 96.8% and a misclassification rate of 3.2%. These results show that DBNs are a potential technology for nutritional deficiency categorization in grape plant leaf tissue.