Abstract: A novel approach based on 3D terrain model is proposed to detect pavement cracks. A 2D pavement surface image is mapped to a 3D spatial surface, and the cracks that are difficult to describe in 2D surface can be regarded well as valleys in it. Then by analyzing the characteristics of valley space curvatures, the valleys are detected. As a result, the majority of crack information is detected successfully.

Key words: road surface image, crack detection, differential geometry, space curvature, valley detection

摘要： 针对现有裂缝检测算法的不足，提出一种新的路面裂缝自动检测算法。将二维平面图像映射到三维空间曲面，使得在二维平面中难以描述的裂缝信息在三维曲面中能通过一条狭长的“山谷”来准确地描述。通过分析三维曲面中“山谷”的曲率特征，采用基于微分几何的空间检测算子准确提取曲面中的“山谷”，并映射到二维图像平面中作为裂缝点，成功检测到绝大部分裂缝信息。

关键词: 路面图像, 裂缝检测, 微分几何, 空间曲率, 谷点检测

CLC Number: 

• U416