基于纹理的复杂环境下道路消失点检测算法

TEXTURE-BASED ROAD VANISHING POINT DETECTION ALGORITHM IN COMPLEX ENVIRONMENT

在乡村或者越野环境中,道路没有明显的边界和行车线,阻碍了一般视觉道路检测算法的应用。采用计算道路消失点的方法,提出一种快速的局部适应软投票FLASV(Fast Local Adaptive Soft Voting)算法。算法首先采用Gabor滤波器计算每个像素点的纹理方向,然后依据像素点的置信水平决定是否参与投票,最后采用快速局部投票的算法估计消失点的位置。实验结果分析表明,该算法能很好地适应复杂多变的环境,具有较低的时间复杂度和较高的准确性。

In rural or off-road environment, there are no obvious boundaries and traffic lanes on the road, which hinders the application of the general road detection algorithms. In this paper we present a fast local adaptive soft voting （FLASV） algorithm l^y adopting the method of road vanishing points calculation. The algorithm first uses Gabor filter to calculate the texture orientation of each pixel. Then, it determines whether or not to take part into the voting process according to the confidence level of the pixels. Finally it uses fast local voting algorithm to estimate the positions of vanishing points. Analysis of experimental results shows that the proposed algorithm can well adapt to the complex and changing environment, and has lower time complexity and higher accuracy.