中文摘要：

针对光照不均匀与水体模糊效应导致水下图像中建筑物裂缝提取难度增加的问题，提出一种基于匀光处理的自适应阈值分割算法。首先，通过对全局图像进行匀光处理。在保护纹理信息的同时消除亮度分布不均的问题。然后，使用Hessian矩阵去除部分非裂缝信息，根据处理前后图像直方图的变化自适应确定分割阈值。最后，使用张量投票对分割结果进行线性增强，以实现去除噪声、连接裂缝片段的效果。实验结果表明：该算法不仅能对水下光照不均匀图像中的弱裂缝进行有效分割提取，同时也适用于常见水上裂缝图像。

英文摘要：

Aim at the problem that uneven illumination distribution and blur eltect ot unoerwater environment increase the difficulty of extracting the crack from the concrete image, an adaptive crack segmentation and extraction algorithm based on the illumination balance process is proposed. In the proposed approach, global illumination balance process is used to deal with the uneven illumination problem, while keeping the texture unchanged. Then, part of the pixels is removed bY the Hessian matrix, which is not belong to the crack. Threshold value is obtained based on the change of the histogram of the image. At last, tensor voting is used to enhance the linearity of the crack and remove noise. The experimental results show that the proposed approach can deal with not only the un- derwater weak crack image with uneven illumination distribution, but also the image out of the water.