This Paper describes a method to recognize and classify complex objects in digital images. To this end, a uniform representation of prototypes is introduced. The notion of a prototype describes a set of local features which allow to recognize objects by their appearance. During a training step a genetic algorithm is applied to the prototypes to optimize them with regard to the classification task. After training the prototypes are compactly stored in a decision tree which allows a fast detection of matches between prototypes and images. The proposed method is tested with natural images of highway scenes, which were divided into 15 classes (including one class for rejection). The learning process is documented and the results show a classification rate of up to 93 percent for the training and test samples.