基于深度自注意力网络的影像智能解译  

Image Intelligent Interpretation based on Transformer

作　　者：李俊 吴长枝[2] 齐晓飞[2] 赵耀 Li Jun;Wu Changzhi;Qi Xiaofei;Zhao Yao(Speed China Technology Co.,Ltd.,Nanjing,China;Xi'an Surveying&Mapping Institute,Xi'an,China;PLA Strategic Support Force Information Engineering University,Zhengzhou,China)

机构地区：[1]速度时空信息科技股份有限公司,江苏南京 [2]西安测绘研究所,陕西西安 [3]中国人民解放军战略支援部队信息工程大学,河南郑州

出　　处：《科学技术创新》2023年第1期124-128,共5页Scientific and Technological Innovation

摘　　要：传统深度神经网络主要关注影像的局部特征,忽视了影像上远距离的语义特征。深度自注意力网络能够打破局部感受野的限制,对影像上的长距离关系实现建模,因此,被广泛应用于近景与遥感影像智能解译。首先介绍了基于深度自注意力网络的影像处理基本原理。然后,重点分析深度自注意力网络在场景分类、目标识别和语义分割任务中的应用。最后,总结了深度自注意力网络在影像智能解译领域的未来发展方向。Traditional deep neural networks mainly focus on local image feature,which ignores the longdistance semantic feature.Transformer has the ability to break the restriction of local receptive field and model the long-distance relationship on the image.Thus,it is widely used in close-range and remote sensing image intelligent interpretation.Firstly,illustrates the basic principle of transformer-based image processing.Then,the applications of transformer models used in scene classification,object detection and semantic segmentation are in-depth analyzed and compared.Finally,future development of transformer in image intelligent interpretation is concluded.

关 键 词：遥感影像 深度自注意力网络 智能解译 场景分类 目标识别 语义分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]