Plant Disease Diagnosis with Color Normalization

摘要

A plant disease diagnosis method based on color histogram invariant features, is evaluated on pear diseases. The employed fuzzy-like classification method is tested with five different normalization methods applied on Red-Green-Blue (RGB), Hue-Saturation-Lightness (HSL), Hue-Saturation-Value (HSV) and $mathbf{L}^{st}mathbf{a}^{st}mathbf{b}$ format. Three metrics (sensitivity, specificity, accuracy) are used to assess the success of each normalization method. The experimental results show that the success in the disease recognition can be improved by up to 4% if appropriate color normalization is employed.