复杂背景的桥梁下部结构裂缝图像分割.doc

┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 装 ┊ ┊ ┊ ┊ ┊ 订 ┊ ┊ ┊ ┊ ┊ 线 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ PAGE I 摘 要 传统的桥梁裂缝检测费时、耗力又影响交通，专业人士通过不断地实验研究提出了一种基于人工视觉的桥梁裂缝检测技术，图像分割作为该技术中一种重要的操作技术备受关注。图像分割有很多种算法，基于数学形态的分水岭算法具有能够对微弱边缘产生良好的响应的优点，因此被广泛的应用。但由于图像具有一定的噪声和图像梯度的局部不规则形，使传统的分水岭算法在分割的过程中存在严重的过分割现象。 本论文着重研究分水岭的改进算法。控制标记符的分水岭算法是一种现存的改进算法，在经过简单的实验分析后发现，该算法由于单阈值调节,对桥梁裂缝图像的分割普遍存在误分割和漏分割现象。针对克服以上两种分水岭算法的缺点，本论文提出了基于双阈值的控制标记符分水岭算法。此算法通过一个总阈值来控制图像的外部标记的提取，另外一个阈值和总阈值共同控制内部标记的提取，最后对经过内外部标记的重构图进行分水岭分割。这种算法对于桥梁裂缝图像分割具有比较普遍的适用性。 关键字：桥梁裂缝 图像分割 分水岭 控制标记符 双阈值  ABSTRACT The traditional crack detection of bridge is time-consuming, consumptive, and it often encumbers the traffic. In order to overcome these disadvantages, some researchers came up with a technology, based on artificial vision to detect the cracks of a bridge. In this intelligent technique, the image segmentation is a certain important operating procedures，and at present it has attracted researchers’ much attention. As one of the image segmentation algorithms, mathematical morphology watershed algorithm is widely used for its advantages to having a good response to weak edges. However it exists serious over-segmentation phenomenon in the process of segmentation, which results from images noise and the images part irregular gradient. This paper emphasizes to study the improved watershed algorithm.Marker controlled watershed algorithm is an improved algorithms based on the traditional watershed algorithm. After the experiment and simple analysis, we find this algorithm exists error and leakage segmentation phenomenon resulting from the single threshold adjustment .In order to overcome the weakness above two kinds of watershed algorithm, this paper raises a marker controlled watershed algorithm based on double threshold. we extract the external markers of the image through a total threshold, and internal markers of the image is extracted through a total threshold combined with another threshold . As to the image segmentation of a bridge crack, thi