摘要
运用THz光谱特性可以对各种化学品以及爆炸物进行识别.由于对原始数据直接进行识别的识别率不理想,本文首先对实验样本的THz频域光谱数据曲线进行归一化和5点平滑,得到了更能表现数据变化趋势和峰值的特征曲线,然后基于该特征曲线利用曲线极值算法,设计并用VC++6.0实现了对THz光谱的分类识别.通过对RDX、DNT、TNT、HMX 4种爆炸物进行对比识别实验,运用样本进行训练,识别率为96%,通过更换2组样本数据进行验证实验,识别率均为98%.实验证明,本文所设计的曲线极值分析法对THz光谱特征曲线具有较高的识别率.
THz spectral characteristics can use a variety of chemicals and explosives identification. As the to identify the recognition rate is not satisfactory, this experimental samples of the THz frequency original data directly spectrum data to the curve of normalization and a smooth of five pointS first, get more performance data have been changes in trends and more characteristics of the peak of the curve, Then, based on the characteristics of the use of extreme algorithm, design them and use VC ＋ ＋ 6.0 achieve the THz spectral classification identification. Through the RDX, DNT, TNT, HMX explosives were compared to identify the four experiments, the use of 10 samples of each type of training, recognition rate was 96%, by replacing two groups the sample data for verification experiment, the identification rates of 98% Experiments show that the design of this extreme THz spectral analysis of this special curve has a high recognition rate
出处
《首都师范大学学报（自然科学版）》
2009年第2期6-9,共4页
Journal of Capital Normal University：Natural Science Edition
关键词
模式识别
太赫兹
特征提取
曲线识别
曲线处理
pattern recognition, THz feature extraction, curve recognition, curve processing
