Abstract: In response to the problem of Frequency-Hopping detection in the HF channel, the paper proposed an algorithm based on the split directed graph which is clustered by K-Means through the image processing method to achieve frequency hopping signal detection. Using the method of correlation interferometer direction completed the broadband signal to generate a directed graph. Then based on the segmentation directed graph by K-Means algorithm, after morphological filtering, preprocessing, and then using the boundary projection algorithm, detect the Frequency-Hopping signals. This algorithm can detect frequency hopping signals, while the course estimated the basic parameters of the frequency hopping signal. Simulation results and performance analysis shows that the algorithm can effectively overcome the complex electromagnetic environment in the signal bandwidth can effectively detect frequency hopping signal, lower computational complexity, easy to project implementation algorithm.