论文部分内容阅读
为了检测混凝土表面裂缝及其宽度,对含有裂缝的数字图像进行k-均值聚类,提取出所有疑似裂缝的像素点,进行了二值化处理。根据像素点的位置关系,提取连通分量,将连通分量作为聚类对象,构造连通分量间的距离函数。利用谱聚类算法将连通分量聚类,根据裂缝特征,去掉伪裂缝部分,得到完整的裂缝对象,并通过局部旋转算法对裂缝的宽度进行了2次数值计算。分析结果表明:与Canny、Sobel算子比较,多级聚类算法在裂缝提取时能去掉较多的噪声,抗噪能力强;通过局部旋转算法计算裂缝宽度时,计算值与实际值的平均相对误差分别为3.86%、2.40%,算法精度高,适用于各种类型裂缝宽度计算。
In order to detect the cracks and their widths on the concrete surface, k-means clustering was performed on the digital images containing the fractures to extract the pixels of all the suspected fractures, and binarization was carried out. According to the position of pixels, the connected components are extracted, and the connected components are regarded as clustering objects to construct the distance function between connected components. The clustering algorithm was used to cluster the connected components. According to the fracture characteristics, the part of the pseudo-crack was removed and the complete crack object was obtained. The crack width was numerically calculated by the local rotation algorithm. The analysis results show that compared with Canny and Sobel operators, the multi-level clustering algorithm can remove more noise during crack extraction and has better anti-noise ability. When calculating the crack width by the local rotation algorithm, the average of the calculated value and the actual value The errors are 3.86% and 2.40%, respectively. The algorithm has high precision and is suitable for various types of crack width calculation.