The purpose of this article was to build a license plates recognition system with high accuracy at night. The system, based on regular PC, catches video frames which include a visible car license plate and processes them. Once a license plate is detected, its digits are recognized, and then checked against a database. The focus is on the modified algorithms to identify the individual characters. In this article, we use the template-matching method and neural net method together, and make some progress on the study before. The result showed that the accuracy is higher at night.