APPLICATION OF NEURAL NETWORK TO CLASSIFICATION OF REMOTE SENSING IMAGE

陈玉敏

DOI: https://doi.org/10.3969/j.issn.1007-3817.2002.03.003

2002-01-01

Abstract:This paper describes the BP neural network approach to the classification of remote sensing images and develops an improved method to train the selected data.

What problem does this paper attempt to address?

Neural Network Classification with Optimization by Genetic Algorithms for Remote Sensing Imagery

Xiaohua Tong,Xue Zhang

DOI: https://doi.org/10.1117/12.760763

2007-01-01

Abstract:On the remote sensing imagery classification, the traditional methods based on statistical principle has the difficulties in distinguishing the objects with similar spectral characteristics, while the back propagation neural network method has the C difficulties in sufficiency and convergence. Therefore, a new method based on neural network classification with optimization by genetic algorithms for remote sensing imagery is proposed in this paper. On the basis of the back propagation( BP)neural network classification, the optimization method by genetic algorithms is presented, including the numbers, the thresholds and the connection weights of nerve nodes of the hide layer in BP neural network. An approach on float coding with alterative length for genetic algorithms is proposed, and the evolution method is improved to obtain an optimal BP neural network. In the end, an experimental test on the remote sensing classification using TM image of Dianshan Lake is carried out, and higher classification accuracy has obtained compared to other methods, which is proved the feasibility and validity of the proposed approach.
Remote Sensing Image Classification Based on Evidence Theory and Neural Networks

Gang Chen,Bi-cheng Li,Zhi-gang Guo

DOI: https://doi.org/10.1007/978-3-540-28647-9_160

2004-01-01

Abstract:An effective method based on evidence theory and neural networks to classify the remote sensing images is brought forward in the paper. Firstly, with the spatial information in consideration, the original image is smoothed with a modified gradient inverse weighting smoothing method, then the classification of the original and smoothed images is performed separately using a BP neural network. Finally, result comes out after fusing the two classification results with D-S evidence theory. Experimental results demonstrate that the proposed method is effective and can improve the classification accuracy.
Image classification of remote sensing based on BP neural networks

LI Guang-lu

2012-01-01

Abstract:Remote sensing image classification technology is an important segment of image analysis and interpretation.B PANN has the characteristic of faster convergence speed and stronger capabilities of self-study and self-adaptation,it's able to utilize the priori knowledge of data sets furthest to automatically determine reasonable models,and model prediction results can be realistic to reflect the real surface features.This paper adopted Landsat TM images for data sources,Dingxiang country of Shanxi province was taken as the research area,principal component analysis was used to compress the data,a comprehensive model of B PANN integrated NDVI with texture features was adopted to classify regional land use information,finally,the result was compared with the classification based on the method of spectrum cell information and Texture Features-based B PANN separately.The result showed that the accuracy of this method is 80.50%,which increases by18.89% compared with spectrum cell information based ANN classification,and 6.23% compared with texture features based ANN classification separately.It could effectively solve the problems of spectral confusion and low classification accuracy.
Research On Classification Of Remote Sensing Image Based On Grey Level Co-Occurrence Matrix And Bp Neural Network

Yuping Jiao,Zhongyang Guo,Shufeng Ye

2014-01-01

Abstract:To improve the classification accuracy of remote sensing image, this research adopted grey level co-occurrence matrix (GLCM) method to extract the textures from the image, and the optimal texture features were screened through a large number of experiments. Then a classification method of BP neural network based on the math software MATLAB was realized. Finally this paper achieved the precision result of 94.78% by combining spectrum features and texture features as the input variables of the neural network. The result of this paper shows that textures of image can significantly improve the classification accuracy of remote sensing image, so it can be applied into other researches for better effect.
BP neural network based on principle component analysis in multi-spectral remote sensing images classification

HU Jian-ce,WU Guo-ping

2009-01-01

Abstract:In order to improve the classification accuracy of multi-spectral remote sensing image,this paper puts forward a new classification method based on principal component analysis.The method is consisted of two steps: reducing the dimensions of multi-spectral remote sensing image with principle component analysis and generating a new image by the three main components of the remote sensing image;performing supervised classification on the new image with BP neural network.The result indicates that this method is superior to traditional algorithms,and its overall accuracy and Kappa coefficient reach 88.5% and 0.862.
Application of Back Propagation Neural Network in the Classification of High Resolution Remote Sensing Image: Take Remote Sensing Image of Beijing for Instance

Jiefeng Jiang,Jing Zhang,Gege Yang,Dapeng Zhang,Lianjun Zhang

DOI: https://doi.org/10.1109/geoinformatics.2010.5568228

2010-01-01

Abstract:In recent years, the development of high-resolution remote sensing image extends the visual field of the terrain features. Quickbird and other high-resolution remote sensing image can show more characteristics such as shape, spectral, texture and so on. Back Propagation neural network is widely used in remote sensing image classification in recent years, it is a self-adaptive dynamical system which is widely connected by large amount of neural units, and it bases on distributing store and parallel processing. It study by exercise and had the capacity of integrating the information, synthesis reasoning, and rapid overall processing capacity. It can solve the regular problem arise from remote sensing image processing, therefore, it is widely used in the application of remote sensing. This paper discusses the Back Propagation neural network method in order to improve the high resolution remote sensing image classification precision. By analyzing the principle and learning algorithms of Back Propagation neural network, we utilize the Quickbird imagery of Beijing with high resolution as experimental data and do the research of road and simple building roof, In this paper, the use of remote sensing image processing software Matlab, and then combined with Back Propagation neural network classifier for the high resolution remote sensing images of their pattern recognition.
The Application of Improved Fuzzy ARTMAP Neural Network in Remote Sensing Classification of Land-use

Yuan Yanbin,Xiong Xianxiao,Zhan Yunjun,Liang Xiao,Zhang Fan,Zhang Xiaopan

DOI: https://doi.org/10.1109/MMIT.2010.28

2010-01-01

Abstract:Land-use change is an important area of global change research, rapid and accurate access to land-use temporal and spatial variation information is a key technology to study land-use change. In this paper proposed a method which utilizes the improved model of fuzzy ARTMAP network - simplified fuzzy ARTMAP neural network for remote sensing land-use classification, and Take the TM remote sensing image of Yiwu as an example to experiment, we compared the classification results with the traditional BP neural network classification results. Tests showed that the improved ARTMAP neural network improved the Accuracy of misclassification; it also shows that the structure of the improved fuzzy ARTMAP network is simple and need less training time. The Simplified Fuzzy ARTMAP network is an effective model to deal with high dimensional remote sensing image classification.
A New Robust Approach for Remote Sensing Image Regional Classification

CHEN Yi-chao,HONG Han-yu,WANG Jun,ZHANG Jian,WANG Cheng-gang

DOI: https://doi.org/10.3969/j.issn.1673-629x.2007.08.040

2007-01-01

Abstract:The main problem of remote sensing image classification is the contradiction of classification precision and algorithm complexity,and algorithm lacking of robust.Therefore,a multi-model robust approach of remote sensing image classification based on non-parameter kernel density estimation of resampling strategy in feature space and edge detection is proposed in this paper.The edge gradient and direction information are obtained by edge detection of remote sensing.Then the new samples sets are weighted mean shift filtering to find kernel density function local maximum of image each region using resampling strategy in the joint spatial-range domain and data points are shifted the local maximum by iterative shifting.Last,the classification image is obtained by combining each region.Experimental results illustrate that it is able to classify remote sensing image effectively and robustly.
Improved Attention Mechanism and Residual Network for Remote Sensing Image Scene Classification

Jiayuan Kong,Yurong Gao,Yanjun Zhang,Huimin Lei,Yao Wang,Hesheng Zhang

DOI: https://doi.org/10.1109/access.2021.3116968

IF: 3.9

2021-01-01

IEEE Access

Abstract:Aiming at the problem that the accuracy of traditional remote sensing image classification model is not ideal, a classification method based on improved attention mechanism and residual network is proposed. In order to prevent overfitting, we choose the network structure of ResNet18 as the framework. Meanwhile, ResNet18 has a small number of parameters and a fast calculation speed. Then, a parallel attention mechanism, ProCBAM, is proposed and added to the BasicBlock of the residual network. By optimizing the representation of the feature map from the spatial dimension and channel dimension of the feature map, more detailed image information is learned and image recognition errors are reduced. The experimental results on the open source dataset show that the accuracy is 93.86%, and a more accurate image classification model is successfully trained.

computer science, information systems,telecommunications,engineering, electrical & electronic
Remote Sensing Images Classification in the City Based on Primary Component Analysis and Fuzzy Neural Network

LH Xu,WZ Yue,WQ Zhang

DOI: https://doi.org/10.1109/igarss.2005.1525400

2005-01-01

Abstract:This paper deals with the using of Fuzzy Neural Networks (FNNs) to classify the Land-sat 7 image. To raise the speed of classification and solve the redundancy problem, the input neural cells are the next three component of Primary Component analysis, and the NDVI image which used of the planetary albedo to calculate. The precision of evaluation shows that Kappa value of classification is 0.865.
A Study of Image Classification of Remote Sensing Based on Back-Propagation Neural Network with Extended Delta Bar Delta

Shi Liang Zhang,Ting Cheng Chang

DOI: https://doi.org/10.1155/2015/178598

IF: 1.43

2015-01-01

Mathematical Problems in Engineering

Abstract:This paper proposes a model to extract feature information quickly and accurately identifying what cannot be achieved through traditional methods of remote sensing image classification. First, process the selected Landsat-8 remote sensing data, including radiometric calibration, geometric correction, optimal band combination, and image cropping. Add the processed remote sensing image to the normalized geographic auxiliary information, digital elevation model (DEM), and normalized difference vegetation index (NDVI), working together to build a neural network that consists of three levels based on the structure of back-propagation neural and extended delta bar delta (BPN-EDBD) algorithm, determining the parameters of the neural network to constitute a good classification model. Then determine classification and standards via field surveys and related geographic information; select training samples BPN-EDBD for algorithm learning and training and, if necessary, revise and improve its parameters using the BPN-EDBD classification algorithm to classify the remote sensing image after pretreatment and DEM data and NDVI as input parameters and output classification results, and run accuracy assessment. Finally, compare with traditional supervised classification algorithms, while adding different auxiliary geographic information to compare classification results to study the advantages and disadvantages of BPN-EDBD classification algorithm.

engineering, multidisciplinary,mathematics, interdisciplinary applications
A new ART neural networks for remote sensing image classification

AnFei Liu,BiCheng Li,Gang Chen,Xianfei Zhang

DOI: https://doi.org/10.1007/11539117_6

2005-01-01

Abstract:A new ART2A-C algorithm based on fuzzy operators to cluster the remote sensing images and aerials is proposed in this paper. By combining two ART ANNs with higher performance, the traditional ART2A-C is developed with the fuzzy operators introduced in matching rule. Then the proposed method is applied to the classification and the new network is implemented as well as two other existed ARTs respectively. Experimental results show that the new method outperforms the traditional ones.
Optimization by genetic algorithms for remote sensing imagery

Xiaohua Tong,Xue Zhang,Miaolong Liu

2008-01-01

Abstract:This paper presents a new method based on genetic algorithms (GA) optimized back propagation neural network classification for remote sensing imagery, which overcomes the defects that traditional approaches based on statistical principle have difficulties in distinguishing the objects with similar spectral characteristics, and the back propagation (BP) neural network method has difficulties in sufficiency and convergence. On the basis of the typical back propagation neural network classification, the optimization for the structure of neural network by genetic algorithms is presented, including the thresholds and connection weights of neural network and the number of neural nodes in the hidden layer of neural network. A coding method on float coding with variable length for genetic algorithms is then introduced, and the evolution method is improved to obtain an optimal back propagation neural network. In the end, an experimental test on remote sensing classification by using Landsat thematic mapper (TM) imagery in Dianshan Lake is conducted. The classification with the proposed approach is the most accurate, which proves its feasibility and validity.
Cross-view Matching Neural Network for Remote Sensing Images.

Tian Hui,Xiaoqing Chen,Ronggang Zhu,YueLei Xu,Zhaoxiang Zhang

DOI: https://doi.org/10.1109/smartcloud52277.2021.00031

2021-01-01

Abstract:In the research of remote sensing image processing, cross-view image matching has always been a researching focus. Due to the changing views of the sensor's observation, the objects feature often have large differences from different views, which lead to the failure of matching method based on image descriptors. Therefore, cross-view remote sensing image matching is a challenging work. We proposed a two-step cross-view image matching method based on deep neural networks: First, the pre-trained Faster RCNN object detection model is used to detect the drone images (search images) and we will get the targets region images through the detection results. Then, the optimized Siamese neural network model is used between the detection results (region images) and the satellite images (template images) to get the results of matching confidence. In this paper, comparative experiments were implemented on the university-1652 dataset, and the experimental results show that our method has excellent results on this benchmark database. The effectiveness of proposed method was demonstrated by evaluated on varying images, the R@10/%, and AP are 88.02% and 67.23% repectively.
BP Neural Network Applied Research in the Air Environment Quality Assessment

Zhenghong Deng

2009-01-01

Abstract:According to ANN theory and method,a BP neural network model for city atmospheric environment quality evaluation was built.The result shows that the application of BP neural network in city atmospheric environment quality evaluation is feasible.This model possesses strong functions of study,association and fault tolerance,moreover,both its analysis results and process approach the metal process and analysis method of human brain,which greatly improves the accuracy for city environment quality evaluation.
REMOTE SENSING IMAGE CLASSIFICATION WITH THE SEN12MS DATASET

M. Schmitt,Y.-L. Wu

DOI: https://doi.org/10.5194/isprs-annals-v-2-2021-101-2021

2021-06-17

Abstract:Abstract. Image classification is one of the main drivers of the rapid developments in deep learning with convolutional neural networks for computer vision. So is the analogous task of scene classification in remote sensing. However, in contrast to the computer vision community that has long been using well-established, large-scale standard datasets to train and benchmark high-capacity models, the remote sensing community still largely relies on relatively small and often application-dependend datasets, thus lacking comparability. With this paper, we present a classification-oriented conversion of the SEN12MS dataset. Using that, we provide results for several baseline models based on two standard CNN architectures and different input data configurations. Our results support the benchmarking of remote sensing image classification and provide insights to the benefit of multi-spectral data and multi-sensor data fusion over conventional RGB imagery.
Improved convolutional neural network in remote sensing image classification

Binghui Xu

DOI: https://doi.org/10.1007/s00521-020-04931-6

2020-04-25

Neural Computing and Applications

Abstract:The classification of land cover is the first step in the analysis and application of remote sensing data in land resources. How to solve the multi-category image recognition and meet certain precision is a key issue in remote sensing image research, which has very important theoretical significance and practical application value. In this study, the algorithm is improved on the basis of convolutional neural network, and experiments are carried out on multi-source remote sensing images with different geomorphologies taken under three different weather conditions to verify the effectiveness and scalability of the improved convolutional neural network. The research results show that the improved algorithm proposed in this paper has certain results in remote sensing image classification and can provide theoretical reference for subsequent related research.

computer science, artificial intelligence
Study on the Design of Remote Image Classifier with HighResolution and Multi-spectrum Based on Neural Net

Qian-xiang ZHOU,Zhong-liang JING

DOI: https://doi.org/10.3321/j.issn:1000-1328.2005.z1.026

2005-01-01

Abstract:As High resolution and multi-spectrum remote image is of complex objects judgement characters, it will be of no ideal results with general supervised or unsupervised classifiers. A BP neural net classifiers has been put forward, which combines the texture and spectral features of object. This method is applied on the classification of some area image of Shanghai. In addition, it is also compared with K-means unsupervised and minimum distance classifier on ENVI/IDL platform. Results show that the method take the image spectral features into consideration completely and can improve the precision of image classification effectively.
The Study of Improved CP Neural Network for Remote Sensing Classification

LI Hong

DOI: https://doi.org/10.3969/j.issn.1004-0323.2000.04.003

2000-01-01

Abstract:The study improves the CP neural network including the weight vectors initial， network learning， the competitive nodes dynamic allocation and the vigilance parameter design etc. It applied to the remote sensing imagery classification. The learning times influence the node numbers and learning accuracy. According its raise， the learning accuracy increases a limit degree. The research boundary is decided by the vigilance parameter ， which decreases with the learning process in order to get the high accuracy and the stability of networks. The two-layer networks improves the overall accuracy by 1.1 ％， Kappa coefficient by 0. 02， which compared to the single-layer networks. It is concluded the improved CP is another classification method for remote sensing.

APPLICATION OF NEURAL NETWORK TO CLASSIFICATION OF REMOTE SENSING IMAGE

Neural Network Classification with Optimization by Genetic Algorithms for Remote Sensing Imagery

Remote Sensing Image Classification Based on Evidence Theory and Neural Networks

Image classification of remote sensing based on BP neural networks

Research On Classification Of Remote Sensing Image Based On Grey Level Co-Occurrence Matrix And Bp Neural Network

BP neural network based on principle component analysis in multi-spectral remote sensing images classification

Application of Back Propagation Neural Network in the Classification of High Resolution Remote Sensing Image: Take Remote Sensing Image of Beijing for Instance

The Application of Improved Fuzzy ARTMAP Neural Network in Remote Sensing Classification of Land-use

A New Robust Approach for Remote Sensing Image Regional Classification

Improved Attention Mechanism and Residual Network for Remote Sensing Image Scene Classification

Remote Sensing Images Classification in the City Based on Primary Component Analysis and Fuzzy Neural Network

A Study of Image Classification of Remote Sensing Based on Back-Propagation Neural Network with Extended Delta Bar Delta

A new ART neural networks for remote sensing image classification

Optimization by genetic algorithms for remote sensing imagery

Cross-view Matching Neural Network for Remote Sensing Images.

BP Neural Network Applied Research in the Air Environment Quality Assessment

REMOTE SENSING IMAGE CLASSIFICATION WITH THE SEN12MS DATASET

Improved convolutional neural network in remote sensing image classification

Study on the Design of Remote Image Classifier with HighResolution and Multi-spectrum Based on Neural Net

The Study of Improved CP Neural Network for Remote Sensing Classification