Eagle Strategy enabled Deep Convolutional Neural Network based Rice Leaf Disease Detection

Abstract

Rice productivity is an essential part of Indian economy, as it is a primary food crop and major and it cropped in major parts. Due to pests and diseases there is a major threat in production of rice and detection of disease at earlier stages may greatly help the production quality and quantity. Image processing with deep learning models plays a vital role in detection of hazardous disease found in plant leaves, as it greatly reduces the monitoring process in a huge farm. The objective of this paper is to develop a enhanced model of rice leaf disease detection at their earlier stages by concentrating on two important processing of image processing. The rice plant image acquired from Kaggle data set is improvised by converting them into HSV representation and the contrast of the leaf image is enhanced by adapting fuzzy joint histogram equalization, handling impreciseness of gray level values of rice plant leaf image in a prominent manner, in this depending on the original image median value it is partitioned into sub-histograms. And by finding probability distribution and cumulative difference the global pixels values are enhanced to make the edges of the image more accurate. After enhancing the rice leaf images then it is fed as input to the deep convolutional neural network where it extracts the significant features using its deep layers and the performance of the CNN is boosted by the knowledge of eagle strategy algorithm, which assigns the optimized parameter value in the fully connected layer of CNN, where the actual classification of leaf disease is done. The performance of Eagle strategy enabled Convolutional Neural Network (CNN+ESA) produce best results in detection of rice leaf diseases at earlier stages while comparing with other traditional detection models.