-
摘要:
针对泥石流灾害沟谷图像分类问题,文章对Resnet18网络进行改进,提出了一种改进的卷积神经网络模型。通过在网络结构中加入残差注意力模块,解决了原模型提取图像特征较差、边缘模糊的问题,改进后的网络能精确捕捉到泥石流灾害沟谷图像中的轮廓和内部山脊信息。此外,文章还对多种注意力机制结构进行了实验对比,分析其差异性,得出最适合泥石流灾害沟谷数据分类的注意力机制网络。实验表明改进后的网络模型在泥石流灾害沟谷图像的分类准确率达到75.42%,其分类性能在Resnet18网络模型的基础上提升了5.1%。
Abstract:For debris flow disasters in valleys image classification problems, this paper improved the Resnet18 network, an improved convolution neural network model is put forward, through adding residual attention in network structure module, solved the original model to extract the image features to solve the problem of poor, edge model accurately capture the debris flow disasters in valleys in the image contour and internal ridge information.In addition, this paper also conducts comparative experiments on various attention mechanism structures, analyzes their differences, and obtains the attention mechanism network most suitable for debris flow disaster gully data.The experimental results show that the classification accuracy of the improved network model in debris flow disaster gullies reaches 75.42%, and its classification performance is improved by 5.1% compared with the Resnet18 network model.
-
Key words:
- Resnet18 /
- attention mechanism /
- remote sensing image /
- debris flow disaster
-
表 1 Alexnet与Alexnet_CBAM结果对比
Table 1. Comparison of Alexnet and Alexnet CBAM results
模型 特异性/% 灵敏度/% 损失 准确率/% Alexnet 60.37 60.08 0.0394 61.29 Alexnet_CBAM 61.81 62.15 0.0373 63.44 表 2 VGG16与VGG16_CBAM结果对比
Table 2. Comparison of VGG16 and VGG16_CBAM results
模型 特异性/% 灵敏度/% 损失 准确率/% VGG16 61.41 61.52 0.0371 62.72 VGG16_CBAM 60.76 62.86 0.0356 65.23 表 3 Resnet18与Resnet18_CBAM结果对比
Table 3. Comparison of Resnet18 and Resnet18_CBAM results
模型 特异性/% 灵敏度/% 损失 准确率/% Resnet18 87.26 69.88 0.0289 70.32 Resnet18_CBAM 87.99 71.35 0.0261 73.12 表 4 不同注意力机制模块结果对比
Table 4. Comparison of results of different attentional mechanism modules
模型 参数量/106 时间 准确度/% 损失 Resnet18_C 11.27 24 m 17 s 70.56 0.0269 Resnet18_S 11.27 24 m 11 s 71.17 0.0280 Resnet18_CS 11.27 22 m 42 s 73.12 0.0266 Resnet18_SC 11.27 21 m 31 s 75.42 0.0248 -
[1] 孙显辰,王保云,刘坤香,等. 云南省泥石流灾害影响因子分析[J]. 人民长江,2020,51(11):121 − 127. [SUN Xianchen,WANG Baoyun,LIU Kunxiang,et al. Analysis on influencing factors of debris flow disasters in Yunnan Province[J]. Yangtze River,2020,51(11):121 − 127. (in Chinese with English abstract) doi: 10.16232/j.cnki.1001-4179.2020.11.021
[2] 曹禄来,徐林荣,陈舒阳,等. 基于模糊神经网络的泥石流危险性评价[J]. 水文地质工程地质,2014,41(2):143 − 147. [CAO Lulai,XU Linrong,CHEN Shuyang,et al. Assessment of debris flow hazard based on Fuzzy Neutral Network[J]. Hydrogeology & Engineering Geology,2014,41(2):143 − 147. (in Chinese with English abstract)
[3] 汪茜,李广杰,郑百功,等. 基于人工神经网络的磐石市泥石流预测[J]. 人民长江,2009,40(3):46 − 48. [WANG Qian,LI Guangjie,ZHENG Baigong,et al. Prediction of debris flow in Panshi City based on artificial neural network[J]. Yangtze River,2009,40(3):46 − 48. (in Chinese) doi: 10.3969/j.issn.1001-4179.2009.03.018
[4] 吉晓玲. 滑坡与泥石流灾害经济损失的粗糙BP神经网络分析[D]. 成都: 西南财经大学, 2012
JI Xiaoling. Analysis for the economic loss of landslid and debris flow on rough set and BP neural network[D]. Chengdu: Southwestern University of Finance and Economics, 2012. (in Chinese with English abstract)
[5] 赵京霞,钱育蓉,张猛,等. 基于改进的卷积神经网络LeNet-5乳腺疾病诊断方法[J]. 东北师大学报(自然科学版),2019,51(2):65 − 70. [ZHAO Jingxia,QIAN Yurong,ZHANG Meng,et al. Diagnosis of breast disease based on an improved convolution neural network LeNet-5[J]. Journal of Northeast Normal University (Natural Science Edition),2019,51(2):65 − 70. (in Chinese with English abstract)
[6] 石翠萍,谭聪,左江,等. 基于改进AlexNet卷积神经网络的人脸表情识别[J]. 电讯技术,2020,60(9):1005 − 1012. [SHI Cuiping,TAN Cong,ZUO Jiang,et al. Expression recognition based on improved AlexNet convolutional neural network[J]. Telecommunication Engineering,2020,60(9):1005 − 1012. (in Chinese with English abstract) doi: 10.3969/j.issn.1001-893x.2020.09.002
[7] 伍思雨,冯骥. 基于改进VGGNet卷积神经网络的鲜花识别[J]. 重庆师范大学学报(自然科学版),2020,37(4):124 − 131. [WU Siyu,FENG Ji. Flower recognition based on improved VGGnet convolutional neural network[J]. Journal of Chongqing Normal University (Natural Science),2020,37(4):124 − 131. (in Chinese with English abstract)
[8] 吕俊超,潘建平,俞社鑫. 基于循环神经网络的缓坡场地液化侧移预测[J]. 中国安全生产科学技术,2021,17(8):18 − 23. [LYU Junchao,PAN Jianping,YU Shexin. Prediction of liquefaction-induced lateral displacement in gentle slope field based on recurrent neural network[J]. Journal of Safety Science and Technology,2021,17(8):18 − 23. (in Chinese with English abstract)
[9] 田亚军,高静怀,王大兴,等. 基于深度神经网络的地震强反射剥离方法[J]. 地球物理学报,2021,64(8):2780 − 2794. [TIAN Yajun,GAO Jinghuai,WANG Daxing,et al. Removing strong seismic reflection based on the deep neural network[J]. Chinese Journal of Geophysics,2021,64(8):2780 − 2794. (in Chinese with English abstract) doi: 10.6038/cjg2021O0165
[10] 刘志远,杨春峰,孟祥翠,等. 人工神经网络波形分类油气有利区带划分可靠性分析—以旬宜上古生界为例[J]. 石油地质与工程,2021,35(1):34 − 39. [LIU Zhiyuan,YANG Chunfeng,MENG Xiangcui,et al. Reliability analysis of favorable oil and gas zone division by artificial neural network waveform classification—by taking the application of upper Paleozoic in Xunyi as an example[J]. Petroleum Geology and Engineering,2021,35(1):34 − 39. (in Chinese with English abstract) doi: 10.3969/j.issn.1673-8217.2021.01.007
[11] Yya B, Zz C, Wz A, et al. Landslide susceptibility mapping using multiscale sampling strategy and convolutional neural network: A case study in Jiuzhaigou region-ScienceDirect[J]. CATENA, 195.
[12] 张珂,牛杰帆,李曦,等. 洪水预报智能模型在中国半干旱半湿润区的应用对比[J]. 水资源保护,2021,37(1):28 − 35. [ZHANG Ke,NIU Jiefan,LI Xi,et al. Comparison of artificial intelligence flood forecasting models in China's semi-arid and semi-humid regions[J]. Water Resources Protection,2021,37(1):28 − 35. (in Chinese with English abstract) doi: 10.3880/j.issn.1004-6933.2021.01.005
[13] 李谷源,罗毅桦,黎焕明,等. 基于BP神经网络的文山里洪水预报模型[J]. 河南水利与南水北调,2020,49(8):28 − 29. [LI Guyuan,LUO Yihua,LI Huanming,et al. Wenshanli flood forecasting model based on BP neural network[J]. Henan Water Resources & South-to-North Water Diversion,2020,49(8):28 − 29. (in Chinese)
[14] 马洁华,孙建奇,汪君,等. 2018年夏季我国极端降水及滑坡泥石流灾害预测[J]. 大气科学学报,2019,42(1):93 − 99. [MA Jiehua,SUN Jianqi,WANG Jun,et al. Real-time prediction for 2018 JJA extreme precipitation and landslides[J]. Transactions of Atmospheric Sciences,2019,42(1):93 − 99. (in Chinese with English abstract)
[15] 朱巍,蔡贺,唐雯,等. 长白山天池火山泥石流展布特征及其灾害研究[J]. 地质与资源,2017,26(6):608 − 615. [ZHU Wei,CAI He,TANG Wen,et al. The Tianchi volcanic mudflow in Changbai Mountains:its distribution and disaster research[J]. Geology and Resources,2017,26(6):608 − 615. (in Chinese with English abstract) doi: 10.3969/j.issn.1671-1947.2017.06.013
[16] 朱君星,刘蕊. 排土场滑坡与泥石流成因机理分析及临界雨量的确定[J]. 有色金属(矿山部分),2017,69(5):75 − 79. [ZHU Junxing,LIU Rui. Mechanism analysis on the dump slide and debris flow and the determination of critical rainfall intensity[J]. Nonferrous Metals (Mining Section),2017,69(5):75 − 79. (in Chinese with English abstract)
[17] 张贺. 岫岩县泥石流趋势预测预报[J]. 农业科技与装备,2017(3):43 − 44. [ZHANG He. Prediction and forecast of debris flow in Xiuyan Country[J]. Agricultural Science & Technology and Equipment,2017(3):43 − 44. (in Chinese with English abstract) doi: 10.16313/j.cnki.nykjyzb.2017.03.021
[18] 杜星,孙永福,宋玉鹏,等. 基于MPL神经网络的地震作用下砂土液化评估及预测[J]. 工程地质学报,2020,28(6):1425 − 1432. [DU Xing,SUN Yongfu,SONG Yupeng,et al. Multilayer perception neural network for assessment and prediction of earthquake-induced sand liquefaction[J]. Journal of Engineering Geology,2020,28(6):1425 − 1432. (in Chinese with English abstract) doi: 10.13544/j.cnki.jeg.2019-321
[19] 刘诗洋,陈祖煜,张云旆,等. 基于卷积神经网络对TBM塌方段的反演分析[J]. 固体力学学报,2021,42(3):287 − 301. [LIU Shiyang,CHEN Zuyu,ZHANG Yunpei,et al. Back analysis of the TBM collapse section based on convolutional neural networks[J]. Chinese Journal of Solid Mechanics,2021,42(3):287 − 301. (in Chinese with English abstract) doi: 10.19636/j.cnki.cjsm42-1250/o3.2021.011
[20] 于国强,张茂省,王根龙. 泥石流平均流速预测模型及敏感因子研究[J]. 工程地质学报,2014,22(3):355 − 360. [YU Guoqiang,ZHANG Maosheng,WANG Genlong. Prediction model and sensitive factors for average speed of debris flows at Jiangjia gully[J]. Journal of Engineering Geology,2014,22(3):355 − 360. (in Chinese with English abstract) doi: 10.13544/j.cnki.jeg.2014.03.001
[21] 周文辉,石敏,朱登明,等. 基于残差注意力网络的地震数据超分辨率方法[J]. 计算机科学,2021,48(8):24 − 31. [ZHOU Wenhui,SHI Min,ZHU Dengming,et al. Seismic data super-resolution method based on residual attention network[J]. Computer Science,2021,48(8):24 − 31. (in Chinese with English abstract) doi: 10.11896/jsjkx.200900034
[22] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. June 27 − 30, 2016, Las Vegas, NV, USA. IEEE, 2016: 770 − 778.
[23] 乔思波,庞善臣,王敏,等. 基于残差混合注意力机制的脑部CT图像分类卷积神经网络模型[J]. 电子学报,2021,49(5):984 − 991. [QIAO Sibo,PANG Shanchen,WANG Min,et al. A convolutional neural network for brain CT image classification based on residual hybrid attention mechanism[J]. Acta Electronica Sinica,2021,49(5):984 − 991. (in Chinese with English abstract) doi: 10.12263/DZXB.20200881
[24] 季鸿坤. 基于卷积神经网络的阿尔茨海默病MRI影像辅助诊断研究[D]. 长春: 长春工业大学, 2021
JI Hongkun. Research on MRI image-assisted diagnosis of Alzheimer’s disease based on convolutional neural network[D]. Changchun: Changchun University of Technology, 2021. (in Chinese with English abstract)
[25] HU J, SHEN L, SUN G. Squeeze and excitationnet works[C]//Proceeding softhe IEEE conferenceon computer vision and pattern recognition. 2018: 7132 − 7141.
[26] WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]//Proceeding softhe European conference oncomputer vision(ECCV). 2018: 3 − 19.