基于轮廓的物体识别与定位方法

Object Identification and Orientation Based on Contours

给出一种通过物体轮廓匹配来识别物体和对物体姿态进行定位的方法。该方法基于提取物体的轮廓信息,然后通过与库存物体的轮廓信息进行匹配来认识物体的类型,最后与库存物体模型的外轮廓进行逐点匹配,获取累计偏差和,通过二分法逐渐逼近最佳相对转角,使累计偏差和变小,从而使当前物体与库存物体模型的相对转角达到较高的精度。本方法可用于生产装配线上的零件识别,具有适应性好,定位准确的特点。

In this paper,a method of identifying object and its attitude by matching contours is presented.First,Contour information of the object is extracted.Then the type of object is identified by matching contour information with stock object models'.Finally its outer contour is matched with the outer contour of the identified stock object model point by point,and the total sum of differences is obtained.The optimal relative rotation angle is approached by dichotomy that makes sum of differences less so that the preciseness of the relative rotation angle of current object with the identified stock object model is improved.The method can be used in identifying parts on assembly line with good adaptability and exactness.