Abstract: With the development of the railway construction and modernization of rail transportation, automated and efficient rail detection received widespread attention. The new method using the phase measuring profile metrology to detect the rail three-dimensional profile is proposed in this paper. By the computer-generated virtual grating, which is projected on to the measured rail, and deformed patterns captured by a Charge-Device (CCD), then the 3D contours of the object is restored by phase shift algorithm, phase unwrapping and other steps. This paper analyzes the advantages and disadvantages of the traditional Stoilov algorithm and in order to reduce the error, an improved algorithm based on probability and statistics is proposed. Experimental results show that using the phase measuring profile metrology to detect the rail contour is practicable.