| 李欣宇,孙传猛,魏 宇,原 玥,武志博,李 勇.融合 Transformer 与残差通道注意力的恶劣场景水
位智能检测方法[J].电子测量与仪器学报,2023,37(1):59-69 |
| 融合 Transformer 与残差通道注意力的恶劣场景水
位智能检测方法 |
| Water level intelligent detection method based on fuse Transformer residualchannel attention mechanism in harsh environments |
| |
| DOI：10.13382/j.issn.1000-7105.2023.01.007 |
| 中文关键词: 水位检测 深度学习 语义分割 Transformer 注意力机制 |
| 英文关键词:water level detection deep learning semantic segmentation Transformer attention mechanism |
| 基金项目:国家重点研发计划青年科学家项目(2022YFC2905700)、山西省高等学校科技创新项目(2020L0294)、山西省科技成果转化引导专项(202104021301061)项目资助 |
|
|
| Author | Institution |
| Li Xinyu | 1. State Key Laboratory of Dynamic Measurement Technology, North University of China,2. School of Electrical and Control Engineering, North University of China |
| Sun Chuanmeng | 1. State Key Laboratory of Dynamic Measurement Technology, North University of China,2. School of Electrical and Control Engineering, North University of China |
| Wei Yu | 1. State Key Laboratory of Dynamic Measurement Technology, North University of China,2. School of Electrical and Control Engineering, North University of China |
| Yuan Yue | 1. State Key Laboratory of Dynamic Measurement Technology, North University of China,2. School of Electrical and Control Engineering, North University of China |
| Wu Zhibo | 1. State Key Laboratory of Dynamic Measurement Technology, North University of China,2. School of Electrical and Control Engineering, North University of China |
| Li Yong | 3. State Key Laboratory of Coal Mine Disaster Dynamics and Control, Chongqing University |
|
| 摘要点击次数: 540 |
| 全文下载次数: 612 |
| 中文摘要: |
| 精准感知水位信息变化是实现精细水务管控和洪涝灾害的关键环节之一,而低照度、雾霾、雨雪、冰冻、波浪、镜头抖动
等恶劣场景给水位检测带来极大挑战。 针对现有方法中难以实现水位精准检测难题,构建一种融合 Transformer 与残差通道注
意力机制的 Unet 模型(TRCAM-Unet),进而提出基于 TRCAM-Unet 的恶劣场景水位智能检测方法。 关键技术包括通过全尺度
连接结构实现多层次特征融合,通过 Transformer 模块强化区域特征的关联性,通过残差通道注意力模块强化有用信息的表达
并削弱无用信息的干扰。 相关试验和实践表明,TRCAM-Unet 取得了 98. 84%MIOU 评分与 99. 42%的 MPA 评分,在约 150 m 距
离外水位检测最大误差不超过 0. 08 m,水位偏差均值(MLD)仅有 1. 609×10
-2 m,优于 Deeplab、PSPNet 等主流语义分割算法。
研究结果对解决恶劣场景下水位精准检测难题及洪涝灾害预警具有重要应用价值。 |
| 英文摘要: |
| Accurate perception of water level changes is one of the key segments to achieve precision water affairs control and flood
disaster, but harsh scenes such as low illumination, haze, rain and snow, freezing, lighting, and waves bring a great challenge to water
level accurate detection. Aiming at the problem of accurate detection of water level in existing methods, this paper constructs a Unet
model fused with transformer residual channel attention mechanism (called “TRCAM-Unet”), then, a water lever intelligent detection
method in harsh environments based on TRCAM-Unet is proposed. The key technologies include that: Multi-level feature fusion is
achieved by full scale connection structure. The relevance of regional feature is strengthened by transformer module. Strengthening the
extraction ability of useful information and weakening the interference of useless information by residual channel attention module. The
experiments and practices of water level semantic segmentation in harsh scenes shows that TRCAM-Unet achieved 98. 84% MIOU scores
and 99. 42% MPA scores, the maximum error of water level detection outside 150 meters was not above 0. 08 m, mean water level
deviation (MLD) had only 1. 609×10
-2 meters, it is much better than the mainstream semantic segmentation models such as Deeplab,
PSPNet, Unet. This study has important application value for water level accurate detection in harsh scenes and flood disaster early
warning. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
