In the construction of large and intricate frame-panel structures, the curvature and twists of the backbone present a challenge in the assessment of the installation quality and the attachment of the panels.The classic method for panel detail design is associated with several drawbacks, such as low efficiency and the need for greater automation. In the context of the Nansha International Finance Forum project, we investigated intelligent construction techniques for large and intricate frame-panel structures, based on point cloud data (PCD) and heuristic algorithms. A backward modelling method applicable to the frame was proposed to meet the demand for the detailed design of the daylighting roof, and to build the reconstruction model for detailed design. An algorithm for extracting the axes of a polygon’s curved and twisted components was developed to assist in the geometric description. An intelligent cell extraction method was introduced to address the low efficiency of manual panel segmentation. Different strategies were used to generate and evaluate panel layout plans, based on the Guillotine algorithm. The results show that the proposed inverse modeling approach allows for the acquisition of profile geometry information with a defined sampling interval. Furthermore, the established frame model exhibits an accuracy better than 6 mm, making it suitable for panel deepening design applications. Panel cell partitioning methods could automatically extract information about various panel sizes, reducing manual point selection efforts. The material waste rate of the panel layout plan generated by the Guillotine algorithm could be limited to 15%.