| 残差网络深度学习在工地扬尘监管中的应用研究 |
| Research on the Application of Residual Network Deep Learning on Construction Sites for Dust Supervision |
| 投稿时间：2024-06-11 修订日期：2025-03-24 |
| DOI：10.19316/j.issn.1002-6002.2025.04.17 |
| 中文关键词: 工地视频 深度学习 扬尘监管 裸土未苫盖 |
| 英文关键词:construction site video deep learning dust supervision uncovered bare land |
| 基金项目:北京市科技计划项目(Z231100003823018) |
|
| 通讯作者:沈秀娥* ;大气颗粒物监测技术北京市重点实验室, 北京 100048 |
| 摘要点击次数: 411 |
| 全文下载次数: 1129 |
| 中文摘要: |
| 施工扬尘是影响城市空气质量的主要因素之一,因此要对施工扬尘开展精细化监管。利用视频监控对全市施工工地进行人工巡查,检查施工工地内的裸土未苫盖情况,存在工作强度高、效率低的问题。对此,将深度学习技术应用在识别施工工地视频图像中的裸土未苫盖问题上。制作了无数据增强样本和经过数据增强处理的样本,利用ResNet18、ResNet34、ResNet50三种算法对两组样本分别进行模型训练并评估模型验证集的准确率。结果显示,图像预处理和数据增强处理可以有效提升识别准确度。基于ResNet50算法形成的未苫盖裸地识别模型效果最佳,识别准确度可达80.65%。将构建形成的未苫盖裸地识别模型应用到工地扬尘问题监管中,可提高问题发现效率。同时,通过对发现的问题进行统计,可以分析出反复出现问题的工地,以及问题高发时段、区域和工地类型,从而为工地扬尘精细化监管提供“标靶”。 |
| 英文摘要: |
| Construction dust is one of the main source affecting urban air quality.It is necessary to carry out refined supervision for construction dust.Traditional manual inspection of uncovered soil at construction sites through video surveillance suffers from high labor intensity and low efficiency.To address this,this study applied deep learning technology to identify uncovered bare land by using video images of construction sites.Raw image samples and data-augmented samples were prepared in this study,then ResNet18,ResNet34,and ResNet50 were used to train model,and the accuracy of the model was evaluated by using these samples separately.The results showed that image preprocessing and data augmentation methods can effectively improve the accuracy of recognition.The ResNet50-based model achieved optimal performance in identifying uncovered soil areas,with accuracy reaching 80.65%.Implementing the uncovered bare land recognition model to the dust supervision of construction sites can improve the efficiency of problem discovery.Furthermore,statistical analysis of the discovered problems can identify construction sites with multiple and repeated problems,as well as the periods,regions,and types of construction sites with high incidence of problems,providing targets for dust supervision on construction sites. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
