Drone detection by neural network using SURF feature and GLCM method

Abstract

This presents a drone detection method using image processing. In the proposed model, a hybrid feature extraction method using SURF and GLCM is utilized. As a machine learning tool, we use a neural network for pattern recognition to train and test. Finally, we measure the performance of the proposed model using cross entropy. For our tested drone dataset, experimental results demonstrate improved performance over state-of-art models by exhibiting less cross entropy and percentage error. It also presents experimented results of drone detection using different combination of methods and the results why it is recommended to use our approach. The combinations we used are SURF, GLCM, SURF with GLCM, and MSER. Here we focused on an optimal combination of performance and error percentage results. The recommended approach is to detect drones with very minimum error percentage and very high performance.