With the rapid advancement of optical fiber technology and optical communication technology, optical cables have been deeply integrated into the process of national construction as fundamental resources. Phase-sensitive optical time domain reflectometer is extensively utilized in domains such as perimeter security, railway train tracking, anti-excavation of oil and gas pipe-lines, and oil and gas exploration, owing to its merits like wide monitoring range and high sensitivity. In recent years, improvements in its optical system by researchers have resulted in longer sensing distances and higher spatial resolutions. Nevertheless, the volume of data has significantly increased, and the diverse types of environmental noises and disturbances have presented challenges to the practical application of the distributed disturbance sensing system. In this paper, by analyzing the signal characteristics of the sensing system, a compressive sensing algorithm is proposed to process the signals of the sensing system, thereby achieving the precise positioning and restoration of vibration signals. The application of compressive sensing algorithms is mainly for sensing systems that require frequency sweep operations or systems that do not require frequency sweep but have a short sensing length. Currently, there is no application of compressive sensing algorithms to the φ-OTDR system that does not require frequency sweep and has a considerable distance. In view of the issues of a large number of data points and high computational complexity, the data dimension for each compressive sensing is reduced through windowing, and then the compressive sensing algorithm is applied to each window. Experiments demonstrate that the segmented compressive sensing can effectively reduce the required reconstruction time, and the reconstructed signal can still maintain a high level of accuracy, ultimately enabling vibration data reconstruction at a low sampling rate. For the similar characteristics of amplitude signals, a representation dictionary of amplitude signals is established through sparse coding and dictionary learning to realize the representational capability of the learning dictionary for signals, and the denoising target of the target signal is accomplished through the block matching algorithm. Experiments indicate that compared with the results obtained by the traditional moving average algorithm, the results obtained after introducing block matching have a superior signal-to-noise ratio.