Due to the complex working conditions around the city water supply pipeline, it is difficult to achieve pattern recognition on leakage and other interference vibration. This study proposes a method to identify leakage vibration detection by utilizing the ФOTDR distributed optical fiber sensor based on fractal box dimension and improved approximate entropy. According to the signal space filling feature of cavitation noise near the leakage, the time domain signal is measured by the fractal box dimension. A new scalefree zone searching method is adopted. Then, the value of the box dimension parameters can be obtained by the high feature of the signal according to the turbulence noise signal chaotic characteristics. According to the turbulence noise signal chaotic characteristics, the approximate entropy is employed to measure the timedomain signal. A new method of selecting a similar margin is used. Hence, the improved approximate entropy can reflect the randomness of signals. The fractal box dimension and improved approximate entropy of denoised signal are utilized as the input of the twodimensional parameter plane. By setting the threshold range in the plane, the distinction between the leakage signals and interfering signals can be realized. By processing signals detected in experiments using this method, the results show that the accuracy rate of leakage recognition can reach up to 96.7%, which indicates that this method can separate water pipeline leakage signals and other signals accurately and efficiently.