改进全卷积神经网络的遥感图像小目标检测  

Detection of Small Targets in Remote Sensing Images UsingImproved Full Convolutional Neural Network

作　　者：徐雪峰 郭广伟 黄余 XU Xue-feng;GUO Guang-wei;HUANG Yu(Jiangsu Electronic Information Vocational College,Jiangsu Huaian 223003,China;Hebei Engineering University,Hebei Handan 056038,China;Chongqing University,Chongqing 400044,China)

机构地区：[1]江苏电子信息职业学院,江苏淮安223003 [2]河北工程大学,河北邯郸056038 [3]重庆大学,重庆400044

出　　处：《机械设计与制造》2024年第10期38-42,共5页Machinery Design & Manufacture

基　　金：江苏省重点技术创新项目导向计划(苏工信创新〔2021〕112-1654)。

摘　　要：对遥感图像中小目标的检测进行研究,提出改进全卷积神经网络的检测新算法。首先,分析了分层概率图模型和深度学习的基本概念和模型。然后,提出分层概率图模型中分层马尔可夫随机场的后验边际模式的递归获取步骤。最后,将全卷积神经网络和分层概率图模型联合,实现对全卷积神经网络的改进,构建遥感图像小目标检测新方法。此外,在所提方法中,选用随机森林技术从分类学习样本中估计每个类和分辨率的后验概率。基于对某地区卫星数据集的处理,将所提出的检测方法与其他四种方法进行了对比。对比实验结果表明,与其他方法相比,所提出的检测方法对低矮植被、车辆、树等遥感图像中的小目标具有更高的检测准确率。A novel method based on improved full convolutional neural network is proposed for the detection of small targets in remote sensing images.Firstly,the basic concepts and models of hierarchical probability graph model and deep learning are analyzed.Then,the recursive steps of obtaining the posterior marginal mode of the layered Markov random field in the layered probability graph model are proposed.In addition,the full convolutional neural network and hierarchical probability graph model are combined to realize the improvement of the full convolutional neural network.Finally,a novel method for small target detection in remote sensing images is constructed based on this improved full convolutional neural network.In addition,random forest technique is used to estimate the posterior probability of each class and resolution from the classification learning samples.The proposed method is compared with the other four methods based on the processing of satellite data set in a certain area.The simulation results show that compared with other methods,the proposed method has higher detection accuracy for small targets in remote sensing images such as low vegetation,vehicles and trees.

关 键 词：小目标检测 遥感图像 全卷积神经网络 分层概率图模型 随机森林 

分 类 号：TH16[机械工程—机械制造及自动化] TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]