To address the challenges faced by existing reservoir water level segmentation methods, such as large computational parameters, insufficient extraction of multi-scale water level features, and inadequate attention to detailed water level features, a reservoir water level segmentation method (DEGNet) is proposed based on multi-scale and multi-attention mechanisms, achieving high-precision segmentation of reservoir water levels. First, a lightweight MobileNetv2 is employed as the backbone network, integrated with a convolutional block attention module (CBAM), for reducing computational parameters and enhancing the extraction capability of water level features. Secondly, an innovative multi-scale information fusion module (ESPP) is proposed to effectively capture and fuse spatial information from multi-scale water level feature maps, significantly enhancing feature spatial expressiveness. Finally, a global attention mechanism (GAM) is incorporated into the encoding region to focus on detailed water level features, thereby realizing high-precision and efficient segmentation of reservoir water levels. Experimental results demonstrate that, compared to methods such as PSPNet, HRNet, and DeepLabv3+, the proposed method achieves improvements of 4.9 and 7.76 percentage points in mean intersection over union (MIoU) and mean average precision (MAP), respectively, reaching 98.75% MIoU and 99.43% MAP, respectively, and exhibits significant superiority in both segmentation accuracy and detection precision.
ZHANGRu, LOUChendi, ZHANGZetian, et al. Utilization and protection of water resources under the background of carbon neutralization[J]. Advanced Engineering Sciences, 2022, 54(1): 69-82. (in Chinese)
LIUKaihua, YANHongping, SHABaoyin, et al. Reconstruction of ship draft and water level identification using weighted directed graph[J]. Journal of Computer-Aided Design & Computer Graphics, 2023, 35(7): 1077-1085. (in Chinese)
LIUChunguo, ZHAOGang, KONGLingchang, et al. Observation error and test method of digital pressure type water level instrument[J]. Journal of Geodesy and Geodynamics, 2018, 38(1): 97-101. (in Chinese)
KUERBANJIANGYiti. Correlation analysis between water resources development and utilization and hydrological water resources monitoring[J]. Energy and Energy Conservation, 2021(12): 79-81. (in Chinese)
LIHailu. Comparative study on radar water level gauge and artificial observation water level at Zongyangzha hydrological station[J]. Scientific and Technological Innovation, 2024(3): 36-39. (in Chinese)
[12]
XUZ, FENGJ, ZHANGZ, et al. Water level estimation based on image of staff gauge in smart city[C]//2018 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computing, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI), IEEE, 2018: 1341-1345.
SUNChuanmeng, WEIYu, LIXinyu, et al. Intelligent detection method of image water level inversion for water level without water scale in complex scenes[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(3): 558-568. (in Chinese)
LINRuifeng, XUHai. Automatic measurement method for canals water level based on imaging sensor[J]. Transducer and Microsystem Technologies, 2013, 32(8): 53-55. (in Chinese)
JIANGGuiting, LIShuiyan. Tidal level analysis based on continuous wavelet transform and regularization method[J]. Journal of Test and Measurement Technology, 2022, 36(4): 308-313. (in Chinese)
NIUZhixing, HUChunjie, RUANCong, et al. The extraction and application of automatic water level monitoring system based on image gauge[J]. Electronic Design Engineering, 2019, 27(23): 103-107. (in Chinese)
LIXinyu, SUNChuanmeng, WEIYu, et al. Water level intelligent detection method based on fuse Transformer residual channel attention mechanism in harsh environments[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(1): 59-69. (in Chinese)
XIAPing, WANGFeng, LEIBangjun, et al. Intelligent visual water level recognition algorithm based on super-pixel and graph cut segmentation[J]. Computer Simulation, 2021, 38(3): 430-436. (in Chinese)
FANGAiyin, WANGYongxian, YINXimeng, et al. DetSegNet: a high-precision water gauge level detection network based on detection and segmentation[J]. Journal of Hohai University (Natural Sciences), 2024, 52(2): 28-34. (in Chinese)
WANGLinlin, MENGLiang, BUBoya, et al. Concrete dam apparent crack characteristic extraction based on improved DeepLabv3+ model[J]. Journal of Southeast University (Natural Science Edition), 2024, 54(4): 929-936. (in Chinese)
[33]
ZHUQ, ZHUANGH, ZHAOM, et al. A study on expression recognition based on improved mobilenetV2 network[J]. Scientific Reports, 2024, 14: 8121.
NIUSiqi, MARui, XUXiaolin, et al. Mobile NetV2 maize seed variety recognition based on improved attention mechanism CBAM[J]. Journal of the Chinese Cereals and Oils Association, 2024, 39(3): 159-165. (in Chinese)
[38]
LIF, MOUY, ZHANGZ, et al. A novel model for the pavement distress segmentation based on multi-level attention DeepLabV3+[J]. Engineering Applications of Artificial Intelligence, 2024, 137: 109175.
[39]
MAHAJANC N, JADHAVA. Improved Yolov7 tiny with global attention mechanism for camouflage object detection[J/OL]. Journal of the Institution of Engineers (India): Series B, 2024.
WANGGuogang, LUOXianglian. Image cascade semantic segmentation algorithm based on deartifacting dilated residual and bidirectional attention pyramid enhancement[J]. Journal of Test and Measurement Technology, 2024, 38(6): 575-585. (in Chinese)
[42]
ZHUF, LID, LIH, et al. Person re-identification based on an improved DeepLabv3+ semantic segmentation network[J]. Engineering Applications of Artificial Intelligence, 2025, 157: 111384.
[43]
SAMUDRALAS, MOHANC K. Semantic segmentation of breast cancer images using DenseNet with proposed PSPNet[J]. Multimedia Tools and Applications, 2024, 83(15): 46037-46063.
[44]
SHIJ, CUIR, WANGZ, et al. Deep learning HRNet FCN for blood vessel identification in laparoscopic pancreatic surgery[J]. NPJ Digital Medicine, 2025, 8: 235.
YANGDongkuan, GAOGuanglei, YINShiyu, et al. Research on optimization algorithm for intelligent detection of road cavity diseases based on DFES-YOLOv5s[J]. Journal of North University of China(Natural Science Edition), 2025, 46(4): 499-507. (in Chinese)