基于双目视觉的工件位姿检测方法

Workpiece Pose Detection Method Based on Binocular Vision

建立了一种基于双目视觉的空间位姿检测模型。采集目标物不同角度的图片以建立目标物离线模板库,通过SURF算法检索模板库以分离目标物所在区域。在该区域利用霍夫变换提取目标物的特征角点。利用基于灰度相关函数的匹配算法完成特征角点的匹配,结合双目视觉模型建立特征角点像素坐标与用户坐标间的映射关系进而求出目标物在用户坐标系下的位姿信息。以检测箱体工件为例进行多组实验,结果表明,该方法可以准确地检测目标物位姿。

A spatial pose detection model based on binocular vision was established.Pictures of the target from different angles were taken to build an offline template library for the target,and the template library was retrieved through the SURF algorithm to separate the area where the target is located.In this area,Hough transform was used to extract the characteristic corner points of the target.The matching algorithm based on gray correlation function was used to complete the matching of feature corners,and the binocular vision model was used to establish the mapping relationship between the pixel coordinates of the feature corners and the user coordinates,and then the posture information of the target object in the user coordinate system was obtained.Taking the detection of box parts as an example,multiple experiments were conducted.The results show that the method can accurately detect the target position and posture.