Intelligent Identification of Sudden Geohazard Bodies Based on Doppler Radar
-
摘要:
针对突发性地质灾害识别的高精度与实时性需求，提出了一种适用于多普勒雷达的轻量级多尺度特征融合网络DRWAF-Net（doppler radar wavelet attention fuse network），通过小波变换与注意力机制协同优化，实现了复杂地表环境下泥石流、滚石等灾害体的实时识别.研究充分利用多普勒雷达动态捕获灾害体距离与速度的能力，整合环境干扰下的泥石流数据集与RDRD数据集的核心要素，针对性构建了突发地质灾害场景的多普勒雷达数据集.实验结果表明，DRWAF-Net以2.38 M参数量、9.27 MB模型大小和6.31 ms推理速度，使准确率（96.77%）、精确率（96.90%）、召回率（96.77%）和F1分数（96.77%）均在测试集上达到最优水平.消融实验验证，结合多输入注意力门控（MIAG）机制的DRWAF-Net较基准模型提升识别率1.87%~3.13%.通过轻量化设计与实时推理优化，为突发性地质灾害应急响应提供了实时、智能的监测方案.
Abstract:To address the demands for high-accuracy and real-time identification of sudden geological hazards, a lightweight multi-scale feature fusion network for Doppler radar, termed DRWAF-Net (doppler radar Wavelet Attention Fuse Network), is proposed. By jointly integrating wavelet transform and attention mechanisms, the proposed method enables real-time recognition of debris flows, rockfalls, and other hazard targets under complex surface conditions. The study fully exploits the capability of Doppler radar to dynamically capture the range and velocity characteristics of moving hazard bodies, and constructs a Doppler radar dataset for sudden geological hazard scenarios by integrating key elements from a debris flow dataset under environmental interference and the RDRD dataset. Experimental results show that DRWAF-Net achieves superior performance on the test set with only 2.38 M parameters, a model size of 9.27 MB, and an inference time of 6.31 ms, attaining an accuracy of 96.77%, precision of 96.90%, recall of 96.77%, and an F1-score of 96.77%. Ablation experiments further demonstrate that the introduction of a multi-input attention gating (MIAG) mechanism improves recognition accuracy by 1.87%-3.13% compared with baseline models. Owing to its lightweight design and real-time inference capability, the proposed approach provides an effective and intelligent monitoring solution for emergency response to sudden geological hazards.
-
图 1 构建面向突发地质灾害应急响应数据集
a.预处理环境干扰下的泥石流数据集；b.预处理RDRD数据集
Fig. 1. Construction of an emergency-response-oriented dataset for sudden geological hazard
图 2 DRWAF-Net总体框架
a.轻量级目标识别网络模型结构图；b.下采样特征提取模块结构图；c.预测头结构图
Fig. 2. Framework of DRWAF-Net
图 3 注意力改进的小波卷积模块
a.wtagconv2dblock的结构；b.wtagconv2dblock对输入特征图的处理管线
Fig. 3. Attention-enhanced wavelet transform convolution module
图 5 各模型每轮测试集的性能对比
a.准确率对比；b.精确率对比；c.召回率对比；d.F1分数对比
Fig. 5. Performance comparison of all models on the test dataset across epochs
图 6 最佳模型评价指标对比
Fig. 6. Comparison of evaluation metrics for the best-performing models
图 7 对测试样本进行多分类的混淆矩阵
Fig. 7. Confusion matrix for multi-class classification on the test samples
表 1 对比模型列表暨边缘设备兼容性分析
Table 1. Comparison of models and analysis of edge-device compatibility
Model (year) Params (M) FLOPs (G) Size (MB) MREI 兼容性 Vgg16 (2014) 134.29 15.47 512.29 102.10 ★☆☆☆☆ ResNet50 (2016) 23.52 4.13 90.06 20.61 ★★☆☆☆ MobileNet-V3-large (2019) 4.21 0.23 16.30 2.51 ★★★★☆ DopplerNet (2020) 100.94 0.14 385.07 17.59 ★★★☆☆ ConvNeXt-Tiny (2022) 27.82 4.46 106.23 23.62 ★★☆☆☆ DRWAF-Net 2.38 0.29 9.27 1.86 ★★★★★ 
下载: 导出CSV
表 2 模型最佳性能与推理耗时对比
Table 2. Comparison of model performance and inference time
Model (year) FLOPs (G) A (%) P (%) R (%) F1 (%) Time (ms) Vgg16 (2014) 15.47 93.26 93.73 93.26 93.09 2.35 ResNet50 (2016) 4.13 95.72 95.83 95.73 95.71 6.37 MobileNet-V3-large (2019) 0.23 95.59 95.63 95.59 95.58 6.42 DopplerNet (2020) 0.14 91.18 91.38 91.19 91.04 1.26 ConvNeXt-Tiny (2022) 4.46 77.73 79.79 77.73 77.02 5.69 DRWAF-Net 0.29 96.77 96.90 96.77 96.77 6.31
下载: 导出CSV
表 3 消融实验性能对比
Table 3. Comparison of performance in ablation experiments
Model (block) Size (MB) A (%) P (%) R (%) F1 (%) Time (ms) DRNet-v1 (None) 6.50 93.68 93.85 93.69 93.64 4.12 DRNet-v2 (with AG) 6.88 94.20 94.44 94.20 94.15 4.93 DRNet-v3 (with MIAG) 8.88 95.56 95.73 95.56 95.54 5.90 DRWAF-Net 9.27 96.77 96.90 96.77 96.77 6.31
下载: 导出CSV
-
Ajit, A., Acharya, K., Samanta, A., 2020. A Review of Convolutional Neural Networks. In: 2020 International Conference on Emerging Trends in Information Technology and Engineering (ic-ETITE). February 24-25, 2020. Vellore, India. IEEE, 1-5. https://doi.org/10.1109/ic-etite47903.2020.049 Boonpook, W., Tan, Y. M., Xu, B., 2021. Deep Learning-Based Multi-Feature Semantic Segmentation in Building Extraction from Images of UAV Photogrammetry. International Journal of Remote Sensing, 42(1): 1-19. https://doi.org/10.1080/01431161.2020.1788742 Cai, Z. W., Fan, Q. F., Feris, R. S., et al., 2016. A Unified Multi-Scale Deep Convolutional Neural Network for Fast Object Detection. In: Computer Vision-ECCV 2016. Springer International Publishing, Cham: 354-370. https://doi.org/10.1007/978-3-319-46493-0_22 Casagli, N., Intrieri, E., Tofani, V., et al., 2023. Landslide Detection, Monitoring and Prediction with Remote-Sensing Techniques. Nature Reviews Earth & Environment, 4(1): 51-64. https://doi.org/10.1038/s43017-022-00373-x Chang, K. X., 2019. Returning to the Team ahead of Time for Rescue, always Falling on the Post—A Record of the Squad Leader of Shuimo Town Government Full-Time Team of Wenchuan County Brigade of Aba Prefecture Fire Brigade, Geng Siqiong. Jinri Xiaofang, 5(15): 38-39 (in Chinese with English abstract). Chiang, H. C., Moses, R. L., Potter, L. C., 2000. Model-Based Classification of Radar Images. IEEE Transactions on Information Theory, 46(5): 1842-1854. https://doi.org/10.1109/18.857795 Cui, P., Liu, S. J., Tan, W. P., 2000. Progress of Debris Flow Forecast in China. Journal of Natural Disasters, 9(2): 10-15 (in Chinese with English abstract). Dubey, A., Santra, A., Fuchs, J., et al., 2021. Bayesradar: Bayesian Metric-Kalman Filter Learning for Improved and Reliable Radar Target Classification. In: 2021 IEEE 31st International Workshop on Machine Learning for Signal Processing (MLSP). October 25-28, 2021, Gold Coast, Australia. IEEE: 1-6. https://doi.org/10.1109/MLSP52302.2021.9596290 Finder, S. E., Amoyal, R., Treister, E., et al., 2024. Wavelet Convolutions for Large Receptive Fields. In: Computer Vision – ECCV 2024. Springer Nature Switzerland, Cham: 363-380. https://doi.org/10.1007/978-3-031-72949-2_21 Gu, J. X., Wang, Z. H., Kuen, J., et al., 2018. Recent Advances in Convolutional Neural Networks. Pattern Recognition, 77: 354-377. https://doi.org/10.1016/j.patcog.2017.10.013 He, K. M., Zhang, X. Y., Ren, S. Q., et al., 2016. Deep Residual Learning for Image Recognition. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). June 27-30, 2016, Las Vegas, NV, USA. IEEE: 770-778. https://doi.org/10.1109/CVPR.2016.90 He, S. M., Wang, D. P., Wu, Y., et al., 2014. Formation Mechanism and Key Prevention Technology of Rockfalls. Chinese Journal of Nature, 36(5): 336-345 (in Chinese with English abstract). Heydarian, M., Doyle, T. E., Samavi, R., 2022. MLCM: Multi-Label Confusion Matrix. IEEE Access, 10: 19083-19095. https://doi.org/10.1109/ACCESS.2022.3151048 Howard, A., Sandler, M., Chen, B., et al., 2019. Searching for MobileNetV3. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). October 27-November 2, 2019. Seoul, Korea. IEEE, 1314-1324. . https://doi.org/10.1109/iccv.2019.00140 Huether, B. M., Gustafson, S. C., Broussard, R. P., 2001. Wavelet Preprocessing for High Range Resolution Radar Classification. IEEE Transactions on Aerospace and Electronic Systems, 37(4): 1321-1332. https://doi.org/10.1109/7.976968 Jakob, M., Lambert, S., 2009. Climate Change Effects on Landslides along the Southwest Coast of British Columbia. Geomorphology, 107(3/4): 275-284. https://doi.org/10.1016/j.geomorph.2008.12.009 Khan, A., Sohail, A., Zahoora, U., et al., 2020. A Survey of the Recent Architectures of Deep Convolutional Neural Networks. Artificial Intelligence Review, 53(8): 5455-5516. https://doi.org/10.1007/s10462-020-09825-6 La, R. F., Lv, T., Bai, P. F., et al., 2022. Research on Collaborative and Optimal Deployment and Decision Making among Major Geological Disaster Rescue Subjects. Geotechnical and Geological Engineering, 40(1): 57-71. https://doi.org/10.1007/s10706-021-01883-z Li, M. W., Tang, C., Chen, M., et al., 2021. Formation and Vulnerability Analysis for Debris Flow Occurred on 20 August 2019 in Banzi Catchment, Wenchuan County, Sichuan Province, China. Journal of Disaster Prevention and Mitigation Engineering, 41(2): 238-245 (in Chinese with English abstract). Li, Q., Qi, Y. C., Zhang, Z., et al., 2024. Nonmeteorological Echoes Identification Method Based on Bayesian Classifier and Echo Physical Characteristics Using C-Band Radar and Its Performance. Chinese Journal of Atmospheric Sciences, 48(3): 823-836 (in Chinese with English abstract). Li, Z. W., Liu, F., Yang, W. J., et al., 2022. A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. IEEE Transactions on Neural Networks and Learning Systems, 33(12): 6999-7019. https://doi.org/10.1109/TNNLS.2021.3084827 Lin, T. Y., Dollár, P., Girshick, R., et al., 2017. Feature Pyramid Networks for Object Detection. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). July 21-26, 2017, Honolulu, HI, USA. IEEE: 936-944. https://doi.org/10.1109/CVPR.2017.106 Litjens, G., Kooi, T., Bejnordi, B. E., et al., 2017. A Survey on Deep Learning in Medical Image Analysis. Medical Image Analysis, 42: 60-88. https://doi.org/10.1016/j.media.2017.07.005 Liu, S. S., 2024. Research on Crack Detection of Concrete Structures based on CNN and Transformer Hybrid Architecture (Dissertation). Hubei University of Technology, Wuhan (in Chinese with English abstract). Liu, S., Hu, K. H., Li, H., et al., 2025. Radar-Based Deep Learning for Debris Flow Identification Amid the Environmental Disturbances. Geophysical Research Letters, 52(2): e2024GL112351. https://doi.org/10.1029/2024gl112351 Liu, Z., Mao, H. Z., Wu, C. Y., et al., 2022. A ConvNet for the 2020s. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 18-24, 2022, New Orleans, LA, USA. IEEE: 11966-11976. https://doi.org/10.1109/CVPR52688.2022.01167 Luo, J., 2020. Slope Dynamic Response and Formation Mechanism of Large-Scale Rockslide Dam in the "8·3" Ludian Earthquake (Dissertation). Chengdu University of Technology, Chengdu (in Chinese with English abstract). Luo, Y. T., 2021. Disaster Chain and Risk Quantitative Evaluation of "8·20" River-Blocking Debris Flow in Wenchuan County (Dissertation). Chengdu University of Technology, Chengdu (in Chinese with English abstract). Ma, T., Zhou, L. F., Li, J. X., 2024. Space Object Recognition Method Based on Wideband Radar RCS Data. Journal of Ordnance Equipment Engineering, 45(7): 275-282 (in Chinese with English abstract). Michelini, A., Viviani, F., Bianchetti, M., et al., 2020. A New Radar-Based System for Detecting and Tracking Rockfall in Open Pit Mines Slope Stability 2020. In: Proceedings of the 2020 International Symposium on Slope Stability in Open Pit Mining and Civil Engineering, Online. https://papers.acg.uwa.edu.au/p/2025_79_Stopka/ Rawat, W., Wang, Z. H., 2017. Deep Convolutional Neural Networks for Image Classification: A Comprehensive Review. Neural Computation, 29(9): 2352-2449. doi: 10.1162/neco_a_00990 Rickenmann, D., 1999. Empirical Relationships for Debris Flows. Natural Hazards, 19(1): 47-77. https://doi.org/10.1023/A:1008064220727 Ristic, B., Kim, D. Y., Rosenberg, L., et al., 2022. Exploiting Doppler in Bernoulli Track-before-Detect for a Scanning Maritime Radar. IEEE Transactions on Aerospace and Electronic Systems, 58(1): 720-728. doi: 10.1109/TAES.2021.3098117 Roldan, I., del-Blanco, C. R., Duque de Quevedo, Á., et al., 2020. DopplerNet: A Convolutional Neural Network for Recognising Targets in Real Scenarios Using a Persistent Range–Doppler Radar. IET Radar, Sonar & Navigation, 14(4): 593-600. https://doi.org/10.1049/iet-rsn.2019.0307 Romeo, S., Cosentino, A., Giani, F., et al., 2021. Combining Ground Based Remote Sensing Tools for Rockfalls Assessment and Monitoring: The Poggio Baldi Landslide Natural Laboratory. Sensors, 21(8): 2632. https://doi.org/10.3390/s21082632 Shen, Y. Y., Huang, X. Y., Huang, S. R., et al., 2020. Identification and Validation of Sea-Wave Echoes Collected by a Doppler Weather Radar Based on a Bayes Classifier. Marine Sciences, 44(6): 83-90 (in Chinese with English abstract). Shi, Y., Meng, X. H., 2014. Documentary of the "8·3" Earthquake in Ludian, Yunnan. China Report, (9): 46-47 (in Chinese). Simonyan, K., Zisserman, A., 2014. Very Deep Convolutional Networks for Large-Scale Image Recognition. Computer Science. https://doi.org/10.48550/arXiv.1409.1556 Szegedy, C., Liu, W., Jia, Y. Q., et al., 2015. Going Deeper with Convolutions. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). June 7-12, 2015, Boston, MA, USA. IEEE: 1-9. https://doi.org/10.1109/CVPR.2015.7298594 Vaswani, A., Shazeer, N., Parmar, N., et al., 2017. Attention is All You Need. Advances in Neural Information Processing Systems, 30. https://doi.org/10.48550/arXiv.1706.03762 Viviani, F., Michelini, A., Mayer, L., 2020. RockSpot: An Interferometric Doppler Radar for Rockfall/Avalanche Detection and Tracking. In: 2020 IEEE Radar Conference (RadarConf20). September 21-25, 2020, Florence, Italy. IEEE: 1-5. https://doi.org/10.1109/RadarConf2043947.2020.9266677 Wang, J., Zhu, H., Lei, P., et al., 2019. CNN Based Classification of Rigid Targets in Space Using Radar Micro-Doppler Signatures. Chinese Journal of Electronics, 28(4): 856-862. doi: 10.1049/cje.2018.08.003 Wang, K. P., Yan, C. L., Yang, H., 2014. Research into Radar Emitter Recognition Based on Template Matching. Shipboard Electronic Countermeasure, 37(5): 31-33, 41 (in Chinese with English abstract). Xiu, J. G., Chen, Y. M., 2014. Ludian Emergency: Insights from the Emergency Rescue Efforts Following the "8·3" M6.5 Earthquake in Ludian, Yunnan. Overview of Disaster Prevention, (5): 28-35 (in Chinese with English abstract). Yu, F., Koltun, V., 2016. Multi-Scale Context Aggregation by Dilated Convolutions. In: 4th International Conference on Learning Representations (ICLR 2016), San Juan, Puerto Rico. http://arxiv.org/abs/1511.07122 Zhang, L., Yang, H. P., Deng, X., et al., 2014. One Method Based on Template Matching to Remove Strong Anomalous Propagation Echo at Changle Site. Meteorological Monthly, 40(3): 364-372 (in Chinese with English abstract). Zhang, P. Y., Shen, L., Huang, X. T., et al., 2022. Ground Penetrating Radar Image Template Matching Based on Symmetrical Structure Features. Progress in Geophysics, 37(6): 2657-2666 (in Chinese with English abstract). Zhao, Y. G., Lei, B., 2022. The Application Issues of Ultrasonic Interface Analyzer. Analytical Instrumentation, (3): 40-43 (in Chinese with English abstract). Zhou, Z., Cao, Z. J., Pi, Y. M., 2019. Subdictionary-Based Joint Sparse Representation for SAR Target Recognition Using Multilevel Reconstruction. IEEE Transactions on Geoscience and Remote Sensing, 57(9): 6877-6887. https://doi.org/10.1109/TGRS.2019.2909121 Zuo, L. Y., Yang, J. F., 2023. Progress, Effectiveness, and Future Trends of Investigation and Monitoring of Sudden Geological Disasters in China. China Mining Magazine, 32(S2): 7-12 (in Chinese with English abstract). 昌开馨, 2019. 为救援提前归队, 永远倒在岗位上: 记阿坝州消防支队汶川县大队水磨镇政府专职队班长更斯穷. 消防界(电子版), 5(15): 38-39. 崔鹏, 刘世建, 谭万沛, 2000. 中国泥石流监测预报研究现状与展望. 自然灾害学报, 9(2): 10-15. 何思明, 王东坡, 吴永, 等, 2014. 崩塌滚石灾害的力学机理与防治技术. 自然杂志, 36(5): 336-345. 李明威, 唐川, 陈明, 等, 2021. 四川省汶川县板子沟"8·20"泥石流成因与易损强度分析. 防灾减灾工程学报, 41(2): 238-245. 李巧, 戚友存, 张哲, 等, 2024. 基于贝叶斯分类器和回波物理特征的C波段雷达非气象回波识别方法和性能分析. 大气科学, 48(3): 823-836. 刘石狮, 2024. 基于CNN和Transformer混合架构的混泥土结构裂缝检测研究(硕士学位论文). 武汉: 湖北工业大学. 罗璟, 2020. "8·3" 鲁甸地震斜坡动力响应及巨型岩质滑坡堵江机制研究(博士学位论文). 成都: 成都理工大学. 罗玉婷, 2021. 汶川县"8·20" 堵江型泥石流灾害链及风险定量评价(硕士学位论文). 成都: 成都理工大学. 马腾, 周兰凤, 李建鑫, 2024. 基于宽带雷达RCS数据的空间物体识别方法. 兵器装备工程学报, 45(7): 275-282. 沈妍琰, 黄兴友, 黄书荣, 等, 2020. 基于贝叶斯分类器的多普勒天气雷达海浪回波识别和效果检验. 海洋科学, 44(6): 83-90. 石岩, 孟宪虎, 2014. 云南鲁甸"8·3" 地震纪实. 中国报道, (9): 46-47. 王琨鹏, 颜春林, 杨辉, 2014. 基于模板匹配的雷达辐射源信号识别研究. 舰船电子对抗, 37(5): 31-33, 41. 修济刚, 陈宇鸣, 2014. 鲁甸应急: 云南鲁甸"8·3" 6.5级地震应急救援启示. 防灾博览, (5): 28-35. 张林, 杨洪平, 邓鑫, 等, 2014. 基于模板匹配法的长乐雷达强超折射回波识别. 气象, 40(3): 364-372. 张鹏宇, 申亮, 黄晓涛, 等, 2022. 基于对称结构特征的探地雷达图像模板匹配算法. 地球物理学进展, 37(6): 2657-2666. 赵延广, 雷斌, 2022. 超声波泥位计应用问题及解决措施. 分析仪器, (3): 40-43. 左力艳, 杨建锋, 2023. 我国突发性地质灾害调查监测进展、成效与未来趋势. 中国矿业, 32(增刊2): 7-12. -
点击查看大图
计量
- 文章访问数: 341
- HTML全文浏览量: 15
- PDF下载量: 52
- 被引次数: 0
