In this article, we propose a vision based unmanned aerial vehicle automatic positioning and tracking system, and to use precise control algorithms to keep the unmanned aerial vehicle directly above the legged robot during the landing process, ultimately achieving precise landing. The research mainly includes the following five aspects: YOLO based object detection and recognition of a quadruped robots with Slicing Aided Hyper Inference (SAHI), relative position estimation between unmanned aerial vehicles (quadrotors) and quadruped robots, motion trajectory prediction and control strategies, combination of MPC (Model Predictive Control) and PID control algorithms, autonomous landing and error correction of quadrotors, all under Image-based visual-servo (IBVS) control.