Image Based Major Disease Recognition and Control Management System for Potato Crop

Abstract

newline Machine learning has progressed dramatically over the past two decades. We introduced an newlineapproach that integrates image processing and machine learning to allow diagnosing diseases newlinefrom potato plant images. Potatoes are very prevalent as vegetable in the world. The cultivation newlineof crop is extremely methodical. Potato crop is affected by different types of diseases. Disease newlinecauses massive loss in crop yield. The disease datasets of potato crop are collected from valid newlinedata source and crop field. These datasets are related with different diseases of leaf, tuber, stem newlineand root of potato. Pre-processing techniques has applied for removing the noises in images. newlineThe seventy-six (76) features are extracted of images regarding colour, texture and shape. All newlineextracted features have not contributed the positive effects in prediction of disease class. To newlineimprove the accuracy of classifiers, a feature selection technique is needed. The Feed Forward newlineNeural Network (FFNN) and K- Nearest Neighbours (KNN) models are used for prediction newlineand classification of disease class of crop. The accuracy of ANN model without feature newlineselection is 97.64%, 84.60%, 96.06%, 93.07% of leaf, tuber, stem and root dataset respectively. newlineThe precision of KNN model without feature selection is 80.76%, 63.33%, 94.38%, 84.41% of newlineleaf, tuber, stem and root dataset respectively. The accuracy of ANN model using existing newlineStepwise feature selection technique is 97.94%, 73.17%, 97.75%, 96.10% of leaf, tuber, stem newlineand root dataset respectively. The precision of KNN model using existing Stepwise feature newlineselection technique is 87.68%, 70.95%, 84.26%, 76.62% of leaf, tuber, stem and root dataset newlinerespectively. The accuracy of ANN model using existing Relieff feature selection technique is newline98.46%, 85.23%, 97.93%, 95.23% of leaf, tuber, stem and root dataset respectively. The newlineprecision of KNN model using existing Relieff feature selection technique is 87.73%, 72.38%, newline94.38%, 89.61% of leaf, tuber, stem and root dataset respectively.