基于CNN的路面裂缝图像分类

Classification of Pavement Crack Images Based on CNN

笔者提出一种基于CNN理论的的裂缝分类模型,共包括七层:输入层、两个卷积层、两个池化层、全连接层和输出层。该模型以Re LU函数作为激活函数,采用最大池化方法进行降维操作,并利用Softmax分类器将裂缝分为横向裂缝、纵向裂缝以及复杂裂缝(包括网状和块状)。实验数据选用4 000幅裂缝图像开展训练,选用1 000幅裂缝图像进行测试。结果表明,该基于CNN的裂缝分类方法具备良好的分类效果。

This paper proposes a crack classification model based on CNN theory.It consists of seven layers:input layer,two convolutional layers,two pooling layers,fully connected layers and output layers.The model uses the ReLU function as the activation function,and uses the maximum pooling method for dimensionality reduction.The Softmax classifier is used to divide the crack into transverse cracks,longitudinal cracks and complex cracks(including mesh and block).Finally,the experimental data was selected using 4000 crack images for training,and 1000 crack images were used for testing.The results show that the proposed CNN-based crack classification method can obtain good classification results.