Map Symbol Recognition Based On Neural Network Research

The thesis is based on the automatic recognition of map symbols and consists of four parts:(1)The research of point map mark and the analysis of recognition methods in existence.(2) Recognizing point map mark through nerve network. The flow of point map mark recognizing is designed in detail. A model of BP neural network for three layers is adopted and the network is improved by the set up of input/output and hidden knots design, the method of momentum item and the method of variable run length. The program is improved by data form analysis. A rule of parameters change is concluded through experiment.(3) Improving BP neural network for automatic recognition through converting map digit symbols to standard point mark by pretreatments.(4) A digits of map recognition system is developed by VC++6.0, which is able to accomplish pretreatments including gray processing, binarization and sharpness of map scanning images, and to recognize digital symbols automaticly.A digits of map recognition system based on the improved BP neural network is built up in this paper. The results of experiment show that this system can recognize the difference scale and angle digits of map and get high correct rate at a certain extent. It is worth being studied later.