For target cognition and data fusion, the sensors are usually considered as synchronous sampling. This assumption is often no longer valid as the data rate of infrared sensors is significantly higher than that of radar. In this paper, a novel algorithm named LS IMM/MSPDAF is proposed for the least squares (LS) virtual fusion method by involving time calibration due to multi-sensors asynchronous sampling. As for the interactive multi-model (IMM) part, it is designed for fusing multimodal observations in order to track a highly maneuvering target with different dimensions. Experimental results of two case studies have demonstrated that our algorithm can conquer the defect of asynchronous sampling, and achieve improvement in the accuracy of track estimation than MSPDAF.