基于启发式图搜索的遥感影像道路半自动提取

启发式图搜索法用于线状目标识别的原理是：用图结构表示边缘点和边缘段，根据启发函数计算顶点权值，在图的路径上建立相应的代价函数，通过在图中搜索对应的最小代价的通道以找到最优路径。图搜索法是一种全局最优方法，它在受噪声影响较大时效果仍然较好。文中使用了启发式图搜索法（A*算法）实现了道路的半自动跟踪。它的基本思路是：首先利用自适应平滑滤波算子进行道路信息增强，然后对传统的道路数学模型进行了进一步的扩展，突出了对道路几何特性和辐射特性的描述，并依此构建图搜索的代价函数，实现了基于启发式图搜索法A*算法的道路半自动跟踪。经实验证明，该方法进行遥感影像的道路半自动提取效果较好。

Semi-automatic Extraction of Roads from Satellite Imagery Based on Heuristic Graph Search

The theory of utilization of Heuristic Graph Search techniques in recognizing line object is to use graph structure to express the edge points and edge segments, to caculate power value to vertices according to the heuristic function, and then to build corresponding cost function on the graph path. Graph Search algorithm is a global optimum operator, so it has good results even if there are influences of noise. In this paper, a semi-automatic road extraction scheme is proposed in detail. At first an adaptive low-pass filter is built up for road enhancement, and then a road model is formulated with six photometric and geometric properties, at last road semi-extraction is realized sequentially by A* algorithm. Some experimental results are also presented in this paper.