A remote sensing imagery-based model for assessment of building damage induced by large-equivalent explosions
-
摘要: 为了研究大当量爆炸建筑物毁伤评估问题，基于遥感影像解译和大数据分析构建了大当量爆炸建筑物毁伤评估模型。首先，基于大当量爆炸的具体历史案例构建了毁伤数据集，具体指基于遥感影像提取建筑物毁伤信息，辅助大数据信息补充毁伤细节，利用地理信息系统空间分析数字化毁伤信息，构成毁伤数据集。然后，基于毁伤数据集中的训练样本修正经验模型参数，构建了适用于大当量爆炸的针对不同类型建筑物的毁伤评估模型，并基于毁伤数据集中的验证样本测试了模型性能。实验证明：所构建模型拟合优度高于96%，检验样本准确度高于84%，整体误差在可接受范围内。所构建模型在一定精度要求下可为大当量爆炸事故评估提供参考。Abstract: To address challenges in the field of large-scale explosive building damage assessment, where the explosion process is too complex for high-precision numerical simulation, and relying solely on change detection from remote sensing imagery cannot capture detailed internal information and lacks the capability of predicting in advance, this paper establishes a building damage assessment model for large-scale explosive events by coupling empirical mechanics models with remote sensing image interpretation and big data analysis. The study initially constructs a damage dataset based on specific historical cases of large-scale explosions. This involves extracting building damage information (including building types and damage levels) from remote sensing imagery and supplementing damage details with additional big data sources such as collected online images, videos, and news reports to enhance the precision of the sampled data. Geographic information systems spatial analysis is employed to digitize the damage information, obtaining data on building types, damage levels, and the distance from the target building to the explosion center, forming the damage dataset. Subsequently, the empirical model parameters are refined based on the training samples from the damage dataset, creating damage assessment models applicable to different building types for large-scale explosive events. The performance of the model is then tested using validation samples from the damage dataset. Experimental results demonstrate a model fitting goodness of over 96%, accuracy on validation samples exceeding 84%, and an overall error within an acceptable range. The model, under certain accuracy requirements, can provide guidance for site selection of storage locations for chemicals and hazardous materials, emergency evacuation of people in the event of a risk of large-scale explosions, critical equipment evacuation during an emergency, resource dispatching for rescue and relief after an accident, and building damage assessment.
-
Key words:
- large equivalent explosion /
- remote sensing imagery /
- damage assessment /
- big data analysis /
- building
-
图 6 连云港化工园爆炸图像
Figure 6. Images of the explosion at Lianyungang Chemical Industry Park
图 7 砖混结构建筑物中度和重度毁伤评估模型拟合曲线
Figure 7. Model fitting curves of moderate and severe damage predictions for brick and concrete structure
图 8 工业厂房中度和重度毁伤评估模型拟合曲线
Figure 8. Model fitting curves of moderate and severe damage predictions for industrial building
表 1 建筑物目视解译毁伤等级标准
Table 1. Standard for visual interpretation of damage grade of buildings
建筑物类型
与毁伤等级毁伤前影像 毁伤后影像 标准细则 砖混结构
中度毁伤
(1)屋顶和楼板出现明显破损或塌陷，
形成明显洞口或凹陷；
(2)墙体出现明显裂缝；
(3)结构柱明显变形，墙体倾斜砖混结构
重度毁伤
(1)屋顶或楼板出现严重破损，楼层部分
或全部垮塌；
(2)墙体部分或全部崩塌；
(3)结构柱断裂或严重变形，墙体严重倾斜
或移位工业厂房
中度毁伤
(1)屋顶破损但整体结构未受到影响；
(2)钢结构轻微变形或扭曲，但并未断裂
或明显变形；
(3)主体结构轻微倾斜或不平衡工业厂房
重度毁伤
(1)屋顶严重破损；
(2)钢结构出现明显脱离、断裂
或腐蚀等情况；
(3)主体结构受到严重破坏，部分
或整体倒塌
下载: 导出CSV
表 2 样本详情
Table 2. Sample details
样本 事故详情 发生时间 经纬度 TNT当量/t 爆炸前成像时间 爆炸后成像时间 训练样本1 河北省张家口市盛华化工有限公司爆炸 2018.11.28 40.75°N 115.00°E 0.5 2018.11 2018.12 训练样本2 山东省青岛市中石化东黄输油管道爆炸 2013.11.22 36.05°N 120.21°E 3 2013.03 2014.02 训练样本3 美国德克萨斯州韦斯特镇的韦斯特化工厂爆炸 2013.04.17 31.81°N 97.09°W 10 2012.10 2013.04 训练样本4 河南省三门峡市义马气化厂爆炸 2019.07.19 34.74°N 111.84°E 30 2019.03 2020.04 训练样本5 江苏省盐城市陈家港镇工业园区爆炸 2019.08.04 34.32°N 119.79°E 260 2018.03 2019.04 训练样本6 黎巴嫩贝鲁特港口爆炸 2020.08.04 33.90°N 35.52°E 620 2020.07 2020.08 检验样本1 江苏省连云港市堆沟港镇化工园爆炸 2017.12.09 34.42°N 119.78°E 10 2017.09 2018.03 检验样本2 天津市滨海新区天津港爆炸 2015.08.12 39.04°N 117.74°E 450 2015.05 2015.08
下载: 导出CSV
表 3 训练样本数据
Table 3. Training sample data
样本 TNT当量/t 砖混结构中度毁伤 砖混结构重度毁伤 工业厂房中度毁伤 工业厂房重度毁伤 y/m ŷ/m p/% y/m ŷ/m p/% y/m ŷ/m p/% y/m ŷ/m p/% 1 0.5 45 38 84.44 30 28 93.33 2 3 141 70 49.65 50 45 90.00 3 10 192 129 67.19 105 82 78.10 263 171 65.02 154 124 80.52 4 30 252 223 88.49 180 142 78.89 343 296 86.30 195 214 90.26 5 260 733 655 89.36 468 417 89.10 980 872 88.98 735 631 85.85 6 770 1065 1128 94.08 677 717 94.09 1417 1500 94.12 1025 1085 94.15 R2/% 96.97 97.80 97.74 97.86 注：y为原始值，ŷ为预测值；p为准确度，p=1−|相对误差|；
拟合优度由R2度量，R2=1−残差平方和/总平方和=1−Σ(原始值−预测值)2/Σ(原始值−原始值均值)2。
下载: 导出CSV
表 4 检验样本数据
Table 4. Test sample data
样本 TNT当量/t 砖混结构中度毁伤 砖混结构重度毁伤 工业厂房中度毁伤 工业厂房重度毁伤 y/m ŷ/m p/% y/m ŷ/m p/% y/m ŷ/m p/% y/m ŷ/m p/% 1 10 130 129 99.23 96 82 85.42 157 171 91.08 107 124 86.29 2 450 895 862 96.31 473 548 84.14 988 1147 83.91 716 830 86.27 pmean/% 97.77 84.78 87.49 86.28 注：pmean为平均准确度。
下载: 导出CSV
-
[1] 张连玉, 汪令羽, 苗瑞生. 爆炸气体动力学基础 [M]. 北京: 北京工业学院出版社, 1987: 425–426. [2] 周子钦. 基于多时相图像的打击效果评估技术研究 [D]. 武汉: 华中科技大学, 2020: 40–45. DOI: 10.27157/d.cnki.ghzku.2020.006392.ZHOU Z Q. Study of damage effect assessment based on multi-temporal image [D]. Wuhan: Huazhong University of Science and Technology, 2020: 40–45. DOI: 10.27157/d.cnki.ghzku.2020.006392. [3] 张军. 多时相图像检测方法及其在毁伤评估系统中的应用 [D]. 上海: 上海交通大学, 2008: 17–18.ZHANG J. Processing method of multi-temporal remote sensing images and the use in damage assessment [D]. Shanghai: Shanghai Jiao Tong University, 2008: 17–18. [4] SIRMACEK B, UNSALAN C. Damaged building detection in aerial images using shadow information [C]//Proceedings of the 4th International Conference on Recent Advances in Space Technologies. Istanbul: IEEE, 2009: 249–252. DOI: 10.1109/RAST.2009.5158206. [5] 王威. 基于图像理解的打击效果评估系统研究与实现 [D]. 武汉: 华中科技大学, 2015: 9–13.WANG W. Research and implementation of battle damage assessment system based on image understanding [D]. Wuhan: Huazhong University of Science and Technology, 2015: 9–13. [6] 勾涛. 基于图像分析的毁伤评估系统关键技术研究 [D]. 长春: 吉林大学, 2019: 5–6.GOU T. Research on key technologies of damage assessment system based on image analysis [D]. Changchun: Jilin University, 2019: 5–6. [7] 孔祥锡, 秦闻远, 苏飘逸, 等. 基于深度学习及模糊层次分析的毁伤评估算法 [J]. 航空学报, 2023, 40: 1–18.KONG X X, QIN W Y, SU P Y, et al. Damage assessment algorithm based on deep learning and fuzzy analytic hierarchy process [J]. Acta Aeronautica et Astronautica Sinica, 2023, 40: 1–18. [8] 中国人民解放军总装备部军事训练教材编辑工作委员会. 核爆炸物理概论 [M]. 北京: 国防工业出版社, 2003: 51–55. [9] 李秦超, 姚成宝, 程帅, 等. 神经网络状态方程在强爆炸冲击波数值模拟中的应用 [J]. 爆炸与冲击, 2023, 43(4): 044202. DOI: 10.11883/bzycj-2022-0222.LI Q C, YAO C B, CHENG S, et al. Application of the neural network equation of state in numerical simulation of intense blast wave [J]. Explosion and Shock Waves, 2023, 43(4): 044202. DOI: 10.11883/bzycj-2022-0222. [10] GLASSTONE S, DOLAN P J. The effects of nuclear weapons [R]. USA: Defense Technical Information Center, 1977: 453–501. DOI: 10.21236/ada087568. [11] ZHANG H M, WANG M C, ZHANG Y X, et al. TDA-Net: a novel transfer deep attention network for rapid response to building damage discovery [J]. Remote Sensing, 2022, 14(15): 3687. DOI: 10.3390/rs14153687. [12] 孙家抦. 遥感原理与应用 [M]. 3版. 武汉: 武汉大学出版社, 2013: 174–176.SUN J B. Principles and applications of remote sensing [M]. 3rd ed. Wuhan: Wuhan University Press, 2013: 174–176. [13] 全国地震标准化技术委员会. 中华人民共和国地震行业标准: DB/T 75—2018 [S]. 北京: 中国标准出版社, 2019. [14] 卢芳云, 李翔宇, 田占东, 等. 武器毁伤与评估 [M]. 北京: 科学出版社, 2021: 93–152. -
图(8) / 表(4)
计量
- 文章访问数: 262
- HTML全文浏览量: 94
- PDF下载量: 140
- 被引次数: 0