A HYBRID APPROACH TO IMAGE CLASSIFICATION BASED ON CONVOLUTIONAL NEURAL NETWORKS AND FRACTAL DIMENSION

Abstract

In this paper, a hybrid approach to tomato leaf image classification is proposed based on the combination of features extracted using a convolutional neural network and fractal characteristics of the image. CNN is used as a feature extractor to obtain local and hierarchical visual characteristics; the fractal dimension of the image is calculated by the box-counting method and reflects the global structural complexity of the leaf surface. The resulting features are combined into a single feature space and used to train a random forest classifier. The results of the experiments show high classification accuracy and confirm the effectiveness of the proposed hybrid approach. Combining deep learning and fractal analysis improves the quality of automatic plant image classification and can be used in crop monitoring systems.