基于1mm精度路面三维图像的裂缝种子自动识别算法

Automatic Recognition Algorithm for Crack Seeds Based on 1mm Resolution 3DPavement Images

为了准确地检测路面裂缝,给路面养护管理、路面性能评价与预测、路面结构和材料设计提供参考,基于1mm·像素-1的路面三维图像对裂缝自动识别进行研究。首先,将源图像划分为8像素×8像素的子块以降低图像维度;其次,根据深度验证和对称性检测将8像素×8像素的图像子块识别为裂缝子块（即裂缝种子）或非裂缝子块;然后,根据深度和方向相似性连接裂缝片段;最后,设计去噪算法消除孤立噪声,获得裂缝图像。结果表明：所提出的算法具有较高的准确率（均值92.75%）、召回率（均值58.93%）和运行速度（平均2~3s·张-1）,以71.15%的F值优于Otsu分割,Canny边缘检测和另一种子识别算法。

In order to detect pavement cracking accurately and provide reference for pavement maintenance and management, pavement performance evaluation and prediction, and pavement structural and material design, the research on automatic pavement cracking recognition was conducted based on 1 mm per pixel 3D pavement images. Firstly, a source image was divided into blocks of 8 pixels × 8 pixels to reduce image size. Secondly, the image blocks of 8 pixels × 8 pixels were classified as crack blocks （crack seeds） or noncrack blocks according to grayscale verification and symmetry check. Then, the crack segments were joined on the basis of depth and direction proximity. Finally, a denoising algorithm was designed to remove noises so as to obtain the crack images. The results show that the proposed algorithm achieves relatively high precision （with an average of 92. 75%）, recall rate （with an average of 58. 93%） and speed （with an average of 2-3 s per image）. It outperforms Otsu segmentation, Canny edge detection and another seeds based approach, with an F score of 71.15%.