Objective In the context of sustainable renewal, vacated industrial land has great potential for the development of urban green spaces and public spaces. The spatial structure of post-industrial landscape is highly unique, and the composition and layout of industrial heritage and new design elements have a significant influence on the visual quality and public perception of landscape. Coherence and legibility are important perceptual dimensions of landscape preference, and are expressed in the degree of harmony, recognisability and visual orientation of post-industrial landscape elements in physical spaces. Establishing the correlation between coherence, legibility and public perception is important for judging the sustainability of post-industrial landscape.

Methods This research selects 21 renovation scenes of Yangpu riverside public space in Shanghai for an eye-movement experiment and questionnaire survey, and conducts a cluster analysis of the scenes selected based on the findings of coherence and legibility perceptions, so as to clarify the correlation mechanisms between different landscape elements, visual attention and perceptual preferences.

Results The results show that the three types of scenes respectively dominated by plants, industrial remains and open spaces have different effects on the public's perceptions of coherence and legibility. Overall, coherence is positively correlated with the total edge density (ED) value of scene images themselves, and negatively correlated with the proportion of iconic industrial remains and the proportion of water bodies. Legibility is only correlated with the proportion of industrial remains. In the sub-category of legibility, the iconic industrial remains are positively correlated with the "iconicity" and "directionality" of scenes. The high area proportion of paving has a positive effect on the discrimination of "scene directionality", while the high area proportion of shrub and herbaceous plants is negatively correlated with the sense of scene direction. In scenes dominated by plants, legibility increases with the increase of industrial remains. The increase of natural elements, especially trees, significantly affects the judgment of "scene directionality" and reduces the legibility. In scenes dominated by industrial remains, the proportion of water bodies is negatively correlated with "landscape hierarchy" and legibility, especially the former. Meanwhile, a higher proportion of shrub and herbaceous plants is positively correlated with "scene iconicity", while the proportion of trees is negatively correlated with "landscape repetitiveness". A large number of industrial remains, especially heterogeneous industrial remains, are not in harmony with the environment, which may increase "landscape complexity" and interfere with the judgment of "scene direction". Compared with other scenes, the scenes dominated by open spaces are more coherent and legible, as there are relatively few landscape elements, and industrial remains such as buildings and chimneys are highly differentiated from natural elements, making it easier to judge "landscape directionality" and structure. Meanwhile, in terms of visual attention, overall, the interest in fixing iconic industrial remains and structures is positively correlated with the perception of legibility and negatively correlated with the perception of coherence, while the first fixation time of paving is positively correlated with legibility. In scenes dominated by plants, the first fixation time of natural elements is negatively correlated with legibility, while the first fixation time of newly installed facilities is positively correlated with coherence. In scenes dominated by industrial remains, the fixation duration time of natural elements is positively correlated with legibility, while the state of fixation on water bodies has a greater impact on the perceived state of coherence. In scenes dominated by open spaces, visual attention to iconic industrial remains plays an important role in the perception of legibility. It can be seen from the comparison of the three types of scenes that, in scenes dominated by plants, people are more likely to fix their eyes on artificial elements, while too much attention to plant elements may reduce the ability to judge the direction of scenes. In scenes dominated by industrial remains, people are more likely to quickly fix their eyes on industrial remains such as chimneys and fix them for a long time, and the guidance of sight by paving and plants can also help judge direction.

Conclusion Based on visual attention, this research establishes how visual attention relates to the perception of coherence and legibility, and explores the different perceptual effects of old and new elements on the public in post-industrial scenes. From the perspective of enhancing the coherence and legibility of the post-industrial landscape, the research endeavors to discuss how to optimize the spatial layout of industrial heritage, natural elements and new design elements. By analyzing how to improve the environmental coherence of the post-industrial landscape, it is possible to better integrate the preservation of historical heritage with the placement of recreational functions. The research provides a reference for the sustainable use of existing resources to optimize the spatial layout of the post-industrial landscape and to reasonably protect the historical industrial landscape. Research results indicate that the sense of spatial hierarchy and the coherence and legibility of scenes can be improved by incorporating an appropriate proportion of plant elements, ensuring adequate identification of industrial remains, and using plant elements to emphasize the structure of the remains. Additionally, the overall coherence and legibility of scenes can be further enhanced by increasing the proportion of paving and water bodies, improving the quality of water bodies, and emphasizing landscape elements such as landmarks, nodes and paths.