摘要
本文描述了一种运用统计模式识别方法自动识别多种车型轮毂的方法。构造了四个具有平移不变性、比例不变性和幅度线性变换不变性的一维不变量,用以表征轮毂图像的灰度特征。利用边缘检测Canny算子提取轮毂外圆轮廓,之后对外圆轮廓运用改进的Hough变换方法计算出轮毂最大直径,最后运用马氏距离方法判别出轮毂类别。实验证明了本文方法的有效性。
This paper describes a method for identifying various wheels which are manufactured in the diversified styles and sizes by the statistical pattern recognition algorithms. Based on four one-dimension invariants, translation invariability, scale invariability and extent invariability, which take the gray character of wheel ：images. Our method extracts the exterior circle contours of wheels by the Canny edge detection operator, calculates the max wheel diameters by a developed Hough Transform algorithm, and identifies the wheel styles by the Mahalanobis distance method. Experiments prove the validity of this method.
出处
《微计算机信息》
2010年第4期164-166,共3页
Control & Automation
基金
项目名称:基于机器视觉的高准确度实时轮毂识别系统
中国科学院"优秀博士论文
院长奖获得者科研启动专项基金"资助(07AR210201)
关键词
轮毂
特征提取
识别分类
wheel
feature extraction
identify and classify
