关键词: 影像瓦片, Siamese卷积神经网络, 深度学习, 变化检测, 天地图

Abstract: In view of the fast change of geographic information and the inefficiency of traditional updating methods, many scholars have proposed various kinds of change detection methods, but most of these methods experiments are based on image data, which require higher image preprocessing requirements, and the detection accuracy is always unstability, which is greatly affected by the quality of data sources. While in Tianditu maps, Baidu maps, Google maps and other electronic maps, several levels of image tiles can be downloaded free of charge. Therefore, this paper proposes to use Tianditu map image tiles for experiment, use Siamese convolution neural network (SCNN) and deep learning technology, develop a high-precision change monitoring algorithm based on SCNN, quickly find the change area, and realize the change information detection of geographic information.

Key words: image map tiles, Siamese convolution neural network, deep learning, change detection, Tianditumaps