机器视觉在物体位姿检测中的应用

研究了机器视觉技术在物体位姿检测中的应用.从两个方位架设工业相机采集目标图片,采用张正友基于2D平面靶标的相机标定方法对相机进行标定.使用目标图像减去背景图像和形态学运算的方法进行粗定位,初步检测目标区域.利用Hough变换检测目标直线边缘,依据极线约束原则、灰度相似性以及直线约束进行边缘直线及其端点的匹配.计算出各直线端点的空间三维坐标及其单位方向向量.在已知目标模板的条件下,目标的几何参数已知,利用简单的几何关系求取目标的位姿数据.

Application of machine vision in detection of objects pose

Application of machine vision technology in detection of objects pose is researched. Set up industrial cameras in two different directions and take pictures of target objectsthen complete camera calibration using Zhang Zhengyou's method based on 2D planar target. Complete preliminary target detection by subtracting background image from the target image and Morphological operation,so the target area is determined. Then use Hough transform to detect line edge,Epipolar constraint,gray similarity and line constraint is adopted to match the lines and their endpoints. Calculate three-dimensional(3D)space coordinates and unit direction vector of endpoint of each straight lines. Since the target template is known,the geometry parameter is available,using simple geometric relationship to calculate pose data of target.