基于车道线提取的智能车横向定位技术

Lane Line Extraction-Based Lateral Positioning Method for Intelligent Vehicle

将视觉图像序列信息与高精度地图结合进行车道线匹配,提出一种基于车道线提取的智能车横向定位技术。首先对定位坐标系进行定义和转换求解;然后利用Hough变换拟合直线检测图像中的车道线,在高精度地图中采用GPS领域搜索的方法得到相应车道线;最后利用P3P(Perspective-3-Point)求解3D到2D的匹配实现定位。与传统的智能车定位方法相比,具有计算量小、实时性好、匹配效率高的特点,且能够有效抑制横向定位误差。

To solve self-positioning problems of vehicles traveling in cities, the paper matches visual image sequential information with high-precision map to match lane lines, and then develops a new lateral positioning method based on lane line extraction. After the definition and transformation of positioning coordinate system, the paper applies Hough transformation to fit lane lines in line-detected images and GPS domain search method to obtain corresponding lines in high-precision map, uses Perspective-3-Point method to solve the match from three dimensions to two dimensions. Compared with traditional methods, this new method features less computation, prompt response and better match, and can effectively suppress lateral positioning errors.