This paper firstly introduces the principle of distributed sensing system based on Polarization sensitive Optical Time-Domain Reflectometer (P-OTDR). As known, the change of State of Polarization (SOP) in fiber is sensitive to environment, consequently this sensing method can extract very little perturbation of extra temperature or strain; another significant advantage is that the sensing information can be achieved nondestructively at one end. Hence, this technique has become a practical and helpful tool in distributed optical fiber sensing system. However backscattered light in single-mode fiber is very weak, and the optical pulse width limits the spatial resolution, so there is a compromise between SNR and sensor spatial resolution. We investigate two signal-process methods to extract the state polarization (SOP) evolution from backscatted signal. From the experiment results, we find that both the digital average and deconvolution can be used for enhancing SNR and system spatial resolution. And we also find hundred of digital average is enough, the more time we average the less improvement of SNR we get. At the same time, we view that the SOP of backscattered signal is the response of sensing system. By mean of deconvolution the actual SOP evolution is not lost, and the spatial resolution of POTDR System doesn’t degrade with a wider pulse. Compared with previous deconvolution case, the deconvolution method can effectively improve the system spatial resolution without decreasing the transmitting pulse width. Finally we give some experiment results with strain perturbation, our system can detect the point where the environment changes, which is suitable for smart structure and optic-fiber sensor application.