Vascular Recognition Method Of Retinal Fundus Images

The fundus image of the retina is the only blood vessel in the body that can directly observe deep blood vessels in a non-invasive manner.Its morphological and structural changes can directly reflect cataracts,diabetes,hypertension,and arteriosclerosis,and can also be used as an important basis for judging cataracts.The computer's automatic fundus image processing and analysis have important value in the auxiliary medical diagnosis,and the extracted blood vessels can be used as a feature for cataract fundus image classification.The existing retinal blood vessel segmentation algorithm is very easy to be limited by the model parameters,the tuning is complex and there is a certain gap with the segmentation result of the ophthalmologist,and there is room for improvement.The main tasks of this paper are as follows:(1)Preprocessing the fundus image data set.The fundus image is converted into a grayscale image by channel selection,and image enhancement is used to highlight the structural features of the blood vessel context to complete the preprocessing work.(2)Using support vector machine combined with fuzzy C-means clustering to extract fundus images of the blood vessels,and then through morphological processing and thresholding and other image processing techniques to further processing of fundus image vessels,has reached the final extracted blood vessels purpose.(3)Use full convolutional neural network for fundus image vessels.The depth network is used to predict the pixels in the image,that is,the pixels are divided into two types of blood vessel backgrounds,so as to achieve the extraction of fundus blood vessels.The experimental verification has achieved good results,and this algorithm has a higher recognition accuracy for cataract blur images than the support vector machine.Through experimental verification,both of the methods in this paper have achieved good results,especially the identification algorithm based on full convolutional neural network.The accuracy of the recognition of normal fundus images has reached 94.91%,which is superior to the existing methods.In addition,the vascular recognition of cataract blurred images has also achieved a good recognition effect.