一种新的光谱特征提取方法

研究了天体光谱的特征提取问题,这是光谱自动处理中的一个关键环节.通过特征提取,不仅能够约简数据、减少冗余,而且亦能抑制噪声干扰,对识别系统的精度和效率均有重要影响.提出了一种基于空间转换和分解的特征分析模型(STP),基于此,可实现对常用光谱特征提取方法的分析,例如,无监督的主成分分析(PCA),小波变换(Wavele...     

基于小波分解的光谱特征提取算法研究

物质的光谱曲线反映了其独特的反射特性,利用光谱可以进行物质的分类识别工作。由于光谱曲线数据量较大、吸收特征不明显等特点,光谱曲线的特征提取是高光谱影像分类识别中的一个关键问题之一。该文利用小波分析技术通过对原始信号的分解,以及测量目标光谱特征的吸收宽度,确定了小波分解尺度,达到突出目标光谱吸收特征而抑制非相关特征及噪声的目的。通过实验表明,该方法可有效地降光谱维数,提高了光谱匹配识别精度。     

小波包特征提取及方差相似度的人脸识别

小波包变换是小波变换的推广,可视为普通小波函数的线性组合,具有良好的时频局部性和正交性,随着分解层数的增加,小波包分解能够在所有的频率范围聚焦。利用图像小波包变换的系数矩阵,能够构造出不同的人脸特征向量。针对人脸识别过程中的图像匹配问题,采用计算人脸特征向量方差的方法,并通过方差与权值的对应关系,转换出用于相似度计算的权值。基于理论推导得到的权值具有很好的稳定性,由这些权值计算出的方差相似度也具有较强的适应性,能够减弱由图像噪声、变形等干扰带来的影响。实验表明,该方法识别率高、实时性好。     

Rapid ultraspectral matching method based on crosscut feature extraction

A rapid matching method for ultraspectral data is proposed in this paper. We have expanded the concept of the histogram in image processing to spectral matching, and the sampled spectrum histogram is presented. On the basis of the sampled spectrum histogram, a spectral feature named the crosscut feature (CF) is developed. Specifically, we make a number of equally spaced horizontal lines on the ultraspectral curve, and count the number of intersections of each line with the spectral curve to obtain the CF vector. This process is simple and the dimensions of the CF vector can be adjusted to obtain better performance. The CF vector is extracted as the feature to implement the spectral matching procedure. Experimental results show that the CF method reduces the operation time significantly while maintaining a high matching accuracy, and is suitable for different ultraspectral matching applications.     

一种利用Gabor小波特征的目标跟踪方法

小波特征跟踪是一种新的目标跟踪方法。这种方法利用Gabor小波提取目标的边界特征来组成特征模板,并由小波滤波器系数组成特征向量。在跟踪过程中不断优化特征模板的旋转、尺度及平移等几何变形参数,使得小波特征向量值在最小平方和的意义上与初始值相匹配,从而实现目标旋转、尺度及平移等几何变形时的自动跟踪。实验表明,该方法能够有效地跟踪飞机、车辆等目标,并且能够抵抗全局亮度的变化和局部遮挡的干扰。     

基于Hough投票空间的椭圆图像特征亚像素提取方法

椭圆检测是机器视觉和模式识别的基本任务。首先在椭圆中心初定位阶段,通过参数空间的少量投票,依据均值漂移技术获取粗略的椭圆中心,然后再由精提策略实现椭圆中心的精确像素级定位,并采用插值方法进一步获取椭圆中心亚像素坐标。仿真及真实图像的实验结果表明,所提出的方法同时具备了椭圆检测的鲁棒性和中心定位的精确性。     

基于非参数特征提取的早期林火烟雾检测

提出了一种基于非参数特征提取的早期森林火灾烟雾检测方法。采用双背景模型进行运动区域检测,得到可疑烟雾区域。再利用早期火灾烟雾颜色特征和运动特性对提取的可疑运动区域进行颜色判别,对增长区域以及上升区域进行检测,排除部分干扰目标。根据提出的非参数特征提取法,计算并科学地选取可疑区域的特征量,将其输入训练好的支持向量机(SVM)进行烟雾判别。对多段视频进行测试表明,该方法能适用于复杂多变的野外环境,排除干扰,有效地检测出早期林火烟雾。     

天空背景下飞行器结构特征提取的新方法

针对天空背景下低信噪比的飞行器,提出了一种基于SUSAN算法、灰色系统理论和数学形态学相结合的飞行器结构特征提取的新方法。在Visual C++6.0平台下,首先利用SUSAN算法从背景中提取飞行器的结构边缘信息,并与原图像相加实现目标增强;然后用灰色系统理论检测出飞行器的结构特征边缘;最后利用条件膨胀和重构算法,实现云层的抑制,并重构出飞行器目标。实验结果表明:该方法对于实现飞行器的跟踪、结构特征提取以及事后判读有重要的意义,同时验证了该方法的可行性。     

Weak fault feature extraction method based on compound tri-stable stochastic resonance

In the presence of strong background noise, in view of the difficulty in extracting weak fault features, a new compound tri-stable stochastic resonance (CTSR) model is proposed by combining the Gaussian Potential model and the mixed bi-stable model. Compared with the traditional tri-stable stochastic resonance (TTSR) method, all parameters of CTSR model have no coupling characteristics. Therefore, the output signal-to-noise ratio (SNR) can be easily optimized by adjusting the system parameters. The CTSR model retains the advantages of constraint and continuity of the Gaussian Potential model, and has a higher utilization rate of noise. Finally, through bearing and engineering experiments, the outstanding advantages of the proposed method in feature extraction of weak faults are verified.     

水下目标亮点拓扑特征提取及自动识别方法

针对水下目标逆合成孔径声呐(Inverse Synthetic Aperture Sonar,ISAS)图像识别问题中观测角度随机多变,目标结构相互遮挡问题,提出一种基于多亮点拓扑矢量特征的ISAS水下目标识别方法。通过分析ISAS成像过程中散射点位置由三维空间向二维成像平面的投影关系,表明了横向定标后的声呐图像中强亮点之间的距离仅由目标散射结构之间的物理距离决定,据此基于强亮点之间的相互距离,构造能稳定描述不同观测角度下目标的拓扑矢量特征。然后通过K-means聚类获取多聚类中心以克服目标结构互相遮挡造成的亮点缺失问题。最终采用最近邻分类器实现目标识别。水池缩比模型实验表明,该方法对于水下目标的识别率达到84.0%。     

Detecting Doppler ultrasound embolic signals using the wavelet-based feature extraction algorithm

The non-invasive detection of circulating emboli with the Doppler ultrasound technique is of active significance in clinical applications. In order to eliminate drawbacks of artifacts brought by the movement of probes or patients and detect emboli accurately, relevant feature parameters are extracted from two angles of the wavelet transform of Doppler signals. The singularity of the signal waveform is analyzed based on its wavelet scalogram; then transverse and longitudinal parameters are extracted to represent the scalogram characteristics. A novel method is proposed based on the adaptive wavelet packet basis, from which several parameters such as the energy, the scale, etc. are extracted to represent the optimized signal approximation features. With all feature parameters in two aspects, a classification system is established for Doppler Ultrasound embolic signals by solving the generalized Fisher discriminant plane. From experiments on 300 simulated and 298 clinical Doppler ultrasound signals of cerebral arteries, it is shown that the proposed system can achieve the emboli detection rates of 99.0% and 98.5% for the training set and the testing set respectively. Therefore this method makes an improve- ment of emboli detection compared to traditional in the automatic detection of clinical emboli. methods and has the possibility to be applied     

汉语塞音选择性特征自动萃取的小波变换方法

本文提出了汉语语音导引特征的概念,讨论了语音导引特征在自动语音识别中用于导引匹配搜索的运用方式和重要作用;提出并设计了汉语塞音选择性特征自动萃取的小波变换方法和数字信号处理算法。本文方法和算法从声学信号处理和检测的角度,将汉语清辅音声波输入信号自动分为塞音子集BDG:{b,d,g}、塞音子集ZZHJGPTcCHQK:{z,zh,j,g,p,t,c,ch,q,k}和擦音集FsSHhX:{f,s,sh,x,h};对输入的合清辅音的音节,计算检测并输出汉语自动语音识别系统可以利用的清辅音类属标记b.d.g、STOP/BD和f.s.sh.x.h以及它们的音段起始时标;从声学信息计算检测的角度为汉语自动语音识别系统提供一种新的\     

Measurement of the axial periodicity of collagen fibrils using an image processing method

In this paper, a method is described based on a computer-aided analysis of electron-optical images of collagen fibrils from various tissues, in order to determine the axial periodicity of such fibrils. The method gives information at a level of 2-3nm.     

采用上下文相关的注意力机制及循环神经网络的语音增强方法

提出了采用上下文相关的注意力机制及循环神经网络的语音增强方法。该方法在训练阶段联合训练计算注意力评分的多层感知机和增强语音的深度循环网络,在测试阶段计算每一帧语音的注意力向量并与该帧语音拼接输入深度循环网络增强。在不同信噪比的实验中,该方法相比基线模型能更好地提高语音质量和可懂度,-6 dB下相对带噪语音短时客观可懂度(STOI)和语音质量感知评估(PESQ)可分别提高0.16和0.77,同时在未知噪声条件下该方法性能仍最优或接近最优。因此注意力机制可以有效强化模型对上下文信息的利用能力,从而提高模型增强性能。     

Thermal feature extraction of servers in a datacenter using thermal image registration

Thermal cameras provide fine-grained thermal information that enhances monitoring and enables automatic thermal management in large datacenters. Recent approaches employing mobile robots or thermal camera networks can already identify the physical locations of hot spots. Other distribution information used to optimize datacenter management can also be obtained automatically using pattern recognition technology. However, most of the features extracted from thermal images, such as shape and gradient, may be affected by changes in the position and direction of the thermal camera. This paper presents a method for extracting the thermal features of a hot spot or a server in a container datacenter. First, thermal and visual images are registered based on textural characteristics extracted from images acquired in datacenters. Then, the thermal distribution of each server is standardized. The features of a hot spot or server extracted from the standard distribution can reduce the impact of camera position and direction. The results of experiments show that image registration is efficient for aligning the corresponding visual and thermal images in the datacenter, and the standardization procedure reduces the impacts of camera position and direction on hot spot or server features.     

融合多尺度特征的短时音频场景识别方法EI北大核心CSCD

为解决短时音频场景识别任务中识别性能差的问题,提出一种融合多尺度特征的音频场景识别方法。首先将双声道音频中左右声道的和差作为输入,并使用长时帧长进行分帧处理,以保证提取出的帧级特征中包含足够多的音频信息。然后将特征逐帧输入到融合多尺度特征的一维卷积神经网络中,以充分利用网络中不同尺度的浅层、中层和深层嵌入特征。最后综合所有帧级软标签得到短时音频的场景分类结果。实验结果表明,该方法在国际声学场景和事件检测与分类挑战赛(DCASE) 2021短时音频场景数据集上的准确率为79.02%,实现了该数据集上目前为止的最优性能。     

基于集合经验模态分解的类星体光变周期及其混沌特性分析

基于密歇根大学射电天文台数据库中从1965年到2012年收集的类星体3C 345,3C 273和3C 279在射电8.0 GHz的光变数据,利用集合经验模态分解方法将这些类星体的光变资料分解为周期项、趋势项和高频项,并对分解后的高频项计算其饱和关联维数、最大Lyapunov指数和Kolmogorov熵,判断是否具有混沌性.结果表明,这些类星体的光变不仅具有周期性,也具有明显的混沌特性,表明类星体光变应为产生周期性运动的物理机制和产生混沌现象的非线性机制的综合结果.     

基于图像扩散去噪的距离特征量提取研究

距离特征量反映了目标运动过程中的距离变化率,该观测量可由目标辐射噪声LOFAR谱图的干涉条纹中提取得到。低信噪比情况下,由于噪声的影响,干涉条纹特征会变模糊,提取的距离特征量值精度会降低。为提高提取精度,提出了一种边缘定向增强型扩散方程去噪方法,对LOFAR谱图进行处理,使条纹特征更加明显。迭代次数对扩散方程去噪效果影响很大,根据相关性原理,提出了一种迭代次数选择方法。为了使条纹特征更加易于区分,提出了一种基于区域分割的图像二值化方法。数值模拟实验和海试数据处理结果表明,LOFAR谱图经过处理后,提取出的距离特征量精度明显提高。