Space non-cooperative targets are often in the state of tumbling. It doesn’t communicate in information and doesn’t cooperate in maneuvering. The difficulty of tumbling spacecraft capture is how to accurately obtain the relative position and attitude information of the target when the target's motion and spatial structure are unknown. Tumbling spacecraft capture is often divided into the following four stages: (1) flying-around; (2) hovering; (3) approach; (4) close-range capture. In this paper, the dynamics modeling, motion state estimation, and capture path planning of a tumbling spacecraft are investigated. Dynamic environmental map updates are then performed, including information on target spacecraft, debris barriers, etc. Since the structures of tumbling spacecraft are typically easy to identify, including the solar wing and its support frame, communication antenna backplane, star-arrow docking ring, and engine nozzle, here we choose the star-arrow docking ring as the capture point. During the hover phase, we propose an adaptive threshold Canny algorithm for tumbling spacecraft profile and star-arrow docking ring detection to achieve accurate identification and body point extraction of tumbling spacecraft and star-arrow docking ring. Specifically, instead of Gaussian filtering, a bilateral filter that considers both the value and spatial domains is used to combine pixel position and brightness information. The effect of the edge information is effectively protected. At the same time, the optimal threshold segmentation method is combined with the Otsu method to segment the gradient image, and the high and low thresholds of the image are calculated separately to obtain a well-effective edge feature point set for accurate identification of tumbling spacecraft. After acquiring the feature set of the tumbling spacecraft, the RANSAC algorithm is used to perform initial point cloud alignment, complete the construction of the tumbling spacecraft and its surrounding environment, and acquire the initial position attitude of the tumbling spacecraft, then the iterative closest point (ICP) algorithm is used to enter the point cloud alignment to achieve accurate estimation of the position attitude of the tumbling spacecraft. On this basis, the dynamic closest point of the tumbling spacecraft is calculated based on the tumbling spacecraft motion model and the capture based on the dynamic closest point is performed. Simulation show that the proposed adaptive threshold Canny algorithm can more accurately identify the target spacecraft and the docking ring, and the improved point cloud alignment algorithm can accurately estimate the position attitude of the tumbling spacecraft, while verifying the effectiveness of the dynamic closest point-based tumbling spacecraft capture algorithm.