A Practice to Search the Summit of a DEM Using Simulated Annealing Technique

Simulated annealing (SA) algorithm is based on iterative improvement but with stochastic acceptance of bad moves. This allows the algorithm to escape local optimal solution in early sampling rounds and the progressive refinement into efficient solutions in later sampling rounds. In this paper, we show how this algorithm is used to find the highest peak of a Digital Elevation Model (DEM). In order to search efficiently in the terrain data, we proposed a reformed cost function which used logarithmic transformation. A set of parameters were confirmed after many experiments. The result shows that SA is not guarantee to find the optimal solution. However, there are 16% chances to find the highest peak during 400 experiments. Specifically, the search process is demonstrated by 3D visualization in GIS, which can help understand the mechanism of the algorithm and provide convenience for teaching.