Human eye vitreous opacity degree grading method based on R-Unet

The invention provides a human eye vitreous opacity degree grading method based on R-Unet. The method comprises the following steps: collecting a human eye vitreous image; preprocessing the human eye vitreous body image, and obtaining a training data set according to the preprocessed human eye vitreous body image; constructing an R-Unet convolutional neural network, and performing training verification on the training set and the verification set by using the R-Unet convolutional neural network to obtain an R-Unet convolutional neural network model; segmenting a turbid focus area and a cavity trapezoid area of the human eye vitreous body image by using an R-Unet convolutional neural network model; and calculating the area ratio of the turbid focus region to the cavity trapezoidal region, and grading the human eye vitreous opacity degree according to the area ratio. The human eye vitreous opacity degree can be automatically graded and judged, a doctor can be helped to quickly make clinical diagnosis and interv