中文摘要：

结合线阵CCD扫描二维图像工作原理，分析了玻璃厚度检测图像特点。精确提取光条纹图像中心是提高玻璃厚度测量精度的关键，提出一种改进的灰度重心法进行中心提取。先用Canny算子进行边缘检测，确定目标区域，再用加权灰度重心法提取玻璃厚度图像条纹中心。该算法可较好地抑制背景噪声，中心定位可达亚像素级，定位精度提高了0．2像素。基于上述算法，通过计算中心坐标差可得到玻璃厚度值。实验结果表明，该算法正确有效，玻璃厚度测量系统分辨率可达1μm，系统测量误差小于0．1％，实现了高精度、高稳定性的玻璃厚度在线检测。

英文摘要：

Associates with the working principles of scanning two-dimensional image using linear array CCD, characteristics of glass thickness image are analyzed. For the key point of improving precision of thickness measurement is to acquire light stripe centre precisely, a new improved barycentre method is put forward. First, Canny operator is employed to detect the image edge to define the target area. Second, weighted gray barycentre algorithm is used to obtain the centre location of glass thickness stripe image. This method can suppress the background noise and extract the center point with a sub-pixel precision, and the accuracy of center allocation can be up by 0.2 pixels. After centre allocation, thickness values are achieved by calculating the center coordinates difference. The feasibility and the validity of this method have been proven. The resolution of the measurement system can be 1μm and the measurement error is less than 0. 1%. The image processing enables an on-line glass thickness measurement with high precision and high stability.