Automatic Text Recognition from Raster Maps

Text labels in raster maps provide valuable geospatial information by associating geospatial locations with geographical names. Although present commercial optical character recognition (OCR) products can achieve a high recognition rate on documents, text recognition on raster maps is still challenging due to the varying text orientations and the overlapping between text labels. This paper presents an automatic text recognition approach which focuses on locating individual text labels in the map and detecting their orientations and leverages horizontal text recognition of commercial OCR software. We show that our approach detects all strings in the test maps and achieves 96.8% precision and 95.7% recall on character recognition.