一种基于多尺度脊边缘的沥青路面裂缝检测算法

发布时间：2018-06-02 07:14

本文选题：道路工程 + 路面裂缝检测　；参考：《中国公路学报》2017年04期

【摘要】：为了提高沥青路面裂缝的检测精度,针对裂缝的复杂性和多态性,提出了一种普适性较好的基于多尺度脊边缘的沥青路面裂缝检测算法。首先,对路面不同形状裂缝的多种脊边缘特征进行分析,并对利用高斯函数及其导数检测脊边缘的原理进行了推导;其次,根据脊边缘的多尺度特性,采用多个尺度的高斯滤波器对图像中的多尺度脊边缘进行检测,并对各个尺度的检测结果进行像素级融合;然后,结合裂缝的长度、宽度、方向、连续性、线性度等特征去除伪裂缝;最后,利用最小生成树算法实现了检测图像上不同位置裂缝的连接,使得检测出的裂缝更加接近实际裂缝的形态。研究结果表明:提出的算法可以对不同形状特征、尺寸和位置的裂缝目标进行有效检测,具有较高的检测精度和较好的检测效果,且该算法抗噪性能好、通用性强,是一种具有工程应用前景的高效的沥青路面裂缝检测方法。
[Abstract]:In order to improve the detection accuracy of asphalt pavement cracks, in view of the complexity and polymorphism of the cracks, an algorithm based on the multi-scale ridge edge based asphalt pavement crack detection is proposed. First, the characteristics of various ridge edges of different cracks in the pavement are analyzed, and the Gauss function and its derivatives are used to detect the edge of the ridge. Secondly, according to the multi-scale characteristics of the ridge edge, multi scale Gauss filter is used to detect the multiscale ridge edge in the image, and the detection results of each scale are fused in pixel level. Then, the pseudo cracks are removed with the characteristics of the length, width, square direction, continuity and linearity of the cracks; finally, the advantages of the fracture are removed. The minimum spanning tree algorithm is used to connect the cracks in different positions on the detected image, making the detected cracks closer to the actual cracks. The results show that the proposed algorithm can effectively detect the crack targets with different shape features, sizes and positions, and has high detection precision and good detection effect. Moreover, the algorithm has good anti noise performance and versatility. It is an effective method for crack detection of asphalt pavement with engineering application prospect.
【作者单位】：安徽工程大学电气工程学院;长安大学信息工程学院;
【基金】：陕西省自然科学基金项目(S2013JC9397) 高节能电机及控制技术国家地方联合工程实验室基金项目(KFKT201510) 安徽工程大学校青年基金项目(2015YQ15)
【分类号】：TP391.41;U416.217;U418.6

【相似文献】