Visual Perception-Driven Renewal Design and Preservation of Traditional Industrial Heritage

Abstract

Abstract Traditional industrial heritage is a non-renewable resource with significant research value. In this paper, the features in traditional industrial heritage images are fused based on the YOLOv5s method driven by computer visual perception. The visual perception speed is improved by adjusting the scaling factor in the attention mechanism module of normalized weights, and the model is further enhanced by designing the lattice loss function. Furthermore, the research employs the fuzzy comprehensive evaluation method to investigate the level of public visual perception of traditional industrial heritage. The results show that the frequency of public perception of the characteristic street category and macro-scale industrial heritage is higher, 25.38% and 40.11%, respectively, and there is a significant difference in terms of the public’s impression of traditional industrial heritage among different permanent residences (p=0.015<0.05). The military-affiliated public had the highest mean perception score of 3.87 for traditional industrial heritage. The analysis of the public’s visual perception preference for industrial heritage sets the foundation for the renewal design of traditional industrial heritage and promotes the conservation process.