基于红外图谱识别坦克漆特征  

Tank paint characteristics identification based on infrared spectrum

作　　者：刘海燕[1] LIU Haiyan(School of Electromechanical and Information Technology,Yiwu Industrial&Commercial College,Yiwu 322000,China)

机构地区：[1]义乌工商职业技术学院机电信息学院,浙江义乌322000

出　　处：《兵器材料科学与工程》2020年第6期133-136,共4页Ordnance Material Science and Engineering

基　　金：浙江省教育厅课题(Y201636873)。

摘　　要：为识别坦克漆特征,用红外光谱仪采集坦克漆的高光谱图像,提取光谱反射率数据;用平均平滑法预处理矫正光谱数据,去除高频噪声;用主成分分析法提取校正后光谱特征,结合人工神经网络完成坦克漆特征识别。结果表明:所提方法的主成分明度、纯度具有较高累计可信度,对不同坦克漆样本均具有良好的聚类性,将坦克漆敏感特征波段作为人工神经网络输入,输出特征识别结果,识别准确率达95%以上,即使存在外部干扰也可保持较高识别准确性。In order to recognize the characteristics of tank paint,the hyperspectral image of tank paint was acquired by infrared spectrometer,and spectral reflectance data was extracted.The spectral data was preprocessed by mean smoothing method to remove high-frequency noise.Principal component analysis(PCA) was used to extract the corrected spectral features and artificial neural network was used to identify the characteristics of tank paint.Results show that the proposed method has the high lightness and purity cumulative credibility with good clustering effect for different tank paint sample.The sensitive characteristic wavelengths are used as artificial neural network input and the characteristic recognition as output,the accuracy reaches 95% or more.Even if there are external disturbance,the results can keep the higher identification accuracy.

关 键 词：红外图谱 主成分分析法 神经网络 光谱数据校正 坦克漆 特征识别 

分 类 号：O433[机械工程—光学工程]