基于TLBO的工程结构表面缺陷图像边缘检测方法

TLBO-based Edge Detection Method for Structural Surface Defect Image in Engineering

非接触式、数字化的工程质量检测方法对于快速发现建筑结构表面缺陷,如裂缝、焊接缺陷等,降低工程质量检测的劳动强度具有重要的工程意义。本文在计算机数字图像处理技术基础上,根据建筑结构表面缺陷图像特征,提出基于TLBO算法的缺陷图像轮廓识别预处理方法,作为进一步缺陷特征判断的依据。本文在TLBO算法基础上,边缘像素点的搜索不需要设定任何算法参数,实现简单;提出基于的8个方向的灰度导数,建立图像边缘强度矩阵,将边缘点附近的小规模局部搜索和大量的全局搜索相结合,TLBO算法保证了所提出的边缘检测方法不会陷入局部边缘点,找到最重要的图像全局边缘特征;将TLBO算法应用于图像边缘检测,以工程质量检测中常见的钢结构焊缝检测为例加以验证和分析,证明了本文方法在缺陷图像轮廓识别预处理中的抗噪性和有效性。

Non-contact and digital engineering quality inspection method is of great engineering significance for quickly discovering surface defects of building structures,such as cracks and welding defects,and reducing labor intensity of engineering quality inspection.In this paper,based on computer digital image processing technology,according to the characteristics of building structure surface defect image,the contour recognition and preprocessing method of defect image based on teaching-learning-based optimization(TLBO)algorithm is proposed as the basis for further defect feature judgment.Based on the TLBO algorithm,the search of edge pixel points does not need to set any algorithm parameters,which is simple to implement;the gray derivative based on eight directions was innovated,the image edge intensity matrix was established,and the small-scale local search near the edge points was combined with a large number of global searches.The TLBO algorithm ensured that the proposed edge detection method would not fall into the local edge points and find the most important global edge features.The TLBO algorithm is applied to image edge detection,and the weld detection of steel structure is taken as an example to verify and analyze.The noise resistance and effectiveness of the proposed method in the contour recognition pretreatment of defect image are proved.