Retinal vessel segmentation method based on structure self-adaption context sensitivity

The invention requests to protect a retinal vessel segmentation method based on structure self-adaption and context sensitivity. The method comprises the following steps: firstly, carrying out preprocessing such as graying, normalization, contrast-limited adaptive histogram equalization, gamma correction and the like on a data set; then, carrying out picture size segmentation on the training set to complete data enhancement; then, constructing a full convolutional neural network architecture composed of a contraction path and an expansion path; and finally, a proposed structure adaptive layer is adopted to replace a convolutional layer to replace common convolution, a proposed hole residual path is introduced into a jumper connection layer, an adaptive feature fusion module is added in a decoder, and a multi-scale depth supervision mechanism is introduced to obtain a structure adaptive context sensitive network. And the test data are input into the network for rapid test to complete image segmentation. Accord