草地背景下伪装材料的高光谱分类方法  

A Hyperspectral Classification Method for Grassland Background Camouflage Materials

作　　者：赵佳乐 周冰 王广龙 应家驹 陈琪 赵润泽 ZHAO Jiale;ZHOU Bing;WANG Guanglong;YING Jiaju;CHEN Qi;ZHAO Runze(Shijiazhuang Campus,Army Engineering University of PLA,Shijiazhuang 050003,China;Army Equipment Department,Shijiazhuang 050000,China)

机构地区：[1]陆军工程大学石家庄校区,河北石家庄050003 [2]陆军装备部,河北石家庄050000

出　　处：《陆军工程大学学报》2024年第3期43-50,共8页Journal of Army Engineering University of PLA

摘　　要：高光谱成像技术能够综合利用光谱信息与空间信息,为目标精确识别提供了一种可行方案,广泛应用于军事目标特性分析与伪装目标分类等方面。针对伪装物与真实目标光谱相似度高、分类与探测精度低的问题,提出了一种基于高光谱成像与极限学习机(extreme learning machine,ELM)的伪装目标分类方法。该方法通过研究伪装物与真实目标之间的光谱特性,寻找识别伪装的“窗口波段”。利用光谱降维方法找出最适合伪装目标分类的特征,进而将提取的特征作为ELM多分类模型的输入,完成分类任务。以草地背景中伪装材料的分类为例,利用野外拍摄的高光谱图像进行实验。结果表明,所提方法的分类精度达到了97.27%,其分类效果优于其他对比分类算法,为伪装目标分类提供了一种有效的方法。Hyperspectral imaging technology enables the comprehensive utilization of spectral and spatial information,providing a viable solution for precise target identification.It is widely used in areas such as military target characteristic analysis and classification of camouflaged targets.In order to solve the problems of high spectral similarity and low classification and detection accuracy between camouflaged objects and real targets,a camouflage target classification method based on hyperspectral imaging and extreme learning machine(ELM)is proposed.This method first searches for the"window band"for identifying camouflage by studying the spectral characteristics between the camouflage object and the real target.Then,the spectral dimensionality reduction method is used to find the most suitable features for camouflage target classification,and the extracted features are used as input to the ELM multi-classification model to complete the classification task.Taking the classification of camouflage materials in grassland backgrounds as an example,experiments were conducted using hyperspectral images captured in the field.The experimental results show that the classification accuracy of the proposed method reaches 97.27%,which is superior to the classification performance of other comparative classification algorithms.The proposed method provides a valuable reference for future camouflage target classification.

关 键 词：高光谱成像 极限学习机 伪装目标 图像分类 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]