Linear Feature Extraction from Scanned Color Maps with Complex Background

Map digitization is one of the most important means of geographic data acquisition. Automatic extraction of geographic features from scanned maps is the key technique to improve the efficiency of map digitization. With regard to linear feature extraction from color topographic map images, it is usually performed by interactive or automatic line tracking based on color separated binary images. But for those color maps with heavy interconnectedness of geographic features and complex background such as vegetation tints and relief shadings, the results of color image segmentation cannot meet the demand of automatic extraction, and time-consuming manual tracking is inevitable. This paper presents a new semi-automatic method to extract linear features directly from original scanned color maps without color layer separation. In the proposed method, a sliding window is added on a user-specified linear feature, and the current line in the window is first segmented adaptively by using color space conversion, k-means clustering and directional region growing. After that, a thinning operation is performed and the line in the window is tracked from the centre to the edge. By moving the window continuously along the line, iterative operations of image segmentation, thinning and line tracking are accomplished, thus the specified line is tracked automatically. Meanwhile, a little manual processing is incorporated in case automatic tracking fails. The performance of the proposed method is tested on a number of color map samples with complex background.