Research on vehicle license plate location algorithm based on convolutional neural network

This paper uses Faster RCNN to compromise license plate location algorithm to improve the poor positioning and recognition effect of multiple license plates in the current complex environment. Considering the information around license plate is important for detecting, this paper uses modified GBD-net after Fast RCNN aiming to take use of this kind of information. The region of license plate is respectively small and the location is misalignment, so this paper uses location sensitive map just like R-FCN to further adjust the coordinates of license plate. The License-net proposed by this paper achieves better location accuracy.