基于YOLOv3深度卷积神经网络的遥感图像飞机目标识别  

Aircraft Target Recognition Based on YOLOv3 Deep Convolution Neural Network in Remote Sensing Image

作　　者：柯青青 李润生[1] 胡庆 牛朝阳[1] 刘伟[1] KE Qingqing;LI Runsheng;HU Qin;NIU Chaoyang;LIU Wei(Information Engineering University, Zhengzhou 450001, China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2020年第5期526-533,共8页Journal of Information Engineering University

基　　金：国家自然科学基金资助项目(41901378)。

摘　　要：如何利用计算机自动、快速、准确地在遥感图像中检测目标是当前遥感信息智能化处理的热点和难点。将从深度学习出发,运用计算机视觉相关理论,研究目标检测算法应用于遥感目标判读,制作了含机场在内的十一类飞机目标数据集,并用yolo mark对这些遥感图片逐一进行标注,设计了基于YOLOv3深度卷积神经网络的遥感图像飞机目标检测实验系统,并利用自建数据集验证其正确性和时效性。结果表明,该系统能够较好地对高分辨力遥感图像中的飞机目标进行识别。How to detect the target in remote sensing image automatically,quickly and accurately is the hot and difficult point in the current intelligent processing of remote sensing information.Starting from deep learning,this paper uses the computer vision theory to study the application of target detection algorithm in remote sensing interpretation.Ten kinds of aircraft targets including the airport dataset are made,and the remote sensing images are annotated one by one by YOLO mark.A remote sensing image plane target recognition system based on deep convolutional neural network YOLOv3 is designed,whose correctness and timeliness are verified through the self-built datasets.The results show that the proposed method can identify aircraft targets in high resolution remote sensing images.

关 键 词：深度学习 计算机视觉 目标识别 遥感图像 YOLOv3 

分 类 号：TN918.1[电子电信—通信与信息系统]