基于灰度分析的路面裂缝图像分类算法研究

Research on Image Classification Algorithm of Pavement Crack Based on Gray Scale Analysis

为了应对传统人工调查方法已不能适应当今道路养护管理发展需求的问题,从提高路面图像分类检测效率的需求角度出发,设计基于图像掩膜的去噪增强算法,采取路面图像的灰度作为分类特征,并且设计路面图像子块灰度方差的最值之差、一幅路面图像中子块图像方差的最值之差超过阈值的子块图像数量,以及子块图像灰度熵的最值之差为特征的分类算法,并引入神经网络进行路面图像的分类.实验结果表明,相对于中值滤波、均值滤波、偏微分方程、掩膜增强算法,所选取的图像去噪增强算法能够保留路面图像中的细微裂缝;且图像分类算法具有高检测率,零误检率,能够实现大规模路面图像的快速、准确分类.此外,所建立的分类算法能够保证分离出的完好图像中,不会掺杂路面裂缝图像,可以降低后续对路面裂缝图像处理的工作量,且其能为公路养护管理的决策提供依据.

In order to solve the problem that the traditional manual survey method can no longer meet the development needs of road maintenance management today, a denoising enhancement algorithm based on image mask was designed from the perspective of improving the efficiency of road image classification and detection. The grayscale of the pavement image is considered as the feature, the variance difference between maximum and minimum of the sub-block image, the number of image （the variance difference of the sub-block image is more than the threshold） and entropy difference between the maximum and minimum of the sub-block image are selected as the features, and the neural network is used to classify the pavement image. The results show that image de-noising algorithm designed in the paper could preserve the fine cracks in the pavement image compared with the median filter, the mean filter, the partial differential equation and the mask enhancement algorithm. And the classification algorithm has high detection rate and zero false detection rate, which can realize the fast and accurate classification of large-scale pavement images. Moreover, the established classification algorithm can ensure that the separated intact images will not be mixed with pavement crack images and reduce the subsequent workload of pavement crack image processing, which can provide a basis for the decision-making of highway maintenance management.