中文摘要：

由于轨道影像具有较高的灰度相似性,传统一阶Harris算法提取轨道影像特征存在时间效率低、特征点聚簇等负面问题,文章针对已有Harris方法的不足,提出了一种基于图像分块自适应闻值的二阶导数Harris特征点检测算法,采用自相关矩阵兴趣值识别轨道图像特征点,并分别对有砟与无砟铁路轨道影像开展特征点提取.实验结果表明,二阶Harris算子在轨道近景影像特征点提取方面具有较高的计算效率,所提取出轨道影像特征点满足均匀分布的空间特征,有效避免了传统方法特征点的聚簇现象.

英文摘要：

The phenomenon of image feature cluster has to be overcome with the traditional Harris op- erator due to the special grey similarity of track image. The improved feature extraction algorithm was pro- posed in this paper with track image segmentation and second derivative Harris operator. The autoeorrela- tion matrix and improved interest operator were introduced into the identification of track feature. The ex- periments were performed on the real ballasted and the ballastless tracks with the proposed feature extrac- tion method. The computation results showed that the proposed second derivative Harris operator would be feasible and reliable to extract track image features with evenly distributed image feature points of the tracks, which could avoid feature clusters from traditional methods.