Accurately capturing the geometric shape and structural information of buildings is crucial for point cloud reconstruction of damaged building scenes. Aiming at the problem of low completeness in point cloud reconstruction of damaged building scenes, a point cloud reconstruction method for damaged building scenes is proposed by integrating line segment descriptors and beam adjustment. Using the Mask RCNN method to detect damaged building scene targets and implementing two-dimensional projection on the obtained point cloud plane. Using principal component analysis method to estimate all point cloud normal vectors, introducing region growth method to cluster and segment all point cloud data, fitting each plane in the damaged building scene image, and obtaining the point cloud plane. Using the two-dimensional line segment detection method to obtain the line segment features of the damaged building plane, projecting the two-dimensional line segment into the three-dimensional space to obtain the three-dimensional line segment, and generating the line segment descriptor corresponding to the damaged building scene. Implement beam adjustment optimization on the obtained line segment descriptors to achieve point cloud reconstruction of damaged building scenes. The experimental results show that the proposed method can effectively reconstruct the point cloud of damaged building scenes, and the minimum relative error of point cloud reconstruction is only 0.117.