The nanoparticle size and distribution information in the SEM images of silicon crystals are generally counted by manual methods. The realization of automatic machine recognition is significant in materials science. This paper proposed a superposition partitioning image recognition method to realize automatic recognition and information statistics of silicon crystal nanoparticle SEM images. Especially for the complex and highly aggregated characteristics of silicon crystal particle size, an accurate recognition step and contour statistics method based on morphological processing are given. This method has technical reference value for the recognition of Monocrystalline silicon surface nanoparticle images under different SEM shooting conditions. Besides, it outperforms other methods in terms of recognition accuracy and algorithm efficiency.