中文摘要：

提出了一种基于正射影像视差图的区域级镶嵌线搜索算法。首先利用半全局约束立体匹配算法（semi-global matching，SGM）分别计算立体像对的左右视差图，并通过自适应阈值化去除细小的噪声区域，再经数学形态学方法进一步削弱噪声影响和填补小的漏洞区域，得到了较为精细的房屋等非地面区域的分割结果，从而分离出地面与非地面区域；然后采用改进的贪婪蛇搜索算法进行镶嵌线搜索，以提高算法的稳健性。试验表明，本文算法能很好地避开房屋等明显突出地表的实体，得到不穿越非地面区域的最优路径。

英文摘要：

This paper proposes a regional seam line detection algorithm based on the disparity image of the orthophoto. Firstly, the disparity images of the stereo pair are generated by using the SGM （semi-global matching） method, and the small noises are removed through the adaptive threshold. Then, the mathematical morphology method is used to further eliminate the noises and to fill the small holes on the disparity image in order to accurately separate out the non-ground area. Lastly, the seam line is detected using the improved greedy snake algorithm which with better robustness. The experimental result has shown that our method is effective and is able to acquire an optimized seam line that passed around the entities above the ground.