Point Symbol Recognition Algorithm based on Improved Generalized Hough Transform and Nonlinear Mapping

Abstract

The point symbol recognition in the color-scan topographic map is a difficult task in the pattern recognition of the map, which has the problems of low quality of the scanned topographic map, small symbol area, diversified shapes, superimposition with other topographic elements in the map and difficulty of separation. Therefore, many traditional recognition methods, which usually based on pre-segmentation and recognition, do not perform well in the face of such a complex condition. In this paper, we present a new method of point symbol recognition based on improved generalized Hough transform and nonlinear mapping. The proposed method contains two steps. At the first step, an improved generalized Hough transform algorithm is used to extract the target symbol from the original topographic map directly. At the second step, through the method of non-linear mapping and template matching, the preliminary identification of the region for further screening and correction. Experimental results prove that our proposed method can directly recognize and locate point symbols in color topographic maps and outperforms many traditional point symbols recognition algorithms both in both the accuracy of the location of the symbols and the accuracy of the symbol recognition.