This paper combines multiple techniques, including novel stereovision, three-dimensional image reconstruction and infrared photography technique, to develop a stereovision-based video surveillance system with multi-lens module. The main function of this system including binocular stereoscopic vision, autofocus with motion tracking 3D image reconstruction, and 3D thermal image. First, the system uses the novel stereovision for obtaining three- dimensional position of the surveilled object. Second, tracing the moving surveilled object and record a clear video by using hybrid autofocus technique. Third, using image capture, camera calibration, stereo calibration, image matching, and parallax depth calculation to reconstruct surface reconstruction of 3D models. Finally, the system provides the 3D thermal information of the field, and it can remove the environment thermal information to show the surveilled object’s only.