Integrated Approach Using Efficient Segmentation and Feature Extraction Techniques for Improved Disease Detection in Plant Leaves

Abstract

Diseases in plants are a major threat to the economy that directly affects the food production and thus detection of diseases in plants becomes an important topic of research. voluntary observation of the disease does not give satisfactory results, because monitoring with the naked eye is an old method that requires extra period to diagnose the disease and also requires expertise and finding experts is a challenge.  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. For early detection of plant diseases, image processing has now become a viable tool which provides various operations for preprocessing and enhancement of the images and thus enables the process of detection of diseases using machine learning.  Machine learning algorithms learn from huge examples of training set images and makes in automatic detection and classification of diseases. Though image processing and machine learning algorithms are useful in detection of diseases, still achieving detection with enhanced accuracy is an open issue.  The accuracy of detection needs to be enhanced with alternate techniques and methods. With this perspective, an integrated approach consisting of clustering-based segmentation, Discrete Wavelet Transformation- Gray Level Co-occurrence Matrix based method for extraction of textures and auto encoder based deep neural network for classification has been proposed in this paper.  The preliminary results obtained using benchmark dataset have been presented and discussed.