ENHANCING AUTONOMOUS DRIVING SYSTEMS WITH AI-BASED IMAGE QUALITY ASSESSMENT FOR LENS DEFECT DETECTION

Abstract

Autonomous driving systems rely heavily on camera-based perception, making image quality a critical factor in ensuring accurate decision-making. Lens defects, such as dirt, cracks, and scratches, can significantly degrade image quality, leading to poor performance in object detection and navigation tasks. This study presents an AI-based framework for detecting and assessing lens defects in autonomous vehicle cameras. By combining traditional image processing techniques with deep learning models, including a custom-trained YOLOv8, we demonstrate an effective approach to identifying image distortions. Our findings highlight the strengths and limitations of different detection methods and provide insights into improving defect detection in future implementations.