基于改进Mask R-CNN的建筑钢筋尺寸检测算法

建筑施工现场钢筋图像背景复杂且干扰较多，传统图像检测算法无法有效利用特征信息，难以满足现阶段建筑智能监理行业中钢筋尺寸检测精度的验收要求。提出一种在Mask R-CNN模型基础上加入自下而上路径和注意力机制的改进模型BU-CS Mask R-CNN。在建筑工地现场拍摄图像后，整理自建钢筋数据集，并在此数据集上进行算法验证。实验结果表明，与Mask R-CNN模型相比，BU-CS Mask R-CNN模型的召回率、交并比和像素准确率分别提升了4.9%、6.8%、7.4%，钢筋直径和间距的尺寸检测精度分别提升了14.9%、4.4%，能得到更加准确的钢筋目标检测框和边缘分割掩膜，达到了行业中实际工程验收的精度要求。

Rebar Size Detection Algorithm for Intelligent Construction Supervision Based on Improved Mask R-CNN

When applied to the rebar images with complex background of construction sites, the traditional image detection algorithms cannot efficiently utilize feature information, and thus fail to meet the existing accuracy requirements of rebar size detection in the intelligent supervision industry.To address the problem, an improved Mask R-CNN model(BU-CS Mask R-CNN) with bottom-up path and attention mechanism is proposed.The model is tested on a self-made data set that consists of the images taken at the construction sites.The experimental results show that compared with the Mask R-CNN model, the proposed BU-CS Mask R-CNN model improves the recall rate by 4.9 percentage, IoU by 6.8 percentage, and accuracy by 7.4 percentage.It also improves the detection accuracy of the diameter by 14.9 percentage, and that of rebar spacing by 4.4 percentage.BU-CS Mask R-CNN can provide a more accurate rebar target detection box and edge segmentation mask, bringing the detection accuracy to the requirements of actual engineering projects.