Abstract: Surrounding rock classification is a fundamental step in the design, construction, and support optimization of drill-and-blast tunnels. Traditional classification systems, such as Q, RMR, and BQ, provide relatively mature engineering frameworks for evaluating surrounding rock quality. However, the acquisition of classification-related information still heavily relies on manual geological logging and test parameters, resulting in delayed updating, strong subjectivity, and limited adaptability to dynamic construction changes. To address these challenges, based on tunnel-face images, while-drilling parameters, and related auxiliary information rapidly accumulated during drill-and-blast tunnel construction, this paper systematically reviews the research progress in intelligent surrounding rock classification. First, bibliometric visualization is used to summarize the research intensity and development trends in this field, and existing studies are categorized into three main directions: image recognition, while-drilling parameter interpretation, and multi-source data fusion. Second, key supporting technologies for intelligent surrounding rock classification are summarized, including tunnel-face lithology identification, weathering degree and groundwater condition recognition, discontinuity and fracture detection and quantification, and the selection, processing, and interpretation of while-drilling parameters, with their intermediate supporting roles in surrounding rock classification clarified. Furthermore, regarding multi-source fusion-based surrounding rock classification, the basic principles, implementation procedures, and application characteristics of data-level, feature-level, and decision-level fusion are summarized. It is indicated that feature-level fusion of tunnel-face images and while-drilling parameters has become a major implementation route in current research. Finally, from the perspective of classification implementation pathways, existing methods are classified into direct discrimination methods, quantitative mapping methods, and feature fusion methods, and their technical procedures, applicable conditions, advantages, and limitations are analyzed. The review shows that intelligent surrounding rock classification is shifting from single-source information recognition toward collaborative driving by image recognition and while-drilling parameters. Future research should further improve relevant theories in terms of quantitative parameter standardization, multi-task collaborative recognition, deep multi-source fusion, and on-site edge deployment and application promotion.

Key words: tunnel engineering, surrounding rock classification, drill-and-blast method, intelligent sensing, multi-source data fusion, feature-level fusion