基于红外神经网络的高压断路器故障诊断方法

为了高效、快速、准确地确定高压断路器的故障,提出了一种基于红外诊断与神经网络相结合的高压断路器新型故障诊断方法。首先,利用红外测量技术采集高压断路器故障样本,并将故障样本进行归一化;其次,构造BP神经网络故障诊断模型,提出一种新型改进BP神经网络算法,将构造的高压断路器故障样本输入到改进BP神经网络中进行训练,得到改进BP神经网络的相关参数;最后,通过仿真研究验证了提出的基于红外神经网络的高压断路器故障诊断方法的合理性与优越性。

CNG气瓶缠绕层缺陷的红外定位和定量识别及实验研究

CNG复合材料气瓶长期处于交变载荷作用,易于发生疲劳损伤,形成内部缺陷,造成强度下降,影响使用安全。含内部缺陷的复合材料气瓶在宏观上无明显形变,表观上难以直接进行缺陷检测。在目前的气瓶检测项目中,缺乏快速有效的缠绕层内部缺陷检测手段,可能造成含有内部缺陷气瓶的漏检。本文针对CNG复合材料气瓶检测存在的关键问题,结合现有气瓶检测标准和工艺,提出了一种基于气瓶内部蒸汽冲洗过程表面热像的缺陷检测方案。该方案以气瓶冲洗过程的蒸汽为气瓶的内部热激励,基于红外热像仪采集的气瓶表面瞬态温度分布,利用人工神经网络实现气瓶缠绕层缺陷的定位和定量识别。实验研究表明,人工神经网络能够精确地进行气瓶缠绕层缺陷的定位和定量识别且识别效率较高,适于气瓶的在线检测。

Multi-sensors Image Fusion via NSCT and GoogLeNet

In order to improve the detail preservation and target information integrity of different sensor fusion images,an image fusion method of different sensors based on non-subsampling contourlet transform(NSCT)and GoogLeNet neural network model is proposed. First,the different sensors images,i. e.,infrared and visible images,are transformed by NSCT to obtain a low frequency sub-band and a series of high frequency sub-bands respectively.Then,the high frequency sub-bands are fused with the max regional energy selection strategy,the low frequency subbands are input into GoogLeNet neural network model to extract feature maps,and the fusion weight matrices are adaptively calculated from the feature maps. Next,the fused low frequency sub-band is obtained with weighted summation. Finally,the fused image is obtained by inverse NSCT. The experimental results demonstrate that the proposed method improves the image visual effect and achieves better performance in both edge retention and mutual information.

Determination of Active Components in a Natural Herb with Near Infrared Spectroscopy Based on Artificial Neural Networks

The non-linear relationships between the contents of ginsenoside Rg 1, Rb 1, Rd and Panax notoginseng saponins(PNS) in Panax notoginseng root herb and the near infrared(NIR) diffuse reflectance spectra of the herb were established by means of artificial neural networks(ANNs). Four three-layered perception feed-forward networks were trained with an error back-propagation algorithm. The significant principal components of the NIR spectral data matrix were utilized as the input of the networks. The networks architecture and parameters were selected so as to offer less prediction errors. Relative prediction errors for Rg 1, Rb 1, Rd and PNS obtained with the optimum ANN models were 8.99%, 6.54%, 8.29%, and 5.17%, respectively, which were superior to those obtained with PLSR methods. It is verified that ANN is a suitable approach to model this complex non-linearity. The developed method is fast, non-destructive and accurate and it provides a new efficient approach for determining the active components in the complex system of natural herbs.

Evolving Neural Network Using Variable String Genetic Algorithm for Color Infrared Aerial Image Classification

Coastal wetlands are characterized by complex patterns both in their geomorphlc and ecological teatures. Besides field observations, it is necessary to analyze the land cover of wetlands through the color infrared （CIR） aerial photography or remote sensing image. In this paper, we designed an evolving neural network classifier using variable string genetic algorithm （VGA） for the land cover classification of CIR aerial image. With the VGA, the classifier that we designed is able to evolve automatically the appropriate number of hidden nodes for modeling the neural network topology optimally and to find a near-optimal set of connection weights globally. Then, with backpropagation algorithm （BP）, it can find the best connection weights. The VGA-BP classifier, which is derived from hybrid algorithms mentioned above, is demonstrated on CIR images classification effectively. Compared with standard classifiers, such as Bayes maximum-likelihood classifier, VGA classifier and BP-MLP （multi-layer perception） classifier, it has shown that the VGA-BP classifier can have better performance on highly resolution land cover classification.

Fusion of VNIR and Thermal Infrared Remote Sensing Data Based on GA-SOFM Neural Network

The multi-source data fusion methods are rarely involved in VNIR and thermal infrared remote sensing at present. Therefore, the potential advantages of the two kinds of data have not yet been adequately tapped, which results in low calculation precision of parameters related with land surface temperature. A new fusion method is put forward where the characteristics of the high spatial resolution of VNIR （visible and near infrared） data and the high temporal resolution of thermal infrared data are fully explored in this paper. Non-linear fusion is implemented to obtain the land surface temperature in high spatial resolution and the high temporal resolution between the land surface parameters estimated from VNIR data and the thermal infrared data by means of GA-SOFM （genetic algorithms ＆ self-organizing feature maps）-ANN （artificial neural net-work）. Finally, the method is verified by ASTER satellite data. The result shows that the method is simple and convenient and can rapidly capture land surface temperature distribution of higher resolution with high precision.