Real-time implementation of moving object detection in UAV videos using GPUs
- PDF / 2,044,794 Bytes
- 17 Pages / 595.276 x 790.866 pts Page_size
- 50 Downloads / 241 Views
ORIGINAL RESEARCH PAPER
Real‑time implementation of moving object detection in UAV videos using GPUs Deepak Jaiswal1 · Praveen Kumar1 Received: 3 November 2018 / Accepted: 4 June 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019
Abstract Unmanned aerial vehicles (UAVs) are being increasingly used for video surveillance and remote sensing. Moving object detection is an important algorithm for many such applications. Real-time processing of moving object detection is required for various decision-making tasks in many of these applications. However, being compute-intensive in nature, it is difficult to process high-resolution UAV-sourced videos in real-time. GPU vendors regularly release newer architectures with new features to speed up various kinds of applications. Hence, it becomes imperative to explore parallel implementations of such algorithms using the new GPU architectures. This paper describes parallel implementation strategies for algorithms like feature detection, feature matching, image transformation, frame differencing, morphological processing and connected component labeling which are used to detect moving objects in UAV-sourced videos. The implementation is tested on different NVIDIA GPU microarchitectures (Fermi, Maxwell, and Pascal). Experimental results show the achieved frame processing rates of 43.1 fps, 35.5 fps and 9.1 fps for 1080p videos on Pascal, Maxwell, and Fermi microarchitectures respectively. Keywords Connected component labelling (CCL) · GPU optimizations · Image registration · Morphology · Moving object detection
1 Introduction Recently, there has been a surge in the usage of unmanned aerial vehicles (UAVs) in fields like video surveillance and remote sensing. UAVs are being widely used in military, law enforcement, agriculture, construction industry, forestry, etc. For many applications in these domains, moving object detection is an important algorithm which serves as a basis for further analysis like object tracking, recognition, and classification. There have been many studies attempting to improve the accuracy of moving object detection for UAV videos. One such early attempt was done in moving object detection and tracking (MODAT) framework [1]. The framework includes image alignment, motion detection, and motion tracking modules. Cheraghi and Sheikh proposed a system which included feature extraction, motion estimation, * Deepak Jaiswal [email protected] Praveen Kumar [email protected] 1
Visvesvaraya National Institute of Technology, Nagpur 440010, India
motion compensation, and background subtraction to detect moving objects [2]. Teutsch et al. used image stacking to improve results of top view aerial videos [3]. Wei et al. also used feature detection using Harris corner detection, feature matching and frame differencing for moving object detection [4]. However, these algorithms are computationally too expensive to process the input video stream in real time, especially for high-resolution videos. Baykara et al. attempted to run moving object
Data Loading...
