改进的基于gLoG滤波实时消失点检测算法

Improved real-time vanishing point detection algorithm based on gLoG filtering

消失点的检测对于非结构化道路检测具有重要意义。基于纹理的消失点检测方法有较好的普适性,其用一系列的具有不同方向、尺寸的滤波器,比如Gabor和gLoG滤波器,和图像进行卷积,根据得到的响应来判局部纹理的方向,然后再使用投票的策略来获得消失点的位置。这个过程需要大量的运算,运行时间较长。本文改进了经典的基于gLoG滤波的消失点检测算法,对图像纹理方向计算进行了优化,合并了大量的卷积操作,大大减少了运算时间。并使用了纹理方向线性邻域空间投票法替换了原文中的软投票策略,极大地减少了投票时间的同时,提高了消失点检测的精度。经过实验证明,本文的方法相比原方法,减少了运算时间,仅需几十毫秒即可完成检测,达到了了实时性,同时又有效地提高了消失点检测的精度。

Detection of vanishing point(VP) is of great significance for unstructured road detection. The vanishing point detection method based on texture has good universality. The researches show that it has obvious effect on general road image detection. Texture based vanishing point detection method has good universality. It uses a series of filters with different directions and sizes, such as Gabor and gLoG filters, and convolved with the image. The directions of the local texture are judged according to the obtained response. Finally voting strategies are used to obtain the position of the vanishing points. This process requires a lot of computation and runs for a long time. This paper improves the classical vanishing point detection algorithm based on gLoG filtering. First of all, this paper optimizes the direction of image texture and merges a lot of convolution operations, which greatly reduces the computation time. Then, the texture direction linear neighborhood space voting method is used to replace the original soft voting strategy, which greatly reduces the voting time and improves the precision of the vanishing point detection. Experiments show that the proposed method reduces the time of operation compared with the original method. It takes only dozens of milliseconds to complete the detection, and achieves the real time. At the same time, it effectively improves the precision of the detection of the vanishing point.