DWTA-Unet: Concrete Crack Segmentation Based on Discrete Wavelet Transform and Unet
【摘要】：In this paper, aiming at the characteristics of complex structure and tiny cracks on the surface cracks of concrete, a novel Discrete Wavelet Transform and Attention(DWTA) module is proposed to precisely predicted the crack region based on the famous skeleton network Unet. More crack contextual information is obtained and irrelevant information is suppressed in the output of DWTA module. The effectiveness of the proposed method is verified on crack dataset. Compared with other methods, this method shows the better performance. The Mean intersection over union and F1-socre reach 0.848 and 0.824, respectively.