Lucky imaging constitutes a pivotal approach for restoring turbulence-degraded imagery. Recent advances have explored method optimizations across diverse application scenarios. However， existing reviews published years ago do not cover breakthroughs from the past decade. This study conducts a thorough investigation of cutting-edge algorithms， first introducing classical methodologies and outlining persistent challenges. We then elaborate on the developments and applications through three dimensions： real-time implementation， multi-target adaptability， and integration with complementary image processing techniques. A dedicated turbulence dataset is released alongside systematic analysis of benchmark datasets， evaluation metrics， and comparative performance of leading methods. Scenario-specific applicability and inherent limitations are analyzed， culminating in six future trajectories： GPU-edge heterogeneous computing， dynamic turbulence modeling with non-stationary compensation， data-driven end-to-end fusion， multi-modal cross-scale restoration， event-camera-based dynamic imaging， and standardized evaluation frameworks with open-source ecosystems.