中文摘要：

在瓷砖视觉检测中,提出了一种新的定位分割方法,该方法可对任意角度的瓷砖目标进行定位分割,为后续的缺陷检测及机器人抓取提供了高精度的处理模块与定位基准。首先通过预处理消除采集图像噪声,然后根据瓷砖自身的灰阶梯度,对其进行轮廓提取,并以轮廓斜外接矩形面积及矩形度双重约束去除复杂背景的干扰。再通过轮廓掩膜获取目标区域图像,设置ROI（Regions of Interest,感兴趣区域）使目标区域从检测背景中分离出来成为一个独立单元。最后采用斜矩形对角线与标准位置的几何偏差作为矫正的反馈信息,以矫正轮廓质心作为定位基准。

英文摘要：

A new method of locating and segmentation is proposed in the visual inspection of ceramic tile, which can be used to locate and segment the object from any angle, as it can provide a high precision processing module and position reference for the subsequent defect detection and robot grasping. Firstly through the pre treatment to eliminate the image noise, we can extract its contour according to the gray level gradient of the tile and remove the interference of complex background base on area and rectangle degree of the external contour. And then acquiring target image through the contour mask, the ROI（Regions of Interest） where target regions lie is set to separate them from the background to become an independent image unit. Finally, setting rectangular diagonal and geometric deviations with the standard location as correction feedback and also correcting the contour centroid as the locating datum.