Abstract: This paper discusses the algorithm of extracting the main skeleton line of polygon.Through the mathematical morphology thinning operation on the grid cell of polygon, we can get the skeleton of the polygon including many short lines.Before extracting the longest skeleton, the short lines must be removed.This paper explores the way of vectorization based on node point of grid data and the way of extracting main skeleton line of polygon based on end point of grid data.The main skeleton line using the extracting algorithm is an optimum one.The above two ways are easily implemented by a recursive program. There are three main steps in the algorithm:①Rasterizing the polygon and then using the thinning operation on the grid data.Through this step, we get the grid skeleton lines.②Vectorizing the grid skeleton lines and building the topological relation of lines.③Extracting main skeleton line from the vector skeleton lines. Step ② and step ③ are the paper's kernel. FK(W14。40ZQ It should be pointed out that the efficiency of mathematical morphology thinning operation directly depends on grid data volume.If the size of grid cell is small, data volume is large, it increases the time of computation.If the size is big, it is not propitious to extract accurately the skeleton line of small polygon.The dynamic grid data structure can solve this problem.The structure stores the start row and start column, the total rows and total columns, the grid cell size, and original point of polygon.Before rasterizing, the total rows and columns are fixed firstly such as 100.Secondly, the size of grid is calculated.Finally, the polygon is rasterized.Because the rows and columns are fixed, for large polygons, the data volume of grid cell becomes less, but for small polygons, the data volume of grid cell is increased.It is obvious that the dynamic raster structure has better adaptability. The result of the experiment shows that the algorithm of extracting skeleton line based on dynamic raster data structure is effective.The algorithm has many advantages such as taking less time and less memory.