Classification of Gerbera Type Flowers Based in Decision Tree Rules

In this work, a 4 Gerbera flowers subtypes are classified according to their color components in petals and in the center. We proposed first the preprocessing of each image acquired in real scenario, the preprocessing is realized in two steps, the first is to crop each flower to create a database with the tag of each subtype, the second consist in removing the background using color space transformation to Hue and filter the image according to some specific values. With all the images a decision tree is created with the 70% of the images in each category, the first rule consists of identify the highest value in the histogram, if this value corresponds to green range or orange range the classification is done, nevertheless if the value corresponds to pink, the enclosing circles are estimated, then the color into the smallest circles defined the classification. The evaluation is performed using the last 30% of the images in each category, the classification will be positive if the correct subtype is predicted, if not the classification will be negative, nevertheless the classifier could predict no class based on any existence of the flower, or different flower, in this case if non-class is predicted over a different flower the classification will be positive, if not the classification will be negative. The results obtained in each category are showed in the next list, first is the name of each subtype, followed by the percentage of positive classifications. Gerbera Renato (95%), Gerbera Marinilla (99%), Gerbera Chiper (85%), Gerbera Rio Negro (78%), None/Different (74%).