摘要
自主研发了一套用于储罐内角焊缝焊接的线结构光视觉引导系统。引入光平面建立了视觉传感器的三维数学模型,并通过定义双参考坐标系对光平面进行标定得到精确的像素坐标与世界坐标的三维转换矩阵。在图像处理上则采用两步定位的方法,第一步,使用基于相关性模板匹配法对焊缝进行匹配,匹配ROI仅为原图的10.66%,可以大幅减少系统处理时间,提高效率;第二步,通过改进的Canny算子自适应提取复杂环境中的焊缝特征边缘并拟合直线,得到精确的二维像素坐标,最后结合三维转换矩阵将其转化为三维世界坐标实现定位。通过焊接测试,该视觉引导系统的纠偏偏差小于0.12 mm,平均处理纠偏时间小于130 ms,具有较高的精度和实时性,可以满足实际的焊接要求。
A line structured light vision guidance system for fillet welding in storage tanks was independently developed.The 3D mathematical model of vision sensor was established by introducing line structured light plane.The light plane is calibrated by defining a double reference coordinate system to obtain a precise three-dimensional transformation matrix of pixel coordinates and world coordinates.The two step positioning method was employed in the image processing section.In the first step,the correlation template matching method was used to locate the weld seam,the matching ROI is only 10.66%of the original image,which can significantly reduce system processing time and improve efficiency;In the second step,the improved Canny operator adaptively extracts the edge of the weld feature in the complex environment and fits the line to obtain accurate two-dimensional pixel coordinates.Finally,it was transformed into a three-dimensional world coordinate in combination with a three-dimensional transformation matrix.By testing the actual welding process,the deviation correction of the vision guidance system is less than 0.12mm,and the average processing correction time is less than 130ms.It has high precision and real-time performance,which can meet the actual welding requirements.
作者
陈思豪
王中任
肖光润
刘德政
刘海生
CHEN Si-hao;WANG Zhong-ren;Xiao Guang-run;Liu De-zheng;Liu Hai-sheng(School of Mechanical Engineering,Hubei University of Arts & Science,Xiangyang 441053,China;School of Automobile and Traffic Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)
出处
《激光与红外》
CAS
CSCD
北大核心
2019年第5期543-548,共6页
Laser & Infrared
基金
"机电汽车"湖北省优势特色学科群开放基金项目(No.XKQ2018009)资助
关键词
视觉引导
数学模型
图像处理
焊缝定位
vision guidance
mathematical model
image processing
weld positioning
