Machine intelligence for interpretation and preservation of built heritage

Abstract

Documenting and characterizing built heritage through digital format are topical issues in the architecture and heritage preservation field. Although digitalized built heritage (DBH) features are complex, they have been intelligently interpreted and perceived by researchers supported by machine learning (ML) models. This paper reviews the mainstream ML models applied in the tasks of quantitative interpreting of formal features and parsing of multi-spatial-element synergy mechanisms, and summarizes their applications in the major issues of DBH characterization research, to show their operation paradigms and demonstrate what gaps still exist. Based on the review, the ML models have been capable of quantitatively extracting the formal features of DBH and parsing the synergy weights of multi-spatial-elements. However, future research still requires advances in 1) Automatically summarizing the DBH formal features; 2) Taking point clouds as an ideal DBH carrier; 3) Forming a computer-autonomous decision-making path for built heritage preservation.