Deep learning model for plant-leaf disease detection in precision agriculture

Abstract

Crop disease in the agricultural is the main factor limiting yield and food quality. It requires timely diagnosis of crop illnesses for the better economy developing country. Manual crop illness assessment is limited due to lesser accuracy and restricted accessibility. It is very difficult to accurately identify and classify plant diseases due to corrupt in the data samples, lesser intensity of foreground and background, and the extreme similarity between unhealthy and healthy leaves in terms of colouring and size of crop leaves. Hence the employments of automated and computerised optimisations are needed. To identify plant leaf diseases, a DNSVM classification strategy fusing DenseNet-201 with support vector machine (SVM) is proposed in this work. Plant-Village dataset that provides good-variations, colour-differences, differences in orientation and size-of-leaves. Sugarcane plants-leaves were used for performance analysis of proposed model and obtained 97.78% of classification accuracy over the existing DenseNet-121-based classifier model (94%).