Application of complex networks for automatic classification of damaging agents in soybean leaflets

Many of the difficulties in managing soybean tillage are related to the identification of insect/pests harmful to the plant, since tillage can be attacked by a wide range of such agents. By identifying the most common agents that cause damages to the leaflets, we can obtain more knowledge about appropriate strategies of control. The proposed work presents an automatic method for classification of the main agents that cause damages to soybean leaflets, i.e., beetles and caterpillars. Acquired images are preprocessed and the contours of the damages are taken. Each contour is modeled as a complex network. Features are extracted for each damage based on the connectivity and the joint degree of this network. These features are then used to train a SVM algorithm. In the experiments, we analyze thresholds which model the network and the proposed method reports accuracy greater than 90% for damaging agent classification.