Based on Convolution Neural Network,this paper proposes a method for automatic recognition and classification of images of a large number of traditional villages.This method uses a self-built traditional village database with balanced sample distribution to extract features including texture,landform,building layout,building material,etc.by which the images of traditional villages are divided into several categories based on provinces.It aims to provide a large sample of material database of villages to planners,architects,urban and rural builders,as well as supporting the establishment and subsequent research of Chinese digital information system of traditional villages.