Drone Detection with Deep Learning and Image Processing: CNN- Based Feature Extraction and Machine Learning Classification

• Published in: International Journal of Applied Methods in Electronics and Computers
• Main Authors: Özyer, Begüm, YASAR, Ali, TASPINAR, Yavuz Selim
• Format: Journal Article
• Language: English
• Published: 31.03.2025
• ISSN: 3023-4409; 3023-4409
• DOI: 10.58190/ijamec.2025.116

Drones are now widely used in the entertainment industry, cargo and transportation, security and logistics. However, drones can be confused with many objects and creatures that are similar to them or that they are in the same environment. Due to their widespread use and the fact that they have similar characteristics with other objects and living things, they pose some problems. These problems become problems that concern and sometimes disturb the society. For this reason, image processing methods have been used to classify drone and non-drone objects. Thanks to image processing, drones can be classified without any contact. Thanks to deep learning, data sets containing a large number of images can be classified quickly and accurately. In this study, it is aimed to classify drone and non-drone objects. In this study, a total of 1081 data sets consisting of 597 drone images, 484 non-drone objects and live images were used. 20% of the dataset was used for testing and 80% for training. Convolutional Neural Network (CNN) method was used to determine the features of these images. For each image, 4608 image features obtained from the CNN model were classified with Artificial Neural Network (ANN), K Nearest Neighbor (KNN) and Random Forest (RF) machine learning models. Precision, recall, F1 Score and accuracy metrics were used to evaluate the performance of the CNN model. Classification of the features obtained from the CNN model with ANN, KNN and RF models resulted in 89.9, 85 and 85 classification successes, respectively. The highest classification success was obtained from the ANN model in the classifications made with the features of the CNN model. With the results obtained, it is seen that the proposed classification and feature extraction models can be used to distinguish between drone and non-drone objects. Drones are now widely used in the entertainment industry, cargo and transportation, security and logistics. However, drones can be confused with many objects and creatures that are similar to them or that they are in the same environment. Due to their widespread use and the fact that they have similar characteristics with other objects and living things, they pose some problems. These problems become problems that concern and sometimes disturb the society. For this reason, image processing methods have been used to classify drone and non-drone objects. Thanks to image processing, drones can be classified without any contact. Thanks to deep learning, data sets containing a large number of images can be classified quickly and accurately. In this study, it is aimed to classify drone and non-drone objects. In this study, a total of 1081 data sets consisting of 597 drone images, 484 non-drone objects and live images were used. 20% of the dataset was used for testing and 80% for training. Convolutional Neural Network (CNN) method was used to determine the features of these images. For each image, 4608 image features obtained from the CNN model were classified with Artificial Neural Network (ANN), K Nearest Neighbor (KNN) and Random Forest (RF) machine learning models. Precision, recall, F1 Score and accuracy metrics were used to evaluate the performance of the CNN model. Classification of the features obtained from the CNN model with ANN, KNN and RF models resulted in 89.9, 85 and 85 classification successes, respectively. The highest classification success was obtained from the ANN model in the classifications made with the features of the CNN model. With the results obtained, it is seen that the proposed classification and feature extraction models can be used to distinguish between drone and non-drone objects.