Optical recognition of hand-printed characters of any size, position, and orientation

This paper deals with the optical recognition of text data in documents such as engineering drawings, land-use and land-register maps, and utility maps. The automatic computer acquisition of these documents is performed through the basic steps of vectorization of the line-structure and recognition of the text data interspersed in the document. The latter data are usually handwritten by professional draftsmen, and may have any size, position, and orientation. We review some of the features appropriate to this particular OCR problem, and suggest a special recognition strategy. Numerous examples are given. The results obtained with a prototype system on actual land-register maps are reported.