This paper proposes a Level-of-Detail (LOD) based method for modeling object movement in wireless sensor network tracking and surveillance applications. This method addresses the dilemma of achieving maximum information with limited power and bandwidth in wireless sensor network. LOD provides the researchers with various fidelity levels of data to perform such tasks as regional traffic statistics analysis, object classification and clustering, and at the highest level, object behavior analysis. A simulation test bed extracted from the sensor network hardware platform is being built to allow users to easily generate objects with various behavioral models, design three-dimensional sensor deployment layouts and related surveillance environments, test the accuracy and efficiency of tracking, and calculate lifetime of system and power consumption. An illustrative experiment with preliminary results is provided at the end of this paper.