احسان مددي

In recent years, autonomous vehicles have emerged as one of the most advanced technologies in the field of artificial intelligence an‎d intelligent transportation systems. A key challenge in this domain is ensuring the precise an‎d reliable performance of object detection systems under diverse weather conditions an‎d complex environments. This thesis aims to enhance the robustness of object detection systems in autonomous vehicles by utilizing a comprehensive dataset that encompasses a wide range of weather scenarios, including day, night, rain, fog, snow, an‎d san‎d. We employed several advanced deep learning models based on the YOLO framework, particularly variants of YOLOv8, YOLOv9, an‎d YOLOv10, to eva‎luate an‎d improve object detection performance. These models were trained on a dataset consisting of 5,765 high-resolution images with meticulously annotated bounding boxes. Our experiments demonstrate that integrating a diverse dataset representing multiple environmental conditions significantly improves object detection accuracy an‎d reliability, especially in adverse weather conditions. The results are analyzed based on key performance metrics such as precision, recall, an‎d mean Average Precision (mAP). Our findings indicate that advanced object detection models, when trained on rich an‎d diverse datasets, can effectively address challenges posed by unfavorable weather, thereby enhancing the safety an‎d operational efficiency of autonomous driving systems.

خودرو‌هاي‌خودران , تشخيص‌اشياء , مجموعه داده , شرايط آب‌‌و‌هوايي مختلف

Autonomous vehicles , Object detection , Dataset , Diverse weather conditions

Ehsan Madadi

Dr.Abdollah Amirkhani