Objective The Grand Canal is an important water conservancy facility that embodies the ecological wisdom of ancient Chinese people in water control and management. It is also a bridge connecting Northern and Southern China in both geological space and civilization. As the only "living" cultural heritage in the world, the Grand Canal still maintains its original functions. The Grand Canal and its surrounding environment, under the influence of urbanization, not only retain their own historical and cultural characteristics, but also add modern functions such as recreation, ecological construction and cultural heritage in combination with the public needs of towns nearby. In the face of many changes faced by the Grand Canal nowadays, this research examines the conservation and development practice of canal heritage from a dynamic and developmental perspective, and recognizes the experience and needs of modern society for canal heritage more reasonably and objectively, in hope of providing more ideas for the conservation and utilization of the Grand Canal heritage, and providing prudent reference for the construction of the Grand Canal National Cultural Park.

Methods The LDA (latent Dirichlet allocation) topic model is a text mining method based on the unsupervised machine learning technology, which is used to analyze and identify potential topic information within a given document. Yangzhou, as the origin city of the Grand Canal that takes the lead in applying for the inclusion of the Grand Canal into the list of world heritage, has an important special position along the Grand Canal. There are 10 sites and 6 river courses in Yangzhou that have been listed as world heritage, which constitute the unique and rich canal cultural resources of Yangzhou. Taking Yangzhou as an example, this research invokes the LDA topic model using the Gensim topic model toolkit in Python, to explore the imagery perception characteristics of the current public about canal heritage and relevant cultural landscape. By making a judgment on the perception dimension attribution of feature words in 25 topics, this research calculates the cumulative probability values of feature words under Level I and Level II dimensions in each topic. On this basis, the research classifies and compares the contents of each topic, and summarizes and extracts the imagery perception of Yangzhou canal urban cultural landscape. The research methods used are helpful to extract the imagery perception characteristics of the canal cultural landscape from online text data in a more standardized and semi-quantitative way, and thus help understand the public's multi-dimensional cognition of Yangzhou Grand Canal heritage from the bottom up. Through the LDA topic model, the research constructs a description system for imagery perception of cultural landscape, and analyzes the topic contents of Level I and Level II perception dimensions, which shows that the public's perception imagery of Yangzhou canal cultural landscape is generally abundant.

Results The data obtained shows that: Under Level I perception feature dimension, the imagery perception of canals in Yangzhou is mainly carried out from three aspects, namely typical canal landscape, canal visiting experience and regional influence of canal cultural landscape; under Level II perception feature dimension, high-frequency words related to cultural landscape, natural scenery, special food, commercial services and service facilities jointly build abundant canal perception imageries, and reflect the capability of the ancient canal to connect different urban resources. These imageries not only reflect the modern society's cognition of various tangible and intangible canal heritage resources, but also depict the modern features of Yangzhou as a whole under the influence of the ancient canal.

Conclusion Based on the above research, three conclusions are drawn. First, in the process of heritage conservation and utilization, a comprehensive review should be conducted from the spatial dimension. It is necessary to consider not only the ecological quality and landscape quality of canal water system, but also the structural level and functional layout of the riverside city area. More importantly, the overall canal heritage resources at the urban scale should be considered to prevent the one-sided interpretation of canal cultural landscape. Second, in the future, it is still necessary to comprehensively consider the specific historical characteristics and material relics of the Grand Canal to open up a wider path for heritage activation and utilization and enhance the interest and information dissemination of relevant heritage displays by a variety of media carriers and display technologies, thus ensuring the true and effective realization of the authenticity of the Grand Canal heritage. Third, the construction of the Grand Canal Cultural Belt and the Grand Canal National Cultural Park should be coordinated with that of canal heritage and local leisure, tourism and service resources, and the changes in the public's perception and experience of the Grand Canal heritage and surrounding environment should be taken as an important reference for adjusting relevant planning policies and management models, so as to inject new vitality into the survival and development of canal heritage. This research can provide experience and data support for the shaping of relevant urban cultural landscape, and the construction of the Grand Canal Cultural Belt and the Grand Canal National Cultural Park.