Abstract: Abstract: Three-dimensional (3D) visualization can greatly improve the quality and efficiency of urban rail lines design. However, the traditional CPU-based method of 3D modeling have some problems in modeling speed, render efficiency, optimization scene, which require to be improved urgently. A fast 3D modeling method of urban rail transport lines based on CPU discrete-GPU modeling is proposed. Firstly, the line is divided into linear models and point models, and then, according to the results of line design, CPU is used to decompose linear models to obtain dataset of discrete boundary conditions. Subsequently, the point models are analysed to obtain dataset of spatial information parameters which can form, independent discrete data packets with a small amount of data respectively. Finally, with these discrete data packets, the fast modeling of 3D line model is achieved owing to the parallel computing ability of GPU. Combining the scene picking ability of CPU with the vertex expansion ability of GPU, a kind of scene optimization method that applies to long linear model display is established. The results show that ① The time-consumed by this method is only 0.55%–1.30% of the time using traditional CPU-based method. ② The proposed method is better than traditional CPU modeling method and scene management in browsing experience, minimum frames per second is higher than 70. ③ The performance indicators such as occupancy of memory and CPU are greatly improved, which help to lessen the pressure of design platform. ④ It provides significant implications for the 3D design of practical urban rail transport lines.

Key words: Keywords: urban rail transport, line design, 3D transport line modeling, fast modeling