A Terrain Elevation Map Generation Method Based on Self-Attention Mechanism and Multifeature Sketch

Area of Science:

• Geographic Information Science
• Computer Vision
• Artificial Intelligence

Background:

• Manual terrain feature map annotation is inefficient.
• Existing terrain elevation map generation lacks realism.
• Need for advanced methods in geospatial data visualization.

Purpose of the Study:

• To develop an efficient and realistic terrain elevation map generation method.
• To improve the annotation process for terrain feature maps.
• To enhance the visualization and interaction with 3D terrain models.

Main Methods:

• Utilized adaptive feature enhancement for terrain elevation map data.
• Integrated a self-attention mechanism within a conditional generative adversarial network (cGAN).
• Employed a level of detail (LOD) method for 3D terrain visualization and interactive editing tools.

Main Results:

• The proposed method achieved high subjective visual performance.
• Objective criteria confirmed the method's effectiveness.
• Demonstrated significant advantages over current typical terrain generation methods.

Conclusions:

• The self-attention mechanism enhances the capture of global spatial features for realistic map generation.
• The developed method offers improved efficiency and realism in terrain mapping.
• The interactive tools facilitate better 3D terrain exploration and editing.