Research of Extraction Method of Road Elements from Road surface Image Based on Morphology.

在路面的维护中，需要检测路面上的白色线和裂缝。在这项研究中，目的是提取结构元素，如，裂缝，一个白色/黄线，一个关节和一个人孔的路面图像。由于道路表面图像的低对比度和野性，这些元素的提取是不容易的。本文首先介绍了一种单独提取各元素的方法，并取得了足够高的提取率。另一方面，例如，将人孔周围的微小沟提取为假裂缝。提出了一种基于规则的松弛方法来解决误提取问题。此外，我们获得了大量的新的路面图像数据，并进行了提取实验。结果，存在先前使用的图像数据集中未包括的“污点”，并且它可能是假裂纹，因此我们改进了一些参数以避免假提取。另外，在提取白色中心线/黄色中心线时，需要利用Jansen Shannon散度对边缘提取阶段的参数进行细化。通过这些改进，提出的路面元素提取方法具有足够高的性能，以新的图像数据集。

In maintenance of a road surface, it is necessary to detect white lines and cracks on a road surface. In this study, the aim is extracting structural elements such as, a crack, a white/yellow line, a joint and a manhole in the road surface image. The extraction of such elements is not easy due to the low contrast and wildness of a road surface image. At first we introduced a method to extract each element individually and achieved high enough extraction rate. On the other hand, for example, a tiny ditch around a manhole is extracted as a false crack. We proposed a rule based relaxation method to resolve the false extraction. In addition, we obtained a large quantity of new road surface image data and performed an extraction experiment. As a result, there was "a stain" that was not included in previously used image data set and it could be a false crack, so we refined some parameters to avoid false extraction. In addition, in extraction of the white center line / yellow center line, a refinement the parameter was necessary about edge extraction stage by the Jansen Shannon divergence. Through these refinements, proposed road surface element extraction method has highly enough performance to new image data set.