A novel technique for contour reconstruction to DEM

Abstract

Contour lines recognition from raster topographic map is an interesting research in Geographic Information System. The majority of previous studies, either associates with manual digitizing, interactive or semi-automated process has several drawbacks. Generally, these approaches are time-consuming, inefficient and produced unreliable digitized data. Due to textured background and diverse information characterize on the map, extraction process has become complicated. This study concentrates on establishing an efficient approach for contour line recognition using local geometry and relationship derived from their neighbourhood pixels locality. The approach consists of three fundamental phrases: (1) preprocessing to obtain initial contour layer, (2) processing for enhancement and reconnection, and (3) post-processing to validate the extracted contour information. The entire methodology is decomposed into a number of tasks: (1) scan paper-based map into raster image, (2) colour quantization to reduce colour, (3) map layer separation and integration to acquire a comprehensive contour layer, (4) contour lines enhancement and thinning, (5) endpoints detection, (6) derivation of local geometries (distance, angle and directional information), (7) determination of best connection with identified endpoints, (8) reconnection of broken contour lines using cubic spline interpolation. In order to evaluate the applicability of this study, several experiments are conducted on topographic map in different scale and resolution. The results indicated that the usage of colour quantization is capable to extract contour layer from topographic map. Alternatively, the employment of cubic spline interpolation for contour lines reconnection has proven a favourable result. As a conclusion, this study has shown an acceptable result with low computation time and cost.