Efficient detection of plant diseases in agriculture is an important topic of research as the diseases in plants will directly influence crop production, quality of crops and agricultural economy. Though image processing techniques and classification algorithms are useful in automatic detection of diseases, still detection with adequate or enhanced accuracy is an open issue. Recently in contrast to traditional machine learning algorithms deep neural networks are being used for more accurate prediction tasks owing to their capability in solving complex problems. In this work an approach based on autoencoder technique is proposed for detection of diseases in plant leaves with an intention of improving the accuracy of detection. After preprocessing the images of plant leaves, the images are segmented into the normal and affected portions of the images using FCM algorithm. Features are extracted from the segmented clusters by constructing Discrete wavelet transform (DWT) based Gray-level Co-occurrence Matrix (GLCM). The extracted features have been detected for the presence of diseases using deep autoencoder technique. With typical datasets, the accuracy obtained using autoencoder technique is found to be higher than that obtained using conventional approach.

No abstract