An estimation and sorting method for frequency-hopping signals based on improved connected region labeling is proposed to tackle the problem that small unmanned aerial vehicles （UAVs） are difficult to identify in a complex electromagnetic environment. Firstly， the energy threshold statistical method based on local windows is used to denoise the signal in the time and frequency domains. Then， the conventional connected region labeling method is improved to design a method to associate segments of the connected region and a method to suppress interference， which solves the problems of connected region breaks and interference mixing. After that， in the case of the mixing of multi-frequency-hopping signals， the connected region is reconstructed by the time–frequency amplitude difference. Finally， parameter extraction and signal sorting and identification are carried out according to the improved label map of the connected region. The simulation shows that under the environment of noise， interference， and multi-frequency-hopping signal aliasing， the estimation accuracy of signal parameters and the correct identification probability of target signals are significantly higher than those of the conventional connected region labeling method.