基于CBAM-DeepLabv3+和迁移学习的施工工地区域智能识别  

Intelligent Recognition of Construction Site Workspace Based on CBAM-DeepLabv3+and Transfer Learning

作　　者：梁嘉韵 温喜廉 杨智诚 陈广浩 杨永民 LIANG Jiayun;WEN Xilian;YANG Zhicheng;CHEN Guanghao;YANG Yongmin(Guangzhou Pearl River Construction Development Co.,Ltd.,Guangzhou Guangdong 510075;School of Civil Engineering,Guangzhou University,Guangzhou Guangdong 510006;College of Urban and Rural Construction,Zhongkai University of Agriculture and Engineering,Guangzhou Guangdong 510225;Guangdong Lingnan Township Green Building Industrialization Engineering Technology Research Center,Guangzhou Guangdong 510225)

机构地区：[1]广州珠江建设发展有限公司,广东广州510075 [2]广州大学土木工程学院,广东广州510006 [3]仲恺农业工程学院城乡建设学院,广东广州510225 [4]广东省岭南乡镇绿色建筑工业化工程技术研究中心,广东广州510225

出　　处：《湖北理工学院学报》2024年第6期47-54,共8页Journal of Hubei Polytechnic University

基　　金：广东省住房和城乡建设厅科技创新计划项目(项目编号:2023-K1-463769);广州市科技计划项目(项目编号:2023A04J0647)。

摘　　要：针对建筑工地工作区的智能识别问题,提出一种基于CBAM-DeepLabv3+和迁移学习的工地工作区智能识别方法,以DeepLabv3+为基础框架,引入CBAM模块和迁移学习策略提升模型的特征提取能力和训练效果,同时采用较小的主干网络提升模型的计算速度。结果表明,该方法能够有效识别施工现场各个工作区,平均识别正确率为86.2%,IoU和F1-Score指标分别为0.85和0.86。与非迁移学习方法相比,该智能识别方法的识别效果显著提升,同时也证实了迁移学习方法能够克服样本量不足的问题。For intelligent identification of construction site workspace,an intelligent recognition method for construction site work areas based on CBAM-DeepLabv3+and transfer learning was proposed.With the DeepLabv3+framework,the model′s feature extraction capabilities and training effectiveness were enhanced by introducing CBAM modules and transfer learning strategies,and the model computational speed was improved by employing a smaller backbone network.The results demonstrate that the proposed method effectively identifies various work areas on construction sites with an average recognition accuracy of 86.2%,and critical parameters of IoU of 0.85 and of F1-Score of 0.86,respectively.Compared to non-transfer learning methods,the proposed approach exhibits significant improvements in recognition performance,which also confirms the ability of transfer learning to overcome challenges posed by limited sample sizes.

关 键 词：施工现场工作区 智能识别 CBAM-DeepLabv3+ 迁移学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]