成排连铸坯端面中心坐标视觉自动提取方法

为了实现成排连铸坯端面机器人贴标时各连铸坯端面中心坐标的快速提取，提出了先提取连铸坯端面图像角点像面坐标，再计算各连铸坯端面中心像面坐标的研究方案。首先，提出了一种改进型SUSAN角点检测算法，解决了图像中相邻连铸坯端面图像边界间距离过小和连铸坯端面图像角为弧形角所造成的角点漏检问题；然后，提取角点的像面坐标，并确定各封闭区域所包含连铸坯端面个数；最后，采用一种倾斜连铸坯端面图像中心像面坐标的提取方法，计算各连铸坯端面中心像面坐标。应用以上方法进行成排连铸坯端面机器人贴标实验，实验结果表明，连铸坯端面水平和竖直方向贴标位置误差范围分别为-0.625～0.850 mm和-0.550～0.875 mm，完全满足企业对多根连铸坯端面自动贴标位置误差允许值1 mm的要求。所提视觉方法在理论和实际应用上都是可行的,不仅为连铸坯端面贴标机器人提供了准确的贴标位置，而且为矩形图像元素的中心坐标提取提供了一种可靠的方法，具有一定的应用价值。

Center-point coordinate vision automatic extraction method for row billet end face

In order to realize rapid extraction of the center coordinates for row billet end faces when labeled by multi-strand billet end face labeling robot, a research program is proposed that the image coordinates in billet end face image are extracted and then the center point''s coordinates of the billets end face image are calculated. Firstly, an improved SUSAN corner detection algorithm is proposed to solve the corner leak detection problem caused by the distance between the adjacent billet image edge being too small and the billet image corner angle being the arc angle. Then, the image plane coordinates of corners point are extracted, and the number of the billets end faces image in each enclosed area is calculated. Finally, a method for extracting the coordinates of slant billet end faces is provided, and the center point image plane coordinates of each billet end face are calculated, respectively. The robot labeling experiment of each billet end face in a row billet is carried out, and experimental results indicate that the horizontal and vertical labeling position error ranges of each billet end face are -0.625~0.850 mm and -0.550~0.875 mm, respectively, meeting the enterprises'' allowed center coordinates position error value 1 mm of the billet end face fully. The visual methods mentioned in this paper are feasible in both theoretical and practical applications, and it not only provides an accurate labeling position for the continuous blank end labeling robot, but also provides a reliable method for extracting the center coordinates of rectangular image elements, and has some application value.