基于类Haar纹理的非结构化道路消失点检测

在非结构化道路识别算法的研究中,因其在复杂道路环境下表现出较强的鲁棒性,基于纹理特征的道路消失点检测成为了学术研究的重点。因此,为了能够准确的检测出非结构化道路中的消失点,提出一种基于类Haar纹理的非结构化道路图像消失点的检测方法,即首先设计多尺度多方向的类Haar纹理的实部与虚部模板,利用积分图技术以及垂直投影法快速计算出图像像素的纹理主方向,再采用改进的局部软投票法选出分数最大的点作为初始消失点,最后通过现有两种不同的消失点检测方法与提出的算法进行对比实验。实验结果表明提出的算法可以显著提高非结构化道路消失点的检测性能。由此得出该方法具有良好的实时性、准确性和鲁棒性。

Unstructured Road Vanishing Point Detection Based on Haar-like Texture

In the research of unstructured road recognition algorithm, due to its strong robustness in complex road environment, road vanishing point detection based on texture features has become the focus of academic research. Therefore, in order to accurately detect the vanishing points in unstructured roads, an unstructured road image vanishing point detection method based on Haar-like texture is proposed. First, the real and virtual part templates of multi-scale and multi-directional Haar-like texture are designed, and the texture principal direction of image pixels is calculated quickly by using integral graph technique and vertical projection method. Then the improved local soft voting method is used to select the point with the largest score as the initial vanishing point. Finally, two different vanishing point detection methods are compared with the algorithm in this paper. Experimental results show that the proposed algorithm can significantly improve the detection performance of unstructured road vanishing points. It is concluded that this method has good real-time, accuracy and robustness.