Please wait a minute...
img

官方微信

遥感技术与应用  2021, Vol. 36 Issue (2): 314-323    DOI: 10.11873/j.issn.1004-0323.2021.2.0314
CNN 专栏     
基于迁移学习再训练模型和高分遥感数据的建筑垃圾自动识别方法
祝一诺1(),高婷1,王术东1,周磊1,2(),杜明义1,2
1.北京建筑大学 测绘与城市空间信息学院,北京 102616
2.北京建筑大学 北京未来城市设计高精尖创新中心,北京 100044
Automatic Recognition Method of Construction Waste based on Transfer Learning and Retraining Model and High-score Remote Sensing Data
Yinuo Zhu1(),Ting Gao1,Shudong Wang1,Lei Zhou1,2(),Mingyi Du1,2
1.School of Geomatics and Urban Spatial Informatics,Beijing University of Civil Engineering and Architecture,Beijing 102616,China
2.Beijing Advanced Innovation Center for Future Urban Design,Beijing University of Civil Engineering and Architecture,Beijing 100044,China
 全文: PDF(4748 KB)   HTML
摘要:

目前城市建筑垃圾大量持续产生且堆积严重,利用率较低同时危害城市生态环境。建筑垃圾的识别是实现建筑垃圾分割、提取以及监测的技术基础,但由于建筑垃圾本身的复杂特征和遥感影像的尺度差异、光谱差异等因素导致其识别和监管困难。提出了一种利用迁移学习再训练模型来实现自动识别建筑垃圾的方法。首先根据建筑垃圾的典型遥感特征构建样本库,样本库包含30 292张建筑垃圾和110 110张典型地物在内的共计140 402张样本。之后基于国际先进的深度学习环境Tensorflow,利用迁移学习在模型的最后一层重新输入了建筑垃圾等6类训练数据集,对Inception-V3模型进行了再训练,在较短时间内得到了建筑垃圾识别模型。随机抽取6 016张样本构成验证集逐个输入建筑垃圾识别模型,统计验证样本的模型识别结果构成混淆矩阵,得出该模型对所有地物的整体识别率K为97.43%, Kappa系数Ka为0.96,模型识别建筑垃圾的识别精确度Pv为99.10%,识别灵敏度为94.88%。与传统的航片监测、实地考察等纯人工识别方法相比,该方法所需时间较短且识别精度较高,有利于实现建筑垃圾的全过程实时监控和精准管理。

关键词: 高分遥感影像建筑垃圾迁移学习自动识别Inception?V3    
Abstract:

At present, a quantity of urban construction waste is constantly produced and seriously accumulated, and its utilization rate is low, which endanger the urban ecological environment. The recognition of construction waste is the technical basis for the segmentation, extraction and monitoring of construction waste. However, it is difficult to identify and monitor construction waste due to its complex characteristics, the scale difference and spectral difference of remote sensing image. In this paper, a method of automatic identification of construction waste based on transfer learning and retraining model is proposed. Firstly, a sample bank is constructed according to the typical remote sensing features of construction waste. Then, based on the advanced international deep learning environment Tensorflow, the Inception-V3 model is retrained by using transfer learning, and the recognition model of construction waste is obtained. After verification, the overall recognition accuracy of construction waste can reach 94.88%. Compared with the traditional manual identification methods such as aerial photo monitoring and field investigation, the method studied in this paper has higher efficiency and recognition accuracy, which can provide a technical basis for real-time monitoring and accurate management of construction waste in the whole process.

Key words: High-resolution remote sensing image    Construction waste    Transfer learning    Automatic recognition    Inception-V3
收稿日期: 2019-09-27 出版日期: 2021-05-24
ZTFLH:  TP75  
基金资助: 国家重点研发计划课题(2018YFC0706003);北京市教委科技计划项目(KM201810016014)
通讯作者: 周磊     E-mail: zhuyanruo7@163.com;zhoulei@bucea.edu.cn
作者简介: 祝一诺(1999-),女,河南安阳人,本科生,主要从事机器学习以及环境遥感研究。E?mail:zhuyanruo7@163.com
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章  
祝一诺
高婷
王术东
周磊
杜明义

引用本文:

祝一诺,高婷,王术东,周磊,杜明义. 基于迁移学习再训练模型和高分遥感数据的建筑垃圾自动识别方法[J]. 遥感技术与应用, 2021, 36(2): 314-323.

Yinuo Zhu,Ting Gao,Shudong Wang,Lei Zhou,Mingyi Du. Automatic Recognition Method of Construction Waste based on Transfer Learning and Retraining Model and High-score Remote Sensing Data. Remote Sensing Technology and Application, 2021, 36(2): 314-323.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2021.2.0314        http://www.rsta.ac.cn/CN/Y2021/V36/I2/314

图1  解译标志库部分训练样本
图2  处理前后遥感影像对比图
图3  融合流程图
图4  建筑垃圾样本类型示例
图5  数据增强前后各类地物数据量对比图
图6  数据增强结果(以建筑垃圾和房屋为例)
图7  卷积神经网络基本结构图
参数数值
迭代次数4 000
初始学习率0.01
训练批量大小100
测试批量大小-1
验证批量大小100
测试集比例10%
验证集比例10%
评估间隔10
表1  Inception-V3网络结构参数设置
图8  建筑垃圾识别模型训练曲线
图9  建筑垃圾自动识别技术流程图
真实值
建筑垃圾水体植被道路裸地房屋
预测值建筑垃圾2 871000179
水体1158325100
植被56573000
道路601592010
裸地9911105832
房屋340070579
表2  模型识别效果混淆矩阵

真实值

预测值

建筑垃圾非建筑垃圾
建筑垃圾2 87126
非建筑垃圾1552 974
表3  模型识别效果二分类矩阵
图10  错分典型图片示例
1 Ministry of Housing and Urban Rural Development of the People's Republic of China. Load Code for Building Structures [S]. Beijing: China Architecture & Building Press, 2012. [中华人民共和国住房和城乡建设部. 建筑结构荷载规范[S]. 北京: 中国建筑工业出版社, 2012.]
2 Strategic Alliance for Technological Innovation of China's Construction Waste Resource Industry. Report on the Development of China's Construction Waste Resource Industry[R]. Beijing: 2014.中国建筑垃圾资源化产业技术创新战略联盟. 中国建筑垃圾资源化产业发展报告[R]. 北京: 2014.
3 Zhao Jun, Liu Qiuxia, Lin Liqing, et.al. Evolution and Comparison of Construction Waste Generation Characteristics in Big Cities[J]. Journal of Central South University(Natural Science Edition), 2013, 44(3):1297-1304.
3 赵军,刘秋霞,林立清, 等.大城市建筑垃圾产生特征演变及比较[J]. 中南大学学报(自然科学版), 2013, 44(3): 1297-1304.
4 Lu Zhongyi. Exploration on the Utilization of Urban Construction Waste [J]. Shanxi Architecture,2019, 45(6): 183-184.
4 卢忠义.城市建筑垃圾资源化利用探讨[J].山西建筑,2019,45(6):183-184.
5 Yang X N. Study on Environmental Impact Assessment of Construction Waste[J]. Chemical Intermediate, 2015, 11(10): 109-110.
5 杨小宁. 关于建筑垃圾的环境影响评价研究[J]. 化工中间体, 2015, 11(10): 109-110.
6 Li Yilong. Brief Analysis of Environmental Pollution Hazards and Management Measures of Construction Waste under the Background of Urbanization[J]. Green Building Materials, 2018(3): 19-20.
6 李熠茏. 浅析城镇化背景下建筑垃圾对于环境的污染危害及管理措施[J]. 绿色环保建材, 2018(3): 19-20.
7 He Jiayan. Beijing Speeds up the Recycling of Construction Waste [J].Invest in Beijing, 2018(6): 35-36.
7 何佳艳. 北京提速建筑垃圾资源化处理[J]. 投资北京, 2018(6): 35-36.
8 Zhikang B, Weisheng L. Developing Efficient Circularity for Construction and Demolition Waste Management in Fast Emerging Economies: Lessons Learned from Shenzhen, China[J]. Science of the Total Environment, 2020, 724. doi:10.1016/j.scitotenv.2020.138264.
doi: 10.1016/j.scitotenv.2020.138264
9 Bagheri S, Hordon R M. Hazardous Waste Site Identification Using Aerial Photography: A Pilot Study in Burlington County, New Jersey, USA[J]. Environmental Management, 1988, 12(3): 411-412. doi: 10.1007/BF01867531.
doi: 10.1007/BF01867531
10 Wu Wenwei, Liu Jing. Application of Remote Sensing Technology in the Investigation of Solid Waste Distribution in Beijing[J]. Environmental Health Engineering, 2000(2): 76-78.
10 吴文伟,刘竞. 北京市固体废弃物分布调查中遥感技术的应用[J]. 环境卫生工程, 2000(2): 76-78.
11 Liu Yalan, Ren Yuhuan, Wei Chengjie, et al. Applied Research on the Monitoring of Irregular Garbage Dumps by Beijing No.1 Small Satellite[J]. Journal of Remote Sensing, 2009, 13(2): 320-326.
11 刘亚岚, 任玉环, 魏成阶,等. 北京1号小卫星监测非正规垃圾场的应用研究[J]. 遥感学报, 2009, 13(2): 320-326.
12 Liang W H, Liu J K, Chen Q, et al. Comparative Analysis of Extraction of Eucalyptus Information from GF-2 Image based on Object-oriented Method[J]. Forestry Resource Management, 2017 (6): 54-59.梁文海, 刘吉凯, 陈琦等. 基于面向对象方法的GF-2影像桉树信息提取对比分析[J]. 林业资源管理, 2017(6): 54-59.
13 Qin Haichun. Research on Urban Domestic Refuse Monitoring Method based on High-resolution Remote Sensing Images Made in China[J]. China Construction Informatization, 2016 (4): 75-77.
13 秦海春. 基于国产高分遥感影像的城镇生活垃圾监管方法研究[J]. 中国建设信息化, 2016(4): 75-77.
14 Huang Huiling, Han Jun, Wu Feibin, et al. Study on Color Feature Extraction and Classification of Construction Waste[J]. Optical and Optoelectronic Technology, 2018,16(1):53-57.
14 黄惠玲, 韩军, 吴飞斌, 等. 建筑垃圾的颜色特征提取与分类研究[J]. 光学与光电技术, 2018,16(1):53-57.
15 Li H K, Xu F, Li Q. Remote Sensing Monitoring of Land Damage and Restoration in Rare Earth Mining Areas in 6 Counties in Southern Jiangxi based on Multisource Sequential Images[J]. Journal of Environmental Management, 2020, 267. doi:10.1016/j.jenvman.2020.110653.
doi: 10.1016/j.jenvman.2020.110653
16 Lecun Y, Boser B, Denker J S, et al. Backpropagation Applied to Handwritten Zip Code Recognition[J]. Neural Computation,1989,1(4):541-551.doi:10.1162/neco.1989.1.4. 541.
doi: 10.1162/neco.1989.1.4. 541
17 Rebecca G, Tim D, Warwick H, et al. A Remote Sensing Approach to Mapping Fire Severity in South-eastern Australia Using Sentinel 2 and Random Forest[J]. Remote Sensing of Environment, 2020, 240. doi: 10.1016/j.rse.2020.111702.
doi: 10.1016/j.rse.2020.111702
18 Zheng Zongsheng, Hu Chenyu, Huang Dongmei, et al. Research on Transfer Learning Methods for Classification of Typhoon Cloud Image[J]. Remote Sensing Technology and Application, 2020 ,35(1): 202-210.
18 郑宗生,胡晨雨,黄冬梅, 等. 基于迁移学习及气象卫星云图的台风等级分类研究[J]. 遥感技术与应用, 2020 ,35(1): 202-210.
19 Wang Xin, Li Ke, Mingjun Xv, et al. An Improved Remote Sensing Image Classification Method based on Depth Learning[J]. Computer Applications, 2019, 32(2): 382-387.
19 王鑫, 李可, 徐明君, 等. 改进的基于深度学习的遥感图像分类方法[J]. 计算机应用, 2019, 32(2):382-387.
20 Huang Z M, Cheng G L, Wang H Z, et al. Building Extraction from Multi-source Remote Sensing Images via Deep Deconvolution Neural Networks[C]∥IEEE International Geoscience and Remote Sensing Symposium. Beijing, 2016. doi: 10.1109/IGARSS.2016.7729471.
doi: 10.1109/IGARSS.2016.7729471
21 Zheng Yuanpan, Li Guangyang, Li Ye. Review on the Application of Deep Learning in Image Recognition [J]. Computer Engineering and Applications, 2019, 55 (12): 20-36.
21 郑远攀, 李广阳, 李晔. 深度学习在图像识别中的应用研究综述[J].计算机工程与应用, 2019, 55(12): 20-36.
22 Lin Y Z, Zhang B M, Xu J F, et al. Building Extraction from High Resolution Remote Sensing Images based on Multi-features and Multi-scales[J]. Surveying and Mapping Bulletin, 2017(12):53-57.林雨准,张保明,徐俊峰,等.多特征多尺度相结合的高分辨率遥影像建筑物提取[J]. 测绘通报,2017(12):53-57.
23 Gao Yang. Building Extraction from High Resolution Remote Sensing Images based on Convolutional Neural Network [D]. Nanjing:Nanjing University, 2018.
23 高扬.基于卷积神经网络的高分辨率遥感影像建筑物提取[D].南京:南京大学, 2018.
24 Yang Zhenzhen, Kuang Nan, Fan Lu, et al. An Overview of Image Classification Algorithms based on Convolutional Neural Networks[J]. Signal Processing, 2018, 34(12): 1474-1489.
24 杨真真, 匡楠, 范露,等.基于卷积神经网络的图像分类算法综述[J]. 信号处理, 2018, 34(12): 1474-1489.
25 Wu Zhengwen. Application of Convolutional Neural Network in Image Classification[D]. Chengdu: University of Electronic Science and Technology, 2015.
25 吴正文.卷积神经网络在图像分类中的应用研究[D].成都:电子科技大学, 2015.
26 Zhao Xujiang. Remote Sensing Image Target Detection and Recognition based on Convolution Neural Network[D]. Beijing:China University of Science and Technology, 2018.
26 赵旭江. 基于卷积神经网络的遥感图像目标检测与识别[D].北京:中国科学技术大学, 2018.
27 Lin Yu, Chen Xiaoyong. Research on Road Traffic Sign Recognition based on Inception V3 Model[J]. Jiangxi Science, 2018, 36(5): 849-852.
27 林宇, 陈晓勇. 基于Inception V3模型的道路交通标志识别研究[J]. 江西科学, 2018, 36(5): 849-852.
28 Abadi M, Agarwal A, Barham P, et al. Tensor Flow: Large-Scale Machine Learning on Heterogeneous Distributed Systems[J]. Preliminary White Paper,2016,9:1-19.
29 Saxena A. Convolutional Neural Networks: An Illustration in Tensor Flow[J]. Crossroads, 2016, 22(4): 56-58. doi: 10.1145/2951024.
doi: 10.1145/2951024
30 Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the Inception Architecture for Computer Vision[J].Computer Science, 2015: 2818 -2826. doi:10.1109/CVPR.2016.308.
doi: 10.1109/CVPR.2016.308
31 Angelova A, Zhu S. Efficient Object Detection and Segmentation for Fine-grained Recognition[J].IEEE Conference on Computer Vision and Pattern Recognition, 2013: 9(4): 811-81. doi: 10.1109/cvpr.2013.110.
doi: 10.1109/cvpr.2013.110
32 Lu K A. Current Situation and Comprehensive Utilization of Construction Waste in China[J]. Construction Technique, 2005, 28(6):15-16.
32 陆凯安. 我国建筑垃圾的现状与综合利用[J]. 施工技术, 2005, 28(6):15-16.
[1] 郑宗生,胡晨雨,黄冬梅,邹国良,刘兆荣,宋巍. 基于迁移学习及气象卫星云图的台风等级分类研究[J]. 遥感技术与应用, 2020, 35(1): 202-210.
[2] 毛召武,程结海,袁占良. 顾及面积和位置差异的高分遥感影像分割质量评价方法[J]. 遥感技术与应用, 2016, 31(1): 186-193.
[3] 丁辉,张茂省,李林. 基于多特征面向对象区域滑坡现象识别[J]. 遥感技术与应用, 2013, 28(6): 1107-1113.
[4] 宋莎莎,赵宇鹏,苏腾飞,马佑军,安 伟,孟俊敏,赵朝方. 海上溢油SAR卫星遥感监测系统研发[J]. 遥感技术与应用, 2013, 28(5): 928-933.
[5] 刘行华 胡宝新 王建武 张渊智 赵炜. 基于知识的遥感信息地物理解模型研究[J]. 遥感技术与应用, 1992, 7(4): 1-9.