A Multi-Stream Attention-Aware Convolutional Neural Network: Monitoring of Sand and Dust Storms from Ordinary Urban Surveillance Cameras

被引:3
作者
Wang, Xing [1 ,2 ,3 ,4 ]
Yang, Zhengwei [1 ]
Feng, Huihui [5 ]
Zhao, Jiuwei [6 ]
Shi, Shuaiyi [7 ]
Cheng, Lu [8 ]
机构
[1] Nanjing Univ, Sch Atmospher Sci, Nanjing 210023, Peoples R China
[2] Nanjing Univ Informat Sci & Technol, Minist Educ, Key Lab Meteorol Disaster Minist Educ KLME, Nanjing 210044, Peoples R China
[3] Nanjing Univ InformationScience & Technol, Collaborat Innovat Ctr Forecast & Evaluat Meteorol, Nanjing 210044, Peoples R China
[4] Nanjing Inst Technol, Sch Comp Engn, Nanjing 211167, Peoples R China
[5] Cent South Univ, Sch Geosci & Info Phys, Changsha 410083, Peoples R China
[6] Nanjing Univ Informat Sci & Technol, Sch Atmospher Sci, Nanjing 210044, Peoples R China
[7] Chinese Acad Sci, Aerosp Informat Res Inst, State Key Lab Remote Sensing Sci, Beijing 100094, Peoples R China
[8] Suzhou Univ Sci & Technol, Sch Geog Sci & Geomat Engn, Suzhou 215009, Peoples R China
基金
中国国家自然科学基金;
关键词
sand and dust storm; surveillance camera; deep learning; attention mechanism; IMAGE-ENHANCEMENT; VISION;
D O I
10.3390/rs15215227
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
Sand and dust storm (SDS) weather has caused several severe hazards in many regions worldwide, e.g., environmental pollution, traffic disruptions, and human casualties. Widespread surveillance cameras show great potential for high spatiotemporal resolution SDS observation. This study explores the possibility of employing the surveillance camera as an alternative SDS monitor. Based on SDS image feature analysis, a Multi-Stream Attention-aware Convolutional Neural Network (MA-CNN), which learns SDS image features at different scales through a multi-stream structure and employs an attention mechanism to enhance the detection performance, is constructed for an accurate SDS observation task. Moreover, a dataset with 13,216 images was built to train and test the MA-CNN. Eighteen algorithms, including nine well-known deep learning models and their variants built on an attention mechanism, were used for comparison. The experimental results showed that the MA-CNN achieved an accuracy performance of 0.857 on the training dataset, while this value changed to 0.945, 0.919, and 0.953 in three different real-world scenarios, which is the optimal performance among the compared algorithms. Therefore, surveillance camera-based monitors can effectively observe the occurrence of SDS disasters and provide valuable supplements to existing SDS observation networks.
引用
收藏
页数:19
相关论文
共 39 条
[1]  
Abdulameer F.S., 2019, Adv. Phys. Theor. Appl, V76, P15
[2]   A review of techniques and technologies for sand and dust storm detection [J].
Akhlaq, Muhammad ;
Sheltami, Tarek R. ;
Mouftah, Hussein T. .
REVIEWS IN ENVIRONMENTAL SCIENCE AND BIO-TECHNOLOGY, 2012, 11 (03) :305-322
[3]  
[Anonymous], 2015, Sand and Dust Storm Warning Advisory and Assessment System (Sds-Was) Science and Implementation Plan: 2015-2020
[4]   A General Survey on Attention Mechanisms in Deep Learning [J].
Brauwers, Gianni ;
Frasincar, Flavius .
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2023, 35 (04) :3279-3298
[5]  
Bukhari S.A., 1993, PhD Thesis
[6]  
Chavez P.S., 2002, Arid Lands Newsletter, V51, P1
[7]   Blue Channel and Fusion for Sandstorm Image Enhancement [J].
Cheng, Yaqiao ;
Jia, Zhenhong ;
Lai, Huicheng ;
Yang, Jie ;
Kasabov, Nikola K. .
IEEE ACCESS, 2020, 8 :66931-66940
[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]   The Spatial Variation of Dust Particulate Matter Concentrations during Two Icelandic Dust Storms in 2015 [J].
Dagsson-Waldhauserova, Pavla ;
Magnusdottir, Agnes Osp ;
Olafsson, Haraldur ;
Arnalds, Olafur .
ATMOSPHERE, 2016, 7 (06)
[10]  
Ding Boxheng, 2023, Journal of Ordnance Industry, V44, P3115