摘要
提出一种将车牌纹理和颜色相结合的车牌定位方法,即基于纹理粗定位得到车牌候选区域后,运用改进的自主确定聚类数和聚类中心的RGB空间k-means聚类算法,而不是定义颜色范围来分割车牌。该方法的优越性在于首先利用纹理排除了颜色干扰区域,其次利用颜色聚类去除了纹理干扰区域,又克服了量化定义颜色适应性不强、稳定性差的缺点。实验表明,该方法可以准确定位复杂背景中任意方向和不同光照下的车牌,具有很强的稳定性和鲁棒性。
In this paper we present a vehicle license plate location method which combines the texture and colour of the license plate, that is, after roughly locating based on texture the candidate region of license plate, the modified k-means clustering algorithm in RGB colour space, which allows the number of cluster and the clustering centrcs to be determined autonomously, is used to segment the license plate instead of defining the colour range. The advantages of this method are that first it rules out the colour interference area with texture, then it rules out the texture i,lterference area t, sing colour clustering, and also overcomes the instability and weak adaptability of quantified definition of colour. Experiment demonstrates that this method can accurately detect the license plates in complex background with arbitrary orientations and different illumination. It has strong stability and robustness.
出处
《计算机应用与软件》
CSCD
北大核心
2013年第10期259-262,316,共5页
Computer Applications and Software
关键词
车牌定位
复杂背景
纹理特性
颜色k-means聚类
句法特征
Lieense plate location Complex background Texture eharacteristic, s Colour k-means clustering Syntactic feature
