基于MDCGAN的裂缝样本扩充及识别研究

Research on Crack Sample Expansion and Recognition Based on MDCGAN

摘要:

针对裂缝图像获取困难导致的样本少、传统数据扩充方法提升样本特征空间能力不足等问题,提出了一种基于改进深度卷积生成对抗网络(MDCGAN)的裂缝样本扩充方法。首先对数据集进行预处理,利用滑窗法进行数据降维和清洗；其次优化激活函数,提高生成特征的多样性,同时引入谱归一化进行权重标准化提升网络结构的稳定性,以生成高质量的裂缝数据集；最后,利用改进的Alexnet网络对扩充后的混合样本集进行特征提取并分类识别。结果表明,MDCGAN网络数据增强性能与传统扩充方法相比均有明显提高,适用于扩充裂缝图像。

Abstract:

In view of the lack of samples caused by the difficulty in obtaining crack images and the insufficient ability of traditional data expansion methods to enhance the sample feature space, a crack sample expansion method based on modified deep convolutional generative adversarial network (MDCGAN) was proposed. Firstly, the data set was preprocessed, and the sliding window method was used for data dimension reduction and cleaning. Secondly, the activation function was optimized to improve the diversity of generation features. At the same time, spectral normalization was introduced for weight standardization to improve the stability of network structure, so as to generate high-quality crack data set. Finally, the improved Alexnet network was used to extract and classify the extended mixed sample set. The results show that the data enhancement performance of MDCGAN is significantly improved compared with the traditional expansion method, which is suitable for expanding crack images.