雷達和相機融合應用於自動駕駛的物件偵測

自動駕駛最近十分熱門，檢測汽車周圍的障礙物對於避免造成事故十分重要。自動駕駛中所使用的傳感器有很多種，例如光達、雷達和相機。這些傳感器可以檢測周圍環境，而如何融合它們是一個重要的問題，以達到更好的結果。在提到的三種傳感器當中，光達的價格比其他兩種傳感器的價格昂貴。此外，光達或相機可能會在惡劣的天氣情形下表現不好，然而雷達在這些天氣條件下具有穩健性。考量到雷達的價格因素以及在惡劣天氣下的穩健性，因此我們在實驗中使用雷達並和相機進行融合。在我們的研究中，我們提出了一種融合方法來融合雷達及相機並進行物體偵測，並比較了不同天氣條件下的表現。我們方法的結果在整個資料集中可以達到與光達接近的效果，甚至在惡劣的天氣情況下可以超越光達，顯示我們方法的穩健性。

關鍵字

雷達；相機；自動駕駛；天氣；物件偵測；傳感器融合

並列摘要

Autonomous driving is popular recently, and the way to detect obstacles around the car is important to avoid accidents. There are many kinds of sensors used in autonomous driving such as LiDAR, radar and camera. These sensors can detect surrounding circumstances, and how to fuse them is a significant problem in order to reach better performance. Among the mentioned three sensors, LiDAR is more expensive than the others. Besides, LiDAR and camera can be failed in some bad weather conditions, while radar has robustness in these conditions. Considering the price factor and the robustness of radar in bad weather conditions, we use radar to be a sensor in our experiment and fuse with camera sensors. In our research, we propose a method to fuse radar and camera for object detection, and evaluate the performance in different weather conditions. The results of our method can get close to the LiDAR-only baseline in whole dataset and even outperform in bad weather conditions that shows the robustness of our method.

並列關鍵字

Radar ； Camera ； Self-driving ； Weather ； Object Detection ； Sensor Fusion

參考文獻

[1] D. Barnes, M. Gadd, P. Murcutt, P. Newman, and I. Posner. The oxford radar robotcar

Google Scholar

dataset: A radar extension to the oxford robotcar dataset. In 2020 IEEE International

Google Scholar

Conference on Robotics and Automation (ICRA), pages 6433–6438. IEEE, 2020.

Google Scholar

[2] H. Caesar, V. Bankiti, A. H. Lang, S. Vora, V. E. Liong, Q. Xu, A. Krishnan, Y. Pan,

Google Scholar

G. Baldan, and O. Beijbom. nuscenes: A multimodal dataset for autonomous driving.

Google Scholar

主題瀏覽