Considering the importance of spatial non-cooperative target intention recognition for spatial situational awareness, this paper proposes a spatial non-cooperative target intention recognition method based on long short term memory networks. This method takes the relative position and velocity of the non-cooperative target as inputs and identifies its intention through a long short term memory network. To this end, the typical intentions of the target are defined and classified based on the CW equation, and the non-cooperative target intention datasets are generated based on the analytical solution of the CW equation. Building an LSTM network and training it with the datasets resulted in an accuracy of 94.26%. This method outperforms traditional intent recognition methods in terms of recognition time and accuracy. This method provides a new approach for identifying non-cooperative target’s intention in space, also, quickly and accurately identifying the non-cooperative target’s intention helps the spacecraft takes reasonable measures to ensure their own safety.