SOM神经网络改进及在遥感图像分类中的应用*

针对自组织特征神经网络自身算法的特点和缺陷,采用遗传算法对网络进行改进,形成了基于遗传算法的自组织特征神经网络,并从输入向量、竞争层神经元数量设置和初始权向量设定三方面,结合遥感图像的特性对自组织特征映射网络遥感图像分类的方法进行了改进。将该方法应用于择西安地区的ETM+卫星遥感图像进行分类试验,结果表明,基于遗传算法的自组织特征映射网络使得遥感图像的分类精度更高,且该算法实现简单,具有一定的工程应用价值。     

基于自组织特征映射神经网络的医学图像分割技术

采用一维自组织特征映射神经网络对医学图像进行聚类分析,实现对不同组织的自动分割.避免了直接使用灰度门限分割方法由于门限值选择不当所导致的分割结果有失准确性的缺点.试验结果表明,利用该方法能够较好地保证分割结果的准确性和完整性.     

基于混合核函数的自组织神经网络遥感图像分类

自组织神经网络SOM作为一种无监督学习的竞争式网络,已经得到了广泛的应用,它通过对输入信号的竞争学习,将样本划为不同的类别,但其分类效果常很难令人满意.提出了一种基于混合核函数的SOM神经网络改进方法,并和传统的SOM网络进行了对比,Iris数据和Wine数据的分类实验表明,该方法可以明显改进SOM网络的分类效果.然后对某地Landsat卫星遥感图像数据进行分类实验,实验结果表明,与传统的SOM网络、基于多项式核的SOM网络以及基于RBF核的SOM网络相比较,基于混合核函数的SOM神经网络方法的分类效果有较明显的提高.     

基于自组织神经网络的多传感器遥感图像融合技术

多传感器遥感图像融合是一种遥感信息综合分析与处理的技术,其研究正成为遥感学科领域的热门课题之一。利用自组织特征映射神经网络具有较好的聚类特性,对多源遥感图像进行高水平的分类与融合。通过对GMS 5卫星图像融合前后性质的比较和融合质量的评估,不难发现:融合后的图像不论在视觉效果上还是在噪声特性上都有了很大的改善。     

自组织特征映射网络在舌边数据分类中的应用

尝试利用自组织特征映射网络较强的聚类功能对分割出的舌体边缘进行分类。通过实验证明它能很好的将舌边数据分成舌根、舌尖、舌左、舌右四类点,达到预定目标。     

神经网络在高光谱图像分类中的应用

高光谱遥感图像作为一种新型的遥感图像,鉴于传统的遥感图像识别方法对这种图像的识别精度较低,该文采用BP神经网络方法对高光谱遥感图像进行识别和分类,并使用赤铁矿等六种矿石的光谱图像对神经网络进行洲练,得到很好的效果。     

行程长度纹理特征在SPOT遥感图像分类中的应用

将行程长度纹理特征与神经网络相结合应用于遥感图像分类中.在特征选择阶段采用类内、类间方差标准与Rough集相结合的方法挑选出有较强分类能力的特征并有效去除冗余特征.针对高分辨率、大尺度的SPOT全色遥感卫星图像,分别基于行程长度纹理特征、共生矩阵纹理特征、灰度-梯度共生矩阵纹理特征和灰度-平滑共生矩阵纹理特征,采用BP、RBF两种类型的神经网络以及最近邻分类算法(K-NN法)对其进行分类,并对分类结果进行对比.实验结果证明本文算法的有效性.     

改进的概率神经网络的语义区域分类

针对彩色图像信息量大,分割效果自适应性差的问题,对图像语义区域的分割精度进行控制,提取图像的纹理特征值,再通过改进后的概率神经网络模型对测试样本做分类测试,达到提高图像语义提取和分类准确性的目的。实验表明,改进后的概率神经网络对彩色图像语义区域分类的正确性由原先的70%提高到90%,具有较好的分类效果。     

基于自组织神经网络的遥感图像分类应用研究

在地形复杂地区,采用传统的遥感分类方法进行土地利用遥感分类很难获得理想的精度。针对遥感图像分类的特点,根据自组织竞争神经网络的生物学基础、基本结构和学习算法,利用Matlab平台构建自组织神经网络,对地形复杂区的ETM 遥感图像通过500次训练使网络收敛后,仿真输出分类图。结果表明,基于自组织神经网络的分类器经过训练后,可用于遥感图像的分类,其分类精度明显高于传统的最大似然法。     

神经网络在高光谱图像分类中的应用

高光谱遥感图像作为一种新型的遥感图像,鉴于传统的遥感图像识别方法对这种图像的识别精度较低,该文采用BP神经网络方法对高光谱遥感图像进行识别和分类,并使用赤铁矿等六种矿石的光谱图像对神经网络进行训练,得到很好的效果。     

Classification of binary document images into textual or nontextual data blocks using neural network models

This paper describes a new method for the classification of binary document images as textual or nontextual data blocks using neural network models. Binary document images are first segmented into blocks by the constrained run-length algorithm (CRLA). The component-labeling procedure is used to label the resulting blocks. The features for each block, calculated from the coordinates of its extremities, are then fed into the input layer of a neural network for classification. Four neural networks were considered, and they include back propagation (BP), radial basis functions (RBF), probabilistic neural network (PNN), and Kohonen's self-organizing feature maps (SOFMs). The performance and behavior of these neural network models are analyzed and compared in terms of training times, memory requirements, and classification accuracy. The experiments carried out on a variety of medical journals show the feasibility of using the neural network approach for textual block classification and indicate that in terms of both accuracy and training time RBF should be preferred.     

Classification of MRI brain images using combined wavelet entropy based spider web plots and probabilistic neural network

Magnetic resonance imaging (MRI) is a non-invasive diagnostic tool very frequently used for brain imaging. The classification of MRI images of normal and pathological brain conditions pose a challenge from technological and clinical point of view, since MR imaging focuses on soft tissue anatomy and generates a large information set and these can act as a mirror reflecting the conditions of the brain. A new approach by integrating wavelet entropy based spider web plots and probabilistic neural network is proposed for the classification of MRI brain images. The two step method for classification uses (1) wavelet entropy based spider web plots for the feature extraction and (2) probabilistic neural network for the classification. The spider web plot is a geometric construction drawn using the entropy of the wavelet approximation components and the areas calculated are used as feature set for classification. Probabilistic neural network provides a general solution to the pattern classification problems and the classification accuracy is found to be 100%.     

Segmentation of ultrasound images by using a hybrid neural network

A hybrid neural network is presented for the segmentation of ultrasound images.

Feature vectors are formed by the discrete cosine transform of pixel intensities in region of interest (ROI). The elements and the dimension of the feature vectors are determined by considering only two parameters: The amount of ignored coefficients, and the dimension of the ROI.

First-layer-nodes of the proposed hybrid network represent hyperspheres (HSs) in the feature space. Feature space is partitioned by intersecting these HSs to represent the distribution of classes. The locations and radii of the HSs are found by the genetic algorithms.

Restricted Coulomb energy (RCE) network, modified RCE network, multi-layer perceptron and the proposed hybrid neural network are examined comparatively for the segmentation of ultrasound images.     

基于细胞神经网络的灰度图像边缘提取

针对细胞神经网络(cellular neural network,CNN),研究了图像边缘提取的过程,给出算法流程,阐述了算法实现过程中的几个关键步骤.对二值图像和灰度图像,分别采用基于CNN的算法和传统算子(prewitt、sobel、canny)进行边缘提取,定性分析比较了两类算法在性能上的优劣.实验结果表明,基于CNN的算法在硬件实现上能够高速并行计算,而且处理速度与图像大小无关,能够实现图像实时处理.     

Classification of EEG signals using neural network and logistic regression

Epileptic seizures are manifestations of epilepsy. Careful analyses of the electroencephalograph (EEG) records can provide valuable insight and improved understanding of the mechanisms causing epileptic disorders. The detection of epileptiform discharges in the EEG is an important component in the diagnosis of epilepsy. As EEG signals are non-stationary, the conventional method of frequency analysis is not highly successful in diagnostic classification. This paper deals with a novel method of analysis of EEG signals using wavelet transform and classification using artificial neural network (ANN) and logistic regression (LR). Wavelet transform is particularly effective for representing various aspects of non-stationary signals such as trends, discontinuities and repeated patterns where other signal processing approaches fail or are not as effective. Through wavelet decomposition of the EEG records, transient features are accurately captured and localized in both time and frequency context. In epileptic seizure classification we used lifting-based discrete wavelet transform (LBDWT) as a preprocessing method to increase the computational speed. The proposed algorithm reduces the computational load of those algorithms that were based on classical wavelet transform (CWT). In this study, we introduce two fundamentally different approaches for designing classification models (classifiers) the traditional statistical method based on logistic regression and the emerging computationally powerful techniques based on ANN. Logistic regression as well as multilayer perceptron neural network (MLPNN) based classifiers were developed and compared in relation to their accuracy in classification of EEG signals. In these methods we used LBDWT coefficients of EEG signals as an input to classification system with two discrete outputs: epileptic seizure or non-epileptic seizure. By identifying features in the signal we want to provide an automatic system that will support a physician in the diagnosing process. By applying LBDWT in connection with MLPNN, we obtained novel and reliable classifier architecture. The comparisons between the developed classifiers were primarily based on analysis of the receiver operating characteristic (ROC) curves as well as a number of scalar performance measures pertaining to the classification. The MLPNN based classifier outperformed the LR based counterpart. Within the same group, the MLPNN based classifier was more accurate than the LR based classifier.     

Pattern classification of solder joint images using a correlation neural network

This paper presents a method of classifying solder joints on printed-circuit boards (PCB), using a neural-network approach. Inherently, the surface of the solder joints is curved, tiny and specularly reflective; it induces a difficulty of taking good images of the solder joints. The shapes of the solder joints tend to vary greatly with soldering conditions; solder joints, even when classified into the same soldering quality, have very different shapes. Furthermore, the position of the joints is not consistent within a registered solder pad on the PCB. Due to these aspects, it has been difficult to determine the visual features and classification criteria for automatic solder-joint inspection. In this research, the solder joints, imaged by using a circular, tiered illumination system of three colored lamps, are represented as red, green and blue colored patterns, showing their surface-slopes. Cross-correlation and auto-correlation of the colored patterns are used to classify the 3D shapes of the solder joints by their soldering qualities. To achieve this, a neural network is proposed, based on a functional link net, with two processing modules. The first preprocessing module is designed to implement the calculation of the correlations in functional terms. The subsequent, trainable module classifies the solder joints, based upon the capability learned from a human supervisor. The practical feasibility of the proposed method is demonstrated by testing numerous commercially manufactured PCBs.     

Labelling logical structures of document images using a dynamic perceptive neural network

This paper proposes a new method for labelling the logical structures of document images. The system starts with digitised images of paper documents, performs a physical layout analysis, runs an OCR and finally exploits the OCR’s outputs to find the meaning of each block of text (i.e. assigns labels like “Title”, “Author”, etc.). The method is an extension of our previous work where a classifier, the perceptive neural network, has been developed to be an analogy of the human perception. We introduce in this connectionist model a temporal dimension by the use of a time-delay neural network with local representation. During the recognition stage, the system performs several recognition cycles and corrections, while keeping track and reusing the previous outputs. This dynamic classifier allows then a better handling of noise and segmentation errors. The experiments have been carried out on two datasets: the public MARG containing more than 1,500 front pages of scientific papers with four zones of interest and another one composed of documents from the Siggraph 2003 conference, where 21 logical structures have been identified. The error rate on MARG is less than 2.5% and 7.3% on the Siggraph dataset.     

卷积神经网络提取风云影像土壤湿度

目的 时空分辨率较高的土壤湿度数据对于生产实践和科学研究具有重要意义。以国产的风云气象卫星为数据源,利用卷积神经网络自主学习输入变量间深层关联的优势,获取高质量土壤湿度数据,为科学研究和生产实践服务。方法 首先构建了一个土壤湿度提取卷积神经网络（soil moisture convolutional neural network,SMCNN）,SMCNN由温度子网络和土壤湿度子网络构成,每个子网络均包含特征提取器和编码器。特征提取器用于为每个像素生成一个特征向量,其中温度子网络的特征提取器由11个卷积层组成,湿度子网络的特征提取器由9个卷积层组成,卷积层均使用1×1的卷积核。编码器用于将提取到的特征拟合为目标变量。两个子网络均使用平均方差作为损失函数。使用随机梯度下降算法对模型进行训练,最后利用训练好的模型提取区域土壤湿度数据。结果 选择宁夏回族自治区为实验区,利用获取的2016-2019年风云3D影像和相应地面站点数据作为实验数据,选择线性回归模型、BP（back propagation）神经网络模型作为对比模型开展数据实验,选择均方根误差作为评价指标。实验结果表明,SMCNN的均方根误差为0.006 7,优于对比模型,SMCNN模型在从风云影像中提取土壤湿度方面具有优势。结论 本文利用卷积神经网络分别构建用于反演地表温度和土壤湿度的子网络,再组成一个完整的土壤湿度反演网络结构,从风云3D数据中获取数值精度、时空分辨率均较高的土壤湿度数据,满足了科学研究和生产实践对大范围高精度土壤湿度数据的需求。     

Associative memory design for 256 gray-level images using a multilayer neural network

A design procedure is presented for neural associative memories storing gray-scale images. It is an evolution of a previous work based on the decomposition of the image with 2/sup L/ gray levels into L binary patterns, stored in L uncoupled neural networks. In this letter, an L-layer neural network is proposed with both intralayer and interlayer connections. The connections between different layers introduce interactions among all the neurons, increasing the recall performance with respect to the uncoupled case. In particular, the proposed network can store images with the commonly used number of 256 gray levels instead of 16, as in the previous approach.