【摘要】 在对既有砌体结构进行外观检测时，需要从采集的外观图像中准确地识别缺陷信息，但由于既有砌体结构所处环境干扰较多，通常无法获取清晰的图像。为此，引入近场动力学微分算子(PDDO)对采集的墙体外观图像进行像素修复。首先读取图像全部信息并将像素点数字化，然后遍历所有像素点并判断其是否有效。将无效像素点(污点)从近场动力学点族成员中去除后，PDDO利用周围的有效像素点恢复被去除的信息。PDDO的实现要求解大型稀疏线性方程组，其计算量会随着图像的精细程度急剧增长。为提高计算效率，采用英特尔数学核心函数库(MKL)来求解。结果显示，上述方法对不同受损程度的图像，均能够保证恢复后的图像变化连续，且可以明显提高复原图像的信噪比和视觉效果，有效恢复被污点掩盖的信息，有利于准确识别构件缺陷。更多还原

【Abstract】 When detecting existing masonry structures, it is necessary to accurately identify deficiencies from the captured photos of the walls’ appearance. However, due to the environment interference of existing masonry structures, it is very difficult to obtain clear images. To address the issue, the Peridynamic Differential Operator(PDDO) is introduced to recover the captured appearance images of masonry walls. First, all information of image is read and the pixels are digitalized. Then, the validity of each pixel point is examined and the invalid points are removed from the peridynamic point family members. The PDDO is used to recover the information of the removed pixels, by using the information of their surrounding valid pixel points. Implementation of the PDDO requires solving large scale sparse linear systems of equations, and the computational demand grows dramatically with the fineness of the images. To improve the computational efficiency, the equations are solved by using Intel Math Kernel Library(MKL). The results show that the above methods can ensure the continuity of the recovered images and significantly improve their signal-to-noise ratio and visual effect. It effectively recovers the information masked by stains and is helpful for accurate identification of deficiencies.更多还原