摘要
目标识别是遥感高分辨率影像时代的重要应用方向。采用深度卷积神经网络对遥感影像学习训练,能够从遥感影像中自动提取出多个具有代表性的典型地物特征以及特征组合,并应用于多变而复杂的遥感影像数据中进行目标分类识别。本研究选用NWPU VHR-10数据应用于Faster R-CNN卷积神经网络模型中,并采用MAP进行评价,研究中得到了较好的检测精度,证明在遥感影像数据中采用深度卷积神经网络进行目标识别有着广阔的应用前景。
Target recognition is an important application of high resolution remote sensing images.Convolutional neural network in depth learning can extract representative and discriminant multi-level features from images,which can be used for multi-target recognition of large remote sensing data in complex scenes.In this study,NWPU VHR-10 data is used in Faster R-CNN convolution neural net work model,and evaluated by MAP.There salts show that the convolution neural net work has broad application prospects in target recognition of remote sensing image data.
作者
雷忠腾
宋杰
LEI Zhongteng;SONG Jie(Qingdao Zhongyou Geo-technical Engineering Co.,Ltd.,Qingdao 266071,China)
出处
《测绘与空间地理信息》
2021年第1期149-151,155,共4页
Geomatics & Spatial Information Technology
