To address the issue of urban monitoring networks being unable to track and monitor targets located in video surveillance blind spots, and unable to comprehensively use target trajectory data obtained from multiple channels to determine abnormal trajectories, this paper designs a target abnormal trajectory recognition system model based on a space-based monitoring platform. This article first designs the architecture and organizational operation process of the system, and elaborates on the functions of the image acquisition front-end, server-side, and client-side; Secondly, in order to achieve the goal of extracting the essence and authenticity of target trajectory data, this paper proposes a trajectory fusion algorithm based on hesitant fuzzy set theory to reduce the impact of incorrect target trajectory data; Finally, in order to achieve prediction of target actions, this paper proposes a trajectory prediction algorithm based on the LSTM prediction model, which provides assistance in early warning of abnormal trajectories of targets.