Improved YOLOv9-Based Objects Detection in Adverse Weather Conditions for Autonomous driving

Adverse weather conditions pose challenges to autonomous driving, increasing the difficulty and complexity of object detection. This study proposes a new object detection method based on the improved YOLOv9 framework for autonomous driving under adverse weather conditions, including techniques such as data augmentation, attention module integration, and loss function modification to enhance the object detection accuracy and robustness. The effectiveness of the proposed strategy has been validated through experiments on the DAWN adverse weather traffic dataset, that is captured for traffic scenes under four adverse weather conditions: rain, fog, sand, and snow. The precision of object detection under adverse weather conditions, measured by mAP50, has significantly improved. The proposed method provides a more robust and reliable solution for object detection in autonomous driving under adverse weather conditions.