基于目标检测算法与迭代阈值分割的道路标线可视度评估

Road Marking Visibility Evaluation Based on Object Detection and Iterative Threshold Segmentation

摘要:

提出了一种基于目标检测与迭代阈值分割的道路标线分割算法。首先采用基于BiFormer改进的YOLOv5目标检测算法对道路标线区域进行快速定位与框选，然后运用快速迭代阈值分割对框选区域内的道路标线进行精细提取，最后对提取后的道路标线采用韦伯对比度进行人眼可视度评估。结果表明：该方法能够完成道路标线的快速准确提取，并实现对道路标线可视度的有效检测。

Abstract:

A road marking segmentation algorithm based on object detection and iterative threshold segmentation was proposed. The BiFormer-improved YOLOv5 was adopted to locate road markings and obtain image patches. Then， the iterative threshold segmentation was used to capture the accurate region of road markings. Finally， the extracted road markings were evaluated for visibility based on Weber contrast. The results show that the proposed method can extract road markings rapidly and accurately， and effectively evaluate the road marking visibility.