基于高光谱影像的干旱区草地光谱特征分析

基于石羊河流域金昌地区Hyperion高光谱影像,通过FLAASH大气校正,利用纯像元指数法提取草地波谱信息,并运用光谱一阶微分法和连续统去除法进行定量化处理。结果表明,相对于生长旺盛期的草地光谱特征,衰退期草地光谱红边左移,斜率降低,蓝边、黄边特征减弱,可见光波段反射值较高,近红外波段反射率较低;不同覆盖度草地的红边、绿峰、蓝光和红光吸收谷位置保持一致,可见光波段的光谱吸收特征(波段深度、宽度、面积、对称性)随覆盖度的增大呈有规律的变化,可作为提取或判定植被覆盖度的依据。     

基于神经网络的图像识别方法研究

图像识别涉及大量的信息运算,要求处理速度快、识别精度高,神经网络的实时性和容错性要符合图像识别的要求。利用改进的BP神经网络算法对旋转畸变图像进行了定位和识别,改进算法将附加动量项与自适应学习速率相结合,有效地抑制了网络陷入局部极小点,提高了网络的训练速度。结果表明,基于神经网络的图像识别方法是有效的、可行的。     

Noncontact blood species identification method based on spatially resolved near-infrared transmission spectroscopy

The inspection and identification of whole blood are crucially significant for import-export ports and inspection and quarantine departments. In our previous research, we proved Near-Infrared diffuse transmitted spectroscopy method was potential for noninvasively identifying three blood species, including macaque, human and mouse, with samples measured in the cuvettes. However, in open sampling cases, inspectors may be endangered by virulence factors in blood samples. In this paper, we explored the noncontact measurement for classification, with blood samples measured in the vacuum blood vessels. Spatially resolved near-infrared spectroscopy was used to improve the prediction accuracy. Results showed that the prediction accuracy of the model built with nine detection points was more than 90% in identification between all five species, including chicken, goat, macaque, pig and rat, far better than the performance of the model built with single-point spectra. The results fully supported the idea that spatially resolved near-infrared spectroscopy method can improve the prediction ability, and demonstrated the feasibility of this method for noncontact blood species identification in practical applications.     

基于LGSPP-Bayes的故障检测与辨识方法

针对主元分析(Principal component analysis, PCA)和局部保持投影(Locality preserving projections, LPP)方法在降维过程中分别只能保留数据集的整体信息和局部信息,提出一种基于局部整体结构保持投影的贝叶斯故障检测与辨识方法(Local and global structure preserving projections and bayes, LGSPP-Bayes)。首先,将正常工况操作下的原始数据通过局部整体结构保持投影方法投影到低维特征空间,得到高维到低维的数据转换矩阵;然后通过设计贝叶斯分类器来进行故障检测;最后当检测到故障后通过计算贝叶斯分类函数的大小来识别故障种类。将LGSPP-Bayes方法应用于TE过程,仿真结果表明对故障的检测优于其他方法,并且可以很好地将故障种类识别出来。     

Distributed Brillouin fiber sensing based on spectrum line fitting and wavelet packet denoising

A novel signal processing method is proposed to improve the spatial resolution, frequency resolution and dynamic characteristics of BOTDR. For the BOTDR system with 50 ns pump pulse, by using spectrum line fitting technology based on Levenberg-Marquardt (LM) algorithm, the spatial resolution is improved from 5 m to 5 cm. Combination of LM fitting algorithm, a large frequency scanning interval is adopted without sacrificing measurement accuracy of the BOTDR system. It reduces the number of sampling points of Brillouin spectrum significantly. So, the fitting speed is improved greatly. This is the first time using a large scan interval to increase the spectrum line fitting speed. To improve the fitting speed, the difference between the reference and measured spectrum is used to estimate the variation of Brillouin frequency shift. The measured amplitude of Brillouin spectrum is used to estimate the width of region strain occurred. Finally, by using wavelet packet denoising technology, the spectra containing noise are fitted successfully.     

非下采样小波变换红外光谱数据去噪

为了降低噪声对实测红外光谱信号的影响,引入了一种非下采样小波变换的红外光谱数据去噪方法。采用非下采样小波变换对原始光谱信号进行多尺度分解,提取信号的多尺度细节特征;根据光谱信号和噪声在不同尺度上的差异,通过应用变分偏微分方程方法调整分解后的各子带系数;重构各子带就可以将原始光谱信号中真实信号和噪声分离,从而达到剔除噪声的目的。通过两组实验对比传统小波和该方法针对红外光谱数据的消噪效果,实验结果表明:非下采样小波变换在红外光谱数据去噪方面具有明显的优势,不仅能够有效地去除噪声,很好地保持信号的形状,并且均方误差较小;在实际的红外光谱数据处理中能够获得较好的去噪效果。     

Nondestructive egg freshness assessment from the equatorial and blunt region based on visible near infrared spectroscopy

This research was to study which orientation was better for freshness prediction of the white-shelled eggs using visible near infrared spectroscopy. The transmission spectra were acquired in the equatorial region and at the blunt end of the eggs. After each spectral measurement, the Haugh unit, yolk index, and albumen pH as the freshness parameters were simultaneously measured using traditional destructive methods. Pretreatment methods containing Savitzky-Golay smoothing, multiplicative scatter correction, the standard normal variate, the first derivative and the second derivative were used. A partial least squares regression was developed to predict the Haugh unit, yolk index, and albumen pH. The best correlation coefficients of prediction were obtained from the equatorial region, and were 0.881, 0.855, and 0.888 for the Haugh unit, yolk index and albumen pH, respectively. And root mean square errors in the prediction set were 7.720, 0.034, and 0.147 for the Haugh unit, yolk index and albumen pH, respectively. The results illustrated that the equatorial region showed better ability than the blunt end to predict freshness of the white-shelled eggs.     

互相关及高阶谱核材料富集度识别方法

针对基于传统核材料识别系统中特征值易受系统噪声影响且精确度不够的问题,将高阶统计特征概念引入252Cf源驱动核系统特征提取和识别之中,提出了一种基于互相关函数与高阶统计特征的252Cf源驱动核材料富集度识别方法。通过互相关运算及高阶统计分析,得到核系统信息的三维特征图像,并在此基础上开展了待测核材料(235U)的富集度实验研究与分析识别工作。实验研究结果表明,该识别方法能够较好地降低背景辐射噪声与电子学系统噪声。与传统识别方法相比,对于核部件富集度变化,该算法的敏感性与鲁棒性均有大幅提高。     

非下采样小波变换红外光谱数据去噪

为了降低噪声对实测红外光谱信号的影响,引入了一种非下采样小波变换的红外光谱数据去噪方法。采用非下采样小波变换对原始光谱信号进行多尺度分解,提取信号的多尺度细节特征;根据光谱信号和噪声在不同尺度上的差异,通过应用变分偏微分方程方法调整分解后的各子带系数;重构各子带就可以将原始光谱信号中真实信号和噪声分离,从而达到剔除噪声的目的。通过两组实验对比传统小波和该方法针对红外光谱数据的消噪效果,实验结果表明:非下采样小波变换在红外光谱数据去噪方面具有明显的优势,不仅能够有效地去除噪声,很好地保持信号的形状,并且均方误差较小;在实际的红外光谱数据处理中能够获得较好的去噪效果。     

HDP-HSMM的磨削声发射砂轮钝化状态识别*

在高精度金属材料磨削加工中,刀具即砂轮的状态对加工效率和加工质量具有重要的影响。钝化程度较高的砂轮不适于加工精密工件,需提前预警并修整更换砂轮。该文提出一种通过磨削声发射信号来检测砂轮钝化状态的方法。首先,对于采集到的信号进行小波软阈值降噪。然后,将其分割成多个有重叠的帧,并提取每帧信号的8个特征组成声发射数据集。最后,通过分层Dirichlet过程-隐半马尔可夫模型来建立声发射数据集和不同的砂轮钝化状态之间的非线性关系,旨在识别砂轮钝化状态。结果表明,上述检测方法能有效识别砂轮的不同钝化状态并能对整个加工过程中的砂轮钝化程度进行自动划分,其在测试数据集上的准确率达到93.7%,可以为实际工业应用提供理论指导。     

基于高光谱数据的呼伦贝尔草原花期物种识别和覆盖度估算

实时准确地监测草地植物种类和覆盖面积对草原物种多样性研究和生态环境的可持续发展具有重要意义。草地植物在花期具有独特的光谱特征,相比营养生长期,通过花更容易识别物种。花期是遥感识别物种的关键时期。本文利用2008年8月和2010年7月呼伦贝尔草原上麻花头、棉团铁线莲、冷蒿、蓬子菜、黄花菜、有斑百合和细叶百合七种花以及裸土的实测高光谱数据,通过其冠层光谱的特征分析和参量化,找出了各物种之间的光谱差异,且得到了参量化特征识别方法,经验证,当花在样方中的覆盖度大于10%时,识别方法的精度在90%以上。在此基础上,采用线性解混模型,计算样方中各种花的覆盖度,与实际数据相比较,误差在4%左右,证明了线性解混模型用于估算草原上花覆盖度的可行性。     

多特征波长窗法检测痕量甲醛气体的研究

为了克服化学检测痕量甲醛气体速度慢、有耗材、采样区域局限等缺点,实现快速准确地检测甲醛气体的浓度,设计了一种多特征波长窗的方法。根据甲醛及主要干扰气体的特征光谱选择相干度最小的多组特征波长,特征波长个数分别选用3,4和5个,同时配合相应的多组窄带滤光片。当光源依次通过这些窗口及气室后,由PCI-2TE-13型红外探测器采集,并经相关算法反演样气中的甲醛浓度。实验针对新装修的房屋、建材商场、超市及公园四种环境下采集的样气中的甲醛进行定量分析。实验结果与ARCSpectro-A-MIR型红外光谱仪的检测结果相比较,结果表明,采用多特征波长窗法在10μg.m-3以上的检测结果均与标准值相近,其平均误差均小于5%,满足实际应用的要求,并且具有实时检测、不中毒等优点。     

Blood species identification using Near-Infrared diffuse transmitted spectra and PLS-DA method

Blood species identification is of great significance for blood supervision and wildlife investigations. The current methods used to identify the blood species are destructive. Near-Infrared spectroscopy method is known for its non-invasive properties. In this research, we combined Near-Infrared diffuse transmitted spectra and Partial Least Square Discrimination Analysis (PLS-DA) to identify three blood species, including macaque, human and mouse. Blind test and external test were used to assess the PLS-DA model. The model performed 100% accuracy in its identification between three blood species. This approach does not require a specific knowledge of biochemical features for each individual class but relies on a spectroscopic statistical differentiation of the whole components. This is the first time to demonstrate Near-Infrared diffuse transmitted spectra have the ability to identify the species of origin of blood samples. The results also support a good potential of absorption and scattering spectroscopy for species identification in practical applications for real-time detection.     

基于最差性能优化的运动声源稳健聚焦定位识别方法研究

基于被动合成孔径原理,提出了一种具有高稳健性的运动声源高分辨聚焦定位识别方法.该方法采用综合优化手段,通过矢量最大似然聚焦定位算法生成虚拟阵列坐标及数据矩阵,进而利用基于最差性能优化的稀疏虚拟阵列聚焦算法,获取稳健的高分辨定位识别效果.理论及仿真研究表明,该方法对于非匀速运动以及与基阵存在运动倾角的复杂情况具有较强的适用性,聚焦空间谱表现出更大的动态范围、更为尖锐的聚焦峰尺度以及更强的背景噪声起伏压制能力.湖上试验进一步验证,在高分辨最小方差信号无畸变响应法(MVDR)聚焦算法动态范围仅为3.5 dB的相 关键词： 稳健性 运动噪声源定位识别 矢量阵 最差性能优化     

小样本字典学习的兰姆波模态识别方法*

针对兰姆波多模态识别问题,提出了基于小样本字典学习的模态识别方法。将多层复合板的频散特性看作一个线性时不变系统,首先,根据频散知识模拟各个模态传播特定距离后的信号,提取走时和能量特征创建字典;其次,获取待测信号的走时特征,通过查询字典来识别兰姆波模态;最后,根据能量参数估计结果,实现待测信号中各模态信号的分离和重构。通过对三层粘接的AAA板(铝板-亚克力板-铝板,每层厚度为2mm)中传播距离为0.3m、0.5m的直达波和反射波的实验验证,结果显示该方法对A0、S0模态的有效识别和各个波包信号的准确重建。     

基于独立成分分析和经验模态分解的混沌信号降噪

基于经验模态分解和独立成分分析去噪的特点,提出了一种联合独立成分分析和经验模态分解的混沌信号降噪方法. 利用经验模态分解对混沌信号进行分解,根据平移不变经验模态分解的思想构造多维输入向量, 通过所构造的多维输入向量和独立成分分析对混沌信号的各层内蕴模态函数进行自适应去噪处理; 将处理后的所有内蕴模态函数进行累加重构,从而得到降噪后的混沌信号. 仿真实验中分别对叠加不同强度高斯噪声的Lorenz混沌信号及实际观测的月太阳黑子混沌序列进行了研究, 结果表明本文方法能够对混沌信号进行有效的降噪,而且能够较好地校正相空间中点的位置, 逼近真实的混沌吸引子轨迹. 关键词： 独立成分分析 经验模态分解 混沌信号 降噪     

Research on cotton fiber qualities evaluation based on optoelectronic techniques

The fine and mature cotton fibers can make it possible to spin a finer yarn, and fineness and maturity of cotton fibers were essential qualitative characteristic, however, there was no direct or indirect measurement method that was both fast and reliable to estimate them. In this paper, a novel method was put forward to assess cotton fiber fineness and maturity based on optoelectronic measuring techniques. The experimental results were compared between the microscope measurement and system measurement and the linear fit results of two measuring methods of that linear fit were 0.98788, and standard deviation was 0.19081. Average cross-sectional widths of cotton samples measured by the microscope and the system had a fairly low correlation with the micronaire data. The linear fit results of the fiber maturity ratio of two measuring methods were 0.95587, and standard deviation was 0.33203. This method can be done in the lab and be used in the industry adherent to the present measurement instruments, for example AFIS, after improved.     

声发射衰减特性管道泄漏定位方法

声发射技术具有灵敏度高、实时性强、覆盖范围大等优点.泄漏产生的声发射波沿管壁传播会发生衰减,通过研究衰减系数与金属晶粒散射和热流损失的关系,建立准确的声发射能量衰减模型.在此基础上,针对声发射频带宽的特点对传统衰减定位模型进行改进,提出宽频带声发射源定位模型,该方法先通过实验确定泄漏信号频带,再将滤波后的信号经过小波包...     

Speech dereverberation method based on spectral subtraction and spectral line enhancement

Speech signals recorded with a distant microphone usually are interfered by the spatial reverberation in the room, which severely degrades the clarity and intelligibility of speech. A speech dereverberation method based on spectral subtraction and spectral line enhancement is proposed in this paper. Following the generalized statistical reverberation model, the power spectrum of late reverberation is estimated and removed from the reverberation speech by the spectral subtraction method. Then, according to the human auditory model, a spectral line enhancement technique based on adaptive post-filtering is adopted to further eliminate the reverberant components between adjacent speech formants. The proposed method can effectively suppress the spatial reverberation and improve the auditory perception of speech. The subjective and objective evaluation results reveal that the perceptual quality of speech is greatly improved by the proposed method.     

小波计盒分形维数核爆电磁脉冲信号模糊识别

为了在现有禁核试验条件下开展核爆电磁脉冲探测技术研究,对实测的非均匀采样核爆电磁脉冲数据进行阈值检波、直流剔除、滤波、插值和归一化处理后,采用小波分析提取信号第四尺度小波系数能谱熵、第一尺度和第四尺度小波系数计盒分形维数特征值,在统计分析的基础上分别建立其专用隶属度函数;利用模糊识别理论,对核爆电磁脉冲信号和作为主要探测干扰信号源的闪电电磁脉冲信号进行了模糊识别,获得了总识别率为99.41%的识别效果。