Airplane Detection via Deep Learning based on VGGs and SVMs

Abstract Object detection is widely utilized in many applications, such as airport surveillance. Automated detection systems effectively contribute to the prevention of potential collisions, aid in airspace management, and enhance overall aviation safety. Additionally, such detection systems play a crucial role in security applications by enabling the rapid identification of aircraft for surveillance purposes in compliance with aviation regulations. In this paper, an algorithm is proposed for airplane detection regardless of the variations in the model, type, or colour of airplanes. The main challenges in automatic airplane detection tasks could be the differences in scale, the orientation of the airplanes and similarity with other objects. Therefore, an airplane detection system needs to be designed so that good discrimination is achieved without the influence of rotation, pose, or resolution of airplanes. Object detection can be performed by considering three major phases, i.e., feature extraction, detection of an airplane and evaluation of the airplane. To extract the plane region, a deep feature extraction method is used with the VGG model. The plane is detected by using the SVM. Two datasets are used to evaluate the effectiveness of the designed system. The obtained results achieved a 99% F1-score using the Caltech-101 dataset and 98% for the FGVC-Aircraft dataset.