Abstract: A raindrop spectrum measurement method based on a single image is proposed. Kernel regression method was used to extract the global rain streaks direction of the image, and luminosity and color properties were used to enhance the expression of the elliptical kernel of the rain part of the image. Based on the morphological characteristics of the rain streaks, an improved structural element adaptive multiple morphological filtering method was proposed to separate the rain part of the image. The physical parameters of the rain streaks were extracted by using the minimum bounding rectangle, and the defocus rain streaks was further separated according to the imaging difference and considering the influence of wind. The accuracy of image target particle size extraction was verified through simulation experiments, and the observed raindrop spectra were constructed based on actual rainfall. The meteorological Gamma model was used for fitting to verify the effectiveness of the proposed method.