In the context of autonomous landing for unmanned aerial vehicles (UAVs), selecting a suitable landing site is crucial. This research presents a new method for automatically identifying safe landing sites based on point cloud data acquired through an airborne laser scanning (ALS) system. The proposed approach begins by detecting flat regions using principal component analysis (PCA) and region-growing algorithms. Subsequently, a terrain complexity assessment is conducted through plane fitting using an enhanced progressive sample consensus (PROSAC) algorithm. This assessment assists in identifying the most suitable landing site within a specified landing zone. The method's effectiveness is demonstrated through experiments conducted on two distinct natural terrains from the DALE dataset. The results show that the proposed approach can accurately classify landing zones and identify preferred sites that meet safety criteria. This study's findings underscore the proposed method's effectiveness and feasibility in improving the safety and reliability of autonomous UAV landings.