摘要
溢洪道是水利工程中重要的泄水结构,保障其稳定地泄流对调节水库水位、保障挡水建筑物安全等具有重要意义。现阶段溢洪道检测自动化水平低、缺少专门的管理人员,发生堆积险情上报难以满足即时性要求。因此,利用计算机视觉技术建立了一种便捷、轻量化的溢洪道块石堆积检测方法。先通过室内试验模拟可能出现的块石堆积情况,建立一套图像采集系统拍摄不同形态块石造成的溢洪道泄槽堵塞;然后以常规DeepLabV3+模型为基础,将模型中backbone替换为Res Net50和Mobile NetV2网络结构。通过对比试验表明,修正后模型的分类准确率和平均交并比均有所提升。综合考虑模型性能和图像分割效率,最终采用Mobile Net V2网络结构作为改进的DeepLabV3+模型的backbone。试验结果表明,改进的DeepLab V3+模型的平均交并比较常规模型提升了12.61%,测试集中的平均图像分割时间为277.93 ms。研究所用方法能够为溢洪道自动化检测提供基础。
The spillway is an important water release structure in water conservancy projects.Ensuring stable water discharge is crucial for regulating the reservoir's water level and ensuring the safety of water-retaining structures.Current-ly,the automation level of spillway detection is low,and there is a lack of specialized management personnel,making it difficult to report accumulation risks promptly.This study utilized computer vision technology to establish a convenient,lightweight method for detecting rock accumulation in spillways.Firstly,indoor experiments were conducted to simulate the possible accumulation of rocks,and an image acquisition system was established to capture the blockage of the spill-way chute caused by rocks of different shapes.Based on the conventional DeepLabV3+model,the backbone was re-placed by ResNet50 and MobileNetV2 network structures.Comparative experiments show that both accuracy and mean Intersection over Union of the revised model are improved.Considering the model performance and image segmentation efficiency,the MobileNetV2 network structure was ultimately used as the backbone of the improved DeepLabV3+model.Experimental results indicate that the mean Intersection over Union of the improved DeepLabV3+model achieves 12.61%higher performance than the conventional model,with an average image segmentation time of 277.93 ms on the test set.The proposed method can provide a basis for the automated detection of spillways.
作者
焦修明
张玉贤
张继勋
王虞清
赵昱
JIAO Xiu-ming;ZHANG Yu-xian;ZHANG Ji-xun;WANG Yu-qing;ZHAO Yu(Zhejiang Institute of Hydraulics and Estuary,Hangzhou 310020,China;College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098,China)
出处
《水电能源科学》
北大核心
2024年第11期126-130,共5页
Water Resources and Power
基金
浙江省水利河口研究院院长科学基金项目(ZIHE21Z004)。
作者简介
焦修明(1980-),男,硕士、高级工程师,研究方向为水利工程管理和信息化,E-mail:476189199@qq.com。
