一种用于钣金零件视觉测量的图像拼接方法

An Image Mosaic Method for Vision Measurement of Sheet Metal Parts

摘要:

在钣金零件视觉测量中，大尺寸零件超出图像采集摄像机的视场范围，不能一次性获得该零件的完整图像，需要将存在一定重叠区域的图像进行图像拼接。针对特征稀少、灰度信息少、轮廓对称的钣金零件图像不易拼接问题进行了分析和研究，提出了一种在待拼接图像中选择包含重叠部分的等大小区域作为待配准区域，并对其提取轮廓，然后对轮廓图利用相位相关法求待配准区域的配准参数，再根据待配准区域图像与待拼接图像的坐标关系转换求待拼接图像的配准参数并对图像进行配准，实现钣金零件图像拼接的方法。通过实验表明，该拼接方法对特征稀少、灰度信息少和轮廓对称的钣金零件图像具有速度快、较好的准确性和抗干扰能力，能够满足工业应用的要求。

Abstract:

In measurement of machine vision on the sheet metal parts, a large-scale part is beyond the scope of camera field of view. It is difficult to get the part’s panoramic image at one time, therefore, image mosaic is needed to join several images that have some overlap areas into a complete image.Focused on the problem that sheet metal part images with a few features, less gray level feature and symmetrical outline information are hard to mosaic, an image mosaic method is proposed in this paper. Firstly, equal overlap areas are selected from part images as registration area, and then the contours are extracted.Secondly, registration parameters of registration area are acquired by using phase correlation on the contour images. Finally, image registration is completed by utilizing the final registration parameters that can be acquired by caculating coordinate?transforming?relations between overlap images and complete images.The sheet metal part images are spliced according to the registration parameters.Experimental result shows that this image mosaic method has the advantages of fast speed, good accuracy and high anti-reference ability for the sheet metal part with a few features, less gray level feature and symmetrical outline information. The proposed algorithm can meet the requirements of industrial applications.