基于纹理特征和神经网络牌照定位方法

License Plate Locating Method Based on Texture Features and Neural Network

汽车牌照的定位是一个较难解决的图像分割问题,神经网络为此问题的解决提供了一个有力的解决工具。这里提出一种基于纹理特征和神经网络的牌照定位方法,首先采用基于纹理特征的方法对车牌图像进行行定位和列定位,确定车牌候选区域,然后采用基于神经网络的方法对候选区域的特征进行分析判断,确定车牌区域。实验结果表明,该方法可以实现汽车牌照的快速定位,并且定位准确率较高,鲁棒性较好。

A fast license plate locating method is problem to be solved in the field of license plate recognition in the back- ground. A license plate locating method based on texture features and neural network is proposed, Firstly a method based on texture features can be used to determine row location and column location, the candidate regions can be found. Then the features of candidate regions is analyzed by determining the license plate region using method based on neural network. The experimental results show that this method is excellent in the accuracy and the robustness.