A Novel Technique for Detecting Crop Diseases with Efficient Feature Extraction

Abstract

Classification and feature extraction are other planned areas of study, since they build upon the basis of computer picture processing technology and have a direct impact on the results of image recognition. We compared low-level features, texture features, and form-based features to generate an original feature set from the input segmented crop photographs. We used the Histogram of Gradient (as colour-based features) to extract low-level features because of its ease of use and effectiveness. Within the realm of shape-based features, EHD was used to obtain defining properties. In order to have a better understanding of textures, we used GLCM to extract features based on them. When all features are added together, a feature vector with several dimensions is produced. To conclude, the data were classified using artificial neural networks (ANN). MATLAB is used for simulation purposes. The current difficulty is to use smart image processing technology to extract picture properties from the original image with a strong representation. If the signs of a disease on a crop are vague or otherwise complicated, it will be quite difficult to identify those qualities.