Machine Vision Based Potato Species Recognition

Abstract

One of the bound vegetable elements of our daily life is the potato. There are 4200 kinds of potato species and among them, 82 kinds are found in Bangladesh. It is usually grown for numerous reasons apart from having as a food. Different potatoes are used in different cases. If there’s a curry-making procedure there will a different potato be used and if there’s a French fried making recipe there also a different potato be used. But the problem is many people in our country fail to understand the right kind of potato for the right use. Thus, a big confusion is created in every genre of these potato consuming sectors especially for the people who are unaware of potato species. This is where our project idea born, using a machine vision-based recognition of these potato species which can help people to recognize them. In this paper, we perform an in-depth exploration of a machine vision approach for recognizing different species of potatoes in Bangladesh. A number of potatoes are classified based on the figures analyzed by their images. For our experiment, we collected the data of 4 verities total of 1200 potato images. In this process, we wanted to identify the real image of potatoes. we have applied machine learning algorithms like Random Forest Classifier (RF), Linear Discriminant Analysis (LDA), Logistic Regression, Support Vector Machine (SVM), CART, NB, and KNN on our datasets. Here we apply our developed algorithms to (color, texture, and shape) features. The data obtained from image processing were classified using extraction, resize, and grayscale convert. After applying all algorithms each of them produced different results. Random Forest Classifier (RF) shows the best result and its accuracy rate of 100% and SVM gives the lowest rate. Its accuracy rate 74.074%, which is not only good but also promising for future research.