摘要
为实现“问题地图”识别,采用预训练的网络模型,从海量网络图片中快速高效地识别出地图图片,通过对Google公司提出的MobileNet网络识别模型改造、迁移学习,实现一个小巧的卷积神经网络模型,快速识别网络地图图片的精简图像识别模型。结果表明:模型可以加快工作效率,是地图审核工作中较方便、快捷的方法之一。
To realize"problem map"recognition,using the training of network model,from massive network images quickly identify the map images,through the Google company proposed MobileNet network recognition model transformation,transfer learning,realize a small convolution neural network model,quickly identify network map images thin image recognition model.The results indicate that models can greatly accelerate work efficiency,it is one of the more convenient and fast methods in map review work.
作者
张秀锦
张秀民
ZHANG Xiujin;ZHANG Xiumin(Shandong province Geographic Information Center,Shandong Jinan 250013 China)
