Time-frequency analysis is a powerful tool for frequency-hopping (FH) signal detection, however, the performance of time-frequency analysis will degrade drastically in impulse noise environment, failing to extract the hopping duration, frequency and timing effectively. Moreover, methods based on fractional lower order statistics and maximum-likelihood (ML) are generally used to improve the performance of FH signal time-frequency distribution, but the performance improvement of the former is limited, and the latter is usually  sensitive to the noise distribution and has high computational complexity. To detect FH signals in the presence of impulse noise, a detection method of FH signal is proposed based on data credibility weighting. In the proposed method, the concept of data credibility is established based on the cloud model theory to analyze the uncertainty of the received signal. On this basis, the weighting process is implemented to the received signal and improves the performance of time-frequency distribution of FH signal in the impulse noise environment. Simulation results show that compared with the fractional lower order statistics as well as the Myriad filter based time-frequency analysis methods, the proposed method can detect the FH parameters with the noise being suppressed effectively, and it is robust in the stable noise environment.