A self-adaptive local maximum algorithm for detecting spectral peaks of semiconductor lasers is proposed by analyzing actually measured spectrum characteristics of Fabry-Perot laser diode (FP-LD) and distributed feedback laser diode (DFB-LD), and the algorithm can be used to detect the spectra of laser diodes. In the algorithm, a local maximum matrix of data series is established and is modified with its row vector characteristics, and the peak is located with its column vector characteristics that are further amended and compensated. The algorithm is free from manual operation during the whole searching process, has higher noise-resistance, self-adaptability and robustness, and meets the requirements of real-time computation. Compared with the comparison algorithm, optimized derivative algorithm and genetic algorithm, the local maximum algorithm has obvious advantages in terms of accuracy rate and computation time. The average test accuracy rate reaches 98% while the average computation time is only 0.12 s. The algorithm can be applied to real-time analysis of actually measured spectrum characteristics of semiconductor lasers.