Moving Object Detection in Aerial Images using DeepSORT with Faster R-CNN

Abstract

Aerial imagery is increasingly utilized in various applications, including surveillance, disaster management, agriculture, and urban planning. Detecting and tracking moving objects within aerial images is a crucial task for these applications. This paper presents a novel approach to moving object detection in aerial images, combining the Faster R-CNN (Region-based Convolutional Neural Network) for object detection and the DeepSORT (Deep Simple Online and Realtime Tracking) algorithm for object tracking. The proposed method leverages the strengths of both techniques, enabling accurate and efficient detection and tracking of moving objects in aerial imagery. First the Faster R-CNN model is employed to detect objects in each frame of the aerial image sequence. The model has been pre-trained on a diverse dataset, making it capable of detecting a wide range of objects. Post-detection, the DeepSORT algorithm is applied to track the detected objects across frames. DeepSORT utilizes deep learning for object appearance and Kalman filtering for state estimation, resulting in robust tracking even in challenging scenarios. The proposed model obtained an overall accuracy of around 86%.