A system for efficient and robust map symbol recognition

Abstract

We present a system that allows the user to input maps into a geographic information system (GIS) by using automatic symbol and line recognition. The system is composed of a user interface, a symbol recognition engine, a knowledge base and a database. The recognition is based on the Hausdorff distance and neural networks, where our main contribution is to make the recognition efficient and robust for handling very large maps and many symbols of different scales and orientations. The system allows for efficient and coherent management of maps, recognition processes and recognition results.