Using the vehicle laser radar point cloud data, a forest archive model with "one tree, one file" information was built, the vector data, such as key protection goals, forest roads and water sources was classified and extracted to provide basic data support for the spread of forest fire data including. Based on the forest archives models, self -adaptive forest fire spread model libraries was built, the model correction coefficient was optimized through machine learning, big database and other means, to enhance the coupling characteristics between each model, and promote the three -dimensional simulation capabilities of forest fires. Relying on environmental information such as terrain, vegetation, climate, etc., the most suitable tactical combat and emergency avoidance measures are given to provide important references for commanders to command decisions.