基于形状匹配及纹理筛选的汽车轮毂型号识别

为了对轮毂型号进行识别,提出一种基于形状匹配及纹理筛选的轮型识别算法。首先,确定一个轮辐形状为标准模板并得出其边缘图,把模板作为移动窗口在待识别轮毂图片中移动,逐一计算模板到轮毂图片各感兴趣区域(ROI)的最小二维欧氏距离。若此距离小于设定阈值,则判定搜索到一个与模板相同的形状;然后对待识别的轮毂图片进行随机游走,得出游走直方图,通过改进对游走直方图相似度的评价方式,得出纹理偏差度;最后通过对纹理偏差度的比较确认正确的轮型。识别过程具有非接触、灵活、准确的优点,实验表明对于干扰较大图片也具有较高的识别率和较好的鲁棒性。

Wheel model identification based on shape recognition and texture filtering

A wheel model identification algorithm based on Shape Recognition and Texture Filtering is proposed. Firstly, a spokes shape of a standard template is defined, and its edge map is obtained. Taking the template as a window to shift matching at an edge detection figure of the identified image, the minimum distance between template and region of interest of the identified image is calculated. If the minimum distance is less than the specific threshold, a shape consistent with the template is determined. Then, Random Walk Operator is used to traverse the identified image and obtain the Wander Histogram. Comparing this histogram with the Wander Histogram of template wheel, the texture deviation degree of two image is calculated, and the right wheel model is determined. Finally, the experimental results show this algorithm has the advantages of non contact, flexibility and accuracy, and has good robustness for the shape recognition with heavy disturbance.