基于机器学习的军事目标检测系统开发研究

Development and Research of Military Target Detection System Based on Machine Learning

遥感侦察已成为各国主要的侦察技术手段之一,复杂战场环境中,如何从海量的遥感侦察图像中快速准确的获取目标信息,是获取打击优势的重要环节,为取代效率低下的人工作业,开发用于检测遥感图像中的军事目标软件进行辅助侦察具有重要的军事意义。搜集整理遥感图像和军事目标(以飞机为例)的样本,提取其浅层特征(LBP特征)和中层特征(HOG特征)进行融合,并运用SVM支持向量机进行训练,得到关于军事目标(以飞机为例)的分类器,成功实现对军事目标的检测,并对软件的交互界面进行开发。

Remote sensing reconnaissance has become one of the main reconnaissance techniques of various countries.In a complex battlefield envi⁃ronment,how to quickly and accurately obtain target information from massive remote sensing reconnaissance images is an important link to gain superiority in strike,and to replace inefficient manual work,the development of software for detecting military targets in remote sensing images for auxiliary reconnaissance has important military significance.Collecting and organizing samples of remote sensing imag⁃es and military targets(taking aircraft as an example),extracting their shallow features(LBP features)and intermediate features(HOG fea⁃tures)for fusion,and using SVM support vector machines for training to obtain information about military targets(with aircraft as an exam⁃ple),the classifier successfully detects military targets and develops the interactive interface of the software.