基于BP神经网络的车牌号码识别

Research on the License Plate Recognition Based on BP Neural Network Algorithm

车牌号码的提取与识别在现代智能交通系统中起着非常重要的作用。采用两个主要步骤对车牌进行处理:首先是图像预处理部分,包括车牌图像的定位与提取、彩色图像灰度化、图像倾斜校正、图像二值化、字符分割、尺寸归一化和紧缩重排;其次是特征提取与识别部分,提取字符中9区域像素数和双横纵像素数共13个特征,然后交由已经离线训练好的三层13-8-7的BP神经网络进行识别。在MATLAB中得以模拟实现,取得良好的识别效果。

The feature extraction and recognition of the license plate plays a very important role in the modern intelligent transportation system. Uses two main steps to deal with the license plate： the first is the image preprocessing part, including the location and extraction of license plate image, grayscale, tilt correction, binarization, character segmentation, size normalization and rearrangement. The second is the fea-ture extraction and recognition part, gets the number of pixels of 9 regions and Double horizontal and vertical lines for the three-layer13-8-7 BP neural network that is trained offline to recognize the characters of the license plate. We achieve good recognition results in the MATLAB simulation environment.