Chinese License Plate Recognition Algorithm Based On UNet3+

Abstract

With the development of intelligent transportation system, license plate recognition technology plays an increasingly important role in our life, and many related technologies have been proposed. However, most of these algorithms have a low success rate in detecting large Angle vehicle images in complex background. In this paper, we propose a Chinese license plate recognition algorithm based on UNet3+ to solve the problem of low license plate recognition rate in large-angle scenarios. The algorithm detects license plate position from three scales to improve the accuracy of license plate position recognition. At the same time, we propose a new loss function to enhance the boundary information of license plate. Finally, on the basis of the Chinese City Parking Dataset (CCPD) Dataset, we added the large Angle vehicle pictures we collected to construct a new Chinese license plate Dataset containing 30,000 images, and increased the amount of data through various data enhancement technologies. Experimental results show that our algorithm is effective in large Angle license plate recognition in complex scenes, and the accuracy rate reaches 94.3%.