摘要
施工初期对钢筋的规范绑扎是安全施工的基础,其中钢筋的间距是衡量绑扎效果的一大指标,为了防止绑扎后的钢筋间距不满足要求规范而引发的相应安全事故,提出一种基于深度估计模型AdaBins及改进YOLOX的钢筋间距检测方法。利用深度估计算法获得深度图,从多层的钢筋架构中分割中最上层钢筋,再利用目标检测算法对钢筋进行检测并判断其间距是否满足要求。针对钢筋的特点对目标检测算法进行改进,使用注意力模块并改进多尺度结构,只保留感受野最大的两个检测头。实验结果表明,在对钢筋检测任务中,改进模型均值平均精度高达85.35%,高于其他目标检测算法,满足钢筋间距检测的准确性要求。
The standard binding of reinforcement in the early stage of construction is the basis of safe construction,and the spacing of reinforcement is a major index to measure the binding effect.In order to prevent the corresponding safety accidents caused by the spacing of reinforcement after binding does not meet the required specifications,a reinforcement spacing detection method based on depth estimation model AdaBins and improved YOLO X is proposed.The depth map is obtained by using the depth estimation algorithm,the top reinforcement is segmented from the multi-layer reinforcement structure,and then the object detection algorithm is used to detect the reinforcement and judge whether its spacing meets the requirements.According to the characteristics of reinforcement,the object detection algorithm is improved,the attention module is used and the multi-scale structure is improved,and only the two detection heads with the largest receptive field are retained.In order to improve the generalization of detection,the grounding truth of training samples is randomly masked to cover some pixels,and the output of MAE pre-training model is fused.The experimental results show that in the reinforcement detection task,the average accuracy of the improved model is as high as 85.35%,which is higher than other object detection algorithms,and meets the accuracy requirements of reinforcement spacing detection.
作者
戴振国
陈国栋
赵志峰
张旭生
陈子健
林进浔
黄明炜
DAI Zhenguo;CHEN Guodong;ZHAO Zhifeng;ZHANG Xusheng;CHEN Zijian;LIN Jinxun;HUANG Mingwei(College of Physics and Information Engineering,Fuzhou University,Fuzhou 350108,China;Fujian Shuboxun Information Technology Co.,Ltd.,Fuzhou 350002,China)
出处
《佳木斯大学学报（自然科学版）》
CAS
2023年第3期127-131,150,共6页
Journal of Jiamusi University：Natural Science Edition
基金
国家自然科学基金资助项目(61471124)
福建省科技计划引导性项目(2021H0013)
福建省科技型中小企业创新资金项目(2021C0019)。
