With the increase of energy demand, the mileage of energy transportation is getting longer and longer. In the field of energy transportation, the petrochemical pipeline is the main mode of transportation, and its safety is of vital importance. A small leakage of the petrochemical pipeline accompanied by internal pressure in the fluid can cause fire and other safety accidents at any time, affecting the safety of the entire section of the pipeline and station yard of the production. Therefore, pipeline safety inspection is necessary and indispensable. In the process of oil and gas transportation, due to the increase of external load, erosion, and other factors, the local stress concentration of the pipeline is caused. The pipe wall is thinner, and the leakage point appears on the surface, which reduces the safety performance of the pipeline. Through the use of an acoustic detection device to establish safety detection of pipeline micro-leakage points, multiple groups of leakage data were collected in the experiment, including Φ0 mm, Φ1 mm, and Φ1.5 mm leak points, respectively. The mathematical model for processing the acoustic signal was built by the Matlab platform, and the time-frequency domain characteristics of the acoustic signal were analyzed. In the frequency domain, acoustic signals appeared echo interference, electromagnetic interference, and other problems. Then, the cepstrum method was used to analyze the acoustic signals in the frequency domain, in which the signal had the problem of frequency domain aliasing to be a distinguished difficulty. Finally, the logarithmic spectral model was proposed to improve the cepstrum method to deal with the acoustic signal. Because different leakage points had different characteristics such as decibels and frequency values, the acoustic signals with spectral aliasing could be distinguished effectively with the help of different characteristic values, and the problems of electromagnetic interference and echo interference in frequency domain analysis could be avoided. The analysis of three groups of data shows that the pore diameter of the pipe leakage has a large impact on the decibels and frequency values of the acoustic detection signal. With the increase of the leakage pore, the greater the decibel of the pipeline detection sound signal, and the greater the frequency. In addition, the pipeline in the state of non-leakage, due to electromagnetic interference within the detection device will also affect the pipeline acoustic detection signal. Furthermore, different decibels values and frequencies are used to distinguish different leakage pore diameter of pipelines decibel values with the decibel values of NdB0= 0.17 dB, NdB1= 0.052 dB, NdB2= 0.24 dB, respectively. And, the frequency values are f0 =603 Hz, f1 = 1879 Hz, f2 = 4049 Hz, respectively. In short, this method has important theoretical significance and application value in the field of pipeline micro leakage.