Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation
Recent developments in Convolutional Neural Networks (CNNs) have allowed for the achievement of solid advances in semantic segmentation of high-resolution remote sensing (HRRS) images. Nevertheless, the problems of poor classification of small objects and unclear boundaries caused by the characteris...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2019-05-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/9/9/1816 |
| _version_ | 1819117125937135616 |
|---|---|
| author | Guangsheng Chen Chao Li Wei Wei Weipeng Jing Marcin Woźniak Tomas Blažauskas Robertas Damaševičius |
| author_facet | Guangsheng Chen Chao Li Wei Wei Weipeng Jing Marcin Woźniak Tomas Blažauskas Robertas Damaševičius |
| author_sort | Guangsheng Chen |
| collection | DOAJ |
| description | Recent developments in Convolutional Neural Networks (CNNs) have allowed for the achievement of solid advances in semantic segmentation of high-resolution remote sensing (HRRS) images. Nevertheless, the problems of poor classification of small objects and unclear boundaries caused by the characteristics of the HRRS image data have not been fully considered by previous works. To tackle these challenging problems, we propose an improved semantic segmentation neural network, which adopts dilated convolution, a fully connected (FC) fusion path and pre-trained encoder for the semantic segmentation task of HRRS imagery. The network is built with the computationally-efficient DeepLabv3 architecture, with added Augmented Atrous Spatial Pyramid Pool and FC Fusion Path layers. Dilated convolution enlarges the receptive field of feature points without decreasing the feature map resolution. The improved neural network architecture enhances HRRS image segmentation, reaching the classification accuracy of 91%, and the precision of recognition of small objects is improved. The applicability of the improved model to the remote sensing image segmentation task is verified. |
| first_indexed | 2024-12-22T05:28:01Z |
| format | Article |
| id | doaj.art-4f52c61fe35f4a27a7f88bb4bd6c4bfe |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-12-22T05:28:01Z |
| publishDate | 2019-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-4f52c61fe35f4a27a7f88bb4bd6c4bfe2022-12-21T18:37:32ZengMDPI AGApplied Sciences2076-34172019-05-0199181610.3390/app9091816app9091816Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image SegmentationGuangsheng Chen0Chao Li1Wei Wei2Weipeng Jing3Marcin Woźniak4Tomas Blažauskas5Robertas Damaševičius6College of Information Science and Technology, North East Forest University, Harbin 150040, ChinaCollege of Information Science and Technology, North East Forest University, Harbin 150040, ChinaCollege of Computer Science and Engineering, Xi’an University of Technology, Xi’an 710048, ChinaCollege of Information Science and Technology, North East Forest University, Harbin 150040, ChinaInstitute of Mathematics, Silesian University of Technology, 44-100 Gliwice, PolandDepartment of Software Engineering, Kaunas University of Technology, 51386 Kaunas, LithuaniaDepartment of Software Engineering, Kaunas University of Technology, 51386 Kaunas, LithuaniaRecent developments in Convolutional Neural Networks (CNNs) have allowed for the achievement of solid advances in semantic segmentation of high-resolution remote sensing (HRRS) images. Nevertheless, the problems of poor classification of small objects and unclear boundaries caused by the characteristics of the HRRS image data have not been fully considered by previous works. To tackle these challenging problems, we propose an improved semantic segmentation neural network, which adopts dilated convolution, a fully connected (FC) fusion path and pre-trained encoder for the semantic segmentation task of HRRS imagery. The network is built with the computationally-efficient DeepLabv3 architecture, with added Augmented Atrous Spatial Pyramid Pool and FC Fusion Path layers. Dilated convolution enlarges the receptive field of feature points without decreasing the feature map resolution. The improved neural network architecture enhances HRRS image segmentation, reaching the classification accuracy of 91%, and the precision of recognition of small objects is improved. The applicability of the improved model to the remote sensing image segmentation task is verified.https://www.mdpi.com/2076-3417/9/9/1816semantic segmentationremote sensingdilated convolutionfully convolutional neural networkdeep learning |
| spellingShingle | Guangsheng Chen Chao Li Wei Wei Weipeng Jing Marcin Woźniak Tomas Blažauskas Robertas Damaševičius Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation Applied Sciences semantic segmentation remote sensing dilated convolution fully convolutional neural network deep learning |
| title | Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation |
| title_full | Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation |
| title_fullStr | Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation |
| title_full_unstemmed | Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation |
| title_short | Fully Convolutional Neural Network with Augmented Atrous Spatial Pyramid Pool and Fully Connected Fusion Path for High Resolution Remote Sensing Image Segmentation |
| title_sort | fully convolutional neural network with augmented atrous spatial pyramid pool and fully connected fusion path for high resolution remote sensing image segmentation |
| topic | semantic segmentation remote sensing dilated convolution fully convolutional neural network deep learning |
| url | https://www.mdpi.com/2076-3417/9/9/1816 |
| work_keys_str_mv | AT guangshengchen fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation AT chaoli fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation AT weiwei fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation AT weipengjing fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation AT marcinwozniak fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation AT tomasblazauskas fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation AT robertasdamasevicius fullyconvolutionalneuralnetworkwithaugmentedatrousspatialpyramidpoolandfullyconnectedfusionpathforhighresolutionremotesensingimagesegmentation |