A Comparative Disease Detection Approach on Potato Leaves with Image Processing Algorithms

Abstract

Potatoes are a well-known vegetable to all of us. If we look over other countries, we can simply find that potatoes are the number one vegetable all over the world behind rice, wheat & corn as claimed by many Agricultural Department. From the record of the world's statistics, we can see that a large number of potatoes are eaten by Americans. So, we can easily understand the importance of potatoes but the fact is potato leaf disease which caused the potatoes damaged badly. Potato leaf disease detection is a vital issue. So, we think that if we are capable of giving accurate and timely detection to the farmers so that they can be aware of the risk factors and also apply the fertilizer to reduce the economic losses. For this reason, we want to develop a system using Python programming Language which can detect potato leaf disease. In Python, we learn about many algorithms and machine learning techniques Linear Regression, Logistic Regression, CNN (Convolutional Neural Network) architectures like Alexnet, LeNet, VggNet, Resnet and KNN (Knearest neighbors) to detect the leaf. Machine Learning techniques allow us to predict the risk level. The purpose of the current study is to predict the risk level of potato leaf disease. After much discussion, we choose CNN (Convolutional Neural Network) architecture. For image classification, CNN architecture performs best so we used a sequential model compared to other algorithms and architecture. To detect our model, we used a large number of image datasets for the train. The results show that our offered model successfully sorts out and finds the potato leaf disease based on image processing methods. Our tentative results show that we got 97% accuracy with our offered model.