톱-햇 및 CLAHE 전처리 기반 이미지 융합을 활용한 신호등 객체 탐지 성능 향상

초록

This paper proposes a method to enhance traffic light object detection performance in autonomous driving systems by applying preprocessing-based image fusion to RGB images. Conventional object detection models often struggle to detect small objects in low-contrast environments due to lighting variations and poor visibility. To address this limitation, CLAHE and White Top-Hat transformations are independently applied to emphasize object features. CLAHE enhances local contrast, while White Top-Hat removes background noise and highlights bright objects, improving the detection of small objects. Experimental results using the YOLOv8 model demonstrate overall performance improvements. This study introduces a novel preprocessing approach for RGB-based object detection.

키워드