The invention belongs to the field of automatic driving and computer vision, and relates to a vision-based track turnout identification method, which comprises the following steps: constructing a turnout classification network model; constructing a monorail segmentation network model; judging whether the current frame of track scene is a turnout scene or not by utilizing a turnout classification network model; segmenting the turnout scene by using a monorail segmentation network model to obtain a monorail region image with a bifurcation point; and utilizing a fork center point detection algorithm to carry out track fork center point identification on the monorail area image, and outputting a fork center point coordinate. According to the invention, pure vision is used for track turnout recognition, color information and image distribution conditions can be obtained, visual scene understanding is easy, and track forks can be positioned on images in advance.
本发明属于自动驾驶及计算机视觉领域,涉及一种基于视觉的轨道道岔识别方法,包括如下步骤:构建道岔分类网络模型;构建单轨分割网络模型;利用道岔分类网络模型判断当前帧轨道场景是否为道岔场景;利用单轨分割网络模型对道岔场景进行分割,得到具有分叉点的单轨区域图像;利用岔心点检测算法对单轨区域图像进行轨道岔心点识别,并输出岔心点坐标。本发明利用纯视觉来进行轨道道岔识别,可以获取到颜色信息和图像分布情况,易于直观的场景理解,可以提前对轨道分岔口在图像上进行定位。
Vision-based track turnout identification method
一种基于视觉的轨道道岔识别方法
2020-10-30
Patent
Electronic Resource
Chinese
Engineering Index Backfile | 1930
|