Embedded Based License Plate Super-resolution Recognition Algorithm
In real shooting scenarios,due to the limitations of imaging equipment performance and long-distance shooting,the number of pixels collected for license plates will decrease.Improving the quality of imaging equipment incurs high costs.Due to involve long-distance scenes,therefore,super-resolution reconstruction of license plates with fewer pixels is crucial.To this end,the super-resolution recognition algorithm for license plates based on the Atlas200DK embedded device was proposed.The inference process of the entire network was conducted on the Atlas200DK embedded development board,and was detected in natural scenes containing long-distance and fuzzy information.The results show that the detection speed is 33 fps,the accuracy of license plate positioning is 99.2％,and the accuracy of license plate recognition is improved by 10.6 percentage point compared to direct recognition,reaching 91.9％.The algorithm can improve the accuracy of license plate recognition without increasing costs and losing accuracy,which is of great significance for the large-scale application of intelligent transportation.
万方数据
维普
