U-Net Based Concrete Honeycomb Planking Segmentation in Sewage Treatment Plant under Low Light Condition

The honeycomb planking occurs when the vibrated concrete is not compacted during the construction process of sewage treatment plants, which leads to leakage if not addressed in time. In order to achieve rapid detection of honeycomb planking under low light and background noise in sewage treatment plants, this study proposes a U-net-based deep learning method for honeycomb planking segmentation, which first annotates the original image of the concrete structure and reorganizes the data set after data enhancement, so that the deep learning model is still robust and has a strong generalization ability in the case of edge blurring and lighting changes. This study is based on a wastewater treatment plant to validate and achieve good results, with an accuracy rate of more than 83%, compared with the traditional manual detection method. In addition, the research in this paper was carried out under complex conditions, such as a low light environment and the challenging situation that it is difficult to manually detect the inner wall of the wastewater treatment plant. Compared with the traditional manual detection methods, the segmentation model based on u-net proposed in this paper can automatically, quickly, and on a large scale, detect cell defects and reduce manual intervention.