ABSTRACT
With the rapid social and economic development and social progress, bridges are no longer purely to meet traffic demand. The huge span, strong body-shape expression, and extraordinary scale of the bridge may even reshape the urban or terrestrial landscape. Therefore, it is urgent to pay more attention to bridge aesthetics in the process of bridge design. In this paper, a neural network for assessing bridge tower aesthetics is proposed, which can classify a bridge tower into low or high aesthetic quality automatically, and better serve the bridge aesthetic design. Firstly, 150 cable-stayed bridges and 118 suspension bridges with the highest spans in the world were selected. For each bridge, several high-quality images were screened from the Internet, and the standard bridge tower diagram of each bridge was obtained through semantic segmentation and projection mapping. Then, a WeChat Mini Program for bridge tower aesthetic assessment was constructed and released, and the aesthetic assessments of the above bridge towers from different groups of people were obtained. Finally, the bridge tower aesthetic assessment network (BTAANet) was trained with the above standard bridge tower diagram and the corresponding assessment label, and the aesthetic quality can be obtained automatically with the well-trained BTAANet. The proposed method can provide an important reference for bridge design and promote the implementation of bridge aesthetics.
