Ultrasonic technique based on neural networks in vehicle modulation recognition

Adapted learning rate η according to the convergence pace has been used in back-propagation neural networks (BPNN). Several vehicle recognition methods are compared and ultrasonic sensor has been given. Vehicle´s portrait section is acquired by ultrasonic sensor and the character been distilled as the BPNN´s input. Three layers BPNN with 6 inputs, 12 hidden nodes and 3 outputs run in the vehicle recognition program, and 6 different kinds of vehicles can be most correctly distinguished. The application in practice shows its recognition rate achieved 93.2%. Data is stored in server to provide different use.