针对现有的车牌识别方法的缺陷,改进了车牌识别系统中的多项关键技术;采用了基于图像二维能量与HSI彩色空间相结合的方法进行车牌定位与提取,并对现有的能量算法与彩色图像分割算法做了改进;在识别过程中,引人了特征提取与多级BP神经网络相结合的分类识别方法,对车牌中部分相似字符采用第二级神经网络进行精细识别;通过上述改进,提高了系统的整体性能;实验表明,这些关键技术的改进可以大大提高车牌识别系统的准确率与鲁棒性.%For the defects of the key technology of the existing license plate recognize system, improve the key technology of the system.The techniques of license plate location and extraction adopt the method combining the image 2D energy and the HSI color space. Besides,improve the existing method of energy and image segmentation. For the tilt correction technology, introduce a new method combining the feature extraction and multiply BP neural network learning. Considering the similarity of the characters in the license, propose a two-level neural network to recognize the characters; according to the above improvements, obtain better system performance. Experiment results show that the system with the key technology of the license plate recognition system is more of robustness and more of precision.
展开▼
