Classification of multispectral remote sensing data using aback-propagation neural network

The suitability of a back-propagation neural network for classification of multispectral image data is explored. A methodology is developed for selection of both training parameters and data sets for the training phase. A new technique is also developed to accelerate the learning phase. To benchmark the network, the results are compared to those obtained using three other algorithms: a statistical contextual technique, a supervised piecewise linear classifier, and an unsupervised multispectral clustering algorithm. All three techniques were applied to simulated and real satellite imagery. Results from the classification of both Monte Carlo simulation and real imagery are summarized     

A novel convolutional neural network architecture of multispectral remote sensing images for automatic material classification

For real-world simulation, terrain models must combine various types of information on material and texture in terrain reconstruction for the three-dimensional numerical simulation of terrain. However, the construction of such models using the conventional method often involves high costs in both manpower and time. Therefore, this study used a convolutional neural network (CNN) architecture to classify material in multispectral remote sensing images to simplify the construction of future models. Visible light (i.e., RGB), near infrared (NIR), normalized difference vegetation index (NDVI), and digital surface model (DSM) images were examined.This paper proposes the use of the robust U-Net (RUNet) model, which integrates multiple CNN architectures, for material classification. This model, which is based on an improved U-Net architecture combined with the shortcut connections in the ResNet model, preserves the features of shallow network extraction. The architecture is divided into an encoding layer and a decoding layer. The encoding layer comprises 10 convolutional layers and 4 pooling layers. The decoding layer contains four upsampling layers, eight convolutional layers, and one classification convolutional layer. The material classification process in this study involved the training and testing of the RUNet model. Because of the large size of remote sensing images, the training process randomly cuts subimages of the same size from the training set and then inputs them into the RUNet model for training. To consider the spatial information of the material, the test process cuts multiple test subimages from the test set through mirror padding and overlapping cropping; RUNet then classifies the subimages. Finally, it merges the subimage classification results back into the original test image.The aerial image labeling dataset of the National Institute for Research in Digital Science and Technology (Inria, abbreviated from the French Institut national de recherche en sciences et technologies du numérique) was used as well as its configured dataset (called Inria-2) and a dataset from the International Society for Photogrammetry and Remote Sensing (ISPRS). Material classification was performed with RUNet. Moreover, the effects of the mirror padding and overlapping cropping were analyzed, as were the impacts of subimage size on classification performance. The Inria dataset achieved the optimal results; after the morphological optimization of RUNet, the overall intersection over union (IoU) and classification accuracy reached 70.82% and 95.66%, respectively. Regarding the Inria-2 dataset, the IoU and accuracy were 75.5% and 95.71%, respectively, after classification refinement. Although the overall IoU and accuracy were 0.46% and 0.04% lower than those of the improved fully convolutional network, the training time of the RUNet model was approximately 10.6 h shorter. In the ISPRS dataset experiment, the overall accuracy of the combined multispectral, NDVI, and DSM images reached 89.71%, surpassing that of the RGB images. NIR and DSM provide more information on material features, reducing the likelihood of misclassification caused by similar features (e.g., in color, shape, or texture) in RGB images. Overall, RUNet outperformed the other models in the material classification of remote sensing images. The present findings indicate that it has potential for application in land use monitoring and disaster assessment as well as in model construction for simulation systems.     

A statistical self-organizing learning system for remote sensing classification

A new learning system called a statistical self-organizing learning system (SSOLS), combining functional-link neural networks, statistical hypothesis testing, and self-organization of a number of enhancement nodes, is introduced for remote sensing applications. Its structure consists of two stages, a mapping stage and a learning stage. The input training vectors are initially mapped to the enhancement vectors in the mapping stage by multiplying with a random matrix, followed by pointwise nonlinear transformations. Starting with only one enhancement node, the enhancement layer incrementally adds an extra node in each iteration. The optimum dimension of the enhancement layer is determined by using an efficient leave-one-out cross-validation method. In this way, the number of enhancement nodes is also learned automatically. A t-test algorithm can also be applied to the mapping stage to mitigate the effect of overfitting and to further reduce the number of enhancement nodes required, resulting in a more compact network. In the learning stage, both the input vectors and the enhancement vectors are fed into a least squares learning module to obtain the estimated output vectors. This is made possible by choosing the output layer linear. In addition, several SSOLSs can be trained independently in parallel to form a consensual SSOLS, whose final output is a linear combination of the outputs of each SSOLS module. The SSOLS is simple, fast to compute, and suitable for remote sensing applications, especially with hyperspectral image data of high dimensionality.     

遥感图像分类中的遗传算法LVQ神经网络运用

对于遥感图像分类过程中的问题,提出遗传算法LVQ神经网络来实现遥感图像的分类。将LVQ神经网络结合遗传算法,使用遗传算法最优阈值与权值实现网络训练,使分类精度得到提高。之后融合相似灰度值创建分类图像特征矢量,使特征矢量在神经网络中输入实现训练。学习矢量量化神经算法对初值非常敏感,对遥感图像分类精度具有一定影响。最后,为了对性能进行测试,在实验过程中对比本文分类方法和SVM决策树分类方法,通过实验结果表示,文中提出的分类方法的遥感图像分类精度为95.82%,与其他分类方法相比,分类精度得到进一步提高。     

基于遥感影像和神经网络的城市用地功能分类

城市用地功能分类的准确识别对精准把握城市现状、优化城市空间结构有重要意义。基于此,利用高分辨力遥感影像,提出一种针对中国城市用地功能分类的模型。设计一种多分辨力特征融合的卷积神经网络识别遥感影像中的特定功能区;针对中国城市功能区分布的特点,建立一个用于城市用地功能分类的新数据集。实验显示,本文算法在6种用地功能类型上的分类精确度达88%,表明算法对城市用地功能分类识别具有较高的准确性。最后,通过对北京部分主要城区的案例研究,验证了所提出的模型在城市规划相关领域提供数据支持的价值和有效性。     

BP神经网络学习算法的改进及其应用

针对标准BP算法的不足给出了典型的改进算法。对两个BP网络的应用实例利用MAT LAB语言编制了仿真程序 ,并对几种算法的学习收敛速度进行了比较。结果表明改进算法的学习收敛速度大大地优于标准BP算法。     

基于卷积神经网络的卫星遥感图像区域识别

为了提高卫星遥感图像的识别与分类效果,提出一种基于卷积神经网络的卫星遥感图像识别与分类方法。该方法通过导向滤波去雾和旋转图像数据提高了模型的泛化能力,同时采用了双全连接层网络结构增强了模型数据表达能力。实验证明,该方法在卫星遥感图像的识别与分类上优于传统图像识别方法和一般卷积神经网络模型。     

Designing optimal spectral indexes for remote sensing applications

Satellite remote sensing data constitute a significant potential source of information on our environment, provided they can be adequately interpreted. Vegetation indexes, a subset of the class of spectral indexes, represent one of the most commonly used approaches to analyze data in the optical domain. An optimal spectral index is very sensitive to the desired information (e.g. the amount of vegetation), and as insensitive as possible to perturbing factors (such as soil color changes or atmospheric effects). Since both the desired signal and the perturbing factors vary spectrally, and since the instruments themselves only provide data for particular spectral bands, optimal indexes should be designed for specific applications and particular instruments. This paper describes a rational approach to the design of an optimal index to estimate vegetation properties on the basis of the red and near-infrared reflectances of the AVHRR instrument, taking into account the perturbing effects of soil brightness changes, atmospheric absorption and scattering. The rationale behind the Global Environment Monitoring index (GEMI) is explained, and this index is proposed as an alternative to the Normalized Difference Vegetation Index (NDVI) for global applications. The techniques described here are generally applicable to any multispectral sensor and application     

A lossless compression algorithm of remote sensing image for space applications

A simple and adaptive lossless compression algorithm is proposed for remote sensing image compression, which includes integer wavelet transform and the Rice entropy coder. By analyzing the probability distribution of integer wavelet transform coefficients and the characteristics of Rice entropy coder, the divide and rule method is used for high-frequency sub-bands and low-frequency one. High-frequency sub-bands are coded by the Rice entropy coder, and low-frequency coefficients are predicted before coding. The role of predictor is to map the low-frequency coefficients into symbols suitable for the entropy coding. Experimental results show that the average Comprcssion Ratio （CR） of our approach is about two, which is close to that of JPEG 2000. The algorithm is simple and easy to be implemented in hardware. Moreover, it has the merits of adaptability, and independent data packet. So the algorithm can adapt to space lossless compression applications.     

多偏移遥感图像的BP神经网络亚像元定位

提出了一种借助多偏移遥感图像来改进基于BP神经网络(BPNN)的亚像元定位新方法.不同于原BPNN方法使用单幅低空间分辨率观测图像,新方法利用多幅带有亚像元偏移的低空间分辨图像来确定亚像元属于各类的概率,然后根据概率值和地物覆盖比例确定亚像元类别,以降低BPNN定位模型中的不确定性和误差.实验表明,提出方法在视觉和定量评价上,均能获得更高精度的亚像元定位结果,验证了提出方法的有效性.     

Handwritten digit recognition: applications of neural network chipsand automatic learning

Two novel methods for achieving handwritten digit recognition are described. The first method is based on a neural network chip that performs line thinning and feature extraction using local template matching. The second method is implemented on a digital signal processor and makes extensive use of constrained automatic learning. Experimental results obtained using isolated handwritten digits taken from postal zip codes, a rather difficult data set, are reported and discussed     

A dynamic acceleration method for remote sensing image processing based on CUDA

Wireless Networks - The incredible increase in the volume of remote sensing data has made the concept of Remote Sensing as Big Data reality with recent technological developments. Remote sensing...     

Meteorological applications of remote sensing from satellites

Satellites provide meteorologists with a data source unmatched at comparable spatial and temporal coverage by any existing or practical alternate source. There are limitations, however, both instrumental and fundamental, imposed on the achievable resolution and accuracy. Current and promising future contributions to meteorology from satellite-borne sensors are discussed, with emphasis on performance and the limitations thereto. The discussion covers 1) synoptic meteorology where satellite observations of clouds provide measures of winds, cyclogenesis, and rainfall estimation; 2) atmospheric profiling wherein vertical profiles of temperature, humidity, and certain gaseous constituents are provided; 3) radiation budget or the energy exchange between the earth and the space-sun environment; and 4) surface features of importance to meteorology-temperature, soil moisture, and sea ice coverage. Satellites will be extensively used as data collectors and relays from in situ instruments on buoys, balloons, and fixed earth sites. The accuracy and coverage of such observations, however, will be determined by the in situ sensors and not by the satellite. They are therefore not discussed here.     

YOLOv5-LR：一种遥感影像旋转目标检测模型

真实遥感图像中,目标呈现任意方向分布的特点,原始YOLOv5网络存在难以准确表达目标的位置和范围、以及检测速度一般的问题。针对上述问题,提出一种遥感影像旋转目标检测模型YOLOv5-Left-Rotation,首先利用Transformer自注意力机制,让模型更加注意感兴趣的目标,并且在图像预处理过程中采用Mosaic数据增强,对后处理过程使用改进后的非极大值抑制算法Non-Maximum Suppression。其次,引入角度损失函数,增加网络的输出维度,得到旋转矩形的预测框。最后,在网络模型的浅层阶段,增加滑动窗口分支,来提高大尺寸遥感稀疏目标的检测效率。实验数据集为自制飞机数据集CASIA-plane78和公开的舰船数据集HRSC2016,结果表明,改进旋转目标检测算法相比于原始YOLOv5网络的平均精度提升了3.175%,在吉林一号某星推扫出的大尺寸多光谱影像中推理速度提升了13.6%,能够尽可能地减少冗余背景信息,更加准确检测出光学遥感图像中排列密集、分布无规律的感兴趣目标的区域。     

The military applications of remote sensing by infrared

Remote sensing is the process of acquiring information from the environment by the use of a sensor that is not in physical contact with the object under study. The military services are experienced practitioners of this old, but newly glamorous, art. Their accomplishments in the infrared, that region lying between visible light on the one hand and microwaves on the other, are both impressive and of increasing importance. Our purpose is to provide an overview of these accomplishments. We begin with a brief treatment of the characteristics and peculiarities of the infrared portion of the spectrum and of the sensors that operate there. Early military experience with remote sensing by infrared is described and an applications matrix is developed in order to provide a perspective from which the reader can view the full panorama of military applications. Specific applications ate discussed. These include strategic systems for early warning of intercontinental ballistic missile launches, methods for the detection of atmospheric contaminants, such as poison gas, under field conditions, aids for the precision delivery of weaponry (including passive, active, and laser designator guidance techniques), and sensor systems for reconnaissance and surveillance. Wherever possible, details of sensor performance are given.     

Training techniques for neural network applications in ATM

The main problems of adaptive ATM quality of service (QoS) control methods using neural networks were the exponentially wide range of the output target and the real-time training data sampling. But new practical techniques to overcome these problems may open new neural network applications. In this article, the framework of connection admission control (CAC) is described as a typical example of neural-network-based QoS estimation and two practical techniques, called relative target method and virtual output buffer method, are presented to enhance the neural network performance in CAC     

Content-based network resource allocation for real time remote laboratory applications

This paper presents a practical solution to make remote laboratories a realizable dream. A remote laboratory is an online laboratory where students can get first-hand experience of engineering labs via Internet. Video transmission can provide hands on experience to the user but the transmission channel or networks typically have variable and low bandwidth that poses a tough constraint for such implementation. This work presents a practical solution to such problems by adaptively transmitting the best available quality of laboratory videos to the user depending on network bandwidth. The concept behind our work is that not all objects or frames of the video have equal importance, and thus bandwidth reduction can be accomplished by intelligently transmitting important parts at relatively higher resolution. A localized Time adaptive mean of Gaussian (L-TAMOG) approach is used to search for moving objects which are then allocated network resources dynamically according to the varying network bandwidth variations. Adaptive motion compensated wavelet-based encoding is used to achieve scalability and high compression. The proposed system tracks the network bandwidth and delivers optimally the most important contents of video to the student. Experimental results over several remote laboratory sequences show the efficiency of the proposed framework.     

A self-organizing neural network for supervised learning,recognition, and prediction

Fuzzy ARTMAP, one of a rapidly growing family of attentive self-organizing learning, hypothesis testing, and prediction systems that have evolved from the biological theory of cognitive information processing of which ART forms an important part is discussed. It is shown that this architecture is capable of fast but stable online recognition learning, hypothesis testing and adaptive naming in response to an arbitrary stream of analog or binary input patterns. The fuzzy ARTMAP neural network combines a unique set of computational abilities that are needed to function autonomously in a changing world and that alternative models have not yet achieved. In particular, fuzzy ARTMAP can autonomously learn, recognize, and make predictions about rare events, large nonstationary databases, morphologically variable types of events, and many-to-one and one-to-many relationships. The system's fast learning of rare events and error-based learning and alternatives are described, and uses for ART systems and the development of unsupervised ART systems are reviewed     

Practical methods for rapid and accurate computation ofinterferometric spectra for remote sensing applications

The apodization of an interferogram corresponds to a linear transformation in spectral space between unapodized and apodized radiances. Many apodization functions have well-behaved numerical inverse transformations, and we show an analytic inverse for the Hamming apodization function. The inverse transformation has many practical uses for remote sensing applications and can also be used theoretically to show the equivalence between unapodized spectra and properly apodized spectra. The inverse transformation, which is a representation of the discrete convolution theorem, can be used to readily convert computed apodized spectra to spectra computed for other symmetric apodization functions (including unapodized), which may have poorer characteristics with regard to calculating channel-transmittance parameters or radiances. We also show a quantitative method for comparing apodization functions of different mathematical forms