NEAR模型的外墙面裂缝连接算法

Exterior Wall Crack Connection Algorithm for NEAR Model

在利用数字图像技术检测大型建筑物外墙面裂缝时,墙面上的裂缝宽度过窄,或是被灰尘填充而变得不清晰,又或是图像预处理过度导致提取到的裂缝出现断裂等情况,均会影响到后续如裂缝面积、长度、宽度、走势等参数的测量和评估.为了解决这个问题,本文提出了一种基于山脊线邻域的评价模型(NEAR)来判断和描述裂缝间的连续性,并根据其判断和描述结果,结合二次贝塞尔插值完成裂缝间的平滑连接.此外,本文还进行了对比试验,利用基于Hausdorff距离的最小二乘法对NEAR算法结果和K-D树算法结果进行评估比较.实验结果表明,该算法能明显提高判断裂缝连续性的准确程度.

There exists undeniable bias during the process of detecting,measuring and assessing wall cracks on buildings.It is because of dusts,the clarity of cracks and image overpreprocessing.In this paper we proposed a NEAR(NEighborhood Assessment of Ridge)model to determine and describe the continuity of cracks,according to which we could stitch cracks by means of quadratic Bezier interpolation.Then the results were compared with that applying k-d tree algorithm by Hausdorff-distance-based least square method,which show an evident superiority of NEAR model.