基于卷积神经网络的传统村落图景分类研究

A STUDY ON THE CLASSIFICATION OF TRADITIONAL VILLAGE IMAGES BASED ON CONVOLUTIONAL NEURAL NETWORK

基于卷积神经网络提出了一种适用于大量传统村落现状图自动识别归类的方法.该方法利用自建的一个样本分布均衡的传统村落数据库来提取村落肌理、地形地貌、建筑形态结构、表皮材质等传统村落特征,把全国传统村落图景以省份为界分为若干类,旨在为规划师、建筑师、城乡建设者等提供一个可以参考的大样本村落现状素材库,也为中国传统村落数字管理信息系统的建设和后续研究提供素材和定量支撑.

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 oftraditional villages.