摘要
针对龙门式等离子零件坡口切割平台视觉系统对待切割零件识别速度慢、轮廓定位信息不准以及误匹配等问题,结合Hu矩和图像配准提出一种高效率的匹配和定位方法。首先通过阈值分割算法对图像进行降噪处理,分离零件背景与前景图,对零件轮廓进行识别,基于Hu矩实现对图像库的快速匹配;其次对读入的CAD图像进行坐标变换,通过定义轮廓最小外包斜矩形的角度,基于轮廓质心寻找到一个旋转角度初值并建立函数模型,利用遗传算法的思想优化迭代最优解;最后用CAD模板图像的轮廓路径,修正零件轮廓定位信息。实验结果表明该方法能够快速且准确地从零件图中搜索到与CAD图库中的对应零件,并实现定位信息修正,修正后的整体图像误差在6个像素左右。
The visual system of gantry ion parts groove cutting platform is slow to recognize the cutting parts,inaccurate contour positioning information and mismatching.An efficient matching and positioning method combining Hu moment and image registration is presented.Firstly,the threshold segmentation algorithm is used to reduce the noise of the picture,separate the background and foreground of the part,identify the outline of the part,and realize the rapid matching of the graphic library based on Hu moment.Secondly,coordinate transformation is carried out on the read CAD graphics,and an initial value of rotation Angle is found based on the contour centroid,and a function model is established.The iterative optimal solution is optimized using the idea of genetic algorithm.Finally,the contour path of the CAD template is used to correct the positioning information of the contour of the part.The experimental results show that the proposed method can quickly and accurately search the corresponding parts from the CAD library and realize the positioning information correction,and the overall image error after correction is about 6 pixels.
作者
赵明明
李文强
张永贵
ZHAO Mingming;LI Wenqiang;ZHANG Yonggui(Lanzhou University of Technology,Lanzhou 730050,China;CRRC Qishuyan Rolling Stock Technology Research Institute Co.,Ltd.,Changzhou 213011,China)
出处
《现代制造工程》
CSCD
北大核心
2024年第8期118-125,共8页
Modern Manufacturing Engineering
关键词
定位
图像匹配
图像配准
零件识别
精度
positioning
pattern matching
graphic registration
part identification
precision
