Abstract: To address the issue of traditional contour interpolation methods failing to handle nonlinear terrain features in complex areas,a multi-layer perceptron (MLP)-based automatic interpolation method is proposed.Control points are generated by densifying contour lines,and free points with elevation constraints are created using a gridding method.Spatial coordinates are processed with positional encoding and input into an MLP model to build a nonlinear mapping from coordinates to elevations,with terrain constraints incorporated to optimize predictions.Experimental comparisons show that the method generates first contours accurately following terrain trends with smooth transitions in sharply curved areas like valleys and ridges,significantly outperforming inverse distance weighting and cubic approximation methods.The latter two exhibit node mismatches and abrupt curves in such regions.Ablation experiments confirm that elevation constraints enhance the continuity of terrain representation.The study demonstrates the effectiveness of MLP-based nonlinear modeling for contour interpolation,but it has efficiency limitations for large-scale data processing.Future research may focus on lightweight network design,constraint optimization,and distributed computing to improve practical applicability.

Key words: contour lines, topographic mapping, spatial interpolation, MLP