MLNet: multichannel feature fusion lozenge network for land segmentation

被引:50
作者
Gao, Jiahong [1 ,2 ]
Weng, Liguo [1 ]
Xia, Min [2 ]
Lin, Haifeng [3 ]
机构
[1] Nanjing Univ Informat Sci & Technol, Jiangsu Collaborat Innovat Ctr Atmospher Environm, Nanjing, Peoples R China
[2] Nanjing Univ Informat Sci & Technol, B DAT, Nanjing, Peoples R China
[3] Nanjing Forestry Univ, Coll Informat Sci & Technol, Nanjing, Peoples R China
关键词
land cover; lozenge net; three-sided network; multiresolution segmentation; extract key features; CLASSIFICATION;
D O I
10.1117/1.JRS.16.016513
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
The use of remote sensing images for land cover analysis has broad prospects. At present, the resolution of aerial remote sensing images is getting higher and higher, and the span of time and space is getting larger and larger, therefore segmenting target objects enconter great difficulties. Convolutional neural networks are widely used in many image semantic segmentation tasks, but existing models often use simple accumulation of various convolutional layers or the direct stacking of interfeature reuse of up- and downs ampling, the network very heavy. To improve the accuracy of land cover segmentation, we propose a multichannel feature fusion lozenge network. The multichannel feature fusion lozenge network (MLNet) is a three-sided network composed of three branches: one branch uses different levels of feature indexes to sample to maintain the integrity of high-frequency information; one branch focuses on contextual information and strengthens the compatibility of information within and between classes; and the last branch uses feature integration to filter redundant information based on multiresolution segmentation to extract key features. Compared with FCN, UNet, PSP, and other serial single road computing models, the MLNet, which performs feature fusion after three-way parallelism structure, can significantly improve the accuracy with only small increase in complexity. Experimental results show that the average accuracy of 85.30% is obtained on the land cover data set, which is much higher than that of 82.98% of FCN, 81.87% of UNet, 77.52% of SegNet, and 83.09% of EspNet, which proves the effectiveness of the model. (C) 2022 Society of Photo-Optical Instrumentation Engineers (SPIE)
引用
收藏
页数:19
相关论文
共 50 条
[1]  
[Anonymous], 2015, 1511 ARXIV
[2]   Advances in Remote Sensing of Agriculture: Context Description, Existing Operational Monitoring Systems and Major Information Needs [J].
Atzberger, Clement .
REMOTE SENSING, 2013, 5 (02) :949-981
[3]   SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation [J].
Badrinarayanan, Vijay ;
Kendall, Alex ;
Cipolla, Roberto .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2017, 39 (12) :2481-2495
[4]   Object based image analysis for remote sensing [J].
Blaschke, T. .
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2010, 65 (01) :2-16
[5]  
Bock Sebastian., 2018, CoRR
[6]  
Chen LB, 2017, IEEE INT SYMP NANO, P1, DOI 10.1109/NANOARCH.2017.8053709
[7]   Remote Sensing Image Ship Detection under Complex Sea Conditions Based on Deep Semantic Segmentation [J].
Chen, Yantong ;
Li, Yuyang ;
Wang, Junsheng ;
Chen, Weinan ;
Zhang, Xianzhong .
REMOTE SENSING, 2020, 12 (04)
[8]   Xception: Deep Learning with Depthwise Separable Convolutions [J].
Chollet, Francois .
30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, :1800-1807
[9]  
Culurciello E, 2016, ARXIV160602147
[10]  
Gordon-Rodriguez E., 2020, USES ABUSES CROSS EN