基于图像拼接的微小零件视觉检测方法

Vision Measurement Method Based on Image Mosaicing for Micro-part

针对毫米级微小尺寸零件的高准确度检测问题,提出了一种基于图像拼接技术的视觉检测方法,即利用亚像素微小位移结合相位相关的自动拼接方法。以两幅图像的拼接为例,这种方法首先对第一幅图像利用三次样条插值,进行亚像素微小位移;然后用处理后的第一幅图像与第二幅图像进行相位相关处理,计算两者的相关功率谱函数分布图;再根据分布图中功率谱最大值求出两幅图像的亚像素配准关系;最后根据配准关系把两幅图像拼接起来。实验表明,使用这种拼接方法,准确度可达到亚象素级,并具有抗噪声、易实现自动化等特点。

In order to increase the detecting accuracy of micro-parts, a vision detecting method lying on the image mosaicing has been brought forward in this paper, which is based on subpixel displacement interpolation and phase correlation. Two images have been taken as a sample. Cubic interpolation has been used to displace one image at first; cross power spectrum of the two images has been computed later; subpixel registration has been found out then, based on the maximum data of cross power spectrum distributing function ; image interpolation is used to end the job. The experiment shows that： high accuracy, anti-noise and automatization can be achieved by this method.