基于Hessian矩阵多尺度滤波的路面裂缝图像检测方法

Pavement crack detection based on Hessian matrix multi-scale filtering

针对路面裂缝特征的快速提取与分类,提出一种基于Hessian矩阵的多尺度滤波算法。该算法通过从不同尺度检测路面图像中的裂缝,抑制路面图像非重要特征,突出路面裂缝特征,利用Hessian矩阵的特征值和特征方向实现裂缝生长方向的跟踪,再根据裂缝曲率特征对裂缝进行快速分类。实验结果表明,该算法可以准确、快速检测出裂缝覆盖区域并进行分类,抗噪声能力强;可以有效提取微弱路面裂缝信号,图像分割精度高,漏检率和误检率很低,适应大部分复杂路面环境。

Considering fast extraction and classification of pavement crack characteristics,an Hessian-based multi-scale filtering algorithm was proposed to detect pavement cracks.The cracks were detected in pavement images in different scales,which inhibited the unimportant features of pavement images,and highlighted the characteristics of pavement cracks.Growth direction can be traced by analyzing the feature and feature direction of Hessian matrix,then the cracks could be classified according to the crack curvature.Experimental results show that the proposed algorithm can detect the coverage area of crack with high accuracy and classify them quickly,it also has strong anti-noise capacity and can effectively extract the weak pavement crack signal,with high image segmentation accuracy,low mistake rate and miss rate,and it can be applied in most of the complex road environment.