中文摘要：

钢轨表面损伤的检测对于铁路运输安全具有重要的意义.不同类型的损伤具有不同的成因和特征,单一的检测算法针对性和鲁棒性不强,应根据不同的损伤类型提出针对性的检测方法.本文针对鱼鳞纹损伤的规则方向性特征,提出了基于转动惯量等效椭圆的区域方向性计算和基于纵向区域直方图的区域方向性筛选算法;针对踏面剥离裂纹和浅层掉块损伤的呈带状分布的不规则、不连续和多凹陷特征,提出了基于多孔洞区域的骨架提取的检测算法.这两种算法均能有效地检测出相应损伤,为缺陷的分类识别提供参考.

英文摘要：

The detection of rail track surface defects is very important for railway transportation. Be- cause different types of defects have their distinct causes and features, a single algorithm is not suitable and robust enough for various defects. We proposed different algorithms for corresponding defects. Ac- cording to the regular orientation feature of cataphracted defects, an algorithm, which computes region directions based on equivalent ellipse of moment of inertia and sifts the directions based on histogram of vertical regions, is employed. For irregular, discontinuous, cupped and distributing band-like features of the tread peeling and drop piece defects, an algorithm based on skeleton extraction of multi-hole re- gions is proposed. The methods can effectively detect those rail surface damages, which provide a ref- erence for rail damage classification.