一种基于视觉的道路宽度测量方法

In the research of automated driving technology, a vision-based road width measurement method is proposed to quickly detect the road and measure the width of the road. Firstly, an improved Deeplabv3+ semantic segmentation algorithm is proposed to improve the accuracy of road edge segmentation by introducing the attention mechanism based on spatial domain into the original algorithm flow. Secondly, a semi-global stereo matching algorithm combining histogram equalization and weighted most minor square filtering is proposed, which can significantly improve the matching accuracy of object edge details in images. Then, a road width measurement system based on the proposed method is proposed. The system is mounted on an autonomous vehicle to complete the segmentation and width measurement of the road in the global image. The experimental results show that the measurement error of the proposed algorithm is less than 5.0% of the real value.