Multiple Moving Object Detection and Tracking with Event Camera
12th International Conference on Automation, Robotics and Applications, ICARA 2026, İstanbul, Türkiye, 5 - 07 Şubat 2026, ss.428-433, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/icara69401.2026.11480340
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.428-433
- Anahtar Kelimeler: computer vision, dynamic vision sensor, event-based camera, high-speed objects, object tracking
- Ankara Yıldırım Beyazıt Üniversitesi Adresli: Evet
Özet
Event-based cameras, also known as dynamic vision sensors (DVS), offer significant advantages for computer vision tasks due to their asynchronous operation, low power consumption, and high temporal resolution. These characteristics make them particularly suitable for robotics and automation systems, where real-time perception and efficiency are crucial. In this study, we present an integrated framework for object detection and tracking using event-based data. First, event streams were converted into frame-like representations and annotated to generate a dataset. A Mask R-CNN model was then trained for object detection and segmentation on the event-based data. The trained model was evaluated using standard metrics and demonstrated robust detection performance. Building upon this, we addressed object tracking by leveraging the event data structure in combination with a Kalman filter. Both single-object and multi-object tracking scenarios were investigated, and performance was analyzed using IoU-based evaluation metrics. Experimental results indicate that the proposed framework achieves high detection accuracy and reliable tracking performance, demonstrating the potential of event-based cameras for real-time applications in robotics, surveillance, and automation.