基于深度学习的车牌定位和识别方法
License Plate Location and Recognition Based on Deep Learning
-
摘要: 针对现有的基于车牌字符分割的车牌识别方法,在光照,阴暗等特定自然场景下存在无法定位且车牌字符无法正确分割,直接影响车牌字符识别效果的问题,提出一种基于深度学习的车牌定位和识别方法.首先采用深度学习FasterR-CNN算法进行车牌定位,利用k-means++算法来选择最佳车牌区域尺寸,解决现有车牌定位方法在某些自然场景下无法正确定位车牌的问题;然后在AlexNet网络模型的基础上进行改进和重新构建,提出一种增强的卷积神经网络模型AlexNet-L,该模型是一种针对车牌字符识别的端对端网络模型,可提高车牌识别准确率,避免现有的基于车牌字符分割的车牌识别方法中因无法正确分割车牌字符对车牌字符识别的影响.实验结果表明,该方法可以更有效地提高车牌定位和车牌字符识别的准确度和效率.Abstract: In some cases, the existing license plate recognition methods based on the character segmentation of license plates may fail in certain natural scenes such as dark illumination environments. In addition, the wrong character segmentation of the license plates directly affects license plate character recognition. In order to solve the above problems, a license plate location and recognition method based on deep learning is proposed in this paper. First, a license plate location method based on Faster R-CNN combining with the best license plate area selection using k-means++ is designed in this paper to solve the problem that the existing methods may fail in some natural scenes. Then, on the basis of the AlexNet network model, this paper reconstructs an enhanced convolution neural network model named AlexNet-L. AlexNet-L is an end-to-end network model for license plate character recognition, which can improve the accuracy of license plate recognition and avoid the problem that the wrong license plate character segmentation affects license plate recognition. The experimental results show that our proposed method can improve the accuracy and performance of license plate recognition.