Semiautomated Planning of Urban Drainage System by CNN-Based Land Use Segmentation and GIS-Based Drainage Generation Algorithm

Abstract In response to the challenges of inefficiency and rigidity in the manual planning of large-scale urban drainage systems, this study proposed a semiautomated urban drainage planning method by integrating convolutional neural network (CNN)-based land use segmentation with geographic information system technology for automated data preparation and drainage generation algorithm. The results showed that the CNN model could accurately detect and segment main urban land use and efficiently assist in the characterization of subcatchments. The drainage generation algorithm can generate not only the spatial structure of the pipeline network but also calculate the hydrologic and hydraulic parameters of each pipeline. The proposed method is particularly beneficial for regional and national drainage system modeling because it tackles common problems faced by water modelers who often lack access to detailed municipal data. Additionally, the semiautomated approach outperformed manual planning in terms of efficiency, offering a valuable practical tool for urban planners to rapidly generate baseline drainage layouts for large-scale planning and significantly improve planning efficiency.