首页 | 本学科首页   官方微博 | 高级检索  
相似文献
 共查询到20条相似文献,搜索用时 15 毫秒
1.
2.
3.
介绍了一种新的自组织聚类人工神经网络(DIGNET)模型和网络的非监督学习算法。针对数据融合和目标识别的特点,提出了基于DIGNET的决策层数据融合目标分类方法。利用仿真数据研究了DIGNET和自组织特征映射网络(SOFM)的聚类性能以及基于DIGNET的决策层数据融合结构,实验结果表明DIGNET较SOFM正确分类率高、抗噪能力好;基于DIGNET的决策层数据融合能够有效地实现融合识别。将该数据融合方法应用于前视红外(FLIR)和可见光摄像机目标跟踪系统,结果表明该方法是可行的。  相似文献   

4.
Hai-Zhu Pan 《中国物理 B》2022,31(12):120701-120701
Benefiting from the development of hyperspectral imaging technology, hyperspectral image (HSI) classification has become a valuable direction in remote sensing image processing. Recently, researchers have found a connection between convolutional neural networks (CNNs) and Gabor filters. Therefore, some Gabor-based CNN methods have been proposed for HSI classification. However, most Gabor-based CNN methods still manually generate Gabor filters whose parameters are empirically set and remain unchanged during the CNN learning process. Moreover, these methods require patch cubes as network inputs. Such patch cubes may contain interference pixels, which will negatively affect the classification results. To address these problems, in this paper, we propose a learnable three-dimensional (3D) Gabor convolutional network with global affinity attention for HSI classification. More precisely, the learnable 3D Gabor convolution kernel is constructed by the 3D Gabor filter, which can be learned and updated during the training process. Furthermore, spatial and spectral global affinity attention modules are introduced to capture more discriminative features between spatial locations and spectral bands in the patch cube, thus alleviating the interfering pixels problem. Experimental results on three well-known HSI datasets (including two natural crop scenarios and one urban scenario) have demonstrated that the proposed network can achieve powerful classification performance and outperforms widely used machine-learning-based and deep-learning-based methods.  相似文献   

5.
We introduce a class of projected entangled pair states (PEPS) which is based on a group symmetry twisted by a 3-cocycle of the group. This twisted symmetry is expressed as a matrix product operator (MPO) with bond dimension greater than 1 and acts on the virtual boundary of a PEPS tensor. We show that it gives rise to a new standard form for PEPS from which we construct a family of local Hamiltonians which are gapped, frustration-free and include fixed points of the renormalization group flow. Based on this insight, we advance the classification of 2D gapped quantum spin systems by showing how this new standard form for PEPS determines the emergent topological order of these local Hamiltonians. Specifically, we identify their universality class as Dijkgraaf–Witten topological quantum field theory (TQFT).  相似文献   

6.
Whistles from five delphinid species in the western Mediterranean Sea (Stenella coeruleoalba, Grampus griseus, Delphinus delphis, Tursiops truncatus, Globicephala melas) were taken from GREC sound archives. FFT contours (window size 512, Hanning, sampling frequency 44.1 kHz) were extracted with custom developed Matlab software: 277 samples of striped dolphins (Sc), 158 whistles of Risso’s dolphins (Gg), 120 of common dolphins (Dd), 76 of bottlenose dolphins (Tt), and 66 of pilot whales (Gm) were selected. Seafox software extracted 15 variables from the digitized contours, including: duration, initial, final, maximal and minimal frequency slopes, frequency range, number of frequency extrema, beginning, ending, maximal and minimal frequencies, presence of harmonics. Four of five species were significantly different (Mann-Whitney test) for average durations (respectively 0.73, 0.65, 0.47 and 0.89 s for Sc, Gg, Dd, Gm) while the average duration of bottlenose dolphins was 0.71 s. Frequency ranges (respectively 7.3, 6.3, 4.6, 3.2 and 6.3 kHz) were significantly different for all species pairs, with the exception of bottlenose and Risso’s dolphins. From a global point of view, pilot whale calls were the most distinct, with 43 significant pair-wise tests out of a total of 52, followed by the common dolphins. Risso’s dolphins were closest to other species whistles. A CART classification method achieved a global classification rate of 62.9%.  相似文献   

7.
对超声DOPPLER声谱图进行分类决策的两种方法   总被引:1,自引:0,他引:1       下载免费PDF全文
本文介绍了两种对超声多普勒血流声谱图进行了波形自动分类法决策的方法,方法一与统计得到的标准波形相比,按适当判据进行自动分类,方法二,送入训练的人工神经网络进行自动分类,待分类波形由计算多普勒信号的最大频率提取声谱包络并经分段,平均和归一化等处理得以,文中给出了计算机模拟的结果。  相似文献   

8.
机器学习势由于具有与第一性原理计算相当的准确性,且低得多的计算成本,在原子模拟中极具前景. 然而原子机器学习势的可靠性、速度和可迁移性在很大程度上取决于原子构型的表示. 适当地选取用作机器学习程序输入的描述符是一个成功的机器学习表示的关键. 本文发展了一种简单有效的方法,可以基于训练数据固有的相关性,从大量待选的描述符中自动选取一组最佳的线性独立原子特征. 通过对几个具有较少冗余线性独立嵌入密度描述符的基准分子构建嵌入原子神经网络势的应用,证明了这种新方法的有效性和准确性. 该算法可以大大简化原子特征的初始选取,并极大地提高原子机器学习势的性能.  相似文献   

9.
闪电光电信号的同步观测与闪电类型模式识别   总被引:2,自引:2,他引:2       下载免费PDF全文
 利用宽带电场仪和光学纹波探头,对2006年夏季广东省从化地区发生的自然闪电的光电信号进行了同步观测,并对结果进行了统计分析。统计结果表明:闪电光信号脉冲的发生时间和峰值到达时间总是滞后于电信号,并且脉冲上升沿和宽度也比电信号宽;云闪的光信号在上述特征参数的滞后量、上升沿时间和脉宽时间差要比地闪回击放电大。从闪电光电信号时域信号特征的角度,利用BP人工神经网络,对闪电类型进行了模式识别研究。总体识别结果基本上能够满足要求,但是由于云闪放电过程比较复杂,获得的数据较少,所以对云闪识别的效果并不理想。  相似文献   

10.
利用宽带电场仪和光学纹波探头,对2006年夏季广东省从化地区发生的自然闪电的光电信号进行了同步观测,并对结果进行了统计分析。统计结果表明:闪电光信号脉冲的发生时间和峰值到达时间总是滞后于电信号,并且脉冲上升沿和宽度也比电信号宽;云闪的光信号在上述特征参数的滞后量、上升沿时间和脉宽时间差要比地闪回击放电大。从闪电光电信号时域信号特征的角度,利用BP人工神经网络,对闪电类型进行了模式识别研究。总体识别结果基本上能够满足要求,但是由于云闪放电过程比较复杂,获得的数据较少,所以对云闪识别的效果并不理想。  相似文献   

11.
星系通常分为正常星系(NG)与活动星系(AG)两类。文章提出了一种自动获取NG红移的快速有效方法: (1) 由NG模板根据红移范围Ⅰ: 0.0~0.3与Ⅱ: 0.3~0.5模拟得到两类星系样本, 进行PCA变换获得样本特征向量; (2) 利用概率神经网络设计两类样本特征向量的Bayes分类器; (3) 对于实际NG光谱数据, 利用Bayes分类器进行分类确定其红移的范围, 然后在此范围内进行模板匹配得到红移的准确值。与在整个红移范围内的模板匹配方法相比, 此方法不但节省了50%的模板匹配运算量, 而且还大大提高了红移值测量的精度。文章研究结果对于大型光谱巡天所产生的海量数据的自动处理具有重要意义。  相似文献   

12.
应力盘智能控制盘面面形的表征方法及检测技术的研究   总被引:2,自引:0,他引:2  
分析了CMAC神经网络应力盘智能控制中面形表征及检测的特点,提出了以Zernike多项式系数用于表征应力盘盘面面形的方法,以微位移阵列传感器检测得到的原始数据经插值拟合重构出盘面面型数据,再经Gram-Schimdt正交化后拟合出的Zernike多项式的系数,以此作为CMAC神经网络的输入样本来完成CMAC神经网络的训练。并在实验中初步验证了此方法的可行性。  相似文献   

13.
王淑青  汤照  陈春雷  王慧 《大学物理》2008,27(1):43-45,55
给出了采用BP神经网络拟合标准光源在棱镜摄谱仪的谱平面位置和相应波长的非线性函数,省去了和标准铁谱比对测量未知谱线波长,简化了光谱的定量分析,提高了仪器的自动化和智能化水平.  相似文献   

14.
张强 《物理学报》2008,57(1):125-128
The dynamics of discrete time delayed Hopfield neural networks is investigated. By using a difference inequality combining with the linear matrix inequality, a sufficient condition ensuring global exponential stability of the unique equilibrium point of the networks is found. The result obtained holds not only for constant delay but also for time-varying delays.  相似文献   

15.
张强 《中国物理 B》2008,17(1):125-128
The dynamics of discrete time delayed Hopfield neural networks is investigated. By using a difference inequality combining with the linear matrix inequality, a sufficient condition ensuring global exponential stability of the unique equilibrium point of the networks is found. The result obtained holds not only for constant delay but also for time-varying delays.  相似文献   

16.
Visualization of bone marrow lesion (BML) can improve the diagnosis of many bone disorders that are associated with it. A quantitative approach in detecting BML could increase the accuracy and efficiency of diagnosing those bone disorders. In this paper, we investigated the feasibility of using magnetic resonance imaging (MRI)-based texture to (a) identify slices and (b) classify subjects with and without BML. A total of 58 subjects were studied; 29 of them were affected by BML. The ages of subjects ranged from 45 to 74 years with a mean age of 59. Texture parameters were calculated for the weight-bearing region of distal femur. The parameters were then analyzed using Mann–Whitney U test and individual feature selection methods to identify potentially discriminantive parameters. Forward feature selection was applied to select features subset for classification. Classification results from eight classifiers were studied. Results show that 98 of the 147 parameters studied are statistically significantly different between the normal and affected marrows: parameters based on co-occurrence matrix are ranked highest in their separability. The classification of subjects achieved an area under the receiver operating characteristic curve (AUC) of 0.914, and the classification of slices achieved an AUC of 0.780. The results show that MRI-texture-based classification can effectively classify subjects/slices with and without BML.  相似文献   

17.
卷积神经网络(CNN)在图像分类识别领域应用广泛,但其在近红外光谱分类中的研究还未见报道,对基于CNN的近红外光谱分类建模方法进行了研究。针对近红外光谱数据的特点,提出了一种改进的卷积神经网络建模方法,对CNN经典模型LeNet-5所做改进:①将方形矩阵卷积核改为适用于一维近红外光谱的向量卷积核;②简化网络结构,将LeNet-5结构中C5,F6及输出层改为单层感知机。同时,采用隔点采样的方法对近红外光谱降维,加快收敛速度;并对卷积核尺寸对建模结果的影响进行了研究。以我国东北、黄淮、西南三大烤烟产区的600个中部烟叶样本的近红外光谱为实验对象,建立烟叶产区分类NIR-CNN模型。该模型对训练集和测试集的判别准确率为98.2%和95%。实验结果表明,应用卷积神经网络可对近红外光谱数据准确、可靠地判别分类;烟叶产区NIR-CNN建模方法可为卷烟企业烟叶原料科学合理利用提供指导,为维护卷烟产品的质量稳定有重要意义;基于卷积神经网络的近红外光谱判别方法也可推广到其他农产品的分类应用中。  相似文献   

18.
人工神经网络分类鉴别苦丁茶红外光谱   总被引:7,自引:2,他引:7  
为了分类鉴别苦丁茶,采用竞争神经网络(CNN)和反向传播人工神经网络(BP网络)两种模式的人工神经网络(ANN)分别分析了各种苦丁茶的红外谱图。作者采用25个样本作训练集,11个样本作检验集,用两种网络进行了训练。结果表明,CNN网络和BP网络均能够有效地实现苦丁茶产地的鉴别,但CNN网络能够进一步地区分苦丁茶的级别。实验表明,CNN速度快,预测结果准确,可望用竞争神经网络(CNN)和红外光谱法结合分类鉴别苦丁茶。  相似文献   

19.
In this Letter, the chaotification for a class of cellular neural networks with distributed delays is studied. On the basis of the largest Lyapunov exponent, the sensitivity to the initial conditions is studied for the distributed delays with kernel being weak and strong. Some theoretical results about the chaotification for the neural network with distributed time delays are derived. Finally, two numerical simulations are presented to illustrate the effectiveness of the theoretical results.  相似文献   

20.
The determination of proteins with 2-(4-chloro-2-phosphonophenylazo)-7-(4-iodophenylazo)-1,8-dihydroxynaphthalene-3,6-disulfonic acid (CPA-pI) by Rayleigh light scattering (RLS) was studied in this paper. The weak RLS of CPA-pI and BSA can be enhanced greatly by the addition of Al3+ at the pH 5.6 and an enhanced RLS signal was produced at 365-385 nm. Based on the reaction of CPA-pI, Al3+ and proteins, a new quantitative determination method for proteins has been developed. The effect of three variables for the determination of proteins was optimized by means of artificial neural networks (ANNs) using extended delta-bar-delta (EDBD) algorithms with the optimal network structure of 3-5-1. This method is very sensitive (2.5-35.4 μg/ml for bovine serum albumin (BSA)), rapid (<2 min), simple (one step) and tolerance of most interfering substances. Six samples of protein in human serum were determined and the maximum relative error is no more than 2% and the recovery is between 95% and 105%.  相似文献   

设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号