License plate recognition is the most important means of determining vehicle identity in intelligent transportation management applications, and its recognition speed is related to the operational efficiency of the entire system.This paper proposes a GPU-based parallel programming method for the traditional license plate location and recognition algorithms, which uses the excellent parallel computing power of GPU and high memory bandwidth to improve image processing speed. A corresponding GPU parallel design model is proposed for character segmentation and recognition algorithm in license plates, and a CUDA program is written for simulation calculations. The simulation results show that, compared with the traditional single-CPU-based algorithm, the speed of the algorithm based on the cooperation of CPU and GPU is 24.94 times faster.