In order to detect frequency hopping signals in complex electromagnetic environment, this paper utilizes the approximate l0 norm algorithm for reconstructing the frequency hopping signal with interference according to the sparsity of frequency hopping signal in frequency domain. The unconstrained multidimensional optimization problem is solved by the quasi Newton method, and the paper utilizes the obtained timefrequency map as an image for performing two value morphological filtering to eliminate the interference and noise. Frequency hopping signal is detected by counting the number of matched signals. In view of improving the adaptability of the algorithm, this paper uses the Otsu method in selecting binarization threshold. The theoretical analysis and the simulation show that the algorithm not only can overcome the interference under low SNR condition but also can accomplish signal detection.