Automatic Grading of Potato Leaf using Machine learning & Computer Vision

Abstract

Abstract Economically, agriculture plays a crucial role but agriculture production decreases with plants or crops diseases. The quantity and quality of the plant are also concerned due to various diseases in plants. Hence early exposure plays a crucial role in reducing the bruise of plant diseases. The highest potato manufacturer in the world is China and together, India & China produce one third of overall potatoes. Potato leaf grading and detection are complex issues that require a lot of human expertise. The manual detection for diseases is ineffective, time-consuming, uncertain, and expensive. The reliable, robust, and scalability factors need to be considered when designing a disease detection method. Machine learning and computer vision advancement led to the development of promising solutions in the agriculture field. In this paper an automated, astute, and efficient detection and grading method for potato leaves is introduced. Firstly, Gaussian filtering is used as pre-processing of the image to improve image quality and noise removal. Secondly, image is segmented using fuzzy c-means technique. Then numerous features, such as geometrical, textural, and statistical are separated (112) and principle component analysis is used to select essential feature (30) for classification. Finally, for classification k-Nearest Neighbour, Logistic Regression, Artificial Neural Network, and Support Vector Machine are benefitted as decision making for potato leaf disease grading. The 10 cross-validation processes have been used to validate the system. The algorithm proposed for the disease detection of potato leaves achieves 83.39% (k-NN), 89.72% (LR), 92.54% (ANN), and 99.75% (SVM). The selecting of the appropriate feature indicates improved system performance. Among the four different classifiers, the SVM results are promising contrast to the reviewed literature.