It is necessary to fuse the identification data from multi-sensor for the effective identity determination in order to meet the identification requirements of target attributes under complex battlefield environment. Based on the fundamental belief evaluation and stability of the sensors, the reliability evaluation of identification evidence is carried out, which would reflect the uncertainty introduced by sensor working mechanism, target distributing and registration association. D-S evidence fusion structure is selected dynamically according to evidence reliability evaluation results and evidence conflict status to decrease the disastrous effect on fusion result from conflict evidence. Furthermore, the history identification evidence source tracking process is performed to integrate asynchronous evidence and this would avoid the false identification caused by non-independent evidence. Finally, the simulation indicates the validity of the method.