Improved rotation kernel transformation directional feature for recognition of wheat stripe rust and powdery mildew

To overcome the problem of lacking apparent features, e.g. color or shape, in the process of identifying wheat stripe rust from powdery mildew using computer vision algorithms, a novel directional feature based on Improved Rotation Kernel Transformation (IRKT) is proposed. IRKT can calculate the statistics of the direction distribution of infected leaf images in spatial domain. The statistics calculated from IRKT are insensitive to noise and can lead to a good description of directional distribution of object, which is suitable for the recognition of wheat stripe rust and powdery mildew and provides a novel method to represent other plant disease. As showed in experimental results, the proposed IRKT directional feature is fit for the recognition of wheat stripe rust and powdery mildew, and the accuracy can achieve 97.5%.