彩色扫描地形图的分色方法

Color segmentation algorithm for scanned color map

本文提出了一种新的地形图分色方法。首先去除背景像素,然后根据灰度梯度值提取主色像素,利用直方图模糊c-均值(FCM)聚类方法对主色像素进行颜色聚类,对去背景后的图像进行Canny算子的边缘检测。最后,利用加壳变换和障碍距离变换工具对符号周围的过渡像素进行聚类,从而实现黑棕兰绿四个分版图的提取。该方法效率较高,受图纸扫描质量的影响不大,分色效果较理想。

An improved color segmentation algorithm was presented in the paper. The background pixels were removed firstly. The main color pixels of elements were extracted by gradient gray value and clustered by histogram fuzzy c-means （FCM） algorithm. The Canny edge detection for the image after removing background pixels was carried out. Based on adding-shell transformation and obstacle distance transformation the transition pixels on symbol brink were clustered, and the black, brown, blue and green element maps were extracted finally. This algorithm was few influenced by scanning quality, and the efficiency and effect could be satisfactory.