Abstract: Target recognition using information fusion techniques becomes an important research direction in the domain of pattern recognition. Temporal spatial information fusion using the DS theory is a research focus in the domain of information fusion. Information fusion techniques include integration of several evidences in the spatial domain and integration of evidences in different moments from the same evidence source in the time domain. This paper introduces three classical temporal spatial information fusion models, namely recursive centralized, recursive distributed without feedback and recursive distributed with feedback firstly. By theoretical analysis and example simulation, this paper presents a new conclusion which is different from the others. Then an effective hybrid model of temporal-spatial information fusion is proposed according to the background of space target Fusion Recognition, which can save system resources and reduce the computational complexity on the condition of ensuring recognition efficiency. Simulation result shows the efficiency of the new model.