【作者】 崔涌静； 陈亮； 陈禾； 庄胤；

【Author】 CUI Yongjing;CHEN Liang;CHEN He;ZHUANG Yin;Radar Technology Research Institute, School of Information and Electronic, Beijing Institute of Technology;

【机构】 北京理工大学信息与电子学院雷达技术研究所；

【摘要】 随着城市化建设不断推进,智慧城市理念不断革新,对城市进行合理精确地规划,提高城市居民的幸福指数以及城市的宜居程度成为了城市管理者的一个新的目标和考察因素。提出基于U-net模型结构的U-netRS模型,参考了U-net3+网络模型结构,将机器学习与遥感图像处理技术相结合,针对遥感图像Vaihingen2D语义数据集,有较好的识别效果,甚至对于像素点较少的目标也能达到较高的像素准确度。此外,模型采用的图像分割技术,对遥感图像进行小块切割,通过减少一次模型训练的数据量,增大总体模型训练的数据量的方式,提高模型预测的准确率,降低所需硬件要求。更多还原

【Abstract】 With the continuous advancement of urbanization and the continuous innovation of the concept of smart city, it has become a new objective and investigation factor for city managers to make reasonable and accurate urban planning, improve the happiness index of urban residents and the livability degree of the city. A U-net RS model based on the U-net model structure is proposed, which refers to the U-net 3+ network model structure, combines machine learning and remote sensing image processing technology, and has a good recognition effect for the Vaihingen 2 D semantic data set of remote sensing images. Even for the target with fewer pixels, it can achieve higher pixel accuracy. In addition, the model adopts image segmentation technology to cut remote sensing images into small pieces. By reducing the amount of data for one model training and increasing the amount of data for the overall model training, the accuracy of model prediction can be improved and the required hardware requirements can be reduced.更多还原