The multi-target fiber spectroscopic telescope can obtain a large number of spectral data of different celestial bodies in one observation. The light detected from the celestial body passes through the slit of the spectrometer, and after passing through the optical fiber, it is transmitted to the CCD sensor to obtain a two-dimensional spectral image. After a series of processing by the fiber optic spectral data processing system, the available spectral data is finally output and stored. The one-dimensional spectrum is the main means by which we obtain information about the target celestial body. The LAMOST telescope is used to obtain the observed celestial information. Taking LAMOST as an example, before obtaining a one-dimensional spectrum, the telescope system first obtains a two-dimensional spectrum consisting of 250 optical fiber spectra after one observation, and then undergoes a series of processing to obtain a one-dimensional spectrum. However, due to the increased use time of the telescope, the components will wear and age, which will cause a certain degree of bending of the fiber trajectory in the two-dimensional spectrum. This bending is particularly evident on both sides of the two-dimensional spectrum. The ordinate direction of a two-dimensional spectrum represents the wavelength direction of the extracted one-dimensional spectrum, and the abscissa direction represents the flow direction of the extracted one-dimensional spectrum. The generation of such deformation affects the subsequent wavelength calibration and flow. The calibration makes the extracted one-dimensional spectrum information inaccurate. The current initial solution is to minimize the impact by comparing with the calibration lamp spectrum. This not only causes a waste of time and manpower, but also has low accuracy and efficiency. In this paper, we propose a method of straightening the curved two-dimensional line based on the curve distance method. Firstly, the gray center of gravity method is used to locate the 250 fiber center trajectories in a two-dimensional spectrum, and the abnormal point is set. The robust local regression method is used to eliminate the curve, and then the center trajectory is curve-fitted to obtain the equation of the fiber trajectory. By simulating the inverse process of the curve bending, that is, keeping the curve distance between the two points on the trajectory unchanged, and then bending the spectrum Map to the vertical normal line to complete the straightening process. At the same time, the gray value of each corresponding point is kept unchanged, and the sparse problem of generating pixel points is solved by edge processing and interpolation operation. Finally, the one-dimensional spectrum extraction is performed by the accumulation method, and the one-dimensional spectrum extracted after straightening is compared with the undimensionally extracted one-dimensional spectrum. The difference between the two ends of the one-dimensional spectrum before and after the straightening is large, and the difference is passed. The value line also illustrates this. The method realizes the automatic alignment of the two-dimensional spectrum, which greatly improves the efficiency and accuracy of extracting the one-dimensional spectrum. The two-dimensional spectral pre-processing and alignment method proposed in this paper is validated on the LAMOST data at first. Considering the similarity of the principle of the multi-target Optical Fiber Spectral Telescope system, this method can also be applied to other multi-target Optical Fiber Spectral Telescope systems, and has good reference and application value.