In response to the challenges posed by conventional historical building conservation techniques, including inadequate 3D information representation, inefficiencies in multi-source datasets, and insufficient modeling precision, this study focuses on the Jianghan Customs Building and proposes a high-accuracy digital modeling method for historical buildings by integrating multi-source spatial information technology and reverse modeling. Through terrestrial 3D laser scanning, oblique photogrammetry, and close-range image acquisition, a framework combining global calibration and a weighted Iterative Closest Point (ICP) algorithm is developed to achieve high-precision registration of massive multi-source heterogeneous point clouds, effectively eliminating information gaps caused by scanning blind zones. Leveraging reverse modeling techniques, a component-based family library is categorized by functional attributes, integrating the geometric structure of point clouds with architectural semantic information to establish a historical building information model with both morphological accuracy and informational depth. The model is applied to current state assessment of damages, restoration and renovation decision-making support, and virtual heritage visualization, providing a scalable technical paradigm for the full lifecycle management of complex historical buildings.