Real-Time Object Detection for Automotive Systems With FMCW Radar-Based Sensor Fusion
27th International Radar Symposium, IRS 2026, Krakow, Polonya, 19 - 21 Mayıs 2026, ss.273-278, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.23919/irs70539.2026.11548965
- Basıldığı Şehir: Krakow
- Basıldığı Ülke: Polonya
- Sayfa Sayıları: ss.273-278
- Anahtar Kelimeler: ADAS, collision risk assessment, deep learning, FMCW automotive radar, LiDAR, multi-sensor perception, radar signal processing, real-time object detection, sensor fusion
- Gazi Üniversitesi Adresli: Hayır
Özet
This paper presents a radar centric multi sensor perception framework for real time collision risk assessment that integrates motion cues from Frequency Modulated Continuous Wave (FMCW) automotive radar with camera based deep learning detections and Light Detection and Ranging (LiDAR) based spatial ranging. The proposed approach explicitly uses Doppler derived radial velocity for detection verification and association stabilization through lightweight motion coherent tracking, instead of relying only on camera and LiDAR alignment and static range consistency. In order to isolate the contribution of motion information, a camera and LiDAR baseline is comparatively evaluated against a radar assisted configuration under the same experimental setup. Range Doppler representations and tracking outputs show clear separation of moving targets from static components near the zero Doppler region, confirming reliable extraction of range and radial velocity through two stage Fast Fourier Transform (FFT) processing. The results indicate improved temporal consistency and association robustness without a meaningful increase in computational load on an embedded platform.