Abstract: A color model between RGB and HSV according to the vehicle license plate (VLP) feature is studied to select a few weights depending on each component contribution so as to facilitate the detection VLP, the VLP edge image can be detected using vertical edge operator, a threshold is designed to complete binary processing, at last the VLP location can be executed accurately by projection method. The experiments on 200 automobiles RGB real color images that were taken under various conditions from traffic stations show the satisfied accuracy and robustness.

Key words: weight, vertical edge operator, threshold, HSV color space

摘要： 考虑车牌本身特征，选择HSV和RGB颜色模型，并对每个分量的贡献值进行分析研究，选取一种有利于车牌提取的颜色空间及一组空间分量的权重值，利用设计的垂直方向算子得到车牌边缘信息，并设计一种阈值选取方法进行二值化处理，最后使用投影积分方法进行车牌细定位。通过各种情况下在卡口获取的200幅车辆实测图像进行测试，车牌定位的准确性和鲁棒性都较好。

关键词: 权重, 垂直边缘算子, 阈值, HSV颜色模型