Image Processing Techniques and Data Mining Algorithms for Coffee Plant’s Leaves Classification

Abstract

Arabica coffee is known for its unique taste and aroma. This coffee variety contributed majority of coffee production in the world. However, arabica coffee and other coffee varieties are prone to extinction because of several reasons including climate change, drought, diseases and issues in identification of nutritional deficiencies. Nutritional deficiencies are identified and classified manually with an expert to validate the visual symptoms occurred in the coffee leaves. On the other hand, the utilization of image processing to analyze images as well as data mining is a strong combination for classification. Therefore, this study was conducted to classify the nutritional deficiencies in arabica coffee plants including Phosphorus (P) and Potassium (K) using image processing and data mining. The images of 2045 instances with 1001 features undergone image processing techniques such as image acquisition, image pre-processing and image analysis. The 70% of data was for training and 30% was for testing using Waikato Environment of Knowledge Analysis (WEKA) and Orange Visual Programming. Random Forest, Support Vector Machine (SVM), Neural Network (ANN) and K-Nearest Neighbors (KNN) served as the classifiers of two classes. Results shows that SVM has the highest AUC of 1.000 and CA, F1, Precision and Recall of 0.983. The Correctly Classified Instances (CCI) is 98.73% and Incorrectly Classified Instances (ICI) is 1.27%. Further, the Kappa statistics of 0.97 shows an almost perfect value of agreement and implies that the classifier is better in coffee plants leave classification together with image processing. © 2020, World Academy of Research in Science and Engineering. All rights reserved.