Research Of Raindrop Spectrum Image Segmentation Algorithm

Traditional raindrop spectrum data detection is mainly through the way of artificial measurement, this method procedure is multifarious, vulnerable to the influence of artificial factors and acquisition equipment, bringing to the raindrop spectrum data of the test result error, and having subjectivities.In this paper, carry out a large of research using machine vision detection method, to the actual production problems, basing the full results of previous studies on the classification problem for the raindrop size. Firstly, the image processing technology is used for raindrop spectrum pretreatment of samples, using the histogram method to average enhancement and sharpening with a gradient method, as a result eliminate noise and improves the image contrast. Then segment the rain image using two-dimensional threshold based on Otsu segmentation algorithm, realize the multi-object identification in image with target marker method. Because raindrops sample image exists in the rain particles adhesion or overlapping problems, put forward a algorithm based on the adhesion of particles boundary contour feature, using the method of approximate polygon and the properties of vector dot product and cross product obtained the dividing point, achieve the goal of the adhesion rain particles segmentation. In the final analysis testing data, obtain raindrop diameter and diameter, filter paper, filter paper splash and the corresponding relationship between pixels and the diameter of raindrop spectrum image of rainfall rate is determined.Experiments show that this paper builds the raindrop spectrum detection method based on machine vision method instead of manual operation, can improve the testing efficiency, have higher precision.