Scene classification of remote sensing image based on deep network grading transferring

Yang, Zhou; Mu, Xiaodong; Zhao, Feng’an

doi:10.1016/j.ijleo.2018.04.092

Cited by 14 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are more and more studies using the deep convolution neural network to classify remote sensing images [68]. Our results have approved that, for big remote sensing data like GF-1 images with high spatial and temporal resolutions, the deep learning method can be used to extract water bodies with accurate results efficiently.…”

Section: Discussionsupporting

confidence: 54%

Water Identification from High-Resolution Remote Sensing Images Based on Multidimensional Densely Connected Convolutional Neural Networks

Wang

Wei

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

The accurate acquisition of water information from remote sensing images has become important in water resources monitoring and protections, and flooding disaster assessment. However, there are significant limitations in the traditionally used index for water body identification. In this study, we have proposed a deep convolutional neural network (CNN), based on the multidimensional densely connected convolutional neural network (DenseNet), for identifying water in the Poyang Lake area. The results from DenseNet were compared with the classical convolutional neural networks (CNNs): ResNet, VGG, SegNet and DeepLab v3+, and also compared with the Normalized Difference Water Index (NDWI). Results have indicated that CNNs are superior to the water index method. Among the five CNNs, the proposed DenseNet requires the shortest training time for model convergence, besides DeepLab v3+. The identification accuracies are evaluated through several error metrics. It is shown that the DenseNet performs much better than the other CNNs and the NDWI method considering the precision of identification results; among those, the NDWI performance is by far the poorest. It is suggested that the DenseNet is much better in distinguishing water from clouds and mountain shadows than other CNNs.

show abstract

Section: Discussionsupporting

confidence: 54%

Water Identification from High-Resolution Remote Sensing Images Based on Multidimensional Densely Connected Convolutional Neural Networks

Wang

Wei

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…A classification structure is built, that checks the abstraction and strength of the network. A combination of deep convolutional neural network (DCNN) and grading transfer is proposed in [5]. In this initially, fine tuning was carried out by applying an already trained dataset to a small dataset.…”

Section: Related Workmentioning

confidence: 99%

Neural Network and Cuckoo Optimization Algorithm for Remote Sensing Image Classification

Janarthanan*¹,

Viswanathan²,

Umamaheswari³

2019

IJRTE

View full text Add to dashboard Cite

Since Remote sensing images are used for a variety of applications, classification of such images is essential for extracting information. However, due to the lack of sufficient training examples, classification of remote sensing images is more complex. Also, in traditional method of classification, the statistical procedure uses only the gray value of the digital images to extract information. Neural network (NN) has a better impact in this domain. The proposed work uses an integrated NN and cuckoo Optimisation Algorithm (COA) for classification. The NN picks, organise and constructs the data to be trained into a network. The network is then trained and tested. The COA is combined with NN to aid the task of classification and to calculate the cost function. The optimization algorithm provides an appropriate feature suitable for the neural network classifier to produce final output. This hybridized technique presents the effective advantages of the neural network and further investigate the classification of the multispectral remote sensible images. The performance and error rate of the system is better when compared with other classical methods. The observed results illustrate the effectiveness of the COA to optimize the cost function. As training parameters in Neural network is found to be NP hard, utilization of COA involve its responsibility in optimizing the learning rate of NN.

show abstract

“…Automatic classification methods have demonstrated high accuracy over satellite imaging (Wright y Gallant, 2007), as well as shorter processing terms and ability to efficiently estimate changes in the coverage compared to conventional techniques (Han et al, 2012); (Munyati, 2000). The use of ANN is one of the classification methods, a model based on a large amount of training data which later can predict to what kind each pixel belongs, achieving accuracies up to 93.4% as in the case of Yang et al (2018). Han et al, (2012) compared different variations of the ANN method, among which are the radial-based neural networks with those that obtained an accuracy of 48 to 86%, and also compared the ANNs to RF, in which reached an accuracy of 96%.…”

Section: Introductionmentioning

confidence: 99%

Coverage Changes Detection at Ciénaga Grande, Santa Marta – Colombia Using Automatic Classification

Vinasco

Velásquez

Quintero

et al. 2020

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. The Ciénaga Grande, Santa Marta is the largest and most diverse ecosystem of its kind in Colombia. Its primary function is acting as a filter for the organic carbon cycle. Recently, this place has been suffering disruptions due to the anthropic activities taking place in its surroundings. The present study, the changes in the surface of Ciénaga Grande, Santa Marta, Magdalena, Colombia between 2013 and 2018 were determined using semiautomatic detection methods with high resolution data from remote sensors (Landsat 8). The zone of studies was classified in six kinds of surfaces: 1) artificial territories, 2) agricultural territories, 3) forests and semi-natural areas, 4) wet areas, 5) deep water surfaces & 6) wich is related to clouds as a masking method. Random Forest classifiers were utilized and the Feed For Ward multilayer perceptron neuronal network (ANN) was simultaneously assessed. The training stage for both methods was performed with 300 samples, distributed in equal quantities, over each coverage class. The semi-automatic classification was carried out with an annual frequency, but the monitoring was carried out throughout the analysis period through the performance of three indicators Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI) and Normalized Difference Water Index (NDWI). It was found from the confusion matrix that the Random Forest method more accurately classified four classes while Neural Networks Analysis (NNA) just three. Finally, taking the Random Forest results into account, it was found that the agricultural expansion increased from 7% to 9% and the urban zone increased from 20% to 30% of the total area. As well as a decrease of damp areas from 27% to 12% and forests from 4% to 3% of the total area of study.

show abstract

Scene classification of remote sensing image based on deep network grading transferring

Cited by 14 publications

References 11 publications

Water Identification from High-Resolution Remote Sensing Images Based on Multidimensional Densely Connected Convolutional Neural Networks

Water Identification from High-Resolution Remote Sensing Images Based on Multidimensional Densely Connected Convolutional Neural Networks

Neural Network and Cuckoo Optimization Algorithm for Remote Sensing Image Classification

Coverage Changes Detection at Ciénaga Grande, Santa Marta – Colombia Using Automatic Classification

Contact Info

Product

Resources

About