Abstract: The vehicle logo background, radiator grid, is diverse in shape size and color which lead to the difficulty of vehicle logo location, therefore accurate radiator grid classification is the basis of accurate vehicle logo location. This paper presents a method for the accurate vehicle logo location based on accurate radiator grid classification. Firstly, the vehicle logo can be rough located by the spatial relationship between license plate and vehicle logo, and then based on grid texture features, the use of Hough transform and gradient gray value is determined radiator grid background category, the background can be ablated respectively by different operator. In order to ensure accurate locationinvariety lighting conditions, adaptive binary method in vehicle logo location is fused by dispersion and Ostu method, and combines with morphology to do further processing for radiator grid background. It is suitable to locate vehicle logo in different light intensity and different vehicle logo background. According to the definition of the accuracy of vehicle logo location in this paper, for 10 categories of vehicle logo, 30 categories of radiator grid, a total of 1, 200 images, the overall accuracy of vehicle location can reach 97.67% in this method.

Key words: vehicle logo location, accurate radiator grid classification, adaptive binarization, background ablation

摘要： 由于车标的背景散热器栅格形状大小不一、颜色不定、背景多样，因此导致了车标定位的困难，故精确分类散热器栅格是准确定位车标的基础。提出了一种基于散热器栅格背景精确分类的车标定位方法，首先依照车牌与车标空间位置关系确定车标粗定位，然后依据栅格纹理特征，利用霍夫变换和灰度值的梯度变化确定散热器栅格背景的类别，进而通过不同算子分别对不同种类栅格背景进行背景消融；为了保证多种光照条件下的准确定位，引入离散度，并将其与大津法进行融合，形成一种适用于车标定位的自适应二值化方法，同时结合形态学对栅格背景进一步处理，得到准确的车标定位。这种方法适用于在不同光照强度和不同类型的车标背景条件下，对车标进行定位。对10类车标、30类散热器栅格共1 200张图像进行车标定位，实验结果显示，图像总体的车标定位准确率可以达到97.67%。

关键词: 车标定位, 散热器栅格精确分类, 自适应二值化, 背景消融