基于形状上下文的车牌字符识别

License Plate Character Recognition Based on Shape Context

针对车牌字符识别问题,该文提出一种基于形状上下文的识别算法,对分割好的车牌字符进行识别。对分割字符进行边缘检测,提取部分边缘点作为样本点,计算每个边缘点的形状上下文描述算子;再对分割字符和字符模版进行相似度比较,得到相应的对数极坐标图,计算相似度是否达到预设的阈值来判定分割字符是否可识别,并判别结果。实验表明,此算法在二维目标的不变性中,具有精确的匹配效果。

For plate character recognition problem,this paper proposes a kind of recognition algorithm based on shape context.First,choosing samples of the character edge as,and calculating its shape context descriptor,which is represented by log-polar coordinates histogram.Then calculating similarity cost between the image and character template,to decide whether the image is recognizable or not,if it is recognizable,then judging which character it is.Experiments show that this algorithm subject to the constraint that the matching be one-to-one,has the advantage of shape invariance,which can achieve an accurate matching effect.