摘要
装配式建筑钢结构表观病害检测过程中,受环境和主观因素影响的细小病害检测性能较低。因此,利用BP算法,设计一种新的装配式建筑钢结构表观病害检测方法。在检测车上搭载工业相机,采集装配式建筑钢结构图像,通过线性灰度拉伸、非线性滤波等方式,进行图像预处理。通过图像灰度共生矩阵,计算出图像纹理特征参数。依托于BP算法,构建包含数个3层神经子网络的表观病害检测模型,并通过自适应调整和训练,实现复杂环境下的装配式建筑钢结构表观病害准确检测。实验结果表明:所提方法输出检测结果的综合指数F1值在0.9以上,满足表观病害检测质量要求,优化了检测质量。
In detecting apparent defects in prefabricated steel structures,the detection of small defects affected by environmental and subjective factors has a relatively poor performance.Therefore,using the BP algorithm,a new method for detecting apparent defects in prefabricated building steel structures is designed.An industrial camera is installed on the inspection vehicle to capture images of prefabricated building steel structures,and preprocess the images through linear grayscale stretching,nonlinear filtering and other methods.The image grayscale co-occurrence matrix is employed to determine the texture feature parameters.Based on the BP algorithm,a surface defect detection model consisting of several 3-layer neural subnetworks is constructed,and accurate detection of surface defects in prefabricated steel structures in complex environments is achieved through adaptive adjustment and training.The experimental results show that the value of the comprehensive index F 1 of the output detection results of the proposed method is above 0.9,which meets quality requirements for apparent disease detection and optimizes detection quality.
作者
曹园
CAO Yuan(School of Architectural Engineering and Surveying and Mapping,Hefei College of Finance&Economics,Hefei 230061,China)
出处
《太原学院学报（自然科学版）》
2024年第1期7-12,共6页
Journal of TaiYuan University:Natural Science Edition
基金
2022年安徽省自然科学研究项目(2022AH053041)。
关键词
BP算法
装配式建筑
钢结构
表观病害
梯度下降法
BP algorithm
fabricated building
steel structure
apparent defect
gradient descent method
