The invention relates to a fluorescent stained image CTC intelligent recognition method. The method is based on small sample data set training and comprises the following steps of training a small sample data set; reading an acquired original fluorescent image; evaluating image quality; preprocessing the image; performing model-based recessive feature matching; performing feature-based dominant feature filtering; marking and counting suspected CTC targets; and generating suspected CTC cell images for interpretation. The method, in view of the size of the data set, designs a reasonable networkmodel matching the data set for training, combines a recessive rule based on an artificial intelligence method with a dominant rule based on a feature method, and is supplemented by a plurality of preprocessing methods. The purpose of the method is to obtain a sufficiently accurate CTC recognition algorithm model for intelligent CTC recognition in the context of insufficient and complex data sets,to solve the missing detection of CTC recognition results, and to effectively reduce the false detection rate of the CTC recognition results.