Detection of Ginseng leaf cicatrices base on K-means clustering algorithm

Diseases like black spot, anthracnose and Cylindrocarpon destructans always appear on ginseng leaf. The traditional way to distinguish the disease in farmland is checking by planter in person. This way takes more time and energy. In this paper, according to cicatrices are different with normal leaf in color, we use K-means clustering algorithm and combine the difference between a and b components in Lab color space to discern the cicatrices in leaf image. Firstly, we can distinguish cicatrices and normal leaves. Secondly, we take out useless part like petiole. Lastly, we calculate the ratio of area of cicatrices and normal leaf. The ratio can be used as reference of the ginseng disease and its degree. The result of experiment shows that K-means clustering algorithm can segment the leaf image, and the segmentation is based on the difference of color, then we can calculate the ratio of pixel in each part, and judge the degree of disease.