Anthracnose disease diagnosis by image processing, support vector machine and correlation with pigments

Abstract

Mango (Mangifera indica L.) has been growing in India since very ancient times and is supposed to be the king of all fruits due to its exceptional taste. In India anthracnose disease is one of the major constraint for mango production and the cause of 90% crop losses every year. Colletotrichum gloeosporioides infects and induces anthracnose in a wide range of plants among which Mangifera indica is more sensitive fruit tree. In the present study, the diagnosis of anthracnose was attempted with a rapid and reliable technique with biochemical and computational approaches. Using MATLAB, disease severity was analyzed by capturing images from a DSLR camera. The quality of images was enhanced by preprocessing and the textural features of the images were obtained using a Gray Level Co-occurrence Matrix (GLCM). The images were classified as healthy or infected (anthracnose) by a Support Vector Machine (SVM). The effect of fungal infection on the pigments including total chlorophyll and carotenoids contents was also estimated and correlated with the computational data. This study provides a quick, targeted and cost-effective detection methodology of anthracnose disease in mango leaves at an early stage of infection. Disease identification in the early stage may be useful for mango farmers, and can help in plant protection and for sustainable agriculture.