摘要
牌照字符识别是车牌识别系统中关键的一步,而字符识别的关键在于有效特征的选取.小波矩是小波多尺度分析与矩相结合的新的视觉不变量,图像的小波矩特征能很好地反映图像的局部和全局特征,并且具有较强的抗干扰能力.但不同的小波矩离散化方法在性能上有很大的差异.在分析小波矩和矩快速算法的基础上,引入了一种新的小波矩离散化算法用于车牌字符识别系统,以车牌字符图像的小波矩作为特征量,结合改进的BP神经网络实现了车牌字符的识别,获得了很好的识别效果.
License plate character recognition is a main step in license plate recognition system, but its emphasis is to choose efficient features. Wavelet motion is a new visual invariant, it is a integration of moment and wavelet multi-scale analysis. Wavelet moment feature of image can reflect the image's part and whole characters and has strong anti-jamming ability. But different method of discretization of wavelet moment shows different performance. On the basis of analysis of wavelet moment, the paper introduces the algorithm of discretization of wavelet moment, and applies it to license plate character recognition system. With the wavelet moment features and improved BP neural network, the recognition of license plate character is realized and good performance gained.
出处
《浙江工业大学学报》
CAS
2005年第2期170-172,共3页
Journal of Zhejiang University of Technology
基金
浙江省自然科学基金资助项目(M603165)
